diff --git a/CMakeLists.txt b/CMakeLists.txt index 8336bc975e3..329dd30bb57 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -213,15 +213,6 @@ if( WIN32 AND NOT CYGWIN ) set(LLVM_LIT_TOOLS_DIR "" CACHE PATH "Path to GnuWin32 tools") endif() -# On Win32 using MS tools, provide an option to set the number of parallel jobs -# to use. -if( MSVC_IDE AND ( MSVC90 OR MSVC10 ) ) - # Only Visual Studio 2008 and 2010 officially supports /MP. Visual Studio - # 2005 supports it but it is experimental. - set(LLVM_COMPILER_JOBS "0" CACHE STRING - "Number of parallel compiler jobs. 0 means use all processors. Default is 0.") -endif() - # Define options to control the inclusion and default build behavior for # components which may not strictly be necessary (tools, runtime, examples, and # tests). @@ -396,7 +387,6 @@ add_subdirectory(utils/FileUpdate) add_subdirectory(utils/count) add_subdirectory(utils/not) add_subdirectory(utils/llvm-lit) -add_subdirectory(utils/json-bench) add_subdirectory(utils/yaml-bench) add_subdirectory(projects) diff --git a/CREDITS.TXT b/CREDITS.TXT index ef471b0887e..bf32a4c565b 100644 --- a/CREDITS.TXT +++ b/CREDITS.TXT @@ -50,9 +50,15 @@ N: Cameron Buschardt E: buschard@uiuc.edu D: The `mem2reg' pass - promotes values stored in memory to registers +N: Brendon Cahoon +E: bcahoon@codeaurora.org +D: Loop unrolling with run-time trip counts. + N: Chandler Carruth E: chandlerc@gmail.com -D: LinkTimeOptimizer for Linux, via binutils integration, and C API +D: Hashing algorithms and interfaces +D: Inline cost analysis +D: Machine block placement pass N: Casey Carter E: ccarter@uiuc.edu @@ -210,6 +216,10 @@ N: Benjamin Kramer E: benny.kra@gmail.com D: Miscellaneous bug fixes +N: Sundeep Kushwaha +E: sundeepk@codeaurora.org +D: Implemented DFA-based target independent VLIW packetizer + N: Christopher Lamb E: christopher.lamb@gmail.com D: aligned load/store support, parts of noalias and restrict support @@ -245,6 +255,10 @@ N: Nick Lewycky E: nicholas@mxc.ca D: PredicateSimplifier pass +N: Tony Linthicum, et. al. +E: tlinth@codeaurora.org +D: Backend for Qualcomm's Hexagon VLIW processor. + N: Bruno Cardoso Lopes E: bruno.cardoso@gmail.com W: http://www.brunocardoso.org @@ -271,6 +285,10 @@ N: Scott Michel E: scottm@aero.org D: Added STI Cell SPU backend. +N: Kai Nacke +E: kai@redstar.de +D: Support for implicit TLS model used with MS VC runtime + N: Takumi Nakamura E: geek4civic@gmail.com E: chapuni@hf.rim.or.jp diff --git a/autoconf/configure.ac b/autoconf/configure.ac index 0a2c8b69ddd..51f89217a9f 100644 --- a/autoconf/configure.ac +++ b/autoconf/configure.ac @@ -838,6 +838,13 @@ AC_ARG_WITH(gcc-toolchain, AC_DEFINE_UNQUOTED(GCC_INSTALL_PREFIX,"$withval", [Directory where gcc is installed.]) +AC_ARG_WITH(default-sysroot, + AS_HELP_STRING([--with-default-sysroot], + [Add --sysroot= to all compiler invocations.]),, + withval="") +AC_DEFINE_UNQUOTED(DEFAULT_SYSROOT,"$withval", + [Default to all compiler invocations for --sysroot=.]) + dnl Allow linking of LLVM with GPLv3 binutils code. AC_ARG_WITH(binutils-include, AS_HELP_STRING([--with-binutils-include], diff --git a/cmake/modules/HandleLLVMOptions.cmake b/cmake/modules/HandleLLVMOptions.cmake index 3a10a861d67..d10e59a0972 100644 --- a/cmake/modules/HandleLLVMOptions.cmake +++ b/cmake/modules/HandleLLVMOptions.cmake @@ -110,9 +110,9 @@ if( CMAKE_SIZEOF_VOID_P EQUAL 8 AND NOT WIN32 ) endif( LLVM_BUILD_32_BITS ) endif( CMAKE_SIZEOF_VOID_P EQUAL 8 AND NOT WIN32 ) -if( MSVC_IDE AND ( MSVC90 OR MSVC10 ) ) - # Only Visual Studio 2008 and 2010 officially supports /MP. - # Visual Studio 2005 do support it but it's experimental there. +# On Win32 using MS tools, provide an option to set the number of parallel jobs +# to use. +if( MSVC_IDE ) set(LLVM_COMPILER_JOBS "0" CACHE STRING "Number of parallel compiler jobs. 0 means use all processors. Default is 0.") if( NOT LLVM_COMPILER_JOBS STREQUAL "1" ) diff --git a/configure b/configure index e87160d75c4..efa80dbeb95 100755 --- a/configure +++ b/configure @@ -1442,6 +1442,7 @@ Optional Packages: --with-c-include-dirs Colon separated list of directories clang will search for headers --with-gcc-toolchain Directory where gcc is installed. + --with-default-sysroot Add --sysroot= to all compiler invocations. --with-binutils-include Specify path to binutils/include/ containing plugin-api.h file for gold plugin. --with-bug-report-url Specify the URL where bug reports should be @@ -3802,7 +3803,7 @@ else llvm_cv_target_os_type="Darwin" ;; *-*-minix*) llvm_cv_target_os_type="Minix" ;; - *-*-freebsd*| *-*-kfreebsd-gnu) + *-*-freebsd* | *-*-kfreebsd-gnu) llvm_cv_target_os_type="FreeBSD" ;; *-*-openbsd*) llvm_cv_target_os_type="OpenBSD" ;; @@ -5583,6 +5584,20 @@ _ACEOF +# Check whether --with-default-sysroot was given. +if test "${with_default_sysroot+set}" = set; then + withval=$with_default_sysroot; +else + withval="" +fi + + +cat >>confdefs.h <<_ACEOF +#define DEFAULT_SYSROOT "$withval" +_ACEOF + + + # Check whether --with-binutils-include was given. if test "${with_binutils_include+set}" = set; then withval=$with_binutils_include; @@ -10386,7 +10401,7 @@ else lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2 lt_status=$lt_dlunknown cat > conftest.$ac_ext <GenRegisterNames.inc file for that architecture. For instance, by - inspecting lib/Target/X86/X86GenRegisterNames.inc we see that the - 32-bit register EAX is denoted by 15, and the MMX register - MM0 is mapped to 48.

+ inspecting lib/Target/X86/X86GenRegisterInfo.inc we see that the + 32-bit register EAX is denoted by 43, and the MMX register + MM0 is mapped to 65.

Some architectures contain registers that share the same physical location. A notable example is the X86 platform. For instance, in the X86 architecture, @@ -1627,7 +1627,7 @@ def : Pat<(i32 imm:$imm), bits. These physical registers are marked as aliased in LLVM. Given a particular architecture, you can check which registers are aliased by inspecting its RegisterInfo.td file. Moreover, the method - TargetRegisterInfo::getAliasSet(p_reg) returns an array containing + MCRegisterInfo::getAliasSet(p_reg) returns an array containing all the physical registers aliased to the register p_reg.

Physical registers, in LLVM, are grouped in Register Classes. @@ -3182,7 +3182,7 @@ MOVSX32rm16 -> movsx, 32-bit register, 16-bit memory Chris Lattner
The LLVM Compiler Infrastructure
- Last modified: $Date: 2012-03-27 13:25:16 +0200 (Tue, 27 Mar 2012) $ + Last modified: $Date: 2012-04-15 22:22:36 +0200 (Sun, 15 Apr 2012) $ diff --git a/docs/CommandGuide/FileCheck.pod b/docs/CommandGuide/FileCheck.pod index dbd626c7cb7..2662cc01288 100644 --- a/docs/CommandGuide/FileCheck.pod +++ b/docs/CommandGuide/FileCheck.pod @@ -67,20 +67,20 @@ This syntax says to pipe the current file ("%s") into llvm-as, pipe that into llc, then pipe the output of llc into FileCheck. This means that FileCheck will be verifying its standard input (the llc output) against the filename argument specified (the original .ll file specified by "%s"). To see how this works, -lets look at the rest of the .ll file (after the RUN line): +let's look at the rest of the .ll file (after the RUN line): define void @sub1(i32* %p, i32 %v) { entry: - ; CHECK: sub1: - ; CHECK: subl + ; CHECK: sub1: + ; CHECK: subl %0 = tail call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %p, i32 %v) ret void } define void @inc4(i64* %p) { entry: - ; CHECK: inc4: - ; CHECK: incq + ; CHECK: inc4: + ; CHECK: incq %0 = tail call i64 @llvm.atomic.load.add.i64.p0i64(i64* %p, i64 1) ret void } @@ -111,18 +111,18 @@ driven from one .ll file. This is useful in many circumstances, for example, testing different architectural variants with llc. Here's a simple example: ; RUN: llvm-as < %s | llc -mtriple=i686-apple-darwin9 -mattr=sse41 \ - ; RUN: | FileCheck %s -check-prefix=X32 + ; RUN: | FileCheck %s -check-prefix=X32> ; RUN: llvm-as < %s | llc -mtriple=x86_64-apple-darwin9 -mattr=sse41 \ - ; RUN: | FileCheck %s -check-prefix=X64 + ; RUN: | FileCheck %s -check-prefix=X64> define <4 x i32> @pinsrd_1(i32 %s, <4 x i32> %tmp) nounwind { %tmp1 = insertelement <4 x i32>; %tmp, i32 %s, i32 1 ret <4 x i32> %tmp1 - ; X32: pinsrd_1: - ; X32: pinsrd $1, 4(%esp), %xmm0 + ; X32: pinsrd_1: + ; X32: pinsrd $1, 4(%esp), %xmm0 - ; X64: pinsrd_1: - ; X64: pinsrd $1, %edi, %xmm0 + ; X64: pinsrd_1: + ; X64: pinsrd $1, %edi, %xmm0 } In this case, we're testing that we get the expected code generation with @@ -147,13 +147,13 @@ example, something like this works as you'd expect: store <2 x double> %tmp9, <2 x double>* %r, align 16 ret void - ; CHECK: t2: - ; CHECK: movl 8(%esp), %eax - ; CHECK-NEXT: movapd (%eax), %xmm0 - ; CHECK-NEXT: movhpd 12(%esp), %xmm0 - ; CHECK-NEXT: movl 4(%esp), %eax - ; CHECK-NEXT: movapd %xmm0, (%eax) - ; CHECK-NEXT: ret + ; CHECK: t2: + ; CHECK: movl 8(%esp), %eax + ; CHECK-NEXT: movapd (%eax), %xmm0 + ; CHECK-NEXT: movhpd 12(%esp), %xmm0 + ; CHECK-NEXT: movl 4(%esp), %eax + ; CHECK-NEXT: movapd %xmm0, (%eax) + ; CHECK-NEXT: ret } CHECK-NEXT: directives reject the input unless there is exactly one newline @@ -177,9 +177,9 @@ can be used: %A = load i8* %P3 ret i8 %A - ; CHECK: @coerce_offset0 - ; CHECK-NOT: load - ; CHECK: ret i8 + ; CHECK: @coerce_offset0 + ; CHECK-NOT: load + ; CHECK: ret i8 } @@ -195,7 +195,7 @@ matching for a majority of what we do, FileCheck has been designed to support mixing and matching fixed string matching with regular expressions. This allows you to write things like this: - ; CHECK: movhpd {{[0-9]+}}(%esp), {{%xmm[0-7]}} + ; CHECK: movhpd {{[0-9]+}}(%esp), {{%xmm[0-7]}} In this case, any offset from the ESP register will be allowed, and any xmm register will be allowed. @@ -217,20 +217,20 @@ allows named variables to be defined and substituted into patterns. Here is a simple example: ; CHECK: test5: - ; CHECK: notw [[REGISTER:%[a-z]+]] - ; CHECK: andw {{.*}}[[REGISTER]] + ; CHECK: notw [[REGISTER:%[a-z]+]] + ; CHECK: andw {{.*}}[REGISTER]] -The first check line matches a regex (%[a-z]+) and captures it into -the variables "REGISTER". The second line verifies that whatever is in REGISTER +The first check line matches a regex (B<%[a-z]+>) and captures it into +the variable "REGISTER". The second line verifies that whatever is in REGISTER occurs later in the file after an "andw". FileCheck variable references are -always contained in [[ ]] pairs, are named, and their names can be -formed with the regex "[a-zA-Z_][a-zA-Z0-9_]*". If a colon follows the +always contained in B<[[ ]]> pairs, are named, and their names can be +formed with the regex "B<[a-zA-Z_][a-zA-Z0-9_]*>". If a colon follows the name, then it is a definition of the variable, if not, it is a use. FileCheck variables can be defined multiple times, and uses always get the latest value. Note that variables are all read at the start of a "CHECK" line and are all defined at the end. This means that if you have something like -"CHECK: [[XYZ:.*]]x[[XYZ]]" that the check line will read the previous +"B", the check line will read the previous value of the XYZ variable and define a new one after the match is performed. If you need to do something like this you can probably take advantage of the fact that FileCheck is not actually line-oriented when it matches, this allows you to diff --git a/docs/DebuggingJITedCode.html b/docs/DebuggingJITedCode.html index 1946fdd9e08..ffb4cd96b81 100644 --- a/docs/DebuggingJITedCode.html +++ b/docs/DebuggingJITedCode.html @@ -8,93 +8,16 @@ -

Debugging JITed Code With GDB

+

Debugging JIT-ed Code With GDB

    -
  1. Example usage
  2. Background
    3. +
  3. GDB Version
    4. +
  4. Debugging MCJIT-ed code
    5. +
-
Written by Reid Kleckner
- - -

Example usage

- -
- -

In order to debug code JITed by LLVM, you need GDB 7.0 or newer, which is -available on most modern distributions of Linux. The version of GDB that Apple -ships with XCode has been frozen at 6.3 for a while. LLDB may be a better -option for debugging JITed code on Mac OS X. -

- -

Consider debugging the following code compiled with clang and run through -lli: -

- -
-#include <stdio.h>
-
-void foo() {
-    printf("%d\n", *(int*)NULL);  // Crash here
-}
-
-void bar() {
-    foo();
-}
-
-void baz() {
-    bar();
-}
-
-int main(int argc, char **argv) {
-    baz();
-}
-
- -

Here are the commands to run that application under GDB and print the stack -trace at the crash: -

- -
-# Compile foo.c to bitcode.  You can use either clang or llvm-gcc with this
-# command line.  Both require -fexceptions, or the calls are all marked
-# 'nounwind' which disables DWARF exception handling info.  Custom frontends
-# should avoid adding this attribute to JITed code, since it interferes with
-# DWARF CFA generation at the moment.
-$ clang foo.c -fexceptions -emit-llvm -c -o foo.bc
-
-# Run foo.bc under lli with -jit-emit-debug.  If you built lli in debug mode,
-# -jit-emit-debug defaults to true.
-$ $GDB_INSTALL/gdb --args lli -jit-emit-debug foo.bc
-...
-
-# Run the code.
-(gdb) run
-Starting program: /tmp/gdb/lli -jit-emit-debug foo.bc
-[Thread debugging using libthread_db enabled]
-
-Program received signal SIGSEGV, Segmentation fault.
-0x00007ffff7f55164 in foo ()
-
-# Print the backtrace, this time with symbols instead of ??.
-(gdb) bt
-#0  0x00007ffff7f55164 in foo ()
-#1  0x00007ffff7f550f9 in bar ()
-#2  0x00007ffff7f55099 in baz ()
-#3  0x00007ffff7f5502a in main ()
-#4  0x00000000007c0225 in llvm::JIT::runFunction(llvm::Function*,
-    std::vector<llvm::GenericValue,
-    std::allocator<llvm::GenericValue> > const&) ()
-#5  0x00000000007d6d98 in
-    llvm::ExecutionEngine::runFunctionAsMain(llvm::Function*,
-    std::vector<std::string,
-    std::allocator<std::string> > const&, char const* const*) ()
-#6  0x00000000004dab76 in main ()
-
- -

As you can see, GDB can correctly unwind the stack and has the appropriate -function names. -

-
+
Written by Reid Kleckner and Eli Bendersky

Background

@@ -107,37 +30,144 @@ debug information from the object file of the code, but for JITed code, there is no such file to look for.

-

Depending on the architecture, this can impact the debugging experience in -different ways. For example, on most 32-bit x86 architectures, you can simply -compile with -fno-omit-frame-pointer for GCC and -disable-fp-elim for LLVM. -When GDB creates a backtrace, it can properly unwind the stack, but the stack -frames owned by JITed code have ??'s instead of the appropriate symbol name. -However, on Linux x86_64 in particular, GDB relies on the DWARF call frame -address (CFA) debug information to unwind the stack, so even if you compile -your program to leave the frame pointer untouched, GDB will usually be unable -to unwind the stack past any JITed code stack frames. -

-

In order to communicate the necessary debug info to GDB, an interface for registering JITed code with debuggers has been designed and implemented for -GDB and LLVM. At a high level, whenever LLVM generates new machine code, it -also generates an object file in memory containing the debug information. LLVM -then adds the object file to the global list of object files and calls a special -function (__jit_debug_register_code) marked noinline that GDB knows about. When +GDB and LLVM MCJIT. At a high level, whenever MCJIT generates new machine code, +it does so in an in-memory object file that contains the debug information in +DWARF format. MCJIT then adds this in-memory object file to a global list of +dynamically generated object files and calls a special function +(__jit_debug_register_code) marked noinline that GDB knows about. When GDB attaches to a process, it puts a breakpoint in this function and loads all -of the object files in the global list. When LLVM calls the registration +of the object files in the global list. When MCJIT calls the registration function, GDB catches the breakpoint signal, loads the new object file from -LLVM's memory, and resumes the execution. In this way, GDB can get the +the inferior's memory, and resumes the execution. In this way, GDB can get the necessary debug information.

- -

At the time of this writing, LLVM only supports architectures that use ELF -object files and it only generates symbols and DWARF CFA information. However, -it would be easy to add more information to the object file, so we don't need to -coordinate with GDB to get better debug information. -

+ +

GDB Version

+ + +

In order to debug code JIT-ed by LLVM, you need GDB 7.0 or newer, which is +available on most modern distributions of Linux. The version of GDB that Apple +ships with XCode has been frozen at 6.3 for a while. LLDB may be a better +option for debugging JIT-ed code on Mac OS X. +

+ + + +

Debugging MCJIT-ed code

+ +
+ +

The emerging MCJIT component of LLVM allows full debugging of JIT-ed code with +GDB. This is due to MCJIT's ability to use the MC emitter to provide full +DWARF debugging information to GDB.

+ +

Note that lli has to be passed the -use-mcjit flag to JIT the code +with MCJIT instead of the old JIT.

+ +

Example

+ +
+ +

Consider the following C code (with line numbers added to make the example +easier to follow):

+ +
+1   int compute_factorial(int n)
+2   {
+3       if (n <= 1)
+4           return 1;
+5
+6       int f = n;
+7       while (--n > 1) 
+8           f *= n;
+9       return f;
+10  }
+11
+12
+13  int main(int argc, char** argv)
+14  {
+15      if (argc < 2)
+16          return -1;
+17      char firstletter = argv[1][0];
+18      int result = compute_factorial(firstletter - '0');
+19  
+20      // Returned result is clipped at 255...
+21      return result;
+22  }
+
+ +

Here is a sample command line session that shows how to build and run this +code via lli inside GDB: +

+ +
+$ $BINPATH/clang -cc1 -O0 -g -emit-llvm showdebug.c
+$ gdb --quiet --args $BINPATH/lli -use-mcjit showdebug.ll 5
+Reading symbols from $BINPATH/lli...done.
+(gdb) b showdebug.c:6
+No source file named showdebug.c.
+Make breakpoint pending on future shared library load? (y or [n]) y
+Breakpoint 1 (showdebug.c:6) pending.
+(gdb) r
+Starting program: $BINPATH/lli -use-mcjit showdebug.ll 5
+[Thread debugging using libthread_db enabled]
+
+Breakpoint 1, compute_factorial (n=5) at showdebug.c:6
+6	    int f = n;
+(gdb) p n
+$1 = 5
+(gdb) p f
+$2 = 0
+(gdb) n
+7	    while (--n > 1) 
+(gdb) p f
+$3 = 5
+(gdb) b showdebug.c:9
+Breakpoint 2 at 0x7ffff7ed404c: file showdebug.c, line 9.
+(gdb) c
+Continuing.
+
+Breakpoint 2, compute_factorial (n=1) at showdebug.c:9
+9	    return f;
+(gdb) p f
+$4 = 120
+(gdb) bt
+#0  compute_factorial (n=1) at showdebug.c:9
+#1  0x00007ffff7ed40a9 in main (argc=2, argv=0x16677e0) at showdebug.c:18
+#2  0x3500000001652748 in ?? ()
+#3  0x00000000016677e0 in ?? ()
+#4  0x0000000000000002 in ?? ()
+#5  0x0000000000d953b3 in llvm::MCJIT::runFunction (this=0x16151f0, F=0x1603020, ArgValues=...) at /home/ebenders_test/llvm_svn_rw/lib/ExecutionEngine/MCJIT/MCJIT.cpp:161
+#6  0x0000000000dc8872 in llvm::ExecutionEngine::runFunctionAsMain (this=0x16151f0, Fn=0x1603020, argv=..., envp=0x7fffffffe040)
+    at /home/ebenders_test/llvm_svn_rw/lib/ExecutionEngine/ExecutionEngine.cpp:397
+#7  0x000000000059c583 in main (argc=4, argv=0x7fffffffe018, envp=0x7fffffffe040) at /home/ebenders_test/llvm_svn_rw/tools/lli/lli.cpp:324
+(gdb) finish
+Run till exit from #0  compute_factorial (n=1) at showdebug.c:9
+0x00007ffff7ed40a9 in main (argc=2, argv=0x16677e0) at showdebug.c:18
+18	    int result = compute_factorial(firstletter - '0');
+Value returned is $5 = 120
+(gdb) p result
+$6 = 23406408
+(gdb) n
+21	    return result;
+(gdb) p result
+$7 = 120
+(gdb) c
+Continuing.
+
+Program exited with code 0170.
+(gdb) 
+
+
+ +
+
+ +
@@ -145,9 +175,10 @@ coordinate with GDB to get better debug information. src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"> Valid HTML 4.01 - Reid Kleckner
+ Reid Kleckner, + Eli Bendersky
The LLVM Compiler Infrastructure
- Last modified: $Date: 2011-10-31 12:21:59 +0100 (Mon, 31 Oct 2011) $ + Last modified: $Date: 2012-05-01 09:58:54 +0200 (Tue, 01 May 2012) $
diff --git a/docs/LLVMBuild.html b/docs/LLVMBuild.html index cce607d611f..f39a8a6a1c0 100644 --- a/docs/LLVMBuild.html +++ b/docs/LLVMBuild.html @@ -101,7 +101,7 @@ $ROOT of project trees for things which can be checked out separately. -->

As mentioned earlier, LLVM projects are organized into - logical components. Every component is typically grouped into it's + logical components. Every component is typically grouped into its own subdirectory. Generally, a component is organized around a coherent group of sources which have some kind of clear API separation from other parts of the code.

@@ -190,7 +190,7 @@ required_libraries = Archive BitReader Core Support TransformUtils number in order for files with multiple components ("component_0", "component_1", and so on).

-

Section names not matches this format (or the "common" section) are +

Section names not matching this format (or the "common" section) are currently unused and are disallowed.

Every component is defined by the properties in the section. The exact list @@ -212,7 +212,7 @@ required_libraries = Archive BitReader Core Support TransformUtils

  • parent [required]

    The name of the logical parent of the component. Components are organized into a logical tree to make it easier to navigate and organize - groups of components. The parent's have no semantics as far as the project + groups of components. The parents have no semantics as far as the project build is concerned, however. Typically, the parent will be the main component of the parent directory.

    @@ -263,7 +263,7 @@ required_libraries = Archive BitReader Core Support TransformUtils

    If given, a list of the names of Library or LibraryGroup components which must also be linked in whenever this library is used. That is, the link time dependencies for this component. When tools are built, - the build system will include the transitive closer of + the build system will include the transitive closure of all required_libraries for the components the tool needs.

  • add_to_library_groups [optional] diff --git a/docs/LangRef.html b/docs/LangRef.html index d1c6e58c31c..5261230efba 100644 --- a/docs/LangRef.html +++ b/docs/LangRef.html @@ -103,7 +103,7 @@
  • Metadata Nodes and Metadata Strings
    1. 'tbaa' Metadata
      2. -
    2. 'fpaccuracy' Metadata
      3. +
    3. 'fpmath' Metadata
    4. 'range' Metadata
    • @@ -3000,16 +3000,16 @@ call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)

    - 'fpaccuracy' Metadata + 'fpmath' Metadata

    -

    fpaccuracy metadata may be attached to any instruction of floating - point type. It expresses the maximum relative error allowed in the result - of that instruction, in ULPs, thus potentially allowing the compiler to use - a more efficient but less accurate method of computing it. - ULP is defined as follows:

    +

    fpmath metadata may be attached to any instruction of floating point + type. It can be used to express the maximum acceptable error in the result of + that instruction, in ULPs, thus potentially allowing the compiler to use a + more efficient but less accurate method of computing it. ULP is defined as + follows:

    @@ -3021,13 +3021,12 @@ call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
    -

    The metadata node shall consist of a single non-negative floating - point number representing the maximum relative error. For example, - 2.5 ULP:

    +

    The metadata node shall consist of a single positive floating point number + representing the maximum relative error, for example:

    -!0 = metadata !{ float 2.5 }
+!0 = metadata !{ float 2.5 } ; maximum acceptable inaccuracy is 2.5 ULPs
    @@ -8506,7 +8505,7 @@ LLVM.

    Chris Lattner
    The LLVM Compiler Infrastructure
    - Last modified: $Date: 2012-04-10 10:22:43 +0200 (Tue, 10 Apr 2012) $ + Last modified: $Date: 2012-04-16 21:39:33 +0200 (Mon, 16 Apr 2012) $ diff --git a/docs/ProgrammersManual.html b/docs/ProgrammersManual.html index 951e2d5766d..625ef9ae841 100644 --- a/docs/ProgrammersManual.html +++ b/docs/ProgrammersManual.html @@ -1734,6 +1734,9 @@ already in the table, and each pair in the map is store in a single allocation

    StringMap also provides query methods that take byte ranges, so it only ever copies a string if a value is inserted into the table.

    + +

    StringMap iteratation order, however, is not guaranteed to be deterministic, +so any uses which require that should instead use a std::map.

    @@ -4125,7 +4128,7 @@ arguments. An argument has a pointer to the parent Function.

    Dinakar Dhurjati and Chris Lattner
    The LLVM Compiler Infrastructure
    - Last modified: $Date: 2012-03-27 13:25:16 +0200 (Tue, 27 Mar 2012) $ + Last modified: $Date: 2012-04-18 22:28:55 +0200 (Wed, 18 Apr 2012) $ diff --git a/docs/ReleaseNotes.html b/docs/ReleaseNotes.html index bcac2939531..3b5c2e65a9d 100644 --- a/docs/ReleaseNotes.html +++ b/docs/ReleaseNotes.html @@ -206,6 +206,36 @@ Release Notes.
    + + +

    +Polly: Polyhedral Optimizer +

    + +
    + +

    Polly is an experimental + optimizer for data locality and parallelism. It currently provides high-level + loop optimizations and automatic parallelisation (using the OpenMP run time). + Work in the area of automatic SIMD and accelerator code generation was + started. + +

    Within the LLVM 3.1 time-frame there were the following highlights:

    + +
      +
    • Polly became an official LLVM project
      • +
    • Polly can be loaded directly into clang (Enabled by '-O3 -mllvm -polly' + )
      • +
    • An automatic scheduling optimizer (derived from Pluto) was integrated. It + performs loop transformations to optimize for data-locality and parallelism. + The transformations include, but are not limited to interchange, fusion, + fission, skewing and tiling. +
      • +
    + +
    + @@ -269,6 +299,8 @@ Release Notes. Support to model instruction bundling / packing.
  • ARM Integrated Assembler, A full featured assembler and direct-to-object support for ARM.
    • +
  • Basic Block Placement + Probability driven basic block placement.
  • ....
    • @@ -292,6 +324,9 @@ Release Notes. module as a whole to LLVM subsystems.
  • Loads can now have range metadata attached to them to describe the possible values being loaded.
    • +
  • Inline cost heuristics have been completely overhauled and now closely + model constant propagation through call sites, disregard trivially dead + code costs, and can model C++ STL iterator patterns.
  • ....
    • @@ -308,6 +343,14 @@ Release Notes. optimizers:

      +
    • The loop unroll pass now is able to unroll loops with run-time trip counts. + This feature is turned off by default, and is enabled with the + -unroll-runtime flag.
      • +
    • A new basic-block autovectorization pass is available. Pass + -vectorize to run this pass along with some associated + post-vectorization cleanup passes. For more information, see the EuroLLVM + 2012 slides: + Autovectorization with LLVM.
    • ....
    @@ -373,6 +416,19 @@ Release Notes. target's schedule description which can be queried to determine legal groupings of instructions in a bundle.

    +

    We have added a new target independent VLIW packetizer based on the + DFA infrastructure to group machine instructions into bundles.

    + + + +

    +Basic Block Placement +

    +
    +

    A probability based block placement and code layout algorithm was added to +LLVM's code generator. This layout pass supports probabilities derived from +static heuristics as well as source code annotations such as +__builtin_expect.

    @@ -391,6 +447,7 @@ Release Notes. times and better support for different calling conventions. The old WINCALL instructions are no longer needed.
  • DW2 Exception Handling is enabled on Cygwin and MinGW.
    • +
  • Support for implicit TLS model used with MS VC runtime
    • @@ -451,6 +508,8 @@ syntax, there are still significant gaps in that support.

    +

    Support for Qualcomm's Hexagon VLIW processor has been added.

    +
    • ....
      • @@ -524,6 +583,8 @@ syntax, there are still significant gaps in that support.

    • llvm::getTrapFunctionName()
    • llvm::EnableSegmentedStacks
    +
  • The MDBuilder class has been added to simplify the creation of + metadata.
  • ....
    • @@ -621,7 +682,7 @@ syntax, there are still significant gaps in that support.

    src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"> LLVM Compiler Infrastructure
    - Last modified: $Date: 2012-04-12 17:17:35 +0200 (Thu, 12 Apr 2012) $ + Last modified: $Date: 2012-04-17 03:13:53 +0200 (Tue, 17 Apr 2012) $ diff --git a/docs/TestingGuide.html b/docs/TestingGuide.html index 805ae776a8d..33ce793b5b2 100644 --- a/docs/TestingGuide.html +++ b/docs/TestingGuide.html @@ -299,15 +299,15 @@ clang/test directory.

    you.

    In order for the regression tests to work, each directory of tests must - have a dg.exp file. Lit looks for this file to determine how to - run the tests. This file is just a Tcl script and it can do anything you want, + have a lit.local.cfg file. Lit looks for this file to determine how + to run the tests. This file is just Python code and thus is very flexible, but we've standardized it for the LLVM regression tests. If you're adding a - directory of tests, just copy dg.exp from another directory to get - running. The standard dg.exp simply loads a Tcl library - (test/lib/llvm.exp) and calls the llvm_runtests function - defined in that library with a list of file names to run. The names are - obtained by using Tcl's glob command. Any directory that contains only - directories does not need the dg.exp file.

    + directory of tests, just copy lit.local.cfg from another directory to + get running. The standard lit.local.cfg simply specifies which files + to look in for tests. Any directory that contains only directories does not + need the lit.local.cfg file. Read the + Lit documentation for more + information.

    The llvm-runtests function looks at each file that is passed to it and gathers any lines together that match "RUN:". These are the "RUN" lines @@ -473,7 +473,7 @@ negatives).

    llc, then pipe the output of llc into FileCheck. This means that FileCheck will be verifying its standard input (the llc output) against the filename argument specified (the original .ll file specified by "%s"). To see how this works, -lets look at the rest of the .ll file (after the RUN line):

    +let's look at the rest of the .ll file (after the RUN line):

    @@ -900,7 +900,7 @@ the Test Suite Makefile Guide.
    
 
   John T. Criswell, Daniel Dunbar, Reid Spencer, and Tanya Lattner
    
   The LLVM Compiler Infrastructure
    
-  Last modified: $Date: 2012-03-27 13:25:16 +0200 (Tue, 27 Mar 2012) $
+  Last modified: $Date: 2012-04-18 10:02:25 +0200 (Wed, 18 Apr 2012) $
 
 
 
diff --git a/include/llvm-c/lto.h b/include/llvm-c/lto.h
index 5d9cecbc515..f43d365e3db 100644
--- a/include/llvm-c/lto.h
+++ b/include/llvm-c/lto.h
@@ -250,12 +250,6 @@ extern void
 lto_codegen_set_assembler_args(lto_code_gen_t cg, const char **args,
                                int nargs);
 
-/**
- * Enables the internalize pass during LTO optimizations.
- */
-extern void
-lto_codegen_set_whole_program_optimization(lto_code_gen_t cg);
-
 /**
  * Adds to a list of all global symbols that must exist in the final
  * generated code.  If a function is not listed, it might be
diff --git a/include/llvm/ADT/SmallPtrSet.h b/include/llvm/ADT/SmallPtrSet.h
index 70693d5b9aa..498a0345d8b 100644
--- a/include/llvm/ADT/SmallPtrSet.h
+++ b/include/llvm/ADT/SmallPtrSet.h
@@ -126,9 +126,6 @@ protected:
 private:
   bool isSmall() const { return CurArray == SmallArray; }
 
-  unsigned Hash(const void *Ptr) const {
-    return static_cast(((uintptr_t)Ptr >> 4) & (CurArraySize-1));
-  }
   const void * const *FindBucketFor(const void *Ptr) const;
   void shrink_and_clear();
 
diff --git a/include/llvm/ADT/StringMap.h b/include/llvm/ADT/StringMap.h
index 097418efc81..b4497a276d0 100644
--- a/include/llvm/ADT/StringMap.h
+++ b/include/llvm/ADT/StringMap.h
@@ -239,7 +239,7 @@ public:
   explicit StringMap(AllocatorTy A)
     : StringMapImpl(static_cast(sizeof(MapEntryTy))), Allocator(A) {}
 
-  explicit StringMap(const StringMap &RHS)
+  StringMap(const StringMap &RHS)
     : StringMapImpl(static_cast(sizeof(MapEntryTy))) {
     assert(RHS.empty() &&
            "Copy ctor from non-empty stringmap not implemented yet!");
diff --git a/include/llvm/CodeGen/DFAPacketizer.h b/include/llvm/CodeGen/DFAPacketizer.h
index 2d2db78144a..ee1ed0779be 100644
--- a/include/llvm/CodeGen/DFAPacketizer.h
+++ b/include/llvm/CodeGen/DFAPacketizer.h
@@ -28,7 +28,6 @@
 
 #include "llvm/CodeGen/MachineBasicBlock.h"
 #include "llvm/ADT/DenseMap.h"
-#include 
 
 namespace llvm {
 
@@ -37,7 +36,7 @@ class MachineInstr;
 class MachineLoopInfo;
 class MachineDominatorTree;
 class InstrItineraryData;
-class DefaultVLIWScheduler;
+class ScheduleDAGInstrs;
 class SUnit;
 
 class DFAPacketizer {
@@ -78,8 +77,6 @@ public:
   // reserveResources - Reserve the resources occupied by a machine
   // instruction and change the current state to reflect that change.
   void reserveResources(llvm::MachineInstr *MI);
-
-  const InstrItineraryData *getInstrItins() const { return InstrItins; }
 };
 
 // VLIWPacketizerList - Implements a simple VLIW packetizer using DFA. The
@@ -90,21 +87,20 @@ public:
 // and machine resource is marked as taken. If any dependency is found, a target
 // API call is made to prune the dependence.
 class VLIWPacketizerList {
-protected:
   const TargetMachine &TM;
   const MachineFunction &MF;
   const TargetInstrInfo *TII;
 
-  // The VLIW Scheduler.
-  DefaultVLIWScheduler *VLIWScheduler;
+  // Encapsulate data types not exposed to the target interface.
+  ScheduleDAGInstrs *SchedulerImpl;
 
+protected:
   // Vector of instructions assigned to the current packet.
   std::vector CurrentPacketMIs;
   // DFA resource tracker.
   DFAPacketizer *ResourceTracker;
-
-  // Generate MI -> SU map.
-  std::map MIToSUnit;
+  // Scheduling units.
+  std::vector SUnits;
 
 public:
   VLIWPacketizerList(
@@ -122,32 +118,17 @@ public:
   DFAPacketizer *getResourceTracker() {return ResourceTracker;}
 
   // addToPacket - Add MI to the current packet.
-  virtual MachineBasicBlock::iterator addToPacket(MachineInstr *MI) {
-    MachineBasicBlock::iterator MII = MI;
-    CurrentPacketMIs.push_back(MI);
-    ResourceTracker->reserveResources(MI);
-    return MII;
-  }
+  void addToPacket(MachineInstr *MI);
 
   // endPacket - End the current packet.
-  void endPacket(MachineBasicBlock *MBB, MachineInstr *MI);
-
-  // initPacketizerState - perform initialization before packetizing
-  // an instruction. This function is supposed to be overrided by
-  // the target dependent packetizer.
-  virtual void initPacketizerState(void) { return; }
+  void endPacket(MachineBasicBlock *MBB, MachineInstr *I);
 
   // ignorePseudoInstruction - Ignore bundling of pseudo instructions.
-  virtual bool ignorePseudoInstruction(MachineInstr *I,
-                                       MachineBasicBlock *MBB) {
-    return false;
-  }
+  bool ignorePseudoInstruction(MachineInstr *I, MachineBasicBlock *MBB);
 
-  // isSoloInstruction - return true if instruction MI can not be packetized
-  // with any other instruction, which means that MI itself is a packet.
-  virtual bool isSoloInstruction(MachineInstr *MI) {
-    return true;
-  }
+  // isSoloInstruction - return true if instruction I must end previous
+  // packet.
+  bool isSoloInstruction(MachineInstr *I);
 
   // isLegalToPacketizeTogether - Is it legal to packetize SUI and SUJ
   // together.
@@ -160,7 +141,6 @@ public:
   virtual bool isLegalToPruneDependencies(SUnit *SUI, SUnit *SUJ) {
     return false;
   }
-
 };
 }
 
diff --git a/include/llvm/CodeGen/Passes.h b/include/llvm/CodeGen/Passes.h
index 3b381998588..e76fe992572 100644
--- a/include/llvm/CodeGen/Passes.h
+++ b/include/llvm/CodeGen/Passes.h
@@ -56,7 +56,7 @@ public:
 
 protected:
   TargetMachine *TM;
-  PassManagerBase &PM;
+  PassManagerBase *PM;
   PassConfigImpl *Impl; // Internal data structures
   bool Initialized;     // Flagged after all passes are configured.
 
diff --git a/include/llvm/CodeGen/ScheduleDAGInstrs.h b/include/llvm/CodeGen/ScheduleDAGInstrs.h
index c8de7bc8f89..4fee108cd2b 100644
--- a/include/llvm/CodeGen/ScheduleDAGInstrs.h
+++ b/include/llvm/CodeGen/ScheduleDAGInstrs.h
@@ -181,6 +181,13 @@ namespace llvm {
     /// the def-side latency only.
     bool UnitLatencies;
 
+    /// The standard DAG builder does not normally include terminators as DAG
+    /// nodes because it does not create the necessary dependencies to prevent
+    /// reordering. A specialized scheduler can overide
+    /// TargetInstrInfo::isSchedulingBoundary then enable this flag to indicate
+    /// it has taken responsibility for scheduling the terminator correctly.
+    bool CanHandleTerminators;
+
     /// State specific to the current scheduling region.
     /// ------------------------------------------------
 
diff --git a/include/llvm/CodeGen/SlotIndexes.h b/include/llvm/CodeGen/SlotIndexes.h
index d868cb8dade..0457e43e6b7 100644
--- a/include/llvm/CodeGen/SlotIndexes.h
+++ b/include/llvm/CodeGen/SlotIndexes.h
@@ -23,6 +23,7 @@
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineFunctionPass.h"
 #include "llvm/ADT/PointerIntPair.h"
+#include "llvm/ADT/ilist.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/Support/Allocator.h"
@@ -33,8 +34,7 @@ namespace llvm {
   /// SlotIndexes pass. It should not be used directly. See the
   /// SlotIndex & SlotIndexes classes for the public interface to this
   /// information.
-  class IndexListEntry {
-    IndexListEntry *next, *prev;
+  class IndexListEntry : public ilist_node {
     MachineInstr *mi;
     unsigned index;
 
@@ -51,31 +51,26 @@ namespace llvm {
     void setIndex(unsigned index) {
       this->index = index;
     }
-    
-    IndexListEntry* getNext() { return next; }
-    const IndexListEntry* getNext() const { return next; }
-    void setNext(IndexListEntry *next) {
-      this->next = next;
-    }
 
-    IndexListEntry* getPrev() { return prev; }
-    const IndexListEntry* getPrev() const { return prev; }
-    void setPrev(IndexListEntry *prev) {
-      this->prev = prev;
-    }
   };
 
-  // Specialize PointerLikeTypeTraits for IndexListEntry.
   template <>
-  class PointerLikeTypeTraits { 
+  struct ilist_traits : public ilist_default_traits {
+  private:
+    mutable ilist_half_node Sentinel;
   public:
-    static inline void* getAsVoidPointer(IndexListEntry *p) {
-      return p;
+    IndexListEntry *createSentinel() const {
+      return static_cast(&Sentinel);
     }
-    static inline IndexListEntry* getFromVoidPointer(void *p) {
-      return static_cast(p);
-    }
-    enum { NumLowBitsAvailable = 3 };
+    void destroySentinel(IndexListEntry *) const {}
+
+    IndexListEntry *provideInitialHead() const { return createSentinel(); }
+    IndexListEntry *ensureHead(IndexListEntry*) const { return createSentinel(); }
+    static void noteHead(IndexListEntry*, IndexListEntry*) {}
+    void deleteNode(IndexListEntry *N) {}
+
+  private:
+    void createNode(const IndexListEntry &);
   };
 
   /// SlotIndex - An opaque wrapper around machine indexes.
@@ -112,13 +107,13 @@ namespace llvm {
     SlotIndex(IndexListEntry *entry, unsigned slot)
       : lie(entry, slot) {}
 
-    IndexListEntry& entry() const {
+    IndexListEntry* listEntry() const {
       assert(isValid() && "Attempt to compare reserved index.");
-      return *lie.getPointer();
+      return lie.getPointer();
     }
 
     int getIndex() const {
-      return entry().getIndex() | getSlot();
+      return listEntry()->getIndex() | getSlot();
     }
 
     /// Returns the slot for this SlotIndex.
@@ -150,8 +145,7 @@ namespace llvm {
     SlotIndex() : lie(0, 0) {}
 
     // Construct a new slot index from the given one, and set the slot.
-    SlotIndex(const SlotIndex &li, Slot s)
-      : lie(&li.entry(), unsigned(s)) {
+    SlotIndex(const SlotIndex &li, Slot s) : lie(li.listEntry(), unsigned(s)) {
       assert(lie.getPointer() != 0 &&
              "Attempt to construct index with 0 pointer.");
     }
@@ -179,7 +173,7 @@ namespace llvm {
     bool operator!=(SlotIndex other) const {
       return lie != other.lie;
     }
-   
+
     /// Compare two SlotIndex objects. Return true if the first index
     /// is strictly lower than the second.
     bool operator<(SlotIndex other) const {
@@ -211,7 +205,7 @@ namespace llvm {
     /// isEarlierInstr - Return true if A refers to an instruction earlier than
     /// B. This is equivalent to A < B && !isSameInstr(A, B).
     static bool isEarlierInstr(SlotIndex A, SlotIndex B) {
-      return A.entry().getIndex() < B.entry().getIndex();
+      return A.listEntry()->getIndex() < B.listEntry()->getIndex();
     }
 
     /// Return the distance from this index to the given one.
@@ -236,25 +230,25 @@ namespace llvm {
     /// is the one associated with the Slot_Block slot for the instruction
     /// pointed to by this index.
     SlotIndex getBaseIndex() const {
-      return SlotIndex(&entry(), Slot_Block);
+      return SlotIndex(listEntry(), Slot_Block);
     }
 
     /// Returns the boundary index for associated with this index. The boundary
     /// index is the one associated with the Slot_Block slot for the instruction
     /// pointed to by this index.
     SlotIndex getBoundaryIndex() const {
-      return SlotIndex(&entry(), Slot_Dead);
+      return SlotIndex(listEntry(), Slot_Dead);
     }
 
     /// Returns the register use/def slot in the current instruction for a
     /// normal or early-clobber def.
     SlotIndex getRegSlot(bool EC = false) const {
-      return SlotIndex(&entry(), EC ? Slot_EarlyClobber : Slot_Register);
+      return SlotIndex(listEntry(), EC ? Slot_EarlyClobber : Slot_Register);
     }
 
     /// Returns the dead def kill slot for the current instruction.
     SlotIndex getDeadSlot() const {
-      return SlotIndex(&entry(), Slot_Dead);
+      return SlotIndex(listEntry(), Slot_Dead);
     }
 
     /// Returns the next slot in the index list. This could be either the
@@ -266,15 +260,15 @@ namespace llvm {
     SlotIndex getNextSlot() const {
       Slot s = getSlot();
       if (s == Slot_Dead) {
-        return SlotIndex(entry().getNext(), Slot_Block);
+        return SlotIndex(listEntry()->getNextNode(), Slot_Block);
       }
-      return SlotIndex(&entry(), s + 1);
+      return SlotIndex(listEntry(), s + 1);
     }
 
     /// Returns the next index. This is the index corresponding to the this
     /// index's slot, but for the next instruction.
     SlotIndex getNextIndex() const {
-      return SlotIndex(entry().getNext(), getSlot());
+      return SlotIndex(listEntry()->getNextNode(), getSlot());
     }
 
     /// Returns the previous slot in the index list. This could be either the
@@ -286,15 +280,15 @@ namespace llvm {
     SlotIndex getPrevSlot() const {
       Slot s = getSlot();
       if (s == Slot_Block) {
-        return SlotIndex(entry().getPrev(), Slot_Dead);
+        return SlotIndex(listEntry()->getPrevNode(), Slot_Dead);
       }
-      return SlotIndex(&entry(), s - 1);
+      return SlotIndex(listEntry(), s - 1);
     }
 
     /// Returns the previous index. This is the index corresponding to this
     /// index's slot, but for the previous instruction.
     SlotIndex getPrevIndex() const {
-      return SlotIndex(entry().getPrev(), getSlot());
+      return SlotIndex(listEntry()->getPrevNode(), getSlot());
     }
 
   };
@@ -315,7 +309,7 @@ namespace llvm {
       return (LHS == RHS);
     }
   };
-  
+
   template <> struct isPodLike { static const bool value = true; };
 
 
@@ -346,8 +340,10 @@ namespace llvm {
   class SlotIndexes : public MachineFunctionPass {
   private:
 
+    typedef ilist IndexList;
+    IndexList indexList;
+
     MachineFunction *mf;
-    IndexListEntry *indexListHead;
     unsigned functionSize;
 
     typedef DenseMap Mi2IndexMap;
@@ -374,84 +370,18 @@ namespace llvm {
       return entry;
     }
 
-    void initList() {
-      assert(indexListHead == 0 && "Zero entry non-null at initialisation.");
-      indexListHead = createEntry(0, ~0U);
-      indexListHead->setNext(0);
-      indexListHead->setPrev(indexListHead);
-    }
-
-    void clearList() {
-      indexListHead = 0;
-      ileAllocator.Reset();
-    }
-
-    IndexListEntry* getTail() {
-      assert(indexListHead != 0 && "Call to getTail on uninitialized list.");
-      return indexListHead->getPrev();
-    }
-
-    const IndexListEntry* getTail() const {
-      assert(indexListHead != 0 && "Call to getTail on uninitialized list.");
-      return indexListHead->getPrev();
-    }
-
-    // Returns true if the index list is empty.
-    bool empty() const { return (indexListHead == getTail()); }
-
-    IndexListEntry* front() {
-      assert(!empty() && "front() called on empty index list.");
-      return indexListHead;
-    }
-
-    const IndexListEntry* front() const {
-      assert(!empty() && "front() called on empty index list.");
-      return indexListHead;
-    }
-
-    IndexListEntry* back() {
-      assert(!empty() && "back() called on empty index list.");
-      return getTail()->getPrev();
-    }
-
-    const IndexListEntry* back() const {
-      assert(!empty() && "back() called on empty index list.");
-      return getTail()->getPrev();
-    }
-
-    /// Insert a new entry before itr.
-    void insert(IndexListEntry *itr, IndexListEntry *val) {
-      assert(itr != 0 && "itr should not be null.");
-      IndexListEntry *prev = itr->getPrev();
-      val->setNext(itr);
-      val->setPrev(prev);
-      
-      if (itr != indexListHead) {
-        prev->setNext(val);
-      }
-      else {
-        indexListHead = val;
-      }
-      itr->setPrev(val);
-    }
-
-    /// Push a new entry on to the end of the list.
-    void push_back(IndexListEntry *val) {
-      insert(getTail(), val);
-    }
-
-    /// Renumber locally after inserting newEntry.
-    void renumberIndexes(IndexListEntry *newEntry);
+    /// Renumber locally after inserting curItr.
+    void renumberIndexes(IndexList::iterator curItr);
 
   public:
     static char ID;
 
-    SlotIndexes() : MachineFunctionPass(ID), indexListHead(0) {
+    SlotIndexes() : MachineFunctionPass(ID) {
       initializeSlotIndexesPass(*PassRegistry::getPassRegistry());
     }
 
     virtual void getAnalysisUsage(AnalysisUsage &au) const;
-    virtual void releaseMemory(); 
+    virtual void releaseMemory();
 
     virtual bool runOnMachineFunction(MachineFunction &fn);
 
@@ -463,22 +393,21 @@ namespace llvm {
 
     /// Returns the zero index for this analysis.
     SlotIndex getZeroIndex() {
-      assert(front()->getIndex() == 0 && "First index is not 0?");
-      return SlotIndex(front(), 0);
+      assert(indexList.front().getIndex() == 0 && "First index is not 0?");
+      return SlotIndex(&indexList.front(), 0);
     }
 
     /// Returns the base index of the last slot in this analysis.
     SlotIndex getLastIndex() {
-      return SlotIndex(back(), 0);
+      return SlotIndex(&indexList.back(), 0);
     }
 
     /// Returns the distance between the highest and lowest indexes allocated
     /// so far.
     unsigned getIndexesLength() const {
-      assert(front()->getIndex() == 0 &&
+      assert(indexList.front().getIndex() == 0 &&
              "Initial index isn't zero?");
-
-      return back()->getIndex();
+      return indexList.back().getIndex();
     }
 
     /// Returns the number of instructions in the function.
@@ -503,19 +432,15 @@ namespace llvm {
     /// Returns the instruction for the given index, or null if the given
     /// index has no instruction associated with it.
     MachineInstr* getInstructionFromIndex(SlotIndex index) const {
-      return index.isValid() ? index.entry().getInstr() : 0;
+      return index.isValid() ? index.listEntry()->getInstr() : 0;
     }
 
     /// Returns the next non-null index.
     SlotIndex getNextNonNullIndex(SlotIndex index) {
-      SlotIndex nextNonNull = index.getNextIndex();
-
-      while (&nextNonNull.entry() != getTail() &&
-             getInstructionFromIndex(nextNonNull) == 0) {
-        nextNonNull = nextNonNull.getNextIndex();
-      }
-
-      return nextNonNull;
+      IndexList::iterator itr(index.listEntry());
+      ++itr;
+      while (itr != indexList.end() && itr->getInstr() == 0) { ++itr; }
+      return SlotIndex(itr, index.getSlot());
     }
 
     /// getIndexBefore - Returns the index of the last indexed instruction
@@ -659,31 +584,31 @@ namespace llvm {
       assert(mi->getParent() != 0 && "Instr must be added to function.");
 
       // Get the entries where mi should be inserted.
-      IndexListEntry *prevEntry, *nextEntry;
+      IndexList::iterator prevItr, nextItr;
       if (Late) {
         // Insert mi's index immediately before the following instruction.
-        nextEntry = &getIndexAfter(mi).entry();
-        prevEntry = nextEntry->getPrev();
+        nextItr = getIndexAfter(mi).listEntry();
+        prevItr = prior(nextItr);
       } else {
         // Insert mi's index immediately after the preceeding instruction.
-        prevEntry = &getIndexBefore(mi).entry();
-        nextEntry = prevEntry->getNext();
+        prevItr = getIndexBefore(mi).listEntry();
+        nextItr = llvm::next(prevItr);
       }
 
       // Get a number for the new instr, or 0 if there's no room currently.
       // In the latter case we'll force a renumber later.
-      unsigned dist = ((nextEntry->getIndex() - prevEntry->getIndex())/2) & ~3u;
-      unsigned newNumber = prevEntry->getIndex() + dist;
+      unsigned dist = ((nextItr->getIndex() - prevItr->getIndex())/2) & ~3u;
+      unsigned newNumber = prevItr->getIndex() + dist;
 
       // Insert a new list entry for mi.
-      IndexListEntry *newEntry = createEntry(mi, newNumber);
-      insert(nextEntry, newEntry);
+      IndexList::iterator newItr =
+        indexList.insert(nextItr, createEntry(mi, newNumber));
 
       // Renumber locally if we need to.
       if (dist == 0)
-        renumberIndexes(newEntry);
+        renumberIndexes(newItr);
 
-      SlotIndex newIndex(newEntry, SlotIndex::Slot_Block);
+      SlotIndex newIndex(&*newItr, SlotIndex::Slot_Block);
       mi2iMap.insert(std::make_pair(mi, newIndex));
       return newIndex;
     }
@@ -694,7 +619,7 @@ namespace llvm {
       // MachineInstr -> index mappings
       Mi2IndexMap::iterator mi2iItr = mi2iMap.find(mi);
       if (mi2iItr != mi2iMap.end()) {
-        IndexListEntry *miEntry(&mi2iItr->second.entry());        
+        IndexListEntry *miEntry(mi2iItr->second.listEntry());
         assert(miEntry->getInstr() == mi && "Instruction indexes broken.");
         // FIXME: Eventually we want to actually delete these indexes.
         miEntry->setInstr(0);
@@ -709,7 +634,7 @@ namespace llvm {
       if (mi2iItr == mi2iMap.end())
         return;
       SlotIndex replaceBaseIndex = mi2iItr->second;
-      IndexListEntry *miEntry(&replaceBaseIndex.entry());
+      IndexListEntry *miEntry(replaceBaseIndex.listEntry());
       assert(miEntry->getInstr() == mi &&
              "Mismatched instruction in index tables.");
       miEntry->setInstr(newMI);
@@ -726,13 +651,13 @@ namespace llvm {
       IndexListEntry *nextEntry = 0;
 
       if (nextMBB == mbb->getParent()->end()) {
-        nextEntry = getTail();
+        nextEntry = indexList.end();
       } else {
-        nextEntry = &getMBBStartIdx(nextMBB).entry();
+        nextEntry = getMBBStartIdx(nextMBB).listEntry();
       }
 
-      insert(nextEntry, startEntry);
-      insert(nextEntry, stopEntry);
+      indexList.insert(nextEntry, startEntry);
+      indexList.insert(nextEntry, stopEntry);
 
       SlotIndex startIdx(startEntry, SlotIndex::Slot_Block);
       SlotIndex endIdx(nextEntry, SlotIndex::Slot_Block);
@@ -766,4 +691,4 @@ namespace llvm {
 
 }
 
-#endif // LLVM_CODEGEN_LIVEINDEX_H 
+#endif // LLVM_CODEGEN_SLOTINDEXES_H
diff --git a/include/llvm/Config/config.h.cmake b/include/llvm/Config/config.h.cmake
index 69e358031eb..95c4d6cac79 100644
--- a/include/llvm/Config/config.h.cmake
+++ b/include/llvm/Config/config.h.cmake
@@ -11,9 +11,6 @@
 /* Relative directory for resource files */
 #define CLANG_RESOURCE_DIR "${CLANG_RESOURCE_DIR}"
 
-/* Directory wherelibstdc++ is installed. */
-#define GCC_INSTALL_PREFIX "${GCC_INSTALL_PREFIX}"
-
 /* Directories clang will search for headers */
 #define C_INCLUDE_DIRS "${C_INCLUDE_DIRS}"
 
diff --git a/include/llvm/Config/config.h.in b/include/llvm/Config/config.h.in
index ccff7da9671..677bf2e40b8 100644
--- a/include/llvm/Config/config.h.in
+++ b/include/llvm/Config/config.h.in
@@ -12,6 +12,9 @@
 /* Directories clang will search for headers */
 #undef C_INCLUDE_DIRS
 
+/* Default  to all compiler invocations for --sysroot=. */
+#undef DEFAULT_SYSROOT
+
 /* Define if position independent code is enabled */
 #undef ENABLE_PIC
 
diff --git a/include/llvm/IntrinsicsX86.td b/include/llvm/IntrinsicsX86.td
index a6fda4a3afc..cb7b3eadc87 100644
--- a/include/llvm/IntrinsicsX86.td
+++ b/include/llvm/IntrinsicsX86.td
@@ -1091,20 +1091,6 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
         GCCBuiltin<"__builtin_ia32_vperm2f128_si256">,
         Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty,
                   llvm_v8i32_ty, llvm_i8_ty], [IntrNoMem]>;
-
-  def int_x86_avx_vpermil_pd :
-        Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx_vpermil_ps :
-        Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-
-  def int_x86_avx_vpermil_pd_256 :
-        Intrinsic<[llvm_v4f64_ty], [llvm_v4f64_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_avx_vpermil_ps_256 :
-        Intrinsic<[llvm_v8f32_ty], [llvm_v8f32_ty,
-                  llvm_i8_ty], [IntrNoMem]>;
 }
 
 // Vector blend
@@ -1659,15 +1645,9 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_avx2_permd : GCCBuiltin<"__builtin_ia32_permvarsi256">,
               Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty, llvm_v8i32_ty],
                         [IntrNoMem]>;
-  def int_x86_avx2_permq : GCCBuiltin<"__builtin_ia32_permdi256">,
-              Intrinsic<[llvm_v4i64_ty], [llvm_v4i64_ty, llvm_i8_ty],
-                        [IntrNoMem]>;
   def int_x86_avx2_permps : GCCBuiltin<"__builtin_ia32_permvarsf256">,
               Intrinsic<[llvm_v8f32_ty], [llvm_v8f32_ty, llvm_v8f32_ty],
                         [IntrNoMem]>;
-  def int_x86_avx2_permpd : GCCBuiltin<"__builtin_ia32_permdf256">,
-              Intrinsic<[llvm_v4f64_ty], [llvm_v4f64_ty, llvm_i8_ty],
-                        [IntrNoMem]>;
   def int_x86_avx2_vperm2i128 : GCCBuiltin<"__builtin_ia32_permti256">,
               Intrinsic<[llvm_v4i64_ty], [llvm_v4i64_ty,
                          llvm_v4i64_ty, llvm_i8_ty], [IntrNoMem]>;
diff --git a/include/llvm/LLVMContext.h b/include/llvm/LLVMContext.h
index 18adcd1e3c2..a8306a9e761 100644
--- a/include/llvm/LLVMContext.h
+++ b/include/llvm/LLVMContext.h
@@ -42,7 +42,7 @@ public:
     MD_dbg = 0,  // "dbg"
     MD_tbaa = 1, // "tbaa"
     MD_prof = 2,  // "prof"
-    MD_fpaccuracy = 3,  // "fpaccuracy"
+    MD_fpmath = 3,  // "fpmath"
     MD_range = 4 // "range"
   };
   
diff --git a/include/llvm/MC/MCParser/AsmLexer.h b/include/llvm/MC/MCParser/AsmLexer.h
index dcecfb6aa01..9a8735f3e72 100644
--- a/include/llvm/MC/MCParser/AsmLexer.h
+++ b/include/llvm/MC/MCParser/AsmLexer.h
@@ -16,14 +16,11 @@
 
 #include "llvm/ADT/StringRef.h"
 #include "llvm/MC/MCParser/MCAsmLexer.h"
-#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/Support/DataTypes.h"
 #include 
-#include 
 
 namespace llvm {
 class MemoryBuffer;
-class SMLoc;
 class MCAsmInfo;
 
 /// AsmLexer - Lexer class for assembly files.
diff --git a/include/llvm/MC/MCParser/MCAsmLexer.h b/include/llvm/MC/MCParser/MCAsmLexer.h
index ac04483ccf1..5e29ad49dd3 100644
--- a/include/llvm/MC/MCParser/MCAsmLexer.h
+++ b/include/llvm/MC/MCParser/MCAsmLexer.h
@@ -15,8 +15,6 @@
 #include "llvm/Support/SMLoc.h"
 
 namespace llvm {
-class MCAsmLexer;
-class MCInst;
 
 /// AsmToken - Target independent representation for an assembler token.
 class AsmToken {
@@ -53,6 +51,7 @@ public:
     Greater, GreaterEqual, GreaterGreater, At
   };
 
+private:
   TokenKind Kind;
 
   /// A reference to the entire token contents; this is always a pointer into
diff --git a/include/llvm/Object/ELF.h b/include/llvm/Object/ELF.h
index 0828985f2e9..e493f5bd929 100644
--- a/include/llvm/Object/ELF.h
+++ b/include/llvm/Object/ELF.h
@@ -33,6 +33,15 @@
 namespace llvm {
 namespace object {
 
+// Subclasses of ELFObjectFile may need this for template instantiation
+inline std::pair
+getElfArchType(MemoryBuffer *Object) {
+  if (Object->getBufferSize() < ELF::EI_NIDENT)
+    return std::make_pair((uint8_t)ELF::ELFCLASSNONE,(uint8_t)ELF::ELFDATANONE);
+  return std::make_pair( (uint8_t)Object->getBufferStart()[ELF::EI_CLASS]
+                       , (uint8_t)Object->getBufferStart()[ELF::EI_DATA]);
+}
+
 // Templates to choose Elf_Addr and Elf_Off depending on is64Bits.
 template
 struct ELFDataTypeTypedefHelperCommon {
diff --git a/include/llvm/Operator.h b/include/llvm/Operator.h
index abd6a1939d7..1e86980cf30 100644
--- a/include/llvm/Operator.h
+++ b/include/llvm/Operator.h
@@ -15,8 +15,9 @@
 #ifndef LLVM_OPERATOR_H
 #define LLVM_OPERATOR_H
 
-#include "llvm/Instruction.h"
 #include "llvm/Constants.h"
+#include "llvm/Instruction.h"
+#include "llvm/Type.h"
 
 namespace llvm {
 
@@ -129,14 +130,15 @@ public:
     IsExact = (1 << 0)
   };
   
+private:
+  ~PossiblyExactOperator(); // do not implement
+
   friend class BinaryOperator;
   friend class ConstantExpr;
   void setIsExact(bool B) {
     SubclassOptionalData = (SubclassOptionalData & ~IsExact) | (B * IsExact);
   }
   
-private:
-  ~PossiblyExactOperator(); // do not implement
 public:
   /// isExact - Test whether this division is known to be exact, with
   /// zero remainder.
@@ -161,7 +163,28 @@ public:
            (isa(V) && classof(cast(V)));
   }
 };
-  
+
+/// FPMathOperator - Utility class for floating point operations which can have
+/// information about relaxed accuracy requirements attached to them.
+class FPMathOperator : public Operator {
+private:
+  ~FPMathOperator(); // do not implement
+
+public:
+
+  /// \brief Get the maximum error permitted by this operation in ULPs.  An
+  /// accuracy of 0.0 means that the operation should be performed with the
+  /// default precision.
+  float getFPAccuracy() const;
+
+  static inline bool classof(const FPMathOperator *) { return true; }
+  static inline bool classof(const Instruction *I) {
+    return I->getType()->isFPOrFPVectorTy();
+  }
+  static inline bool classof(const Value *V) {
+    return isa(V) && classof(cast(V));
+  }
+};
 
   
 /// ConcreteOperator - A helper template for defining operators for individual
diff --git a/include/llvm/Support/IRBuilder.h b/include/llvm/Support/IRBuilder.h
index 782800173f4..ef00e8ec24e 100644
--- a/include/llvm/Support/IRBuilder.h
+++ b/include/llvm/Support/IRBuilder.h
@@ -17,6 +17,7 @@
 
 #include "llvm/Instructions.h"
 #include "llvm/BasicBlock.h"
+#include "llvm/LLVMContext.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/Twine.h"
@@ -331,49 +332,63 @@ template >
 class IRBuilder : public IRBuilderBase, public Inserter {
   T Folder;
+  MDNode *DefaultFPMathTag;
 public:
-  IRBuilder(LLVMContext &C, const T &F, const Inserter &I = Inserter())
-    : IRBuilderBase(C), Inserter(I), Folder(F) {
+  IRBuilder(LLVMContext &C, const T &F, const Inserter &I = Inserter(),
+            MDNode *FPMathTag = 0)
+    : IRBuilderBase(C), Inserter(I), Folder(F), DefaultFPMathTag(FPMathTag) {
   }
 
-  explicit IRBuilder(LLVMContext &C) : IRBuilderBase(C), Folder() {
+  explicit IRBuilder(LLVMContext &C, MDNode *FPMathTag = 0) : IRBuilderBase(C),
+    Folder(), DefaultFPMathTag(FPMathTag) {
   }
 
-  explicit IRBuilder(BasicBlock *TheBB, const T &F)
-    : IRBuilderBase(TheBB->getContext()), Folder(F) {
+  explicit IRBuilder(BasicBlock *TheBB, const T &F, MDNode *FPMathTag = 0)
+    : IRBuilderBase(TheBB->getContext()), Folder(F),
+      DefaultFPMathTag(FPMathTag) {
     SetInsertPoint(TheBB);
   }
 
-  explicit IRBuilder(BasicBlock *TheBB)
-    : IRBuilderBase(TheBB->getContext()), Folder() {
+  explicit IRBuilder(BasicBlock *TheBB, MDNode *FPMathTag = 0)
+    : IRBuilderBase(TheBB->getContext()), Folder(),
+      DefaultFPMathTag(FPMathTag) {
     SetInsertPoint(TheBB);
   }
 
-  explicit IRBuilder(Instruction *IP)
-    : IRBuilderBase(IP->getContext()), Folder() {
+  explicit IRBuilder(Instruction *IP, MDNode *FPMathTag = 0)
+    : IRBuilderBase(IP->getContext()), Folder(), DefaultFPMathTag(FPMathTag) {
     SetInsertPoint(IP);
     SetCurrentDebugLocation(IP->getDebugLoc());
   }
 
-  explicit IRBuilder(Use &U)
-    : IRBuilderBase(U->getContext()), Folder() {
+  explicit IRBuilder(Use &U, MDNode *FPMathTag = 0)
+    : IRBuilderBase(U->getContext()), Folder(), DefaultFPMathTag(FPMathTag) {
     SetInsertPoint(U);
     SetCurrentDebugLocation(cast(U.getUser())->getDebugLoc());
   }
 
-  IRBuilder(BasicBlock *TheBB, BasicBlock::iterator IP, const T& F)
-    : IRBuilderBase(TheBB->getContext()), Folder(F) {
+  IRBuilder(BasicBlock *TheBB, BasicBlock::iterator IP, const T& F,
+            MDNode *FPMathTag = 0)
+    : IRBuilderBase(TheBB->getContext()), Folder(F),
+      DefaultFPMathTag(FPMathTag) {
     SetInsertPoint(TheBB, IP);
   }
 
-  IRBuilder(BasicBlock *TheBB, BasicBlock::iterator IP)
-    : IRBuilderBase(TheBB->getContext()), Folder() {
+  IRBuilder(BasicBlock *TheBB, BasicBlock::iterator IP, MDNode *FPMathTag = 0)
+    : IRBuilderBase(TheBB->getContext()), Folder(),
+      DefaultFPMathTag(FPMathTag) {
     SetInsertPoint(TheBB, IP);
   }
 
   /// getFolder - Get the constant folder being used.
   const T &getFolder() { return Folder; }
 
+  /// getDefaultFPMathTag - Get the floating point math metadata being used.
+  MDNode *getDefaultFPMathTag() const { return DefaultFPMathTag; }
+
+  /// SetDefaultFPMathTag - Set the floating point math metadata to be used.
+  void SetDefaultFPMathTag(MDNode *FPMathTag) { DefaultFPMathTag = FPMathTag; }
+
   /// isNamePreserving - Return true if this builder is configured to actually
   /// add the requested names to IR created through it.
   bool isNamePreserving() const { return preserveNames; }
@@ -496,6 +511,14 @@ private:
     if (HasNSW) BO->setHasNoSignedWrap();
     return BO;
   }
+
+  Instruction *AddFPMathTag(Instruction *I, MDNode *FPMathTag) const {
+    if (!FPMathTag)
+      FPMathTag = DefaultFPMathTag;
+    if (FPMathTag)
+      I->setMetadata(LLVMContext::MD_fpmath, FPMathTag);
+    return I;
+  }
 public:
   Value *CreateAdd(Value *LHS, Value *RHS, const Twine &Name = "",
                    bool HasNUW = false, bool HasNSW = false) {
@@ -511,11 +534,13 @@ public:
   Value *CreateNUWAdd(Value *LHS, Value *RHS, const Twine &Name = "") {
     return CreateAdd(LHS, RHS, Name, true, false);
   }
-  Value *CreateFAdd(Value *LHS, Value *RHS, const Twine &Name = "") {
+  Value *CreateFAdd(Value *LHS, Value *RHS, const Twine &Name = "",
+                    MDNode *FPMathTag = 0) {
     if (Constant *LC = dyn_cast(LHS))
       if (Constant *RC = dyn_cast(RHS))
         return Insert(Folder.CreateFAdd(LC, RC), Name);
-    return Insert(BinaryOperator::CreateFAdd(LHS, RHS), Name);
+    return Insert(AddFPMathTag(BinaryOperator::CreateFAdd(LHS, RHS),
+                               FPMathTag), Name);
   }
   Value *CreateSub(Value *LHS, Value *RHS, const Twine &Name = "",
                    bool HasNUW = false, bool HasNSW = false) {
@@ -531,11 +556,13 @@ public:
   Value *CreateNUWSub(Value *LHS, Value *RHS, const Twine &Name = "") {
     return CreateSub(LHS, RHS, Name, true, false);
   }
-  Value *CreateFSub(Value *LHS, Value *RHS, const Twine &Name = "") {
+  Value *CreateFSub(Value *LHS, Value *RHS, const Twine &Name = "",
+                    MDNode *FPMathTag = 0) {
     if (Constant *LC = dyn_cast(LHS))
       if (Constant *RC = dyn_cast(RHS))
         return Insert(Folder.CreateFSub(LC, RC), Name);
-    return Insert(BinaryOperator::CreateFSub(LHS, RHS), Name);
+    return Insert(AddFPMathTag(BinaryOperator::CreateFSub(LHS, RHS),
+                               FPMathTag), Name);
   }
   Value *CreateMul(Value *LHS, Value *RHS, const Twine &Name = "",
                    bool HasNUW = false, bool HasNSW = false) {
@@ -551,11 +578,13 @@ public:
   Value *CreateNUWMul(Value *LHS, Value *RHS, const Twine &Name = "") {
     return CreateMul(LHS, RHS, Name, true, false);
   }
-  Value *CreateFMul(Value *LHS, Value *RHS, const Twine &Name = "") {
+  Value *CreateFMul(Value *LHS, Value *RHS, const Twine &Name = "",
+                    MDNode *FPMathTag = 0) {
     if (Constant *LC = dyn_cast(LHS))
       if (Constant *RC = dyn_cast(RHS))
         return Insert(Folder.CreateFMul(LC, RC), Name);
-    return Insert(BinaryOperator::CreateFMul(LHS, RHS), Name);
+    return Insert(AddFPMathTag(BinaryOperator::CreateFMul(LHS, RHS),
+                               FPMathTag), Name);
   }
   Value *CreateUDiv(Value *LHS, Value *RHS, const Twine &Name = "",
                     bool isExact = false) {
@@ -581,11 +610,13 @@ public:
   Value *CreateExactSDiv(Value *LHS, Value *RHS, const Twine &Name = "") {
     return CreateSDiv(LHS, RHS, Name, true);
   }
-  Value *CreateFDiv(Value *LHS, Value *RHS, const Twine &Name = "") {
+  Value *CreateFDiv(Value *LHS, Value *RHS, const Twine &Name = "",
+                    MDNode *FPMathTag = 0) {
     if (Constant *LC = dyn_cast(LHS))
       if (Constant *RC = dyn_cast(RHS))
         return Insert(Folder.CreateFDiv(LC, RC), Name);
-    return Insert(BinaryOperator::CreateFDiv(LHS, RHS), Name);
+    return Insert(AddFPMathTag(BinaryOperator::CreateFDiv(LHS, RHS),
+                               FPMathTag), Name);
   }
   Value *CreateURem(Value *LHS, Value *RHS, const Twine &Name = "") {
     if (Constant *LC = dyn_cast(LHS))
@@ -599,11 +630,13 @@ public:
         return Insert(Folder.CreateSRem(LC, RC), Name);
     return Insert(BinaryOperator::CreateSRem(LHS, RHS), Name);
   }
-  Value *CreateFRem(Value *LHS, Value *RHS, const Twine &Name = "") {
+  Value *CreateFRem(Value *LHS, Value *RHS, const Twine &Name = "",
+                    MDNode *FPMathTag = 0) {
     if (Constant *LC = dyn_cast(LHS))
       if (Constant *RC = dyn_cast(RHS))
         return Insert(Folder.CreateFRem(LC, RC), Name);
-    return Insert(BinaryOperator::CreateFRem(LHS, RHS), Name);
+    return Insert(AddFPMathTag(BinaryOperator::CreateFRem(LHS, RHS),
+                               FPMathTag), Name);
   }
 
   Value *CreateShl(Value *LHS, Value *RHS, const Twine &Name = "",
@@ -729,10 +762,10 @@ public:
   Value *CreateNUWNeg(Value *V, const Twine &Name = "") {
     return CreateNeg(V, Name, true, false);
   }
-  Value *CreateFNeg(Value *V, const Twine &Name = "") {
+  Value *CreateFNeg(Value *V, const Twine &Name = "", MDNode *FPMathTag = 0) {
     if (Constant *VC = dyn_cast(V))
       return Insert(Folder.CreateFNeg(VC), Name);
-    return Insert(BinaryOperator::CreateFNeg(V), Name);
+    return Insert(AddFPMathTag(BinaryOperator::CreateFNeg(V), FPMathTag), Name);
   }
   Value *CreateNot(Value *V, const Twine &Name = "") {
     if (Constant *VC = dyn_cast(V))
diff --git a/include/llvm/Support/JSONParser.h b/include/llvm/Support/JSONParser.h
deleted file mode 100644
index 11149f1e47b..00000000000
--- a/include/llvm/Support/JSONParser.h
+++ /dev/null
@@ -1,448 +0,0 @@
-//===--- JSONParser.h - Simple JSON parser ----------------------*- C++ -*-===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-//  This file implements a JSON parser.
-//
-//  See http://www.json.org/ for an overview.
-//  See http://www.ietf.org/rfc/rfc4627.txt for the full standard.
-//
-//  FIXME: Currently this supports a subset of JSON. Specifically, support
-//  for numbers, booleans and null for values is missing.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef LLVM_SUPPORT_JSON_PARSER_H
-#define LLVM_SUPPORT_JSON_PARSER_H
-
-#include "llvm/ADT/StringRef.h"
-#include "llvm/Support/Allocator.h"
-#include "llvm/Support/Casting.h"
-#include "llvm/Support/ErrorHandling.h"
-#include "llvm/Support/SourceMgr.h"
-
-namespace llvm {
-
-class JSONContainer;
-class JSONString;
-class JSONValue;
-class JSONKeyValuePair;
-
-/// \brief Base class for a parsable JSON atom.
-///
-/// This class has no semantics other than being a unit of JSON data which can
-/// be parsed out of a JSON document.
-class JSONAtom {
-public:
-  /// \brief Possible types of JSON objects.
-  enum Kind { JK_KeyValuePair, JK_Array, JK_Object, JK_String };
-
-  /// \brief Returns the type of this value.
-  Kind getKind() const { return MyKind; }
-
-  static bool classof(const JSONAtom *Atom) { return true; }
-
-protected:
-  JSONAtom(Kind MyKind) : MyKind(MyKind) {}
-
-private:
-  Kind MyKind;
-};
-
-/// \brief A parser for JSON text.
-///
-/// Use an object of JSONParser to iterate over the values of a JSON text.
-/// All objects are parsed during the iteration, so you can only iterate once
-/// over the JSON text, but the cost of partial iteration is minimized.
-/// Create a new JSONParser if you want to iterate multiple times.
-class JSONParser {
-public:
-  /// \brief Create a JSONParser for the given input.
-  ///
-  /// Parsing is started via parseRoot(). Access to the object returned from
-  /// parseRoot() will parse the input lazily.
-  JSONParser(StringRef Input, SourceMgr *SM);
-
-  /// \brief Returns the outermost JSON value (either an array or an object).
-  ///
-  /// Can return NULL if the input does not start with an array or an object.
-  /// The object is not parsed yet - the caller must iterate over the
-  /// returned object to trigger parsing.
-  ///
-  /// A JSONValue can be either a JSONString, JSONObject or JSONArray.
-  JSONValue *parseRoot();
-
-  /// \brief Parses the JSON text and returns whether it is valid JSON.
-  ///
-  /// In case validate() return false, failed() will return true and
-  /// getErrorMessage() will return the parsing error.
-  bool validate();
-
-  /// \brief Returns true if an error occurs during parsing.
-  ///
-  /// If there was an error while parsing an object that was created by
-  /// iterating over the result of 'parseRoot', 'failed' will return true.
-  bool failed() const;
-
-private:
-  /// \brief These methods manage the implementation details of parsing new JSON
-  /// atoms.
-  /// @{
-  JSONString *parseString();
-  JSONValue *parseValue();
-  JSONKeyValuePair *parseKeyValuePair();
-  /// @}
-
-  /// \brief Helpers to parse the elements out of both forms of containers.
-  /// @{
-  const JSONAtom *parseElement(JSONAtom::Kind ContainerKind);
-  StringRef::iterator parseFirstElement(JSONAtom::Kind ContainerKind,
-                                        char StartChar, char EndChar,
-                                        const JSONAtom *&Element);
-  StringRef::iterator parseNextElement(JSONAtom::Kind ContainerKind,
-                                       char EndChar,
-                                       const JSONAtom *&Element);
-  /// @}
-
-  /// \brief Whitespace parsing.
-  /// @{
-  void nextNonWhitespace();
-  bool isWhitespace();
-  /// @}
-
-  /// \brief These methods are used for error handling.
-  /// {
-  void setExpectedError(StringRef Expected, StringRef Found);
-  void setExpectedError(StringRef Expected, char Found);
-  bool errorIfAtEndOfFile(StringRef Message);
-  bool errorIfNotAt(char C, StringRef Message);
-  /// }
-
-  /// \brief Skips all elements in the given container.
-  bool skipContainer(const JSONContainer &Container);
-
-  /// \brief Skips to the next position behind the given JSON atom.
-  bool skip(const JSONAtom &Atom);
-
-  /// All nodes are allocated by the parser and will be deallocated when the
-  /// parser is destroyed.
-  BumpPtrAllocator ValueAllocator;
-
-  /// \brief The original input to the parser.
-  MemoryBuffer *InputBuffer;
-
-  /// \brief The source manager used for diagnostics and buffer management.
-  SourceMgr *SM;
-
-  /// \brief The current position in the parse stream.
-  StringRef::iterator Position;
-
-  /// \brief The end position for fast EOF checks without introducing
-  /// unnecessary dereferences.
-  StringRef::iterator End;
-
-  /// \brief If true, an error has occurred.
-  bool Failed;
-
-  friend class JSONContainer;
-};
-
-
-/// \brief Base class for JSON value objects.
-///
-/// This object represents an abstract JSON value. It is the root node behind
-/// the group of JSON entities that can represent top-level values in a JSON
-/// document. It has no API, and is just a placeholder in the type hierarchy of
-/// nodes.
-class JSONValue : public JSONAtom {
-protected:
-  JSONValue(Kind MyKind) : JSONAtom(MyKind) {}
-
-public:
-  /// \brief dyn_cast helpers
-  ///@{
-  static bool classof(const JSONAtom *Atom) {
-    switch (Atom->getKind()) {
-      case JK_Array:
-      case JK_Object:
-      case JK_String:
-        return true;
-      case JK_KeyValuePair:
-        return false;
-    }
-    llvm_unreachable("Invalid JSONAtom kind");
-  }
-  static bool classof(const JSONValue *Value) { return true; }
-  ///@}
-};
-
-/// \brief Gives access to the text of a JSON string.
-///
-/// FIXME: Implement a method to return the unescaped text.
-class JSONString : public JSONValue {
-public:
-  /// \brief Returns the underlying parsed text of the string.
-  ///
-  /// This is the unescaped content of the JSON text.
-  /// See http://www.ietf.org/rfc/rfc4627.txt for details.
-  StringRef getRawText() const { return RawText; }
-
-private:
-  JSONString(StringRef RawText) : JSONValue(JK_String), RawText(RawText) {}
-
-  StringRef RawText;
-
-  friend class JSONParser;
-
-public:
-  /// \brief dyn_cast helpers
-  ///@{
-  static bool classof(const JSONAtom *Atom) {
-    return Atom->getKind() == JK_String;
-  }
-  static bool classof(const JSONString *String) { return true; }
-  ///@}
-};
-
-/// \brief A (key, value) tuple of type (JSONString *, JSONValue *).
-///
-/// Note that JSONKeyValuePair is not a JSONValue, it is a bare JSONAtom.
-/// JSONKeyValuePairs can be elements of a JSONObject, but not of a JSONArray.
-/// They are not viable as top-level values either.
-class JSONKeyValuePair : public JSONAtom {
-public:
-  const JSONString * const Key;
-  const JSONValue * const Value;
-
-private:
-  JSONKeyValuePair(const JSONString *Key, const JSONValue *Value)
-      : JSONAtom(JK_KeyValuePair), Key(Key), Value(Value) {}
-
-  friend class JSONParser;
-
-public:
-  /// \brief dyn_cast helpers
-  ///@{
-  static bool classof(const JSONAtom *Atom) {
-    return Atom->getKind() == JK_KeyValuePair;
-  }
-  static bool classof(const JSONKeyValuePair *KeyValuePair) { return true; }
-  ///@}
-};
-
-/// \brief Implementation of JSON containers (arrays and objects).
-///
-/// JSONContainers drive the lazy parsing of JSON arrays and objects via
-/// forward iterators.
-class JSONContainer : public JSONValue {
-private:
-  /// \brief An iterator that parses the underlying container during iteration.
-  ///
-  /// Iterators on the same collection use shared state, so when multiple copies
-  /// of an iterator exist, only one is allowed to be used for iteration;
-  /// iterating multiple copies of an iterator of the same collection will lead
-  /// to undefined behavior.
-  class AtomIterator {
-  public:
-    AtomIterator(const AtomIterator &I) : Container(I.Container) {}
-
-    /// \brief Iterator interface.
-    ///@{
-    bool operator==(const AtomIterator &I) const {
-      if (isEnd() || I.isEnd())
-        return isEnd() == I.isEnd();
-      return Container->Position == I.Container->Position;
-    }
-    bool operator!=(const AtomIterator &I) const {
-      return !(*this == I);
-    }
-    AtomIterator &operator++() {
-      Container->parseNextElement();
-      return *this;
-    }
-    const JSONAtom *operator*() {
-      return Container->Current;
-    }
-    ///@}
-
-  private:
-    /// \brief Create an iterator for which 'isEnd' returns true.
-    AtomIterator() : Container(0) {}
-
-    /// \brief Create an iterator for the given container.
-    AtomIterator(const JSONContainer *Container) : Container(Container) {}
-
-    bool isEnd() const {
-      return Container == 0 || Container->Position == StringRef::iterator();
-    }
-
-    const JSONContainer * const Container;
-
-    friend class JSONContainer;
-  };
-
-protected:
-  /// \brief An iterator for the specified AtomT.
-  ///
-  /// Used for the implementation of iterators for JSONArray and JSONObject.
-  template 
-  class IteratorTemplate : public std::iterator {
-  public:
-    explicit IteratorTemplate(const AtomIterator& AtomI)
-      : AtomI(AtomI) {}
-
-    bool operator==(const IteratorTemplate &I) const {
-      return AtomI == I.AtomI;
-    }
-    bool operator!=(const IteratorTemplate &I) const { return !(*this == I); }
-
-    IteratorTemplate &operator++() {
-      ++AtomI;
-      return *this;
-    }
-
-    const AtomT *operator*() { return dyn_cast(*AtomI); }
-
-  private:
-    AtomIterator AtomI;
-  };
-
-  JSONContainer(JSONParser *Parser, char StartChar, char EndChar,
-                JSONAtom::Kind ContainerKind)
-    : JSONValue(ContainerKind), Parser(Parser),
-      Position(), Current(0), Started(false),
-      StartChar(StartChar), EndChar(EndChar) {}
-
-  /// \brief Returns a lazy parsing iterator over the container.
-  ///
-  /// As the iterator drives the parse stream, begin() must only be called
-  /// once per container.
-  AtomIterator atom_begin() const {
-    if (Started)
-      report_fatal_error("Cannot parse container twice.");
-    Started = true;
-    // Set up the position and current element when we begin iterating over the
-    // container.
-    Position = Parser->parseFirstElement(getKind(), StartChar, EndChar, Current);
-    return AtomIterator(this);
-  }
-  AtomIterator atom_end() const {
-    return AtomIterator();
-  }
-
-private:
-  AtomIterator atom_current() const {
-    if (!Started)
-      return atom_begin();
-
-    return AtomIterator(this);
-  }
-
-  /// \brief Parse the next element in the container into the Current element.
-  ///
-  /// This routine is called as an iterator into this container walks through
-  /// its elements. It mutates the container's internal current node to point to
-  /// the next atom of the container.
-  void parseNextElement() const {
-    Parser->skip(*Current);
-    Position = Parser->parseNextElement(getKind(), EndChar, Current);
-  }
-
-  // For parsing, JSONContainers call back into the JSONParser.
-  JSONParser * const Parser;
-
-  // 'Position', 'Current' and 'Started' store the state of the parse stream
-  // for iterators on the container, they don't change the container's elements
-  // and are thus marked as mutable.
-  mutable StringRef::iterator Position;
-  mutable const JSONAtom *Current;
-  mutable bool Started;
-
-  const char StartChar;
-  const char EndChar;
-
-  friend class JSONParser;
-
-public:
-  /// \brief dyn_cast helpers
-  ///@{
-  static bool classof(const JSONAtom *Atom) {
-    switch (Atom->getKind()) {
-      case JK_Array:
-      case JK_Object:
-        return true;
-      case JK_KeyValuePair:
-      case JK_String:
-        return false;
-    }
-    llvm_unreachable("Invalid JSONAtom kind");
-  }
-  static bool classof(const JSONContainer *Container) { return true; }
-  ///@}
-};
-
-/// \brief A simple JSON array.
-class JSONArray : public JSONContainer {
-public:
-  typedef IteratorTemplate const_iterator;
-
-  /// \brief Returns a lazy parsing iterator over the container.
-  ///
-  /// As the iterator drives the parse stream, begin() must only be called
-  /// once per container.
-  const_iterator begin() const { return const_iterator(atom_begin()); }
-  const_iterator end() const { return const_iterator(atom_end()); }
-
-private:
-  JSONArray(JSONParser *Parser)
-    : JSONContainer(Parser, '[', ']', JSONAtom::JK_Array) {}
-
-public:
-  /// \brief dyn_cast helpers
-  ///@{
-  static bool classof(const JSONAtom *Atom) {
-    return Atom->getKind() == JSONAtom::JK_Array;
-  }
-  static bool classof(const JSONArray *Array) { return true; }
-  ///@}
-
-  friend class JSONParser;
-};
-
-/// \brief A JSON object: an iterable list of JSON key-value pairs.
-class JSONObject : public JSONContainer {
-public:
-  typedef IteratorTemplate const_iterator;
-
-  /// \brief Returns a lazy parsing iterator over the container.
-  ///
-  /// As the iterator drives the parse stream, begin() must only be called
-  /// once per container.
-  const_iterator begin() const { return const_iterator(atom_begin()); }
-  const_iterator end() const { return const_iterator(atom_end()); }
-
-private:
-  JSONObject(JSONParser *Parser)
-    : JSONContainer(Parser, '{', '}', JSONAtom::JK_Object) {}
-
-public:
-  /// \brief dyn_cast helpers
-  ///@{
-  static bool classof(const JSONAtom *Atom) {
-    return Atom->getKind() == JSONAtom::JK_Object;
-  }
-  static bool classof(const JSONObject *Object) { return true; }
-  ///@}
-
-  friend class JSONParser;
-};
-
-} // end namespace llvm
-
-#endif // LLVM_SUPPORT_JSON_PARSER_H
diff --git a/include/llvm/Support/Locale.h b/include/llvm/Support/Locale.h
new file mode 100644
index 00000000000..b0f12958029
--- /dev/null
+++ b/include/llvm/Support/Locale.h
@@ -0,0 +1,17 @@
+#ifndef LLVM_SUPPORT_LOCALE
+#define LLVM_SUPPORT_LOCALE
+
+#include "llvm/ADT/StringRef.h"
+
+namespace llvm {
+namespace sys {
+namespace locale {
+
+int columnWidth(StringRef s);
+bool isPrint(int c);
+
+}
+}
+}
+
+#endif // LLVM_SUPPORT_LOCALE
diff --git a/include/llvm/Support/MDBuilder.h b/include/llvm/Support/MDBuilder.h
new file mode 100644
index 00000000000..40f028a4327
--- /dev/null
+++ b/include/llvm/Support/MDBuilder.h
@@ -0,0 +1,118 @@
+//===---- llvm/Support/MDBuilder.h - Builder for LLVM metadata --*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file defines the MDBuilder class, which is used as a convenient way to
+// create LLVM metadata with a consistent and simplified interface.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_SUPPORT_MDBUILDER_H
+#define LLVM_SUPPORT_MDBUILDER_H
+
+#include "llvm/Constants.h"
+#include "llvm/DerivedTypes.h"
+#include "llvm/LLVMContext.h"
+#include "llvm/Metadata.h"
+#include "llvm/ADT/APInt.h"
+
+namespace llvm {
+
+  class MDBuilder {
+    LLVMContext &Context;
+
+  public:
+    MDBuilder(LLVMContext &context) : Context(context) {}
+
+    /// \brief Return the given string as metadata.
+    MDString *createString(StringRef Str) {
+      return MDString::get(Context, Str);
+    }
+
+    //===------------------------------------------------------------------===//
+    // FPMath metadata.
+    //===------------------------------------------------------------------===//
+
+    /// \brief Return metadata with the given settings.  The special value 0.0
+    /// for the Accuracy parameter indicates the default (maximal precision)
+    /// setting.
+    MDNode *createFPMath(float Accuracy) {
+      if (Accuracy == 0.0)
+        return 0;
+      assert(Accuracy > 0.0 && "Invalid fpmath accuracy!");
+      Value *Op = ConstantFP::get(Type::getFloatTy(Context), Accuracy);
+      return MDNode::get(Context, Op);
+    }
+
+
+    //===------------------------------------------------------------------===//
+    // Range metadata.
+    //===------------------------------------------------------------------===//
+
+    /// \brief Return metadata describing the range [Lo, Hi).
+    MDNode *createRange(const APInt &Lo, const APInt &Hi) {
+      assert(Lo.getBitWidth() == Hi.getBitWidth() && "Mismatched bitwidths!");
+      // If the range is everything then it is useless.
+      if (Hi == Lo)
+        return 0;
+
+      // Return the range [Lo, Hi).
+      Type *Ty = IntegerType::get(Context, Lo.getBitWidth());
+      Value *Range[2] = { ConstantInt::get(Ty, Lo), ConstantInt::get(Ty, Hi) };
+      return MDNode::get(Context, Range);
+    }
+
+
+    //===------------------------------------------------------------------===//
+    // TBAA metadata.
+    //===------------------------------------------------------------------===//
+
+    /// \brief Return metadata appropriate for a TBAA root node.  Each returned
+    /// node is distinct from all other metadata and will never be identified
+    /// (uniqued) with anything else.
+    MDNode *createAnonymousTBAARoot() {
+      // To ensure uniqueness the root node is self-referential.
+      MDNode *Dummy = MDNode::getTemporary(Context, ArrayRef());
+      MDNode *Root = MDNode::get(Context, Dummy);
+      // At this point we have
+      //   !0 = metadata !{}            <- dummy
+      //   !1 = metadata !{metadata !0} <- root
+      // Replace the dummy operand with the root node itself and delete the dummy.
+      Root->replaceOperandWith(0, Root);
+      MDNode::deleteTemporary(Dummy);
+      // We now have
+      //   !1 = metadata !{metadata !1} <- self-referential root
+      return Root;
+    }
+
+    /// \brief Return metadata appropriate for a TBAA root node with the given
+    /// name.  This may be identified (uniqued) with other roots with the same
+    /// name.
+    MDNode *createTBAARoot(StringRef Name) {
+      return MDNode::get(Context, createString(Name));
+    }
+
+    /// \brief Return metadata for a non-root TBAA node with the given name,
+    /// parent in the TBAA tree, and value for 'pointsToConstantMemory'.
+    MDNode *createTBAANode(StringRef Name, MDNode *Parent,
+                           bool isConstant = false) {
+      if (isConstant) {
+        Constant *Flags = ConstantInt::get(Type::getInt64Ty(Context), 1);
+        Value *Ops[3] = { createString(Name), Parent, Flags };
+        return MDNode::get(Context, Ops);
+      } else {
+        Value *Ops[2] = { createString(Name), Parent };
+        return MDNode::get(Context, Ops);
+      }
+    }
+
+  };
+
+} // end namespace llvm
+
+#endif
diff --git a/include/llvm/Support/Process.h b/include/llvm/Support/Process.h
index 33799229ff3..d796b7906d3 100644
--- a/include/llvm/Support/Process.h
+++ b/include/llvm/Support/Process.h
@@ -136,6 +136,10 @@ namespace sys {
       /// Same as OutputColor, but only enables the bold attribute.
       static const char *OutputBold(bool bg);
 
+      /// This function returns the escape sequence to reverse forground and
+      /// background colors.
+      static const char *OutputReverse();
+
       /// Resets the terminals colors, or returns an escape sequence to do so.
       static const char *ResetColor();
     /// @}
diff --git a/include/llvm/Support/SourceMgr.h b/include/llvm/Support/SourceMgr.h
index 58b8fab5240..76967dbf78a 100644
--- a/include/llvm/Support/SourceMgr.h
+++ b/include/llvm/Support/SourceMgr.h
@@ -128,8 +128,11 @@ public:
   /// PrintMessage - Emit a message about the specified location with the
   /// specified string.
   ///
+  /// @param ShowColors - Display colored messages if output is a terminal and
+  /// the default error handler is used.
   void PrintMessage(SMLoc Loc, DiagKind Kind, const Twine &Msg,
-                    ArrayRef Ranges = ArrayRef()) const;
+                    ArrayRef Ranges = ArrayRef(),
+                    bool ShowColors = true) const;
 
 
   /// GetMessage - Return an SMDiagnostic at the specified location with the
@@ -188,7 +191,7 @@ public:
   const std::vector > &getRanges() const {
     return Ranges;
   }
-  void print(const char *ProgName, raw_ostream &S) const;
+  void print(const char *ProgName, raw_ostream &S, bool ShowColors = true) const;
 };
 
 }  // end llvm namespace
diff --git a/include/llvm/Support/YAMLParser.h b/include/llvm/Support/YAMLParser.h
index b24cacd3c32..47206b3c6d9 100644
--- a/include/llvm/Support/YAMLParser.h
+++ b/include/llvm/Support/YAMLParser.h
@@ -516,8 +516,11 @@ public:
   document_iterator() : Doc(NullDoc) {}
   document_iterator(OwningPtr &D) : Doc(D) {}
 
+  bool operator ==(const document_iterator &Other) {
+    return Doc == Other.Doc;
+  }
   bool operator !=(const document_iterator &Other) {
-    return Doc != Other.Doc;
+    return !(*this == Other);
   }
 
   document_iterator operator ++() {
diff --git a/include/llvm/Support/raw_ostream.h b/include/llvm/Support/raw_ostream.h
index 6bfae5e2982..6c5d4787e0f 100644
--- a/include/llvm/Support/raw_ostream.h
+++ b/include/llvm/Support/raw_ostream.h
@@ -222,6 +222,9 @@ public:
   /// outputting colored text, or before program exit.
   virtual raw_ostream &resetColor() { return *this; }
 
+  /// Reverses the forground and background colors.
+  virtual raw_ostream &reverseColor() { return *this; }
+
   /// This function determines if this stream is connected to a "tty" or
   /// "console" window. That is, the output would be displayed to the user
   /// rather than being put on a pipe or stored in a file.
@@ -379,6 +382,8 @@ public:
                                    bool bg=false);
   virtual raw_ostream &resetColor();
 
+  virtual raw_ostream &reverseColor();
+
   virtual bool is_displayed() const;
 
   /// has_error - Return the value of the flag in this raw_fd_ostream indicating
diff --git a/include/llvm/TableGen/Error.h b/include/llvm/TableGen/Error.h
index c01b32b1c2d..fd5f805ffc9 100644
--- a/include/llvm/TableGen/Error.h
+++ b/include/llvm/TableGen/Error.h
@@ -29,6 +29,11 @@ public:
   const std::string &getMessage() const { return Message; }
 };
 
+void PrintWarning(SMLoc WarningLoc, const Twine &Msg);
+void PrintWarning(const char *Loc, const Twine &Msg);
+void PrintWarning(const Twine &Msg);
+void PrintWarning(const TGError &Warning);
+
 void PrintError(SMLoc ErrorLoc, const Twine &Msg);
 void PrintError(const char *Loc, const Twine &Msg);
 void PrintError(const Twine &Msg);
diff --git a/include/llvm/TableGen/Record.h b/include/llvm/TableGen/Record.h
index 5e68c10a47a..3aea1aeaead 100644
--- a/include/llvm/TableGen/Record.h
+++ b/include/llvm/TableGen/Record.h
@@ -454,7 +454,7 @@ public:
   /// without adding quote markers.  This primaruly affects
   /// StringInits where we will not surround the string value with
   /// quotes.
-  virtual std::string getAsUnquotedString() const { return getAsString(); }  
+  virtual std::string getAsUnquotedString() const { return getAsString(); }
 
   /// dump - Debugging method that may be called through a debugger, just
   /// invokes print on stderr.
@@ -1529,7 +1529,7 @@ struct MultiClass {
 
   void dump() const;
 
-  MultiClass(const std::string &Name, SMLoc Loc, RecordKeeper &Records) : 
+  MultiClass(const std::string &Name, SMLoc Loc, RecordKeeper &Records) :
     Rec(Name, Loc, Records) {}
 };
 
diff --git a/include/llvm/Target/TargetLibraryInfo.h b/include/llvm/Target/TargetLibraryInfo.h
index 70e26bf3c5a..c8cacf284d0 100644
--- a/include/llvm/Target/TargetLibraryInfo.h
+++ b/include/llvm/Target/TargetLibraryInfo.h
@@ -83,7 +83,7 @@ namespace llvm {
       /// long double expm1l(long double x);
       expm1l,
       /// float expm1f(float x);
-      expl1f,
+      expm1f,
       /// double fabs(double x);
       fabs,
       /// long double fabsl(long double x);
@@ -159,8 +159,14 @@ namespace llvm {
       rint,
       /// float rintf(float x);
       rintf,
-      /// long dobule rintl(long double x);
+      /// long double rintl(long double x);
       rintl,
+      /// double round(double x);
+      round,
+      /// float roundf(float x);
+      roundf,
+      /// long double roundl(long double x);
+      roundl,
       /// double sin(double x);
       sin,
       /// long double sinl(long double x);
diff --git a/include/llvm/Target/TargetRegisterInfo.h b/include/llvm/Target/TargetRegisterInfo.h
index 7d8a46b49ac..6ddd36451b5 100644
--- a/include/llvm/Target/TargetRegisterInfo.h
+++ b/include/llvm/Target/TargetRegisterInfo.h
@@ -205,7 +205,7 @@ struct TargetRegisterInfoDesc {
 /// Each TargetRegisterClass has a per register weight, and weight
 /// limit which must be less than the limits of its pressure sets.
 struct RegClassWeight {
-  unsigned RegWeigt;
+  unsigned RegWeight;
   unsigned WeightLimit;
 };
 
diff --git a/include/llvm/Transforms/Utils/BasicBlockUtils.h b/include/llvm/Transforms/Utils/BasicBlockUtils.h
index 867b9e43849..2f9dc54541b 100644
--- a/include/llvm/Transforms/Utils/BasicBlockUtils.h
+++ b/include/llvm/Transforms/Utils/BasicBlockUtils.h
@@ -110,7 +110,8 @@ bool isCriticalEdge(const TerminatorInst *TI, unsigned SuccNum,
 ///
 BasicBlock *SplitCriticalEdge(TerminatorInst *TI, unsigned SuccNum,
                               Pass *P = 0, bool MergeIdenticalEdges = false,
-                              bool DontDeleteUselessPHIs = false);
+                              bool DontDeleteUselessPHIs = false,
+                              bool SplitLandingPads = false);
 
 inline BasicBlock *SplitCriticalEdge(BasicBlock *BB, succ_iterator SI,
                                      Pass *P = 0) {
diff --git a/include/llvm/Transforms/Vectorize.h b/include/llvm/Transforms/Vectorize.h
index 7701ceb4d0b..652916c26c2 100644
--- a/include/llvm/Transforms/Vectorize.h
+++ b/include/llvm/Transforms/Vectorize.h
@@ -34,6 +34,9 @@ struct VectorizeConfig {
   /// @brief Vectorize floating-point values.
   bool VectorizeFloats;
 
+  /// @brief Vectorize pointer values.
+  bool VectorizePointers;
+
   /// @brief Vectorize casting (conversion) operations.
   bool VectorizeCasts;
 
@@ -43,6 +46,12 @@ struct VectorizeConfig {
   /// @brief Vectorize the fused-multiply-add intrinsic.
   bool VectorizeFMA;
 
+  /// @brief Vectorize select instructions.
+  bool VectorizeSelect;
+
+  /// @brief Vectorize getelementptr instructions.
+  bool VectorizeGEP;
+
   /// @brief Vectorize loads and stores.
   bool VectorizeMemOps;
 
diff --git a/lib/Analysis/ConstantFolding.cpp b/lib/Analysis/ConstantFolding.cpp
index 7a0a4e1e824..783c32e6669 100644
--- a/lib/Analysis/ConstantFolding.cpp
+++ b/lib/Analysis/ConstantFolding.cpp
@@ -681,6 +681,7 @@ static Constant *SymbolicallyEvaluateGEP(ArrayRef Ops,
   // This makes it easy to determine if the getelementptr is "inbounds".
   // Also, this helps GlobalOpt do SROA on GlobalVariables.
   Type *Ty = Ptr->getType();
+  assert(Ty->isPointerTy() && "Forming regular GEP of non-pointer type");
   SmallVector NewIdxs;
   do {
     if (SequentialType *ATy = dyn_cast(Ty)) {
@@ -711,10 +712,17 @@ static Constant *SymbolicallyEvaluateGEP(ArrayRef Ops,
       }
       Ty = ATy->getElementType();
     } else if (StructType *STy = dyn_cast(Ty)) {
-      // Determine which field of the struct the offset points into. The
-      // getZExtValue is at least as safe as the StructLayout API because we
-      // know the offset is within the struct at this point.
+      // If we end up with an offset that isn't valid for this struct type, we
+      // can't re-form this GEP in a regular form, so bail out. The pointer
+      // operand likely went through casts that are necessary to make the GEP
+      // sensible.
       const StructLayout &SL = *TD->getStructLayout(STy);
+      if (Offset.uge(SL.getSizeInBytes()))
+        break;
+
+      // Determine which field of the struct the offset points into. The
+      // getZExtValue is fine as we've already ensured that the offset is
+      // within the range representable by the StructLayout API.
       unsigned ElIdx = SL.getElementContainingOffset(Offset.getZExtValue());
       NewIdxs.push_back(ConstantInt::get(Type::getInt32Ty(Ty->getContext()),
                                          ElIdx));
diff --git a/lib/Analysis/ScalarEvolution.cpp b/lib/Analysis/ScalarEvolution.cpp
index 1d55642079a..205227ca0b7 100644
--- a/lib/Analysis/ScalarEvolution.cpp
+++ b/lib/Analysis/ScalarEvolution.cpp
@@ -3187,7 +3187,7 @@ const SCEV *ScalarEvolution::createNodeForGEP(GEPOperator *GEP) {
 
   // Add the total offset from all the GEP indices to the base.
   return getAddExpr(BaseS, TotalOffset,
-                    isInBounds ? SCEV::FlagNUW : SCEV::FlagAnyWrap);
+                    isInBounds ? SCEV::FlagNSW : SCEV::FlagAnyWrap);
 }
 
 /// GetMinTrailingZeros - Determine the minimum number of zero bits that S is
diff --git a/lib/Analysis/ValueTracking.cpp b/lib/Analysis/ValueTracking.cpp
index a430f6281ef..1418e01d7c8 100644
--- a/lib/Analysis/ValueTracking.cpp
+++ b/lib/Analysis/ValueTracking.cpp
@@ -564,7 +564,7 @@ void llvm::ComputeMaskedBits(Value *V, APInt &KnownZero, APInt &KnownOne,
                         Depth+1);
       // If it's known zero, our sign bit is also zero.
       if (LHSKnownZero.isNegative())
-        KnownZero |= LHSKnownZero;
+        KnownZero.setBit(BitWidth - 1);
     }
 
     break;
diff --git a/lib/CodeGen/AsmPrinter/DwarfAccelTable.cpp b/lib/CodeGen/AsmPrinter/DwarfAccelTable.cpp
index 660684d1bea..454a923c13e 100644
--- a/lib/CodeGen/AsmPrinter/DwarfAccelTable.cpp
+++ b/lib/CodeGen/AsmPrinter/DwarfAccelTable.cpp
@@ -36,35 +36,20 @@ const char *DwarfAccelTable::Atom::AtomTypeString(enum AtomType AT) {
   llvm_unreachable("invalid AtomType!");
 }
 
-// The general case would need to have a less hard coded size for the
-// length of the HeaderData, however, if we're constructing based on a
-// single Atom then we know it will always be: 4 + 4 + 2 + 2.
-DwarfAccelTable::DwarfAccelTable(DwarfAccelTable::Atom atom) :
-  Header(12),
-  HeaderData(atom) {
-}
-
 // The length of the header data is always going to be 4 + 4 + 4*NumAtoms.
-DwarfAccelTable::DwarfAccelTable(std::vector &atomList) :
+DwarfAccelTable::DwarfAccelTable(ArrayRef atomList) :
   Header(8 + (atomList.size() * 4)),
-  HeaderData(atomList) {
-}
+  HeaderData(atomList),
+  Entries(Allocator) { }
 
-DwarfAccelTable::~DwarfAccelTable() {
-  for (size_t i = 0, e = Data.size(); i < e; ++i)
-    delete Data[i];
-  for (StringMap::iterator
-         EI = Entries.begin(), EE = Entries.end(); EI != EE; ++EI)
-    for (DataArray::iterator DI = EI->second.begin(),
-           DE = EI->second.end(); DI != DE; ++DI)
-      delete (*DI);
-}
+DwarfAccelTable::~DwarfAccelTable() { }
 
 void DwarfAccelTable::AddName(StringRef Name, DIE* die, char Flags) {
+  assert(Data.empty() && "Already finalized!");
   // If the string is in the list already then add this die to the list
   // otherwise add a new one.
   DataArray &DIEs = Entries[Name];
-  DIEs.push_back(new HashDataContents(die, Flags));
+  DIEs.push_back(new (Allocator) HashDataContents(die, Flags));
 }
 
 void DwarfAccelTable::ComputeBucketCount(void) {
@@ -85,31 +70,23 @@ void DwarfAccelTable::ComputeBucketCount(void) {
   Header.hashes_count = num;
 }
 
-namespace {
-  // DIESorter - comparison predicate that sorts DIEs by their offset.
-  struct DIESorter {
-    bool operator()(const struct DwarfAccelTable::HashDataContents *A,
-                    const struct DwarfAccelTable::HashDataContents *B) const {
-      return A->Die->getOffset() < B->Die->getOffset();
-    }
-  };
+// compareDIEs - comparison predicate that sorts DIEs by their offset.
+static bool compareDIEs(const DwarfAccelTable::HashDataContents *A,
+                        const DwarfAccelTable::HashDataContents *B) {
+  return A->Die->getOffset() < B->Die->getOffset();
 }
 
 void DwarfAccelTable::FinalizeTable(AsmPrinter *Asm, const char *Prefix) {
   // Create the individual hash data outputs.
   for (StringMap::iterator
          EI = Entries.begin(), EE = Entries.end(); EI != EE; ++EI) {
-    struct HashData *Entry = new HashData((*EI).getKeyData());
 
     // Unique the entries.
-    std::stable_sort(EI->second.begin(), EI->second.end(), DIESorter());
+    std::stable_sort(EI->second.begin(), EI->second.end(), compareDIEs);
     EI->second.erase(std::unique(EI->second.begin(), EI->second.end()),
                        EI->second.end());
 
-    for (DataArray::const_iterator DI = EI->second.begin(),
-           DE = EI->second.end();
-         DI != DE; ++DI)
-      Entry->addData((*DI));
+    HashData *Entry = new (Allocator) HashData(EI->getKey(), EI->second);
     Data.push_back(Entry);
   }
 
@@ -216,7 +193,7 @@ void DwarfAccelTable::EmitData(AsmPrinter *Asm, DwarfDebug *D) {
                              D->getStringPool());
       Asm->OutStreamer.AddComment("Num DIEs");
       Asm->EmitInt32((*HI)->Data.size());
-      for (std::vector::const_iterator
+      for (ArrayRef::const_iterator
              DI = (*HI)->Data.begin(), DE = (*HI)->Data.end();
            DI != DE; ++DI) {
         // Emit the DIE offset
diff --git a/lib/CodeGen/AsmPrinter/DwarfAccelTable.h b/lib/CodeGen/AsmPrinter/DwarfAccelTable.h
index 2278d4c784f..963b8cdf342 100644
--- a/lib/CodeGen/AsmPrinter/DwarfAccelTable.h
+++ b/lib/CodeGen/AsmPrinter/DwarfAccelTable.h
@@ -15,6 +15,7 @@
 #define CODEGEN_ASMPRINTER_DWARFACCELTABLE_H__
 
 #include "llvm/ADT/StringMap.h"
+#include "llvm/ADT/ArrayRef.h"
 #include "llvm/MC/MCSymbol.h"
 #include "llvm/Support/Dwarf.h"
 #include "llvm/Support/DataTypes.h"
@@ -164,22 +165,12 @@ public:
 
  private:
   struct TableHeaderData {
-    
     uint32_t die_offset_base;
-    std::vector Atoms;
+    SmallVector Atoms;
+
+    TableHeaderData(ArrayRef AtomList, uint32_t offset = 0)
+      : die_offset_base(offset), Atoms(AtomList.begin(), AtomList.end()) { }
 
-    TableHeaderData(std::vector &AtomList,
-                    uint32_t offset = 0) :
-      die_offset_base(offset) {
-      for (size_t i = 0, e = AtomList.size(); i != e; ++i)
-        Atoms.push_back(AtomList[i]);
-    }
-    
-    TableHeaderData(DwarfAccelTable::Atom Atom, uint32_t offset = 0)
-    : die_offset_base(offset) {
-      Atoms.push_back(Atom);
-    }
-    
 #ifndef NDEBUG
     void print (raw_ostream &O) {
       O << "die_offset_base: " << die_offset_base << "\n";
@@ -221,11 +212,11 @@ private:
     StringRef Str;
     uint32_t HashValue;
     MCSymbol *Sym;
-    std::vector Data; // offsets
-    HashData(StringRef S) : Str(S) {
+    ArrayRef Data; // offsets
+    HashData(StringRef S, ArrayRef Data)
+      : Str(S), Data(Data) {
       HashValue = DwarfAccelTable::HashDJB(S);
     }
-    void addData(struct HashDataContents *Datum) { Data.push_back(Datum); }
     #ifndef NDEBUG
     void print(raw_ostream &O) {
       O << "Name: " << Str << "\n";
@@ -255,15 +246,18 @@ private:
   void EmitHashes(AsmPrinter *);
   void EmitOffsets(AsmPrinter *, MCSymbol *);
   void EmitData(AsmPrinter *, DwarfDebug *D);
-  
+
+  // Allocator for HashData and HashDataContents.
+  BumpPtrAllocator Allocator;
+
   // Output Variables
   TableHeader Header;
   TableHeaderData HeaderData;
   std::vector Data;
 
   // String Data
-  typedef std::vector DataArray;
-  typedef StringMap StringEntries;
+  typedef std::vector DataArray;
+  typedef StringMap StringEntries;
   StringEntries Entries;
 
   // Buckets/Hashes/Offsets
@@ -274,8 +268,7 @@ private:
   
   // Public Implementation
  public:
-  DwarfAccelTable(DwarfAccelTable::Atom);
-  DwarfAccelTable(std::vector &);
+  DwarfAccelTable(ArrayRef);
   ~DwarfAccelTable();
   void AddName(StringRef, DIE*, char = 0);
   void FinalizeTable(AsmPrinter *, const char *);
diff --git a/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp b/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp
index 69dc454ae1d..cc5b6424d73 100644
--- a/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp
+++ b/lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp
@@ -1032,9 +1032,10 @@ DIE *CompileUnit::getOrCreateSubprogramDIE(DISubprogram SP) {
   // Add function template parameters.
   addTemplateParams(*SPDie, SP.getTemplateParams());
 
-  // Unfortunately this code needs to stay here to work around
-  // a bug in older gdbs that requires the linkage name to resolve
-  // multiple template functions.
+  // Unfortunately this code needs to stay here instead of below the
+  // AT_specification code in order to work around a bug in older
+  // gdbs that requires the linkage name to resolve multiple template
+  // functions.
   StringRef LinkageName = SP.getLinkageName();
   if (!LinkageName.empty())
     addString(SPDie, dwarf::DW_AT_MIPS_linkage_name,
diff --git a/lib/CodeGen/DFAPacketizer.cpp b/lib/CodeGen/DFAPacketizer.cpp
index bfbe7790998..5ff641c7c84 100644
--- a/lib/CodeGen/DFAPacketizer.cpp
+++ b/lib/CodeGen/DFAPacketizer.cpp
@@ -23,10 +23,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "llvm/CodeGen/ScheduleDAGInstrs.h"
 #include "llvm/CodeGen/DFAPacketizer.h"
 #include "llvm/CodeGen/MachineInstr.h"
 #include "llvm/CodeGen/MachineInstrBundle.h"
+#include "llvm/CodeGen/ScheduleDAGInstrs.h"
 #include "llvm/Target/TargetInstrInfo.h"
 #include "llvm/MC/MCInstrItineraries.h"
 using namespace llvm;
@@ -100,17 +100,17 @@ void DFAPacketizer::reserveResources(llvm::MachineInstr *MI) {
   reserveResources(&MID);
 }
 
-namespace llvm {
+namespace {
 // DefaultVLIWScheduler - This class extends ScheduleDAGInstrs and overrides
 // Schedule method to build the dependence graph.
 class DefaultVLIWScheduler : public ScheduleDAGInstrs {
 public:
   DefaultVLIWScheduler(MachineFunction &MF, MachineLoopInfo &MLI,
-                   MachineDominatorTree &MDT, bool IsPostRA);
+                       MachineDominatorTree &MDT, bool IsPostRA);
   // Schedule - Actual scheduling work.
   void schedule();
 };
-}
+} // end anonymous namespace
 
 DefaultVLIWScheduler::DefaultVLIWScheduler(
   MachineFunction &MF, MachineLoopInfo &MLI, MachineDominatorTree &MDT,
@@ -129,25 +129,49 @@ VLIWPacketizerList::VLIWPacketizerList(
   bool IsPostRA) : TM(MF.getTarget()), MF(MF)  {
   TII = TM.getInstrInfo();
   ResourceTracker = TII->CreateTargetScheduleState(&TM, 0);
-  VLIWScheduler = new DefaultVLIWScheduler(MF, MLI, MDT, IsPostRA);
+  SchedulerImpl = new DefaultVLIWScheduler(MF, MLI, MDT, IsPostRA);
 }
 
 // VLIWPacketizerList Dtor
 VLIWPacketizerList::~VLIWPacketizerList() {
-  if (VLIWScheduler)
-    delete VLIWScheduler;
+  delete SchedulerImpl;
+  delete ResourceTracker;
+}
 
-  if (ResourceTracker)
-    delete ResourceTracker;
+// ignorePseudoInstruction - ignore pseudo instructions.
+bool VLIWPacketizerList::ignorePseudoInstruction(MachineInstr *MI,
+                                                 MachineBasicBlock *MBB) {
+  if (MI->isDebugValue())
+    return true;
+
+  if (TII->isSchedulingBoundary(MI, MBB, MF))
+    return true;
+
+  return false;
+}
+
+// isSoloInstruction - return true if instruction I must end previous
+// packet.
+bool VLIWPacketizerList::isSoloInstruction(MachineInstr *I) {
+  if (I->isInlineAsm())
+    return true;
+
+  return false;
+}
+
+// addToPacket - Add I to the current packet and reserve resource.
+void VLIWPacketizerList::addToPacket(MachineInstr *MI) {
+  CurrentPacketMIs.push_back(MI);
+  ResourceTracker->reserveResources(MI);
 }
 
 // endPacket - End the current packet, bundle packet instructions and reset
 // DFA state.
 void VLIWPacketizerList::endPacket(MachineBasicBlock *MBB,
-                                         MachineInstr *MI) {
+                                         MachineInstr *I) {
   if (CurrentPacketMIs.size() > 1) {
     MachineInstr *MIFirst = CurrentPacketMIs.front();
-    finalizeBundle(*MBB, MIFirst, MI);
+    finalizeBundle(*MBB, MIFirst, I);
   }
   CurrentPacketMIs.clear();
   ResourceTracker->clearResources();
@@ -157,36 +181,31 @@ void VLIWPacketizerList::endPacket(MachineBasicBlock *MBB,
 void VLIWPacketizerList::PacketizeMIs(MachineBasicBlock *MBB,
                                       MachineBasicBlock::iterator BeginItr,
                                       MachineBasicBlock::iterator EndItr) {
-  assert(VLIWScheduler && "VLIW Scheduler is not initialized!");
-  VLIWScheduler->enterRegion(MBB, BeginItr, EndItr, MBB->size());
-  VLIWScheduler->schedule();
-  VLIWScheduler->exitRegion();
+  assert(MBB->end() == EndItr && "Bad EndIndex");
 
-  // Generate MI -> SU map.
-  //std::map  MIToSUnit;
-  MIToSUnit.clear();
-  for (unsigned i = 0, e = VLIWScheduler->SUnits.size(); i != e; ++i) {
-    SUnit *SU = &VLIWScheduler->SUnits[i];
-    MIToSUnit[SU->getInstr()] = SU;
-  }
+  SchedulerImpl->enterRegion(MBB, BeginItr, EndItr, MBB->size());
+
+  // Build the DAG without reordering instructions.
+  SchedulerImpl->schedule();
+
+  // Remember scheduling units.
+  SUnits = SchedulerImpl->SUnits;
 
   // The main packetizer loop.
   for (; BeginItr != EndItr; ++BeginItr) {
     MachineInstr *MI = BeginItr;
 
-    this->initPacketizerState();
+    // Ignore pseudo instructions.
+    if (ignorePseudoInstruction(MI, MBB))
+      continue;
 
     // End the current packet if needed.
-    if (this->isSoloInstruction(MI)) {
+    if (isSoloInstruction(MI)) {
       endPacket(MBB, MI);
       continue;
     }
 
-    // Ignore pseudo instructions.
-    if (this->ignorePseudoInstruction(MI, MBB))
-      continue;
-
-    SUnit *SUI = MIToSUnit[MI];
+    SUnit *SUI = SchedulerImpl->getSUnit(MI);
     assert(SUI && "Missing SUnit Info!");
 
     // Ask DFA if machine resource is available for MI.
@@ -196,13 +215,13 @@ void VLIWPacketizerList::PacketizeMIs(MachineBasicBlock *MBB,
       for (std::vector::iterator VI = CurrentPacketMIs.begin(),
            VE = CurrentPacketMIs.end(); VI != VE; ++VI) {
         MachineInstr *MJ = *VI;
-        SUnit *SUJ = MIToSUnit[MJ];
+        SUnit *SUJ = SchedulerImpl->getSUnit(MJ);
         assert(SUJ && "Missing SUnit Info!");
 
         // Is it legal to packetize SUI and SUJ together.
-        if (!this->isLegalToPacketizeTogether(SUI, SUJ)) {
+        if (!isLegalToPacketizeTogether(SUI, SUJ)) {
           // Allow packetization if dependency can be pruned.
-          if (!this->isLegalToPruneDependencies(SUI, SUJ)) {
+          if (!isLegalToPruneDependencies(SUI, SUJ)) {
             // End the packet if dependency cannot be pruned.
             endPacket(MBB, MI);
             break;
@@ -215,9 +234,11 @@ void VLIWPacketizerList::PacketizeMIs(MachineBasicBlock *MBB,
     }
 
     // Add MI to the current packet.
-    BeginItr = this->addToPacket(MI);
+    addToPacket(MI);
   } // For all instructions in BB.
 
   // End any packet left behind.
   endPacket(MBB, EndItr);
+
+  SchedulerImpl->exitRegion();
 }
diff --git a/lib/CodeGen/LiveIntervalAnalysis.cpp b/lib/CodeGen/LiveIntervalAnalysis.cpp
index 3ade66097cb..934cc124c77 100644
--- a/lib/CodeGen/LiveIntervalAnalysis.cpp
+++ b/lib/CodeGen/LiveIntervalAnalysis.cpp
@@ -1068,9 +1068,9 @@ public:
 
 #ifndef NDEBUG
     LIValidator validator;
-    std::for_each(Entering.begin(), Entering.end(), validator);
-    std::for_each(Internal.begin(), Internal.end(), validator);
-    std::for_each(Exiting.begin(), Exiting.end(), validator);
+    validator = std::for_each(Entering.begin(), Entering.end(), validator);
+    validator = std::for_each(Internal.begin(), Internal.end(), validator);
+    validator = std::for_each(Exiting.begin(), Exiting.end(), validator);
     assert(validator.rangesOk() && "moveAllOperandsFrom broke liveness.");
 #endif
 
@@ -1115,9 +1115,9 @@ public:
 
 #ifndef NDEBUG
     LIValidator validator;
-    std::for_each(Entering.begin(), Entering.end(), validator);
-    std::for_each(Internal.begin(), Internal.end(), validator);
-    std::for_each(Exiting.begin(), Exiting.end(), validator);
+    validator = std::for_each(Entering.begin(), Entering.end(), validator);
+    validator = std::for_each(Internal.begin(), Internal.end(), validator);
+    validator = std::for_each(Exiting.begin(), Exiting.end(), validator);
     assert(validator.rangesOk() && "moveAllOperandsInto broke liveness.");
 #endif
   }
diff --git a/lib/CodeGen/MachineBasicBlock.cpp b/lib/CodeGen/MachineBasicBlock.cpp
index 6c8a1072697..1abb8f240cc 100644
--- a/lib/CodeGen/MachineBasicBlock.cpp
+++ b/lib/CodeGen/MachineBasicBlock.cpp
@@ -392,22 +392,44 @@ void MachineBasicBlock::updateTerminator() {
         TII->InsertBranch(*this, TBB, 0, Cond, dl);
       }
     } else {
+      // Walk through the successors and find the successor which is not
+      // a landing pad and is not the conditional branch destination (in TBB)
+      // as the fallthrough successor.
+      MachineBasicBlock *FallthroughBB = 0;
+      for (succ_iterator SI = succ_begin(), SE = succ_end(); SI != SE; ++SI) {
+        if ((*SI)->isLandingPad() || *SI == TBB)
+          continue;
+        assert(!FallthroughBB && "Found more than one fallthrough successor.");
+        FallthroughBB = *SI;
+      }
+      if (!FallthroughBB && canFallThrough()) {
+        // We fallthrough to the same basic block as the conditional jump
+        // targets. Remove the conditional jump, leaving unconditional
+        // fallthrough.
+        // FIXME: This does not seem like a reasonable pattern to support, but it
+        // has been seen in the wild coming out of degenerate ARM test cases.
+        TII->RemoveBranch(*this);
+
+        // Finally update the unconditional successor to be reached via a branch
+        // if it would not be reached by fallthrough.
+        if (!isLayoutSuccessor(TBB))
+          TII->InsertBranch(*this, TBB, 0, Cond, dl);
+        return;
+      }
+
       // The block has a fallthrough conditional branch.
-      MachineBasicBlock *MBBA = *succ_begin();
-      MachineBasicBlock *MBBB = *llvm::next(succ_begin());
-      if (MBBA == TBB) std::swap(MBBB, MBBA);
       if (isLayoutSuccessor(TBB)) {
         if (TII->ReverseBranchCondition(Cond)) {
           // We can't reverse the condition, add an unconditional branch.
           Cond.clear();
-          TII->InsertBranch(*this, MBBA, 0, Cond, dl);
+          TII->InsertBranch(*this, FallthroughBB, 0, Cond, dl);
           return;
         }
         TII->RemoveBranch(*this);
-        TII->InsertBranch(*this, MBBA, 0, Cond, dl);
-      } else if (!isLayoutSuccessor(MBBA)) {
+        TII->InsertBranch(*this, FallthroughBB, 0, Cond, dl);
+      } else if (!isLayoutSuccessor(FallthroughBB)) {
         TII->RemoveBranch(*this);
-        TII->InsertBranch(*this, TBB, MBBA, Cond, dl);
+        TII->InsertBranch(*this, TBB, FallthroughBB, Cond, dl);
       }
     }
   }
diff --git a/lib/CodeGen/MachineBlockPlacement.cpp b/lib/CodeGen/MachineBlockPlacement.cpp
index 22d7212007f..5ba68517b7a 100644
--- a/lib/CodeGen/MachineBlockPlacement.cpp
+++ b/lib/CodeGen/MachineBlockPlacement.cpp
@@ -102,13 +102,13 @@ public:
   }
 
   /// \brief Iterator over blocks within the chain.
-  typedef SmallVectorImpl::const_iterator iterator;
+  typedef SmallVectorImpl::iterator iterator;
 
   /// \brief Beginning of blocks within the chain.
-  iterator begin() const { return Blocks.begin(); }
+  iterator begin() { return Blocks.begin(); }
 
   /// \brief End of blocks within the chain.
-  iterator end() const { return Blocks.end(); }
+  iterator end() { return Blocks.end(); }
 
   /// \brief Merge a block chain into this one.
   ///
@@ -211,12 +211,15 @@ class MachineBlockPlacement : public MachineFunctionPass {
   void buildChain(MachineBasicBlock *BB, BlockChain &Chain,
                   SmallVectorImpl &BlockWorkList,
                   const BlockFilterSet *BlockFilter = 0);
-  MachineBasicBlock *findBestLoopTop(MachineFunction &F,
-                                     MachineLoop &L,
+  MachineBasicBlock *findBestLoopTop(MachineLoop &L,
                                      const BlockFilterSet &LoopBlockSet);
+  MachineBasicBlock *findBestLoopExit(MachineFunction &F,
+                                      MachineLoop &L,
+                                      const BlockFilterSet &LoopBlockSet);
   void buildLoopChains(MachineFunction &F, MachineLoop &L);
+  void rotateLoop(BlockChain &LoopChain, MachineBasicBlock *ExitingBB,
+                  const BlockFilterSet &LoopBlockSet);
   void buildCFGChains(MachineFunction &F);
-  void AlignLoops(MachineFunction &F);
 
 public:
   static char ID; // Pass identification, replacement for typeid
@@ -540,13 +543,74 @@ void MachineBlockPlacement::buildChain(
 
 /// \brief Find the best loop top block for layout.
 ///
+/// Look for a block which is strictly better than the loop header for laying
+/// out at the top of the loop. This looks for one and only one pattern:
+/// a latch block with no conditional exit. This block will cause a conditional
+/// jump around it or will be the bottom of the loop if we lay it out in place,
+/// but if it it doesn't end up at the bottom of the loop for any reason,
+/// rotation alone won't fix it. Because such a block will always result in an
+/// unconditional jump (for the backedge) rotating it in front of the loop
+/// header is always profitable.
+MachineBasicBlock *
+MachineBlockPlacement::findBestLoopTop(MachineLoop &L,
+                                       const BlockFilterSet &LoopBlockSet) {
+  // Check that the header hasn't been fused with a preheader block due to
+  // crazy branches. If it has, we need to start with the header at the top to
+  // prevent pulling the preheader into the loop body.
+  BlockChain &HeaderChain = *BlockToChain[L.getHeader()];
+  if (!LoopBlockSet.count(*HeaderChain.begin()))
+    return L.getHeader();
+
+  DEBUG(dbgs() << "Finding best loop top for: "
+               << getBlockName(L.getHeader()) << "\n");
+
+  BlockFrequency BestPredFreq;
+  MachineBasicBlock *BestPred = 0;
+  for (MachineBasicBlock::pred_iterator PI = L.getHeader()->pred_begin(),
+                                        PE = L.getHeader()->pred_end();
+       PI != PE; ++PI) {
+    MachineBasicBlock *Pred = *PI;
+    if (!LoopBlockSet.count(Pred))
+      continue;
+    DEBUG(dbgs() << "    header pred: " << getBlockName(Pred) << ", "
+                 << Pred->succ_size() << " successors, "
+                 << MBFI->getBlockFreq(Pred) << " freq\n");
+    if (Pred->succ_size() > 1)
+      continue;
+
+    BlockFrequency PredFreq = MBFI->getBlockFreq(Pred);
+    if (!BestPred || PredFreq > BestPredFreq ||
+        (!(PredFreq < BestPredFreq) &&
+         Pred->isLayoutSuccessor(L.getHeader()))) {
+      BestPred = Pred;
+      BestPredFreq = PredFreq;
+    }
+  }
+
+  // If no direct predecessor is fine, just use the loop header.
+  if (!BestPred)
+    return L.getHeader();
+
+  // Walk backwards through any straight line of predecessors.
+  while (BestPred->pred_size() == 1 &&
+         (*BestPred->pred_begin())->succ_size() == 1 &&
+         *BestPred->pred_begin() != L.getHeader())
+    BestPred = *BestPred->pred_begin();
+
+  DEBUG(dbgs() << "    final top: " << getBlockName(BestPred) << "\n");
+  return BestPred;
+}
+
+
+/// \brief Find the best loop exiting block for layout.
+///
 /// This routine implements the logic to analyze the loop looking for the best
 /// block to layout at the top of the loop. Typically this is done to maximize
 /// fallthrough opportunities.
 MachineBasicBlock *
-MachineBlockPlacement::findBestLoopTop(MachineFunction &F,
-                                       MachineLoop &L,
-                                       const BlockFilterSet &LoopBlockSet) {
+MachineBlockPlacement::findBestLoopExit(MachineFunction &F,
+                                        MachineLoop &L,
+                                        const BlockFilterSet &LoopBlockSet) {
   // We don't want to layout the loop linearly in all cases. If the loop header
   // is just a normal basic block in the loop, we want to look for what block
   // within the loop is the best one to layout at the top. However, if the loop
@@ -557,11 +621,11 @@ MachineBlockPlacement::findBestLoopTop(MachineFunction &F,
   // header and only rotate if safe.
   BlockChain &HeaderChain = *BlockToChain[L.getHeader()];
   if (!LoopBlockSet.count(*HeaderChain.begin()))
-    return L.getHeader();
+    return 0;
 
   BlockFrequency BestExitEdgeFreq;
+  unsigned BestExitLoopDepth = 0;
   MachineBasicBlock *ExitingBB = 0;
-  MachineBasicBlock *LoopingBB = 0;
   // If there are exits to outer loops, loop rotation can severely limit
   // fallthrough opportunites unless it selects such an exit. Keep a set of
   // blocks where rotating to exit with that block will reach an outer loop.
@@ -584,15 +648,10 @@ MachineBlockPlacement::findBestLoopTop(MachineFunction &F,
     // successor isn't found.
     MachineBasicBlock *OldExitingBB = ExitingBB;
     BlockFrequency OldBestExitEdgeFreq = BestExitEdgeFreq;
-    // We also compute and store the best looping successor for use in layout.
-    MachineBasicBlock *BestLoopSucc = 0;
+    bool HasLoopingSucc = false;
     // FIXME: Due to the performance of the probability and weight routines in
-    // the MBPI analysis, we use the internal weights. This is only valid
-    // because it is purely a ranking function, we don't care about anything
-    // but the relative values.
-    uint32_t BestLoopSuccWeight = 0;
-    // FIXME: We also manually compute the probabilities to avoid quadratic
-    // behavior.
+    // the MBPI analysis, we use the internal weights and manually compute the
+    // probabilities to avoid quadratic behavior.
     uint32_t WeightScale = 0;
     uint32_t SumWeight = MBPI->getSumForBlock(*I, WeightScale);
     for (MachineBasicBlock::succ_iterator SI = (*I)->succ_begin(),
@@ -604,10 +663,8 @@ MachineBlockPlacement::findBestLoopTop(MachineFunction &F,
         continue;
       BlockChain &SuccChain = *BlockToChain[*SI];
       // Don't split chains, either this chain or the successor's chain.
-      if (&Chain == &SuccChain || *SI != *SuccChain.begin()) {
-        DEBUG(dbgs() << "    " << (LoopBlockSet.count(*SI) ? "looping: "
-                                                           : "exiting: ")
-                     << getBlockName(*I) << " -> "
+      if (&Chain == &SuccChain) {
+        DEBUG(dbgs() << "    exiting: " << getBlockName(*I) << " -> "
                      << getBlockName(*SI) << " (chain conflict)\n");
         continue;
       }
@@ -616,60 +673,103 @@ MachineBlockPlacement::findBestLoopTop(MachineFunction &F,
       if (LoopBlockSet.count(*SI)) {
         DEBUG(dbgs() << "    looping: " << getBlockName(*I) << " -> "
                      << getBlockName(*SI) << " (" << SuccWeight << ")\n");
-        if (BestLoopSucc && BestLoopSuccWeight >= SuccWeight)
-          continue;
-
-        BestLoopSucc = *SI;
-        BestLoopSuccWeight = SuccWeight;
+        HasLoopingSucc = true;
         continue;
       }
 
+      unsigned SuccLoopDepth = 0;
+      if (MachineLoop *ExitLoop = MLI->getLoopFor(*SI)) {
+        SuccLoopDepth = ExitLoop->getLoopDepth();
+        if (ExitLoop->contains(&L))
+          BlocksExitingToOuterLoop.insert(*I);
+      }
+
       BranchProbability SuccProb(SuccWeight / WeightScale, SumWeight);
       BlockFrequency ExitEdgeFreq = MBFI->getBlockFreq(*I) * SuccProb;
       DEBUG(dbgs() << "    exiting: " << getBlockName(*I) << " -> "
-                   << getBlockName(*SI) << " (" << ExitEdgeFreq << ")\n");
+                   << getBlockName(*SI) << " [L:" << SuccLoopDepth
+                   << "] (" << ExitEdgeFreq << ")\n");
       // Note that we slightly bias this toward an existing layout successor to
       // retain incoming order in the absence of better information.
       // FIXME: Should we bias this more strongly? It's pretty weak.
-      if (!ExitingBB || ExitEdgeFreq > BestExitEdgeFreq ||
+      if (!ExitingBB || BestExitLoopDepth < SuccLoopDepth ||
+          ExitEdgeFreq > BestExitEdgeFreq ||
           ((*I)->isLayoutSuccessor(*SI) &&
            !(ExitEdgeFreq < BestExitEdgeFreq))) {
         BestExitEdgeFreq = ExitEdgeFreq;
         ExitingBB = *I;
       }
-
-      if (MachineLoop *ExitLoop = MLI->getLoopFor(*SI))
-        if (ExitLoop->contains(&L))
-          BlocksExitingToOuterLoop.insert(*I);
     }
 
     // Restore the old exiting state, no viable looping successor was found.
-    if (!BestLoopSucc) {
+    if (!HasLoopingSucc) {
       ExitingBB = OldExitingBB;
       BestExitEdgeFreq = OldBestExitEdgeFreq;
       continue;
     }
-
-    // If this was best exiting block thus far, also record the looping block.
-    if (ExitingBB == *I)
-      LoopingBB = BestLoopSucc;
   }
-  // Without a candidate exitting block or with only a single block in the
+  // Without a candidate exiting block or with only a single block in the
   // loop, just use the loop header to layout the loop.
   if (!ExitingBB || L.getNumBlocks() == 1)
-    return L.getHeader();
+    return 0;
 
   // Also, if we have exit blocks which lead to outer loops but didn't select
   // one of them as the exiting block we are rotating toward, disable loop
   // rotation altogether.
   if (!BlocksExitingToOuterLoop.empty() &&
       !BlocksExitingToOuterLoop.count(ExitingBB))
-    return L.getHeader();
+    return 0;
 
-  assert(LoopingBB && "All successors of a loop block are exit blocks!");
   DEBUG(dbgs() << "  Best exiting block: " << getBlockName(ExitingBB) << "\n");
-  DEBUG(dbgs() << "  Best top block: " << getBlockName(LoopingBB) << "\n");
-  return LoopingBB;
+  return ExitingBB;
+}
+
+/// \brief Attempt to rotate an exiting block to the bottom of the loop.
+///
+/// Once we have built a chain, try to rotate it to line up the hot exit block
+/// with fallthrough out of the loop if doing so doesn't introduce unnecessary
+/// branches. For example, if the loop has fallthrough into its header and out
+/// of its bottom already, don't rotate it.
+void MachineBlockPlacement::rotateLoop(BlockChain &LoopChain,
+                                       MachineBasicBlock *ExitingBB,
+                                       const BlockFilterSet &LoopBlockSet) {
+  if (!ExitingBB)
+    return;
+
+  MachineBasicBlock *Top = *LoopChain.begin();
+  bool ViableTopFallthrough = false;
+  for (MachineBasicBlock::pred_iterator PI = Top->pred_begin(),
+                                        PE = Top->pred_end();
+       PI != PE; ++PI) {
+    BlockChain *PredChain = BlockToChain[*PI];
+    if (!LoopBlockSet.count(*PI) &&
+        (!PredChain || *PI == *llvm::prior(PredChain->end()))) {
+      ViableTopFallthrough = true;
+      break;
+    }
+  }
+
+  // If the header has viable fallthrough, check whether the current loop
+  // bottom is a viable exiting block. If so, bail out as rotating will
+  // introduce an unnecessary branch.
+  if (ViableTopFallthrough) {
+    MachineBasicBlock *Bottom = *llvm::prior(LoopChain.end());
+    for (MachineBasicBlock::succ_iterator SI = Bottom->succ_begin(),
+                                          SE = Bottom->succ_end();
+         SI != SE; ++SI) {
+      BlockChain *SuccChain = BlockToChain[*SI];
+      if (!LoopBlockSet.count(*SI) &&
+          (!SuccChain || *SI == *SuccChain->begin()))
+        return;
+    }
+  }
+
+  BlockChain::iterator ExitIt = std::find(LoopChain.begin(), LoopChain.end(),
+                                          ExitingBB);
+  if (ExitIt == LoopChain.end())
+    return;
+
+  std::rotate(LoopChain.begin(), llvm::next(ExitIt), LoopChain.end());
 }
 
 /// \brief Forms basic block chains from the natural loop structures.
@@ -688,8 +788,20 @@ void MachineBlockPlacement::buildLoopChains(MachineFunction &F,
   SmallVector BlockWorkList;
   BlockFilterSet LoopBlockSet(L.block_begin(), L.block_end());
 
-  MachineBasicBlock *LayoutTop = findBestLoopTop(F, L, LoopBlockSet);
-  BlockChain &LoopChain = *BlockToChain[LayoutTop];
+  // First check to see if there is an obviously preferable top block for the
+  // loop. This will default to the header, but may end up as one of the
+  // predecessors to the header if there is one which will result in strictly
+  // fewer branches in the loop body.
+  MachineBasicBlock *LoopTop = findBestLoopTop(L, LoopBlockSet);
+
+  // If we selected just the header for the loop top, look for a potentially
+  // profitable exit block in the event that rotating the loop can eliminate
+  // branches by placing an exit edge at the bottom.
+  MachineBasicBlock *ExitingBB = 0;
+  if (LoopTop == L.getHeader())
+    ExitingBB = findBestLoopExit(F, L, LoopBlockSet);
+
+  BlockChain &LoopChain = *BlockToChain[LoopTop];
 
   // FIXME: This is a really lame way of walking the chains in the loop: we
   // walk the blocks, and use a set to prevent visiting a particular chain
@@ -721,7 +833,8 @@ void MachineBlockPlacement::buildLoopChains(MachineFunction &F,
       BlockWorkList.push_back(*Chain.begin());
   }
 
-  buildChain(LayoutTop, LoopChain, BlockWorkList, &LoopBlockSet);
+  buildChain(LoopTop, LoopChain, BlockWorkList, &LoopBlockSet);
+  rotateLoop(LoopChain, ExitingBB, LoopBlockSet);
 
   DEBUG({
     // Crash at the end so we get all of the debugging output first.
@@ -733,7 +846,8 @@ void MachineBlockPlacement::buildLoopChains(MachineFunction &F,
              << "  Chain header: " << getBlockName(*LoopChain.begin()) << "\n";
     }
     for (BlockChain::iterator BCI = LoopChain.begin(), BCE = LoopChain.end();
-         BCI != BCE; ++BCI)
+         BCI != BCE; ++BCI) {
+      dbgs() << "          ... " << getBlockName(*BCI) << "\n";
       if (!LoopBlockSet.erase(*BCI)) {
         // We don't mark the loop as bad here because there are real situations
         // where this can occur. For example, with an unanalyzable fallthrough
@@ -743,6 +857,7 @@ void MachineBlockPlacement::buildLoopChains(MachineFunction &F,
                << "  Chain header: " << getBlockName(*LoopChain.begin()) << "\n"
                << "  Bad block:    " << getBlockName(*BCI) << "\n";
       }
+    }
 
     if (!LoopBlockSet.empty()) {
       BadLoop = true;
@@ -882,28 +997,33 @@ void MachineBlockPlacement::buildCFGChains(MachineFunction &F) {
   MachineBasicBlock *TBB = 0, *FBB = 0; // For AnalyzeBranch.
   if (!TII->AnalyzeBranch(F.back(), TBB, FBB, Cond))
     F.back().updateTerminator();
-}
 
-/// \brief Recursive helper to align a loop and any nested loops.
-static void AlignLoop(MachineFunction &F, MachineLoop *L, unsigned Align) {
-  // Recurse through nested loops.
-  for (MachineLoop::iterator I = L->begin(), E = L->end(); I != E; ++I)
-    AlignLoop(F, *I, Align);
-
-  L->getTopBlock()->setAlignment(Align);
-}
-
-/// \brief Align loop headers to target preferred alignments.
-void MachineBlockPlacement::AlignLoops(MachineFunction &F) {
+  // Walk through the backedges of the function now that we have fully laid out
+  // the basic blocks and align the destination of each backedge. We don't rely
+  // on the loop info here so that we can align backedges in unnatural CFGs and
+  // backedges that were introduced purely because of the loop rotations done
+  // during this layout pass.
+  // FIXME: This isn't quite right, we shouldn't align backedges that result
+  // from blocks being sunken below the exit block for the function.
   if (F.getFunction()->hasFnAttr(Attribute::OptimizeForSize))
     return;
-
   unsigned Align = TLI->getPrefLoopAlignment();
   if (!Align)
     return;  // Don't care about loop alignment.
 
-  for (MachineLoopInfo::iterator I = MLI->begin(), E = MLI->end(); I != E; ++I)
-    AlignLoop(F, *I, Align);
+  SmallPtrSet PreviousBlocks;
+  for (BlockChain::iterator BI = FunctionChain.begin(),
+                            BE = FunctionChain.end();
+       BI != BE; ++BI) {
+    PreviousBlocks.insert(*BI);
+    // Set alignment on the destination of all the back edges in the new
+    // ordering.
+    for (MachineBasicBlock::succ_iterator SI = (*BI)->succ_begin(),
+                                          SE = (*BI)->succ_end();
+         SI != SE; ++SI)
+      if (PreviousBlocks.count(*SI))
+        (*SI)->setAlignment(Align);
+  }
 }
 
 bool MachineBlockPlacement::runOnMachineFunction(MachineFunction &F) {
@@ -919,7 +1039,6 @@ bool MachineBlockPlacement::runOnMachineFunction(MachineFunction &F) {
   assert(BlockToChain.empty());
 
   buildCFGChains(F);
-  AlignLoops(F);
 
   BlockToChain.clear();
   ChainAllocator.DestroyAll();
diff --git a/lib/CodeGen/Passes.cpp b/lib/CodeGen/Passes.cpp
index 53d1fcf7377..490547bbb87 100644
--- a/lib/CodeGen/Passes.cpp
+++ b/lib/CodeGen/Passes.cpp
@@ -37,8 +37,9 @@ static cl::opt DisableTailDuplicate("disable-tail-duplicate", cl::Hidden,
     cl::desc("Disable tail duplication"));
 static cl::opt DisableEarlyTailDup("disable-early-taildup", cl::Hidden,
     cl::desc("Disable pre-register allocation tail duplication"));
-static cl::opt EnableBlockPlacement("enable-block-placement",
-    cl::Hidden, cl::desc("Enable probability-driven block placement"));
+static cl::opt DisableBlockPlacement("disable-block-placement",
+    cl::Hidden, cl::desc("Disable the probability-driven block placement, and "
+                         "re-enable the old code placement pass"));
 static cl::opt EnableBlockPlacementStats("enable-block-placement-stats",
     cl::Hidden, cl::desc("Collect probability-driven block placement stats"));
 static cl::opt DisableCodePlace("disable-code-place", cl::Hidden,
@@ -206,7 +207,7 @@ TargetPassConfig::~TargetPassConfig() {
 // Out of line constructor provides default values for pass options and
 // registers all common codegen passes.
 TargetPassConfig::TargetPassConfig(TargetMachine *tm, PassManagerBase &pm)
-  : ImmutablePass(ID), TM(tm), PM(pm), Impl(0), Initialized(false),
+  : ImmutablePass(ID), TM(tm), PM(&pm), Impl(0), Initialized(false),
     DisableVerify(false),
     EnableTailMerge(true) {
 
@@ -233,7 +234,7 @@ TargetPassConfig *LLVMTargetMachine::createPassConfig(PassManagerBase &PM) {
 }
 
 TargetPassConfig::TargetPassConfig()
-  : ImmutablePass(ID), PM(*(PassManagerBase*)0) {
+  : ImmutablePass(ID), PM(0) {
   llvm_unreachable("TargetPassConfig should not be constructed on-the-fly");
 }
 
@@ -268,16 +269,16 @@ AnalysisID TargetPassConfig::addPass(char &ID) {
   Pass *P = Pass::createPass(FinalID);
   if (!P)
     llvm_unreachable("Pass ID not registered");
-  PM.add(P);
+  PM->add(P);
   return FinalID;
 }
 
 void TargetPassConfig::printAndVerify(const char *Banner) const {
   if (TM->shouldPrintMachineCode())
-    PM.add(createMachineFunctionPrinterPass(dbgs(), Banner));
+    PM->add(createMachineFunctionPrinterPass(dbgs(), Banner));
 
   if (VerifyMachineCode)
-    PM.add(createMachineVerifierPass(Banner));
+    PM->add(createMachineVerifierPass(Banner));
 }
 
 /// Add common target configurable passes that perform LLVM IR to IR transforms
@@ -287,46 +288,46 @@ void TargetPassConfig::addIRPasses() {
   // Add TypeBasedAliasAnalysis before BasicAliasAnalysis so that
   // BasicAliasAnalysis wins if they disagree. This is intended to help
   // support "obvious" type-punning idioms.
-  PM.add(createTypeBasedAliasAnalysisPass());
-  PM.add(createBasicAliasAnalysisPass());
+  PM->add(createTypeBasedAliasAnalysisPass());
+  PM->add(createBasicAliasAnalysisPass());
 
   // Before running any passes, run the verifier to determine if the input
   // coming from the front-end and/or optimizer is valid.
   if (!DisableVerify)
-    PM.add(createVerifierPass());
+    PM->add(createVerifierPass());
 
   // Run loop strength reduction before anything else.
   if (getOptLevel() != CodeGenOpt::None && !DisableLSR) {
-    PM.add(createLoopStrengthReducePass(getTargetLowering()));
+    PM->add(createLoopStrengthReducePass(getTargetLowering()));
     if (PrintLSR)
-      PM.add(createPrintFunctionPass("\n\n*** Code after LSR ***\n", &dbgs()));
+      PM->add(createPrintFunctionPass("\n\n*** Code after LSR ***\n", &dbgs()));
   }
 
-  PM.add(createGCLoweringPass());
+  PM->add(createGCLoweringPass());
 
   // Make sure that no unreachable blocks are instruction selected.
-  PM.add(createUnreachableBlockEliminationPass());
+  PM->add(createUnreachableBlockEliminationPass());
 }
 
 /// Add common passes that perform LLVM IR to IR transforms in preparation for
 /// instruction selection.
 void TargetPassConfig::addISelPrepare() {
   if (getOptLevel() != CodeGenOpt::None && !DisableCGP)
-    PM.add(createCodeGenPreparePass(getTargetLowering()));
+    PM->add(createCodeGenPreparePass(getTargetLowering()));
 
-  PM.add(createStackProtectorPass(getTargetLowering()));
+  PM->add(createStackProtectorPass(getTargetLowering()));
 
   addPreISel();
 
   if (PrintISelInput)
-    PM.add(createPrintFunctionPass("\n\n"
-                                   "*** Final LLVM Code input to ISel ***\n",
-                                   &dbgs()));
+    PM->add(createPrintFunctionPass("\n\n"
+                                    "*** Final LLVM Code input to ISel ***\n",
+                                    &dbgs()));
 
   // All passes which modify the LLVM IR are now complete; run the verifier
   // to ensure that the IR is valid.
   if (!DisableVerify)
-    PM.add(createVerifierPass());
+    PM->add(createVerifierPass());
 }
 
 /// Add the complete set of target-independent postISel code generator passes.
@@ -404,7 +405,7 @@ void TargetPassConfig::addMachinePasses() {
   // GC
   addPass(GCMachineCodeAnalysisID);
   if (PrintGCInfo)
-    PM.add(createGCInfoPrinter(dbgs()));
+    PM->add(createGCInfoPrinter(dbgs()));
 
   // Basic block placement.
   if (getOptLevel() != CodeGenOpt::None)
@@ -521,7 +522,7 @@ void TargetPassConfig::addFastRegAlloc(FunctionPass *RegAllocPass) {
   addPass(PHIEliminationID);
   addPass(TwoAddressInstructionPassID);
 
-  PM.add(RegAllocPass);
+  PM->add(RegAllocPass);
   printAndVerify("After Register Allocation");
 }
 
@@ -563,7 +564,7 @@ void TargetPassConfig::addOptimizedRegAlloc(FunctionPass *RegAllocPass) {
     printAndVerify("After Machine Scheduling");
 
   // Add the selected register allocation pass.
-  PM.add(RegAllocPass);
+  PM->add(RegAllocPass);
   printAndVerify("After Register Allocation");
 
   // FinalizeRegAlloc is convenient until MachineInstrBundles is more mature,
@@ -610,10 +611,10 @@ void TargetPassConfig::addMachineLateOptimization() {
 /// Add standard basic block placement passes.
 void TargetPassConfig::addBlockPlacement() {
   AnalysisID ID = &NoPassID;
-  if (EnableBlockPlacement) {
-    // MachineBlockPlacement is an experimental pass which is disabled by
-    // default currently. Eventually it should subsume CodePlacementOpt, so
-    // when enabled, the other is disabled.
+  if (!DisableBlockPlacement) {
+    // MachineBlockPlacement is a new pass which subsumes the functionality of
+    // CodPlacementOpt. The old code placement pass can be restored by
+    // disabling block placement, but eventually it will be removed.
     ID = addPass(MachineBlockPlacementID);
   } else {
     ID = addPass(CodePlacementOptID);
diff --git a/lib/CodeGen/ScheduleDAGInstrs.cpp b/lib/CodeGen/ScheduleDAGInstrs.cpp
index 6be1ab7f5b0..d46eb896e54 100644
--- a/lib/CodeGen/ScheduleDAGInstrs.cpp
+++ b/lib/CodeGen/ScheduleDAGInstrs.cpp
@@ -39,8 +39,8 @@ ScheduleDAGInstrs::ScheduleDAGInstrs(MachineFunction &mf,
                                      LiveIntervals *lis)
   : ScheduleDAG(mf), MLI(mli), MDT(mdt), MFI(mf.getFrameInfo()),
     InstrItins(mf.getTarget().getInstrItineraryData()), LIS(lis),
-    IsPostRA(IsPostRAFlag), UnitLatencies(false), LoopRegs(MLI, MDT),
-    FirstDbgValue(0) {
+    IsPostRA(IsPostRAFlag), UnitLatencies(false), CanHandleTerminators(false),
+    LoopRegs(MLI, MDT), FirstDbgValue(0) {
   assert((IsPostRA || LIS) && "PreRA scheduling requires LiveIntervals");
   DbgValues.clear();
   assert(!(IsPostRA && MRI.getNumVirtRegs()) &&
@@ -554,7 +554,7 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA) {
       continue;
     }
 
-    assert(!MI->isTerminator() && !MI->isLabel() &&
+    assert((!MI->isTerminator() || CanHandleTerminators) && !MI->isLabel() &&
            "Cannot schedule terminators or labels!");
 
     SUnit *SU = MISUnitMap[MI];
diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index d1b998f8d84..0914c662766 100644
--- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -1080,6 +1080,7 @@ void DAGCombiner::Run(CombineLevel AtLevel) {
 
   // If the root changed (e.g. it was a dead load, update the root).
   DAG.setRoot(Dummy.getValue());
+  DAG.RemoveDeadNodes();
 }
 
 SDValue DAGCombiner::visit(SDNode *N) {
diff --git a/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp b/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp
index 3ae8345bd19..9fe4480d113 100644
--- a/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp
+++ b/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp
@@ -417,7 +417,8 @@ SDValue VectorLegalizer::ExpandVSELECT(SDValue Op) {
 
   Op1 = DAG.getNode(ISD::AND, DL, VT, Op1, Mask);
   Op2 = DAG.getNode(ISD::AND, DL, VT, Op2, NotMask);
-  return DAG.getNode(ISD::OR, DL, VT, Op1, Op2);
+  SDValue Val = DAG.getNode(ISD::OR, DL, VT, Op1, Op2);
+  return DAG.getNode(ISD::BITCAST, DL, Op.getValueType(), Val);
 }
 
 SDValue VectorLegalizer::ExpandUINT_TO_FLOAT(SDValue Op) {
diff --git a/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp b/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
index 69dd813b24e..8ec1ae8620f 100644
--- a/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
+++ b/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
@@ -138,9 +138,11 @@ static void AddGlue(SDNode *N, SDValue Glue, bool AddGlue, SelectionDAG *DAG) {
   // Don't add glue from a node to itself.
   if (GlueDestNode == N) return;
 
-  // Don't add glue to something which already has glue.
-  if (N->getValueType(N->getNumValues() - 1) == MVT::Glue) return;
-
+  // Don't add glue to something that already has it, either as a use or value.
+  if (N->getOperand(N->getNumOperands()-1).getValueType() == MVT::Glue ||
+      N->getValueType(N->getNumValues() - 1) == MVT::Glue) {
+    return;
+  }
   for (unsigned I = 0, E = N->getNumValues(); I != E; ++I)
     VTs.push_back(N->getValueType(I));
 
diff --git a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index 94cb95804f6..f1e879be956 100644
--- a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -5050,7 +5050,7 @@ SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I, unsigned Intrinsic) {
   }
   case Intrinsic::gcroot:
     if (GFI) {
-      const Value *Alloca = I.getArgOperand(0);
+      const Value *Alloca = I.getArgOperand(0)->stripPointerCasts();
       const Constant *TypeMap = cast(I.getArgOperand(1));
 
       FrameIndexSDNode *FI = cast(getValue(Alloca).getNode());
diff --git a/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index 09a2b1f3d7a..e341e15e41a 100644
--- a/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -1367,8 +1367,9 @@ bool TargetLowering::SimplifyDemandedBits(SDValue Op,
     // bits on that side are also known to be set on the other side, turn this
     // into an AND, as we know the bits will be cleared.
     //    e.g. (X | C1) ^ C2 --> (X | C1) & ~C2 iff (C1&C2) == C2
-    if ((NewMask & (KnownZero|KnownOne)) == NewMask) { // all known
-      if ((KnownOne & KnownOne2) == KnownOne) {
+    // NB: it is okay if more bits are known than are requested
+    if ((NewMask & (KnownZero|KnownOne)) == NewMask) { // all known on one side 
+      if (KnownOne == KnownOne2) { // set bits are the same on both sides
         EVT VT = Op.getValueType();
         SDValue ANDC = TLO.DAG.getConstant(~KnownOne & NewMask, VT);
         return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::AND, dl, VT,
diff --git a/lib/CodeGen/SlotIndexes.cpp b/lib/CodeGen/SlotIndexes.cpp
index c5bd3a3cae6..26cf2594416 100644
--- a/lib/CodeGen/SlotIndexes.cpp
+++ b/lib/CodeGen/SlotIndexes.cpp
@@ -34,7 +34,8 @@ void SlotIndexes::releaseMemory() {
   mi2iMap.clear();
   MBBRanges.clear();
   idx2MBBMap.clear();
-  clearList();
+  indexList.clear();
+  ileAllocator.Reset();
 }
 
 bool SlotIndexes::runOnMachineFunction(MachineFunction &fn) {
@@ -45,17 +46,15 @@ bool SlotIndexes::runOnMachineFunction(MachineFunction &fn) {
   // iterator in lock-step (though skipping it over indexes which have
   // null pointers in the instruction field).
   // At each iteration assert that the instruction pointed to in the index
-  // is the same one pointed to by the MI iterator. This 
+  // is the same one pointed to by the MI iterator. This
 
   // FIXME: This can be simplified. The mi2iMap_, Idx2MBBMap, etc. should
   // only need to be set up once after the first numbering is computed.
 
   mf = &fn;
-  initList();
 
   // Check that the list contains only the sentinal.
-  assert(indexListHead->getNext() == 0 &&
-         "Index list non-empty at initial numbering?");
+  assert(indexList.empty() && "Index list non-empty at initial numbering?");
   assert(idx2MBBMap.empty() &&
          "Index -> MBB mapping non-empty at initial numbering?");
   assert(MBBRanges.empty() &&
@@ -68,7 +67,7 @@ bool SlotIndexes::runOnMachineFunction(MachineFunction &fn) {
   MBBRanges.resize(mf->getNumBlockIDs());
   idx2MBBMap.reserve(mf->size());
 
-  push_back(createEntry(0, index));
+  indexList.push_back(createEntry(0, index));
 
   // Iterate over the function.
   for (MachineFunction::iterator mbbItr = mf->begin(), mbbEnd = mf->end();
@@ -76,7 +75,7 @@ bool SlotIndexes::runOnMachineFunction(MachineFunction &fn) {
     MachineBasicBlock *mbb = &*mbbItr;
 
     // Insert an index for the MBB start.
-    SlotIndex blockStartIndex(back(), SlotIndex::Slot_Block);
+    SlotIndex blockStartIndex(&indexList.back(), SlotIndex::Slot_Block);
 
     for (MachineBasicBlock::iterator miItr = mbb->begin(), miEnd = mbb->end();
          miItr != miEnd; ++miItr) {
@@ -85,20 +84,20 @@ bool SlotIndexes::runOnMachineFunction(MachineFunction &fn) {
         continue;
 
       // Insert a store index for the instr.
-      push_back(createEntry(mi, index += SlotIndex::InstrDist));
+      indexList.push_back(createEntry(mi, index += SlotIndex::InstrDist));
 
       // Save this base index in the maps.
-      mi2iMap.insert(std::make_pair(mi, SlotIndex(back(),
+      mi2iMap.insert(std::make_pair(mi, SlotIndex(&indexList.back(),
                                                   SlotIndex::Slot_Block)));
- 
+
       ++functionSize;
     }
 
     // We insert one blank instructions between basic blocks.
-    push_back(createEntry(0, index += SlotIndex::InstrDist));
+    indexList.push_back(createEntry(0, index += SlotIndex::InstrDist));
 
     MBBRanges[mbb->getNumber()].first = blockStartIndex;
-    MBBRanges[mbb->getNumber()].second = SlotIndex(back(),
+    MBBRanges[mbb->getNumber()].second = SlotIndex(&indexList.back(),
                                                    SlotIndex::Slot_Block);
     idx2MBBMap.push_back(IdxMBBPair(blockStartIndex, mbb));
   }
@@ -119,38 +118,37 @@ void SlotIndexes::renumberIndexes() {
 
   unsigned index = 0;
 
-  for (IndexListEntry *curEntry = front(); curEntry != getTail();
-       curEntry = curEntry->getNext()) {
-    curEntry->setIndex(index);
+  for (IndexList::iterator I = indexList.begin(), E = indexList.end();
+       I != E; ++I) {
+    I->setIndex(index);
     index += SlotIndex::InstrDist;
   }
 }
 
-// Renumber indexes locally after curEntry was inserted, but failed to get a new
+// Renumber indexes locally after curItr was inserted, but failed to get a new
 // index.
-void SlotIndexes::renumberIndexes(IndexListEntry *curEntry) {
+void SlotIndexes::renumberIndexes(IndexList::iterator curItr) {
   // Number indexes with half the default spacing so we can catch up quickly.
   const unsigned Space = SlotIndex::InstrDist/2;
   assert((Space & 3) == 0 && "InstrDist must be a multiple of 2*NUM");
 
-  IndexListEntry *start = curEntry->getPrev();
-  unsigned index = start->getIndex();
-  IndexListEntry *tail = getTail();
+  IndexList::iterator startItr = prior(curItr);
+  unsigned index = startItr->getIndex();
   do {
-    curEntry->setIndex(index += Space);
-    curEntry = curEntry->getNext();
+    curItr->setIndex(index += Space);
+    ++curItr;
     // If the next index is bigger, we have caught up.
-  } while (curEntry != tail && curEntry->getIndex() <= index);
+  } while (curItr != indexList.end() && curItr->getIndex() <= index);
 
-  DEBUG(dbgs() << "\n*** Renumbered SlotIndexes " << start->getIndex() << '-'
+  DEBUG(dbgs() << "\n*** Renumbered SlotIndexes " << startItr->getIndex() << '-'
                << index << " ***\n");
   ++NumLocalRenum;
 }
 
 
 void SlotIndexes::dump() const {
-  for (const IndexListEntry *itr = front(); itr != getTail();
-       itr = itr->getNext()) {
+  for (IndexList::const_iterator itr = indexList.begin();
+       itr != indexList.end(); ++itr) {
     dbgs() << itr->getIndex() << " ";
 
     if (itr->getInstr() != 0) {
@@ -168,7 +166,7 @@ void SlotIndexes::dump() const {
 // Print a SlotIndex to a raw_ostream.
 void SlotIndex::print(raw_ostream &os) const {
   if (isValid())
-    os << entry().getIndex() << "Berd"[getSlot()];
+    os << listEntry()->getIndex() << "Berd"[getSlot()];
   else
     os << "invalid";
 }
diff --git a/lib/ExecutionEngine/RuntimeDyld/CMakeLists.txt b/lib/ExecutionEngine/RuntimeDyld/CMakeLists.txt
index 002e63cd3b6..cbf7cf14d49 100644
--- a/lib/ExecutionEngine/RuntimeDyld/CMakeLists.txt
+++ b/lib/ExecutionEngine/RuntimeDyld/CMakeLists.txt
@@ -1,5 +1,6 @@
 add_llvm_library(LLVMRuntimeDyld
+  GDBRegistrar.cpp
   RuntimeDyld.cpp
-  RuntimeDyldMachO.cpp
   RuntimeDyldELF.cpp
+  RuntimeDyldMachO.cpp
   )
diff --git a/lib/ExecutionEngine/RuntimeDyld/GDBRegistrar.cpp b/lib/ExecutionEngine/RuntimeDyld/GDBRegistrar.cpp
new file mode 100644
index 00000000000..8b501014224
--- /dev/null
+++ b/lib/ExecutionEngine/RuntimeDyld/GDBRegistrar.cpp
@@ -0,0 +1,214 @@
+//===-- GDBRegistrar.cpp - Registers objects with GDB ---------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "JITRegistrar.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/Support/MutexGuard.h"
+#include "llvm/Support/Mutex.h"
+#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/Compiler.h"
+
+using namespace llvm;
+
+// This must be kept in sync with gdb/gdb/jit.h .
+extern "C" {
+
+  typedef enum {
+    JIT_NOACTION = 0,
+    JIT_REGISTER_FN,
+    JIT_UNREGISTER_FN
+  } jit_actions_t;
+
+  struct jit_code_entry {
+    struct jit_code_entry *next_entry;
+    struct jit_code_entry *prev_entry;
+    const char *symfile_addr;
+    uint64_t symfile_size;
+  };
+
+  struct jit_descriptor {
+    uint32_t version;
+    // This should be jit_actions_t, but we want to be specific about the
+    // bit-width.
+    uint32_t action_flag;
+    struct jit_code_entry *relevant_entry;
+    struct jit_code_entry *first_entry;
+  };
+
+  // We put information about the JITed function in this global, which the
+  // debugger reads.  Make sure to specify the version statically, because the
+  // debugger checks the version before we can set it during runtime.
+  static struct jit_descriptor __jit_debug_descriptor = { 1, 0, 0, 0 };
+
+  // Debuggers puts a breakpoint in this function.
+  LLVM_ATTRIBUTE_NOINLINE void __jit_debug_register_code() { }
+
+}
+
+namespace {
+
+// Buffer for an in-memory object file in executable memory
+typedef llvm::DenseMap< const char*,
+                        std::pair >
+  RegisteredObjectBufferMap;
+
+/// Global access point for the JIT debugging interface designed for use with a
+/// singleton toolbox. Handles thread-safe registration and deregistration of
+/// object files that are in executable memory managed by the client of this
+/// class.
+class GDBJITRegistrar : public JITRegistrar {
+  /// A map of in-memory object files that have been registered with the
+  /// JIT interface.
+  RegisteredObjectBufferMap ObjectBufferMap;
+
+public:
+  /// Instantiates the JIT service.
+  GDBJITRegistrar() : ObjectBufferMap() {}
+
+  /// Unregisters each object that was previously registered and releases all
+  /// internal resources.
+  virtual ~GDBJITRegistrar();
+
+  /// Creates an entry in the JIT registry for the buffer @p Object,
+  /// which must contain an object file in executable memory with any
+  /// debug information for the debugger.
+  void registerObject(const MemoryBuffer &Object);
+
+  /// Removes the internal registration of @p Object, and
+  /// frees associated resources.
+  /// Returns true if @p Object was found in ObjectBufferMap.
+  bool deregisterObject(const MemoryBuffer &Object);
+
+private:
+  /// Deregister the debug info for the given object file from the debugger
+  /// and delete any temporary copies.  This private method does not remove
+  /// the function from Map so that it can be called while iterating over Map.
+  void deregisterObjectInternal(RegisteredObjectBufferMap::iterator I);
+};
+
+/// Lock used to serialize all jit registration events, since they
+/// modify global variables.
+llvm::sys::Mutex JITDebugLock;
+
+/// Acquire the lock and do the registration.
+void NotifyDebugger(jit_code_entry* JITCodeEntry) {
+  llvm::MutexGuard locked(JITDebugLock);
+  __jit_debug_descriptor.action_flag = JIT_REGISTER_FN;
+
+  // Insert this entry at the head of the list.
+  JITCodeEntry->prev_entry = NULL;
+  jit_code_entry* NextEntry = __jit_debug_descriptor.first_entry;
+  JITCodeEntry->next_entry = NextEntry;
+  if (NextEntry != NULL) {
+    NextEntry->prev_entry = JITCodeEntry;
+  }
+  __jit_debug_descriptor.first_entry = JITCodeEntry;
+  __jit_debug_descriptor.relevant_entry = JITCodeEntry;
+  __jit_debug_register_code();
+}
+
+GDBJITRegistrar::~GDBJITRegistrar() {
+  // Free all registered object files.
+ for (RegisteredObjectBufferMap::iterator I = ObjectBufferMap.begin(), E = ObjectBufferMap.end();
+       I != E; ++I) {
+    // Call the private method that doesn't update the map so our iterator
+    // doesn't break.
+    deregisterObjectInternal(I);
+  }
+  ObjectBufferMap.clear();
+}
+
+void GDBJITRegistrar::registerObject(const MemoryBuffer &Object) {
+
+  const char *Buffer = Object.getBufferStart();
+  size_t      Size = Object.getBufferSize();
+
+  assert(Buffer && "Attempt to register a null object with a debugger.");
+  assert(ObjectBufferMap.find(Buffer) == ObjectBufferMap.end() &&
+         "Second attempt to perform debug registration.");
+  jit_code_entry* JITCodeEntry = new jit_code_entry();
+
+  if (JITCodeEntry == 0) {
+    llvm::report_fatal_error(
+      "Allocation failed when registering a JIT entry!\n");
+  }
+  else {
+    JITCodeEntry->symfile_addr = Buffer;
+    JITCodeEntry->symfile_size = Size;
+
+    ObjectBufferMap[Buffer] = std::make_pair(Size, JITCodeEntry);
+    NotifyDebugger(JITCodeEntry);
+  }
+}
+
+bool GDBJITRegistrar::deregisterObject(const MemoryBuffer& Object) {
+  const char *Buffer = Object.getBufferStart();
+  RegisteredObjectBufferMap::iterator I = ObjectBufferMap.find(Buffer);
+
+  if (I != ObjectBufferMap.end()) {
+    deregisterObjectInternal(I);
+    ObjectBufferMap.erase(I);
+    return true;
+  }
+  return false;
+}
+
+void GDBJITRegistrar::deregisterObjectInternal(
+    RegisteredObjectBufferMap::iterator I) {
+
+  jit_code_entry*& JITCodeEntry = I->second.second;
+
+  // Acquire the lock and do the unregistration.
+  {
+    llvm::MutexGuard locked(JITDebugLock);
+    __jit_debug_descriptor.action_flag = JIT_UNREGISTER_FN;
+
+    // Remove the jit_code_entry from the linked list.
+    jit_code_entry* PrevEntry = JITCodeEntry->prev_entry;
+    jit_code_entry* NextEntry = JITCodeEntry->next_entry;
+
+    if (NextEntry) {
+      NextEntry->prev_entry = PrevEntry;
+    }
+    if (PrevEntry) {
+      PrevEntry->next_entry = NextEntry;
+    }
+    else {
+      assert(__jit_debug_descriptor.first_entry == JITCodeEntry);
+      __jit_debug_descriptor.first_entry = NextEntry;
+    }
+
+    // Tell the debugger which entry we removed, and unregister the code.
+    __jit_debug_descriptor.relevant_entry = JITCodeEntry;
+    __jit_debug_register_code();
+  }
+
+  delete JITCodeEntry;
+  JITCodeEntry = NULL;
+}
+
+} // end namespace
+
+namespace llvm {
+
+JITRegistrar& JITRegistrar::getGDBRegistrar() {
+  static GDBJITRegistrar* sRegistrar = NULL;
+  if (sRegistrar == NULL) {
+    // The mutex is here so that it won't slow down access once the registrar
+    // is instantiated
+    llvm::MutexGuard locked(JITDebugLock);
+    // Check again to be sure another thread didn't create this while we waited
+    if (sRegistrar == NULL) {
+      sRegistrar = new GDBJITRegistrar;
+    }
+  }
+  return *sRegistrar;
+}
+
+} // namespace llvm
diff --git a/lib/ExecutionEngine/RuntimeDyld/JITRegistrar.h b/lib/ExecutionEngine/RuntimeDyld/JITRegistrar.h
new file mode 100644
index 00000000000..f964bc61829
--- /dev/null
+++ b/lib/ExecutionEngine/RuntimeDyld/JITRegistrar.h
@@ -0,0 +1,43 @@
+//===-- JITRegistrar.h - Registers objects with a debugger ----------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_EXECUTION_ENGINE_JIT_REGISTRAR_H
+#define LLVM_EXECUTION_ENGINE_JIT_REGISTRAR_H
+
+#include "llvm/Support/MemoryBuffer.h"
+
+namespace llvm {
+
+/// Global access point for the JIT debugging interface.
+class JITRegistrar {
+public:
+  /// Instantiates the JIT service.
+  JITRegistrar() {}
+
+  /// Unregisters each object that was previously registered and releases all
+  /// internal resources.
+  virtual ~JITRegistrar() {}
+
+  /// Creates an entry in the JIT registry for the buffer @p Object,
+  /// which must contain an object file in executable memory with any
+  /// debug information for the debugger.
+  virtual void registerObject(const MemoryBuffer &Object) = 0;
+
+  /// Removes the internal registration of @p Object, and
+  /// frees associated resources.
+  /// Returns true if @p Object was previously registered.
+  virtual bool deregisterObject(const MemoryBuffer &Object) = 0;
+
+  /// Returns a reference to a GDB JIT registrar singleton
+  static JITRegistrar& getGDBRegistrar();
+};
+
+} // end namespace llvm
+
+#endif // LLVM_EXECUTION_ENGINE_JIT_REGISTRAR_H
diff --git a/lib/ExecutionEngine/RuntimeDyld/ObjectImage.h b/lib/ExecutionEngine/RuntimeDyld/ObjectImage.h
new file mode 100644
index 00000000000..8206eadbd25
--- /dev/null
+++ b/lib/ExecutionEngine/RuntimeDyld/ObjectImage.h
@@ -0,0 +1,59 @@
+//===---- ObjectImage.h - Format independent executuable object image -----===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file declares a file format independent ObjectImage class.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_RUNTIMEDYLD_OBJECT_IMAGE_H
+#define LLVM_RUNTIMEDYLD_OBJECT_IMAGE_H
+
+#include "llvm/Object/ObjectFile.h"
+
+namespace llvm {
+
+class ObjectImage {
+  ObjectImage(); // = delete
+  ObjectImage(const ObjectImage &other); // = delete
+protected:
+  object::ObjectFile *ObjFile;
+
+public:
+  ObjectImage(object::ObjectFile *Obj) { ObjFile = Obj; }
+  virtual ~ObjectImage() {}
+
+  virtual object::symbol_iterator begin_symbols() const
+              { return ObjFile->begin_symbols(); }
+  virtual object::symbol_iterator end_symbols() const
+              { return ObjFile->end_symbols(); }
+
+  virtual object::section_iterator begin_sections() const
+              { return ObjFile->begin_sections(); }
+  virtual object::section_iterator end_sections() const
+              { return ObjFile->end_sections(); }
+
+  virtual /* Triple::ArchType */ unsigned getArch() const
+              { return ObjFile->getArch(); }
+
+  // Subclasses can override these methods to update the image with loaded
+  // addresses for sections and common symbols
+  virtual void updateSectionAddress(const object::SectionRef &Sec,
+                                    uint64_t Addr) {}
+  virtual void updateSymbolAddress(const object::SymbolRef &Sym, uint64_t Addr)
+              {}
+
+  // Subclasses can override this method to provide JIT debugging support
+  virtual void registerWithDebugger() {}
+  virtual void deregisterWithDebugger() {}
+};
+
+} // end namespace llvm
+
+#endif // LLVM_RUNTIMEDYLD_OBJECT_IMAGE_H
+
diff --git a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp
index 63cec1aca3b..1b1840ae066 100644
--- a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp
+++ b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp
@@ -59,11 +59,17 @@ void RuntimeDyldImpl::mapSectionAddress(void *LocalAddress,
   llvm_unreachable("Attempting to remap address of unknown section!");
 }
 
+// Subclasses can implement this method to create specialized image instances
+// The caller owns the the pointer that is returned.
+ObjectImage *RuntimeDyldImpl::createObjectImage(const MemoryBuffer *InputBuffer) {
+  ObjectFile *ObjFile = ObjectFile::createObjectFile(const_cast
+                                                                 (InputBuffer));
+  ObjectImage *Obj = new ObjectImage(ObjFile);
+  return Obj;
+}
+
 bool RuntimeDyldImpl::loadObject(const MemoryBuffer *InputBuffer) {
-  // FIXME: ObjectFile don't modify MemoryBuffer.
-  //        It should use const MemoryBuffer as parameter.
-  OwningPtr obj(ObjectFile::createObjectFile(
-                                       const_cast(InputBuffer)));
+  OwningPtr obj(createObjectImage(InputBuffer));
   if (!obj)
     report_fatal_error("Unable to create object image from memory buffer!");
 
@@ -110,7 +116,8 @@ bool RuntimeDyldImpl::loadObject(const MemoryBuffer *InputBuffer) {
                                 (uintptr_t)FileOffset;
         uintptr_t SectOffset = (uintptr_t)(SymPtr - (const uint8_t*)sData.begin());
         unsigned SectionID =
-          findOrEmitSection(*si,
+          findOrEmitSection(*obj,
+                            *si,
                             SymType == object::SymbolRef::ST_Function,
                             LocalSections);
         bool isGlobal = flags & SymbolRef::SF_Global;
@@ -128,7 +135,7 @@ bool RuntimeDyldImpl::loadObject(const MemoryBuffer *InputBuffer) {
 
   // Allocate common symbols
   if (CommonSize != 0)
-    emitCommonSymbols(CommonSymbols, CommonSize, LocalSymbols);
+    emitCommonSymbols(*obj, CommonSymbols, CommonSize, LocalSymbols);
 
   // Parse and proccess relocations
   DEBUG(dbgs() << "Parse relocations:\n");
@@ -145,7 +152,7 @@ bool RuntimeDyldImpl::loadObject(const MemoryBuffer *InputBuffer) {
 
       // If it's first relocation in this section, find its SectionID
       if (isFirstRelocation) {
-        SectionID = findOrEmitSection(*si, true, LocalSections);
+        SectionID = findOrEmitSection(*obj, *si, true, LocalSections);
         DEBUG(dbgs() << "\tSectionID: " << SectionID << "\n");
         isFirstRelocation = false;
       }
@@ -164,10 +171,14 @@ bool RuntimeDyldImpl::loadObject(const MemoryBuffer *InputBuffer) {
       processRelocationRef(RI, *obj, LocalSections, LocalSymbols, Stubs);
     }
   }
+
+  handleObjectLoaded(obj.take());
+
   return false;
 }
 
-unsigned RuntimeDyldImpl::emitCommonSymbols(const CommonSymbolMap &Map,
+unsigned RuntimeDyldImpl::emitCommonSymbols(ObjectImage &Obj,
+                                            const CommonSymbolMap &Map,
                                             uint64_t TotalSize,
                                             LocalSymbolMap &LocalSymbols) {
   // Allocate memory for the section
@@ -191,6 +202,7 @@ unsigned RuntimeDyldImpl::emitCommonSymbols(const CommonSymbolMap &Map,
     uint64_t Size = it->second;
     StringRef Name;
     it->first.getName(Name);
+    Obj.updateSymbolAddress(it->first, (uint64_t)Addr);
     LocalSymbols[Name.data()] = SymbolLoc(SectionID, Offset);
     Offset += Size;
     Addr += Size;
@@ -199,7 +211,8 @@ unsigned RuntimeDyldImpl::emitCommonSymbols(const CommonSymbolMap &Map,
   return SectionID;
 }
 
-unsigned RuntimeDyldImpl::emitSection(const SectionRef &Section,
+unsigned RuntimeDyldImpl::emitSection(ObjectImage &Obj,
+                                      const SectionRef &Section,
                                       bool IsCode) {
 
   unsigned StubBufSize = 0,
@@ -257,6 +270,7 @@ unsigned RuntimeDyldImpl::emitSection(const SectionRef &Section,
                  << " StubBufSize: " << StubBufSize
                  << " Allocate: " << Allocate
                  << "\n");
+    Obj.updateSectionAddress(Section, (uint64_t)Addr);
   }
   else {
     // Even if we didn't load the section, we need to record an entry for it
@@ -277,7 +291,8 @@ unsigned RuntimeDyldImpl::emitSection(const SectionRef &Section,
   return SectionID;
 }
 
-unsigned RuntimeDyldImpl::findOrEmitSection(const SectionRef &Section,
+unsigned RuntimeDyldImpl::findOrEmitSection(ObjectImage &Obj,
+                                            const SectionRef &Section,
                                             bool IsCode,
                                             ObjSectionToIDMap &LocalSections) {
 
@@ -286,7 +301,7 @@ unsigned RuntimeDyldImpl::findOrEmitSection(const SectionRef &Section,
   if (i != LocalSections.end())
     SectionID = i->second;
   else {
-    SectionID = emitSection(Section, IsCode);
+    SectionID = emitSection(Obj, Section, IsCode);
     LocalSections[Section] = SectionID;
   }
   return SectionID;
diff --git a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp
index 57fefee5ded..db6da8c8ef2 100644
--- a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp
+++ b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.cpp
@@ -20,11 +20,176 @@
 #include "llvm/Object/ObjectFile.h"
 #include "llvm/Support/ELF.h"
 #include "llvm/ADT/Triple.h"
+#include "llvm/Object/ELF.h"
+#include "JITRegistrar.h"
 using namespace llvm;
 using namespace llvm::object;
 
+namespace {
+
+template
+class DyldELFObject : public ELFObjectFile {
+  LLVM_ELF_IMPORT_TYPES(target_endianness, is64Bits)
+
+  typedef Elf_Shdr_Impl Elf_Shdr;
+  typedef Elf_Sym_Impl Elf_Sym;
+  typedef Elf_Rel_Impl Elf_Rel;
+  typedef Elf_Rel_Impl Elf_Rela;
+
+  typedef typename ELFObjectFile::
+    Elf_Ehdr Elf_Ehdr;
+
+  typedef typename ELFDataTypeTypedefHelper<
+          target_endianness, is64Bits>::value_type addr_type;
+
+protected:
+  // This duplicates the 'Data' member in the 'Binary' base class
+  // but it is necessary to workaround a bug in gcc 4.2
+  MemoryBuffer *InputData;
+
+public:
+  DyldELFObject(MemoryBuffer *Object, error_code &ec);
+
+  void updateSectionAddress(const SectionRef &Sec, uint64_t Addr);
+  void updateSymbolAddress(const SymbolRef &Sym, uint64_t Addr);
+
+  const MemoryBuffer& getBuffer() const { return *InputData; }
+
+  // Methods for type inquiry through isa, cast, and dyn_cast
+  static inline bool classof(const Binary *v) {
+    return (isa >(v)
+            && classof(cast >(v)));
+  }
+  static inline bool classof(
+      const ELFObjectFile *v) {
+    return v->isDyldType();
+  }
+  static inline bool classof(const DyldELFObject *v) {
+    return true;
+  }
+};
+
+template
+class ELFObjectImage : public ObjectImage {
+  protected:
+    DyldELFObject *DyldObj;
+    bool Registered;
+
+  public:
+    ELFObjectImage(DyldELFObject *Obj)
+    : ObjectImage(Obj),
+      DyldObj(Obj),
+      Registered(false) {}
+
+    virtual ~ELFObjectImage() {
+      if (Registered)
+        deregisterWithDebugger();
+    }
+
+    // Subclasses can override these methods to update the image with loaded
+    // addresses for sections and common symbols
+    virtual void updateSectionAddress(const SectionRef &Sec, uint64_t Addr)
+    {
+      DyldObj->updateSectionAddress(Sec, Addr);
+    }
+
+    virtual void updateSymbolAddress(const SymbolRef &Sym, uint64_t Addr)
+    {
+      DyldObj->updateSymbolAddress(Sym, Addr);
+    }
+
+    virtual void registerWithDebugger()
+    {
+      JITRegistrar::getGDBRegistrar().registerObject(DyldObj->getBuffer());
+      Registered = true;
+    }
+    virtual void deregisterWithDebugger()
+    {
+      JITRegistrar::getGDBRegistrar().deregisterObject(DyldObj->getBuffer());
+    }
+};
+
+template
+DyldELFObject::DyldELFObject(MemoryBuffer *Object,
+                                                          error_code &ec)
+  : ELFObjectFile(Object, ec),
+    InputData(Object) {
+  this->isDyldELFObject = true;
+}
+
+template
+void DyldELFObject::updateSectionAddress(
+                                                       const SectionRef &Sec,
+                                                       uint64_t Addr) {
+  DataRefImpl ShdrRef = Sec.getRawDataRefImpl();
+  Elf_Shdr *shdr = const_cast(
+                          reinterpret_cast(ShdrRef.p));
+
+  // This assumes the address passed in matches the target address bitness
+  // The template-based type cast handles everything else.
+  shdr->sh_addr = static_cast(Addr);
+}
+
+template
+void DyldELFObject::updateSymbolAddress(
+                                                       const SymbolRef &SymRef,
+                                                       uint64_t Addr) {
+
+  Elf_Sym *sym = const_cast(
+                                 ELFObjectFile::
+                                   getSymbol(SymRef.getRawDataRefImpl()));
+
+  // This assumes the address passed in matches the target address bitness
+  // The template-based type cast handles everything else.
+  sym->st_value = static_cast(Addr);
+}
+
+} // namespace
+
+
 namespace llvm {
 
+ObjectImage *RuntimeDyldELF::createObjectImage(
+                                         const MemoryBuffer *ConstInputBuffer) {
+  MemoryBuffer *InputBuffer = const_cast(ConstInputBuffer);
+  std::pair Ident = getElfArchType(InputBuffer);
+  error_code ec;
+
+  if (Ident.first == ELF::ELFCLASS32 && Ident.second == ELF::ELFDATA2LSB) {
+    DyldELFObject *Obj =
+           new DyldELFObject(InputBuffer, ec);
+    return new ELFObjectImage(Obj);
+  }
+  else if (Ident.first == ELF::ELFCLASS32 && Ident.second == ELF::ELFDATA2MSB) {
+    DyldELFObject *Obj =
+           new DyldELFObject(InputBuffer, ec);
+    return new ELFObjectImage(Obj);
+  }
+  else if (Ident.first == ELF::ELFCLASS64 && Ident.second == ELF::ELFDATA2MSB) {
+    DyldELFObject *Obj =
+           new DyldELFObject(InputBuffer, ec);
+    return new ELFObjectImage(Obj);
+  }
+  else if (Ident.first == ELF::ELFCLASS64 && Ident.second == ELF::ELFDATA2LSB) {
+    DyldELFObject *Obj =
+           new DyldELFObject(InputBuffer, ec);
+    return new ELFObjectImage(Obj);
+  }
+  else
+    llvm_unreachable("Unexpected ELF format");
+}
+
+void RuntimeDyldELF::handleObjectLoaded(ObjectImage *Obj)
+{
+  Obj->registerWithDebugger();
+  // Save the loaded object.  It will deregister itself when deleted
+  LoadedObject = Obj;
+}
+
+RuntimeDyldELF::~RuntimeDyldELF() {
+  if (LoadedObject)
+    delete LoadedObject;
+}
 
 void RuntimeDyldELF::resolveX86_64Relocation(uint8_t *LocalAddress,
                                              uint64_t FinalAddress,
@@ -167,7 +332,7 @@ void RuntimeDyldELF::resolveRelocation(uint8_t *LocalAddress,
 }
 
 void RuntimeDyldELF::processRelocationRef(const ObjRelocationInfo &Rel,
-                                          const ObjectFile &Obj,
+                                          ObjectImage &Obj,
                                           ObjSectionToIDMap &ObjSectionToID,
                                           LocalSymbolMap &Symbols,
                                           StubMap &Stubs) {
@@ -206,7 +371,7 @@ void RuntimeDyldELF::processRelocationRef(const ObjRelocationInfo &Rel,
           if (si == Obj.end_sections())
             llvm_unreachable("Symbol section not found, bad object file format!");
           DEBUG(dbgs() << "\t\tThis is section symbol\n");
-          Value.SectionID = findOrEmitSection((*si), true, ObjSectionToID);
+          Value.SectionID = findOrEmitSection(Obj, (*si), true, ObjSectionToID);
           Value.Addend = Addend;
           break;
         }
@@ -236,7 +401,7 @@ void RuntimeDyldELF::processRelocationRef(const ObjRelocationInfo &Rel,
     //  Look up for existing stub.
     StubMap::const_iterator i = Stubs.find(Value);
     if (i != Stubs.end()) {
-      resolveRelocation(Target, Section.LoadAddress, (uint64_t)Section.Address +
+      resolveRelocation(Target, (uint64_t)Target, (uint64_t)Section.Address +
                         i->second, RelType, 0);
       DEBUG(dbgs() << " Stub function found\n");
     } else {
@@ -247,7 +412,7 @@ void RuntimeDyldELF::processRelocationRef(const ObjRelocationInfo &Rel,
                                                    Section.StubOffset);
       AddRelocation(Value, Rel.SectionID,
                     StubTargetAddr - Section.Address, ELF::R_ARM_ABS32);
-      resolveRelocation(Target, Section.LoadAddress, (uint64_t)Section.Address +
+      resolveRelocation(Target, (uint64_t)Target, (uint64_t)Section.Address +
                         Section.StubOffset, RelType, 0);
       Section.StubOffset += getMaxStubSize();
     }
diff --git a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.h b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.h
index 36566da57a5..e7f6fab16fc 100644
--- a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.h
+++ b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldELF.h
@@ -22,6 +22,8 @@ using namespace llvm;
 namespace llvm {
 class RuntimeDyldELF : public RuntimeDyldImpl {
 protected:
+  ObjectImage *LoadedObject;
+
   void resolveX86_64Relocation(uint8_t *LocalAddress,
                                uint64_t FinalAddress,
                                uint64_t Value,
@@ -47,12 +49,18 @@ protected:
                                  int64_t Addend);
 
   virtual void processRelocationRef(const ObjRelocationInfo &Rel,
-                                    const ObjectFile &Obj,
+                                    ObjectImage &Obj,
                                     ObjSectionToIDMap &ObjSectionToID,
                                     LocalSymbolMap &Symbols, StubMap &Stubs);
 
+  virtual ObjectImage *createObjectImage(const MemoryBuffer *InputBuffer);
+  virtual void handleObjectLoaded(ObjectImage *Obj);
+
 public:
-  RuntimeDyldELF(RTDyldMemoryManager *mm) : RuntimeDyldImpl(mm) {}
+  RuntimeDyldELF(RTDyldMemoryManager *mm)
+      : RuntimeDyldImpl(mm), LoadedObject(0) {}
+
+  virtual ~RuntimeDyldELF();
 
   bool isCompatibleFormat(const MemoryBuffer *InputBuffer) const;
 };
diff --git a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldImpl.h b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldImpl.h
index bf678af6ece..2dea13f15ce 100644
--- a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldImpl.h
+++ b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldImpl.h
@@ -29,6 +29,7 @@
 #include "llvm/ADT/Triple.h"
 #include 
 #include "llvm/Support/Format.h"
+#include "ObjectImage.h"
 
 using namespace llvm;
 using namespace llvm::object;
@@ -154,7 +155,8 @@ protected:
 
   /// \brief Emits a section containing common symbols.
   /// \return SectionID.
-  unsigned emitCommonSymbols(const CommonSymbolMap &Map,
+  unsigned emitCommonSymbols(ObjectImage &Obj,
+                             const CommonSymbolMap &Map,
                              uint64_t TotalSize,
                              LocalSymbolMap &Symbols);
 
@@ -162,14 +164,18 @@ protected:
   /// \param IsCode if it's true then allocateCodeSection() will be
   ///        used for emmits, else allocateDataSection() will be used.
   /// \return SectionID.
-  unsigned emitSection(const SectionRef &Section, bool IsCode);
+  unsigned emitSection(ObjectImage &Obj,
+                       const SectionRef &Section,
+                       bool IsCode);
 
   /// \brief Find Section in LocalSections. If the secton is not found - emit
   ///        it and store in LocalSections.
   /// \param IsCode if it's true then allocateCodeSection() will be
   ///        used for emmits, else allocateDataSection() will be used.
   /// \return SectionID.
-  unsigned findOrEmitSection(const SectionRef &Section, bool IsCode,
+  unsigned findOrEmitSection(ObjectImage &Obj,
+                             const SectionRef &Section,
+                             bool IsCode,
                              ObjSectionToIDMap &LocalSections);
 
   /// \brief If Value.SymbolName is NULL then store relocation to the
@@ -200,11 +206,18 @@ protected:
   /// \brief Parses the object file relocation and store it to Relocations
   ///        or SymbolRelocations. Its depend from object file type.
   virtual void processRelocationRef(const ObjRelocationInfo &Rel,
-                                    const ObjectFile &Obj,
+                                    ObjectImage &Obj,
                                     ObjSectionToIDMap &ObjSectionToID,
                                     LocalSymbolMap &Symbols, StubMap &Stubs) = 0;
 
   void resolveSymbols();
+  virtual ObjectImage *createObjectImage(const MemoryBuffer *InputBuffer);
+  virtual void handleObjectLoaded(ObjectImage *Obj)
+  {
+    // Subclasses may choose to retain this image if they have a use for it
+    delete Obj;
+  }
+
 public:
   RuntimeDyldImpl(RTDyldMemoryManager *mm) : MemMgr(mm), HasError(false) {}
 
diff --git a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.cpp b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.cpp
index 1318b445425..b7f515d6ce6 100644
--- a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.cpp
+++ b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.cpp
@@ -205,7 +205,7 @@ resolveARMRelocation(uint8_t *LocalAddress,
 }
 
 void RuntimeDyldMachO::processRelocationRef(const ObjRelocationInfo &Rel,
-                                            const ObjectFile &Obj,
+                                            ObjectImage &Obj,
                                             ObjSectionToIDMap &ObjSectionToID,
                                             LocalSymbolMap &Symbols,
                                             StubMap &Stubs) {
@@ -246,7 +246,7 @@ void RuntimeDyldMachO::processRelocationRef(const ObjRelocationInfo &Rel,
         break;
     }
     assert(si != se && "No section containing relocation!");
-    Value.SectionID = findOrEmitSection(*si, true, ObjSectionToID);
+    Value.SectionID = findOrEmitSection(Obj, *si, true, ObjSectionToID);
     Value.Addend = *(const intptr_t *)Target;
     if (Value.Addend) {
       // The MachO addend is offset from the current section, we need set it
diff --git a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.h b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.h
index 898b85190e7..418d130f635 100644
--- a/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.h
+++ b/lib/ExecutionEngine/RuntimeDyld/RuntimeDyldMachO.h
@@ -49,7 +49,7 @@ protected:
                             int64_t Addend);
 
   virtual void processRelocationRef(const ObjRelocationInfo &Rel,
-                                    const ObjectFile &Obj,
+                                    ObjectImage &Obj,
                                     ObjSectionToIDMap &ObjSectionToID,
                                     LocalSymbolMap &Symbols, StubMap &Stubs);
 
@@ -59,7 +59,7 @@ public:
                                  uint64_t Value,
                                  uint32_t Type,
                                  int64_t Addend);
-                                 
+
   RuntimeDyldMachO(RTDyldMemoryManager *mm) : RuntimeDyldImpl(mm) {}
 
   bool isCompatibleFormat(const MemoryBuffer *InputBuffer) const;
diff --git a/lib/MC/MCParser/AsmParser.cpp b/lib/MC/MCParser/AsmParser.cpp
index 2d61cac6258..8aef43cb0b4 100644
--- a/lib/MC/MCParser/AsmParser.cpp
+++ b/lib/MC/MCParser/AsmParser.cpp
@@ -1527,11 +1527,11 @@ bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
     }
     Lex();
   }
-  // If there weren't any arguments, erase the token vector so everything
-  // else knows that. Leaving around the vestigal empty token list confuses
-  // things.
-  if (MacroArguments.size() == 1 && MacroArguments.back().empty())
-    MacroArguments.clear();
+  // If the last argument didn't end up with any tokens, it's not a real
+  // argument and we should remove it from the list. This happens with either
+  // a tailing comma or an empty argument list.
+  if (MacroArguments.back().empty())
+    MacroArguments.pop_back();
 
   // Macro instantiation is lexical, unfortunately. We construct a new buffer
   // to hold the macro body with substitutions.
diff --git a/lib/Object/ELFObjectFile.cpp b/lib/Object/ELFObjectFile.cpp
index ab5f8108af1..663b84ec8b1 100644
--- a/lib/Object/ELFObjectFile.cpp
+++ b/lib/Object/ELFObjectFile.cpp
@@ -17,16 +17,6 @@ namespace llvm {
 
 using namespace object;
 
-namespace {
-  std::pair
-  getElfArchType(MemoryBuffer *Object) {
-    if (Object->getBufferSize() < ELF::EI_NIDENT)
-      return std::make_pair((uint8_t)ELF::ELFCLASSNONE,(uint8_t)ELF::ELFDATANONE);
-    return std::make_pair( (uint8_t)Object->getBufferStart()[ELF::EI_CLASS]
-                         , (uint8_t)Object->getBufferStart()[ELF::EI_DATA]);
-  }
-}
-
 // Creates an in-memory object-file by default: createELFObjectFile(Buffer)
 ObjectFile *ObjectFile::createELFObjectFile(MemoryBuffer *Object) {
   std::pair Ident = getElfArchType(Object);
diff --git a/lib/Support/CMakeLists.txt b/lib/Support/CMakeLists.txt
index 9b3b6c801dd..9103327dffa 100644
--- a/lib/Support/CMakeLists.txt
+++ b/lib/Support/CMakeLists.txt
@@ -32,7 +32,7 @@ add_llvm_library(LLVMSupport
   IntrusiveRefCntPtr.cpp
   IsInf.cpp
   IsNAN.cpp
-  JSONParser.cpp
+  Locale.cpp
   LockFileManager.cpp
   ManagedStatic.cpp
   MemoryBuffer.cpp
diff --git a/lib/Support/JSONParser.cpp b/lib/Support/JSONParser.cpp
deleted file mode 100644
index 5dfcf297a7e..00000000000
--- a/lib/Support/JSONParser.cpp
+++ /dev/null
@@ -1,302 +0,0 @@
-//===--- JSONParser.cpp - Simple JSON parser ------------------------------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-//  This file implements a JSON parser.
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/Support/JSONParser.h"
-
-#include "llvm/ADT/Twine.h"
-#include "llvm/Support/Casting.h"
-#include "llvm/Support/MemoryBuffer.h"
-
-using namespace llvm;
-
-JSONParser::JSONParser(StringRef Input, SourceMgr *SM)
-  : SM(SM), Failed(false) {
-  InputBuffer = MemoryBuffer::getMemBuffer(Input, "JSON");
-  SM->AddNewSourceBuffer(InputBuffer, SMLoc());
-  End = InputBuffer->getBuffer().end();
-  Position = InputBuffer->getBuffer().begin();
-}
-
-JSONValue *JSONParser::parseRoot() {
-  if (Position != InputBuffer->getBuffer().begin())
-    report_fatal_error("Cannot reuse JSONParser.");
-  if (isWhitespace())
-    nextNonWhitespace();
-  if (errorIfAtEndOfFile("'[' or '{' at start of JSON text"))
-    return 0;
-  switch (*Position) {
-    case '[':
-      return new (ValueAllocator.Allocate(1)) JSONArray(this);
-    case '{':
-      return new (ValueAllocator.Allocate(1)) JSONObject(this);
-    default:
-      setExpectedError("'[' or '{' at start of JSON text", *Position);
-      return 0;
-  }
-}
-
-bool JSONParser::validate() {
-  JSONValue *Root = parseRoot();
-  if (Root == NULL) {
-    return false;
-  }
-  return skip(*Root);
-}
-
-bool JSONParser::skip(const JSONAtom &Atom) {
-  switch(Atom.getKind()) {
-    case JSONAtom::JK_Array:
-    case JSONAtom::JK_Object:
-      return skipContainer(*cast(&Atom));
-    case JSONAtom::JK_String:
-      return true;
-    case JSONAtom::JK_KeyValuePair:
-      return skip(*cast(&Atom)->Value);
-  }
-  llvm_unreachable("Impossible enum value.");
-}
-
-// Sets the current error to:
-// "expected , but found ".
-void JSONParser::setExpectedError(StringRef Expected, StringRef Found) {
-  SM->PrintMessage(SMLoc::getFromPointer(Position), SourceMgr::DK_Error,
-    "expected " + Expected + ", but found " + Found + ".", ArrayRef());
-  Failed = true;
-}
-
-// Sets the current error to:
-// "expected , but found ".
-void JSONParser::setExpectedError(StringRef Expected, char Found) {
-  setExpectedError(Expected, ("'" + StringRef(&Found, 1) + "'").str());
-}
-
-// If there is no character available, returns true and sets the current error
-// to: "expected , but found EOF.".
-bool JSONParser::errorIfAtEndOfFile(StringRef Expected) {
-  if (Position == End) {
-    setExpectedError(Expected, "EOF");
-    return true;
-  }
-  return false;
-}
-
-// Sets the current error if the current character is not C to:
-// "expected 'C', but got ".
-bool JSONParser::errorIfNotAt(char C, StringRef Message) {
-  if (*Position != C) {
-    std::string Expected =
-      ("'" + StringRef(&C, 1) + "' " + Message).str();
-    if (Position == End)
-      setExpectedError(Expected, "EOF");
-    else
-      setExpectedError(Expected, *Position);
-    return true;
-  }
-  return false;
-}
-
-// Forbidding inlining improves performance by roughly 20%.
-// FIXME: Remove once llvm optimizes this to the faster version without hints.
-LLVM_ATTRIBUTE_NOINLINE static bool
-wasEscaped(StringRef::iterator First, StringRef::iterator Position);
-
-// Returns whether a character at 'Position' was escaped with a leading '\'.
-// 'First' specifies the position of the first character in the string.
-static bool wasEscaped(StringRef::iterator First,
-                       StringRef::iterator Position) {
-  assert(Position - 1 >= First);
-  StringRef::iterator I = Position - 1;
-  // We calulate the number of consecutive '\'s before the current position
-  // by iterating backwards through our string.
-  while (I >= First && *I == '\\') --I;
-  // (Position - 1 - I) now contains the number of '\'s before the current
-  // position. If it is odd, the character at 'Positon' was escaped.
-  return (Position - 1 - I) % 2 == 1;
-}
-
-// Parses a JSONString, assuming that the current position is on a quote.
-JSONString *JSONParser::parseString() {
-  assert(Position != End);
-  assert(!isWhitespace());
-  if (errorIfNotAt('"', "at start of string"))
-    return 0;
-  StringRef::iterator First = Position + 1;
-
-  // Benchmarking shows that this loop is the hot path of the application with
-  // about 2/3rd of the runtime cycles. Since escaped quotes are not the common
-  // case, and multiple escaped backslashes before escaped quotes are very rare,
-  // we pessimize this case to achieve a smaller inner loop in the common case.
-  // We're doing that by having a quick inner loop that just scans for the next
-  // quote. Once we find the quote we check the last character to see whether
-  // the quote might have been escaped. If the last character is not a '\', we
-  // know the quote was not escaped and have thus found the end of the string.
-  // If the immediately preceding character was a '\', we have to scan backwards
-  // to see whether the previous character was actually an escaped backslash, or
-  // an escape character for the quote. If we find that the current quote was
-  // escaped, we continue parsing for the next quote and repeat.
-  // This optimization brings around 30% performance improvements.
-  do {
-    // Step over the current quote.
-    ++Position;
-    // Find the next quote.
-    while (Position != End && *Position != '"')
-      ++Position;
-    if (errorIfAtEndOfFile("'\"' at end of string"))
-      return 0;
-    // Repeat until the previous character was not a '\' or was an escaped
-    // backslash.
-  } while (*(Position - 1) == '\\' && wasEscaped(First, Position));
-
-  return new (ValueAllocator.Allocate())
-      JSONString(StringRef(First, Position - First));
-}
-
-
-// Advances the position to the next non-whitespace position.
-void JSONParser::nextNonWhitespace() {
-  do {
-    ++Position;
-  } while (isWhitespace());
-}
-
-// Checks if there is a whitespace character at the current position.
-bool JSONParser::isWhitespace() {
-  return *Position == ' ' || *Position == '\t' ||
-         *Position == '\n' || *Position == '\r';
-}
-
-bool JSONParser::failed() const {
-  return Failed;
-}
-
-// Parses a JSONValue, assuming that the current position is at the first
-// character of the value.
-JSONValue *JSONParser::parseValue() {
-  assert(Position != End);
-  assert(!isWhitespace());
-  switch (*Position) {
-    case '[':
-      return new (ValueAllocator.Allocate(1)) JSONArray(this);
-    case '{':
-      return new (ValueAllocator.Allocate(1)) JSONObject(this);
-    case '"':
-      return parseString();
-    default:
-      setExpectedError("'[', '{' or '\"' at start of value", *Position);
-      return 0;
-  }
-}
-
-// Parses a JSONKeyValuePair, assuming that the current position is at the first
-// character of the key, value pair.
-JSONKeyValuePair *JSONParser::parseKeyValuePair() {
-  assert(Position != End);
-  assert(!isWhitespace());
-
-  JSONString *Key = parseString();
-  if (Key == 0)
-    return 0;
-
-  nextNonWhitespace();
-  if (errorIfNotAt(':', "between key and value"))
-    return 0;
-
-  nextNonWhitespace();
-  const JSONValue *Value = parseValue();
-  if (Value == 0)
-    return 0;
-
-  return new (ValueAllocator.Allocate(1))
-    JSONKeyValuePair(Key, Value);
-}
-
-/// \brief Parses the first element of a JSON array or object, or closes the
-/// array.
-///
-/// The method assumes that the current position is before the first character
-/// of the element, with possible white space in between. When successful, it
-/// returns the new position after parsing the element. Otherwise, if there is
-/// no next value, it returns a default constructed StringRef::iterator.
-StringRef::iterator JSONParser::parseFirstElement(JSONAtom::Kind ContainerKind,
-                                                  char StartChar, char EndChar,
-                                                  const JSONAtom *&Element) {
-  assert(*Position == StartChar);
-  Element = 0;
-  nextNonWhitespace();
-  if (errorIfAtEndOfFile("value or end of container at start of container"))
-    return StringRef::iterator();
-
-  if (*Position == EndChar)
-    return StringRef::iterator();
-
-  Element = parseElement(ContainerKind);
-  if (Element == 0)
-    return StringRef::iterator();
-
-  return Position;
-}
-
-/// \brief Parses the next element of a JSON array or object, or closes the
-/// array.
-///
-/// The method assumes that the current position is before the ',' which
-/// separates the next element from the current element. When successful, it
-/// returns the new position after parsing the element. Otherwise, if there is
-/// no next value, it returns a default constructed StringRef::iterator.
-StringRef::iterator JSONParser::parseNextElement(JSONAtom::Kind ContainerKind,
-                                                 char EndChar,
-                                                 const JSONAtom *&Element) {
-  Element = 0;
-  nextNonWhitespace();
-  if (errorIfAtEndOfFile("',' or end of container for next element"))
-    return 0;
-
-  if (*Position == ',') {
-    nextNonWhitespace();
-    if (errorIfAtEndOfFile("element in container"))
-      return StringRef::iterator();
-
-    Element = parseElement(ContainerKind);
-    if (Element == 0)
-      return StringRef::iterator();
-
-    return Position;
-  } else if (*Position == EndChar) {
-      return StringRef::iterator();
-  } else {
-    setExpectedError("',' or end of container for next element", *Position);
-    return StringRef::iterator();
-  }
-}
-
-const JSONAtom *JSONParser::parseElement(JSONAtom::Kind ContainerKind) {
-  switch (ContainerKind) {
-    case JSONAtom::JK_Array:
-      return parseValue();
-    case JSONAtom::JK_Object:
-      return parseKeyValuePair();
-    default:
-      llvm_unreachable("Impossible code path");
-  }
-}
-
-bool JSONParser::skipContainer(const JSONContainer &Container) {
-  for (JSONContainer::AtomIterator I = Container.atom_current(),
-                                   E = Container.atom_end();
-       I != E; ++I) {
-    assert(*I != 0);
-    if (!skip(**I))
-      return false;
-  }
-  return !failed();
-}
diff --git a/lib/Support/Locale.cpp b/lib/Support/Locale.cpp
new file mode 100644
index 00000000000..17b9b6c47d6
--- /dev/null
+++ b/lib/Support/Locale.cpp
@@ -0,0 +1,10 @@
+#include "llvm/Support/Locale.h"
+#include "llvm/Config/config.h"
+
+#ifdef __APPLE__
+#include "LocaleXlocale.inc"
+#elif LLVM_ON_WIN32
+#include "LocaleWindows.inc"
+#else
+#include "LocaleGeneric.inc"
+#endif
diff --git a/lib/Support/LocaleGeneric.inc b/lib/Support/LocaleGeneric.inc
new file mode 100644
index 00000000000..278deee3e4d
--- /dev/null
+++ b/lib/Support/LocaleGeneric.inc
@@ -0,0 +1,17 @@
+#include 
+
+namespace llvm {
+namespace sys {
+namespace locale {
+
+int columnWidth(StringRef s) {
+    return s.size();
+}
+
+bool isPrint(int c) {
+    return iswprint(c);
+}
+
+}
+}
+}
diff --git a/lib/Support/LocaleWindows.inc b/lib/Support/LocaleWindows.inc
new file mode 100644
index 00000000000..6827ac15a1a
--- /dev/null
+++ b/lib/Support/LocaleWindows.inc
@@ -0,0 +1,15 @@
+namespace llvm {
+namespace sys {
+namespace locale {
+
+int columnWidth(StringRef s) {
+    return s.size();
+}
+
+bool isPrint(int c) {
+    return ' ' <= c && c <= '~';
+}
+
+}
+}
+}
\ No newline at end of file
diff --git a/lib/Support/LocaleXlocale.inc b/lib/Support/LocaleXlocale.inc
new file mode 100644
index 00000000000..f595e7c582c
--- /dev/null
+++ b/lib/Support/LocaleXlocale.inc
@@ -0,0 +1,61 @@
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/SmallString.h"
+#include "llvm/Support/ManagedStatic.h"
+#include 
+#include 
+
+
+namespace {
+  struct locale_holder {
+    locale_holder()
+    : l(newlocale(LC_CTYPE_MASK,"en_US.UTF-8",LC_GLOBAL_LOCALE))
+    {
+      assert(NULL!=l);
+    }
+    ~locale_holder() {
+      freelocale(l);
+    }
+
+    int mbswidth(llvm::SmallString<16> s) const {
+       // this implementation assumes no '\0' in s
+      assert(s.size()==strlen(s.c_str()));
+
+      size_t size = mbstowcs_l(NULL,s.c_str(),0,l);
+      assert(size!=(size_t)-1);
+      if (size==0)
+        return 0;
+      llvm::SmallVector ws(size);
+      size = mbstowcs_l(&ws[0],s.c_str(),ws.size(),l);
+      assert(ws.size()==size);
+      return wcswidth_l(&ws[0],ws.size(),l);
+    }
+
+    int isprint(int c) const {
+      return iswprint_l(c,l);
+    }
+
+  private:
+
+    locale_t l;
+  };
+
+  llvm::ManagedStatic l;
+}
+
+namespace llvm {
+namespace sys {
+namespace locale {
+
+int columnWidth(StringRef s) {
+  int width = l->mbswidth(s);
+  assert(width>=0);
+  return width;
+}
+
+bool isPrint(int c) {
+  return l->isprint(c);
+}
+
+}
+}
+}
diff --git a/lib/Support/SmallPtrSet.cpp b/lib/Support/SmallPtrSet.cpp
index 68d9c29411f..3b53e9ff49f 100644
--- a/lib/Support/SmallPtrSet.cpp
+++ b/lib/Support/SmallPtrSet.cpp
@@ -13,6 +13,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/ADT/SmallPtrSet.h"
+#include "llvm/ADT/DenseMapInfo.h"
 #include "llvm/Support/MathExtras.h"
 #include 
 #include 
@@ -102,7 +103,7 @@ bool SmallPtrSetImpl::erase_imp(const void * Ptr) {
 }
 
 const void * const *SmallPtrSetImpl::FindBucketFor(const void *Ptr) const {
-  unsigned Bucket = Hash(Ptr);
+  unsigned Bucket = DenseMapInfo::getHashValue(Ptr) & (CurArraySize-1);
   unsigned ArraySize = CurArraySize;
   unsigned ProbeAmt = 1;
   const void *const *Array = CurArray;
diff --git a/lib/Support/SourceMgr.cpp b/lib/Support/SourceMgr.cpp
index bbe36b260b9..15278c598e5 100644
--- a/lib/Support/SourceMgr.cpp
+++ b/lib/Support/SourceMgr.cpp
@@ -193,7 +193,8 @@ SMDiagnostic SourceMgr::GetMessage(SMLoc Loc, SourceMgr::DiagKind Kind,
 }
 
 void SourceMgr::PrintMessage(SMLoc Loc, SourceMgr::DiagKind Kind,
-                             const Twine &Msg, ArrayRef Ranges) const {
+                             const Twine &Msg, ArrayRef Ranges,
+                             bool ShowColors) const {
   SMDiagnostic Diagnostic = GetMessage(Loc, Kind, Msg, Ranges);
   
   // Report the message with the diagnostic handler if present.
@@ -208,7 +209,7 @@ void SourceMgr::PrintMessage(SMLoc Loc, SourceMgr::DiagKind Kind,
   assert(CurBuf != -1 && "Invalid or unspecified location!");
   PrintIncludeStack(getBufferInfo(CurBuf).IncludeLoc, OS);
 
-  Diagnostic.print(0, OS);
+  Diagnostic.print(0, OS, ShowColors);
 }
 
 //===----------------------------------------------------------------------===//
@@ -225,7 +226,14 @@ SMDiagnostic::SMDiagnostic(const SourceMgr &sm, SMLoc L, const std::string &FN,
 }
 
 
-void SMDiagnostic::print(const char *ProgName, raw_ostream &S) const {
+void SMDiagnostic::print(const char *ProgName, raw_ostream &S,
+                         bool ShowColors) const {
+  // Display colors only if OS goes to a tty.
+  ShowColors &= S.is_displayed();
+
+  if (ShowColors)
+    S.changeColor(raw_ostream::SAVEDCOLOR, true);
+
   if (ProgName && ProgName[0])
     S << ProgName << ": ";
 
@@ -244,13 +252,33 @@ void SMDiagnostic::print(const char *ProgName, raw_ostream &S) const {
   }
 
   switch (Kind) {
-  case SourceMgr::DK_Error: S << "error: "; break;
-  case SourceMgr::DK_Warning: S << "warning: "; break;
-  case SourceMgr::DK_Note: S << "note: "; break;
+  case SourceMgr::DK_Error:
+    if (ShowColors)
+      S.changeColor(raw_ostream::RED, true);
+    S << "error: ";
+    break;
+  case SourceMgr::DK_Warning:
+    if (ShowColors)
+      S.changeColor(raw_ostream::MAGENTA, true);
+    S << "warning: ";
+    break;
+  case SourceMgr::DK_Note:
+    if (ShowColors)
+      S.changeColor(raw_ostream::BLACK, true);
+    S << "note: ";
+    break;
   }
-  
+
+  if (ShowColors) {
+    S.resetColor();
+    S.changeColor(raw_ostream::SAVEDCOLOR, true);
+  }
+
   S << Message << '\n';
 
+  if (ShowColors)
+    S.resetColor();
+
   if (LineNo == -1 || ColumnNo == -1)
     return;
 
@@ -292,6 +320,9 @@ void SMDiagnostic::print(const char *ProgName, raw_ostream &S) const {
   }
   S << '\n';
 
+  if (ShowColors)
+    S.changeColor(raw_ostream::GREEN, true);
+
   // Print out the caret line, matching tabs in the source line.
   for (unsigned i = 0, e = CaretLine.size(), OutCol = 0; i != e; ++i) {
     if (i >= LineContents.size() || LineContents[i] != '\t') {
@@ -306,6 +337,9 @@ void SMDiagnostic::print(const char *ProgName, raw_ostream &S) const {
       ++OutCol;
     } while (OutCol & 7);
   }
+
+  if (ShowColors)
+    S.resetColor();
   
   S << '\n';
 }
diff --git a/lib/Support/Unix/Process.inc b/lib/Support/Unix/Process.inc
index 2d7fd384e8b..f640462a451 100644
--- a/lib/Support/Unix/Process.inc
+++ b/lib/Support/Unix/Process.inc
@@ -290,6 +290,10 @@ const char *Process::OutputBold(bool bg) {
   return "\033[1m";
 }
 
+const char *Process::OutputReverse() {
+  return "\033[7m";
+}
+
 const char *Process::ResetColor() {
   return "\033[0m";
 }
diff --git a/lib/Support/Windows/Process.inc b/lib/Support/Windows/Process.inc
index 913b0734ddc..9a388b4efc6 100644
--- a/lib/Support/Windows/Process.inc
+++ b/lib/Support/Windows/Process.inc
@@ -215,6 +215,38 @@ const char *Process::OutputColor(char code, bool bold, bool bg) {
   return 0;
 }
 
+static WORD GetConsoleTextAttribute(HANDLE hConsoleOutput) {
+  CONSOLE_SCREEN_BUFFER_INFO info;
+  GetConsoleScreenBufferInfo(GetStdHandle(STD_OUTPUT_HANDLE), &info);
+  return info.wAttributes;
+}
+
+const char *Process::OutputReverse() {
+  const WORD attributes
+   = GetConsoleTextAttribute(GetStdHandle(STD_OUTPUT_HANDLE));
+
+  const WORD foreground_mask = FOREGROUND_BLUE | FOREGROUND_GREEN |
+    FOREGROUND_RED | FOREGROUND_INTENSITY;
+  const WORD background_mask = BACKGROUND_BLUE | BACKGROUND_GREEN |
+    BACKGROUND_RED | BACKGROUND_INTENSITY;
+  const WORD color_mask = foreground_mask | background_mask;
+
+  WORD new_attributes =
+    ((attributes & FOREGROUND_BLUE     )?BACKGROUND_BLUE     :0) |
+    ((attributes & FOREGROUND_GREEN    )?BACKGROUND_GREEN    :0) |
+    ((attributes & FOREGROUND_RED      )?BACKGROUND_RED      :0) |
+    ((attributes & FOREGROUND_INTENSITY)?BACKGROUND_INTENSITY:0) |
+    ((attributes & BACKGROUND_BLUE     )?FOREGROUND_BLUE     :0) |
+    ((attributes & BACKGROUND_GREEN    )?FOREGROUND_GREEN    :0) |
+    ((attributes & BACKGROUND_RED      )?FOREGROUND_RED      :0) |
+    ((attributes & BACKGROUND_INTENSITY)?FOREGROUND_INTENSITY:0) |
+    0;
+  new_attributes = (attributes & ~color_mask) | (new_attributes & color_mask);
+
+  SetConsoleTextAttribute(GetStdHandle(STD_OUTPUT_HANDLE), new_attributes);
+  return 0;
+}
+
 const char *Process::ResetColor() {
   SetConsoleTextAttribute(GetStdHandle(STD_OUTPUT_HANDLE), defaultColors());
   return 0;
diff --git a/lib/Support/YAMLParser.cpp b/lib/Support/YAMLParser.cpp
index 330519f3019..d38b51b0b28 100644
--- a/lib/Support/YAMLParser.cpp
+++ b/lib/Support/YAMLParser.cpp
@@ -1732,7 +1732,7 @@ StringRef ScalarNode::unescapeDoubleQuoted( StringRef UnquotedValue
           if (UnquotedValue.size() < 3)
             // TODO: Report error.
             break;
-          unsigned int UnicodeScalarValue;
+          unsigned int UnicodeScalarValue = 0;
           UnquotedValue.substr(1, 2).getAsInteger(16, UnicodeScalarValue);
           encodeUTF8(UnicodeScalarValue, Storage);
           UnquotedValue = UnquotedValue.substr(2);
@@ -1742,7 +1742,7 @@ StringRef ScalarNode::unescapeDoubleQuoted( StringRef UnquotedValue
           if (UnquotedValue.size() < 5)
             // TODO: Report error.
             break;
-          unsigned int UnicodeScalarValue;
+          unsigned int UnicodeScalarValue = 0;
           UnquotedValue.substr(1, 4).getAsInteger(16, UnicodeScalarValue);
           encodeUTF8(UnicodeScalarValue, Storage);
           UnquotedValue = UnquotedValue.substr(4);
@@ -1752,7 +1752,7 @@ StringRef ScalarNode::unescapeDoubleQuoted( StringRef UnquotedValue
           if (UnquotedValue.size() < 9)
             // TODO: Report error.
             break;
-          unsigned int UnicodeScalarValue;
+          unsigned int UnicodeScalarValue = 0;
           UnquotedValue.substr(1, 8).getAsInteger(16, UnicodeScalarValue);
           encodeUTF8(UnicodeScalarValue, Storage);
           UnquotedValue = UnquotedValue.substr(8);
diff --git a/lib/Support/raw_ostream.cpp b/lib/Support/raw_ostream.cpp
index 72d3986f41d..86cdca15727 100644
--- a/lib/Support/raw_ostream.cpp
+++ b/lib/Support/raw_ostream.cpp
@@ -633,6 +633,19 @@ raw_ostream &raw_fd_ostream::resetColor() {
   return *this;
 }
 
+raw_ostream &raw_fd_ostream::reverseColor() {
+  if (sys::Process::ColorNeedsFlush())
+    flush();
+  const char *colorcode = sys::Process::OutputReverse();
+  if (colorcode) {
+    size_t len = strlen(colorcode);
+    write(colorcode, len);
+    // don't account colors towards output characters
+    pos -= len;
+  }
+  return *this;
+}
+
 bool raw_fd_ostream::is_displayed() const {
   return sys::Process::FileDescriptorIsDisplayed(FD);
 }
diff --git a/lib/TableGen/Error.cpp b/lib/TableGen/Error.cpp
index 5071ee77ac4..1463b68144a 100644
--- a/lib/TableGen/Error.cpp
+++ b/lib/TableGen/Error.cpp
@@ -20,6 +20,22 @@ namespace llvm {
 
 SourceMgr SrcMgr;
 
+void PrintWarning(SMLoc WarningLoc, const Twine &Msg) {
+  SrcMgr.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg);
+}
+
+void PrintWarning(const char *Loc, const Twine &Msg) {
+  SrcMgr.PrintMessage(SMLoc::getFromPointer(Loc), SourceMgr::DK_Warning, Msg);
+}
+
+void PrintWarning(const Twine &Msg) {
+  errs() << "warning:" << Msg << "\n";
+}
+
+void PrintWarning(const TGError &Warning) {
+  PrintWarning(Warning.getLoc(), Warning.getMessage());
+}
+
 void PrintError(SMLoc ErrorLoc, const Twine &Msg) {
   SrcMgr.PrintMessage(ErrorLoc, SourceMgr::DK_Error, Msg);
 }
diff --git a/lib/Target/ARM/ARMCallingConv.td b/lib/Target/ARM/ARMCallingConv.td
index d33364bb287..b9a25126ba6 100644
--- a/lib/Target/ARM/ARMCallingConv.td
+++ b/lib/Target/ARM/ARMCallingConv.td
@@ -9,10 +9,6 @@
 // This describes the calling conventions for ARM architecture.
 //===----------------------------------------------------------------------===//
 
-/// CCIfSubtarget - Match if the current subtarget has a feature F.
-class CCIfSubtarget:
-  CCIf().", F), A>;
-
 /// CCIfAlign - Match of the original alignment of the arg
 class CCIfAlign:
   CCIf;
diff --git a/lib/Target/ARM/ARMInstrFormats.td b/lib/Target/ARM/ARMInstrFormats.td
index 1d38bcf9e84..f04926aaceb 100644
--- a/lib/Target/ARM/ARMInstrFormats.td
+++ b/lib/Target/ARM/ARMInstrFormats.td
@@ -532,6 +532,7 @@ class AIswp pattern>
   let Inst{11-4} = 0b00001001;
   let Inst{3-0} = Rt2;
 
+  let Unpredictable{11-8} = 0b1111;
   let DecoderMethod = "DecodeSwap";
 }
 
diff --git a/lib/Target/ARM/ARMInstrInfo.td b/lib/Target/ARM/ARMInstrInfo.td
index 3caaa236612..1eb561d6901 100644
--- a/lib/Target/ARM/ARMInstrInfo.td
+++ b/lib/Target/ARM/ARMInstrInfo.td
@@ -219,8 +219,11 @@ def UseFPVMLx        : Predicate<"Subtarget->useFPVMLx()">;
 
 // Prefer fused MAC for fp mul + add over fp VMLA / VMLS if they are available.
 // But only select them if more precision in FP computation is allowed.
-def UseFusedMAC      : Predicate<"!TM.Options.NoExcessFPPrecision">;
-def DontUseFusedMAC  : Predicate<"!Subtarget->hasVFP4()">;
+// Do not use them for Darwin platforms.
+def UseFusedMAC      : Predicate<"!TM.Options.NoExcessFPPrecision && "
+                                 "!Subtarget->isTargetDarwin()">;
+def DontUseFusedMAC  : Predicate<"!Subtarget->hasVFP4() || "
+                                 "Subtarget->isTargetDarwin()">;
 
 //===----------------------------------------------------------------------===//
 // ARM Flag Definitions.
@@ -905,6 +908,11 @@ def p_imm : Operand {
   let DecoderMethod = "DecodeCoprocessor";
 }
 
+def pf_imm : Operand {
+  let PrintMethod = "printPImmediate";
+  let ParserMatchClass = CoprocNumAsmOperand;
+}
+
 def CoprocRegAsmOperand : AsmOperandClass {
   let Name = "CoprocReg";
   let ParserMethod = "parseCoprocRegOperand";
@@ -1184,6 +1192,8 @@ multiclass AI1_cmp_irs opcod, string opc,
     let Inst{19-16} = Rn;
     let Inst{15-12} = 0b0000;
     let Inst{11-0} = imm;
+
+    let Unpredictable{15-12} = 0b1111;
   }
   def rr : AI1 opcod, string opc,
     let Inst{15-12} = 0b0000;
     let Inst{11-4} = 0b00000000;
     let Inst{3-0} = Rm;
+
+    let Unpredictable{15-12} = 0b1111;
   }
   def rsi : AI1 opcod, string opc,
     let Inst{11-5} = shift{11-5};
     let Inst{4} = 0;
     let Inst{3-0} = shift{3-0};
+
+    let Unpredictable{15-12} = 0b1111;
   }
   def rsr : AI1 {
+               [(opnode GPRnopc:$Rn, so_reg_reg:$shift)]> {
     bits<4> Rn;
     bits<12> shift;
     let Inst{25} = 0;
@@ -1227,6 +1241,8 @@ multiclass AI1_cmp_irs opcod, string opc,
     let Inst{6-5} = shift{6-5};
     let Inst{4} = 1;
     let Inst{3-0} = shift{3-0};
+
+    let Unpredictable{15-12} = 0b1111;
   }
 
 }
@@ -4103,7 +4119,7 @@ def ISB : AInoP<(outs), (ins memb_opt:$opt), MiscFrm, NoItinerary,
   let Inst{3-0} = opt;
 }
 
-// Pseudo isntruction that combines movs + predicated rsbmi
+// Pseudo instruction that combines movs + predicated rsbmi
 // to implement integer ABS
 let usesCustomInserter = 1, Defs = [CPSR] in {
 def ABS : ARMPseudoInst<
@@ -4264,9 +4280,9 @@ def CLREX : AXI<(outs), (ins), MiscFrm, NoItinerary, "clrex", []>,
 
 // SWP/SWPB are deprecated in V6/V7.
 let mayLoad = 1, mayStore = 1 in {
-def SWP : AIswp<0, (outs GPR:$Rt), (ins GPR:$Rt2, addr_offset_none:$addr),
+def SWP : AIswp<0, (outs GPRnopc:$Rt), (ins GPRnopc:$Rt2, addr_offset_none:$addr),
                 "swp", []>;
-def SWPB: AIswp<1, (outs GPR:$Rt), (ins GPR:$Rt2, addr_offset_none:$addr),
+def SWPB: AIswp<1, (outs GPRnopc:$Rt), (ins GPRnopc:$Rt2, addr_offset_none:$addr),
                 "swpb", []>;
 }
 
@@ -4295,7 +4311,7 @@ def CDP : ABI<0b1110, (outs), (ins p_imm:$cop, imm0_15:$opc1,
   let Inst{23-20} = opc1;
 }
 
-def CDP2 : ABXI<0b1110, (outs), (ins p_imm:$cop, imm0_15:$opc1,
+def CDP2 : ABXI<0b1110, (outs), (ins pf_imm:$cop, imm0_15:$opc1,
                c_imm:$CRd, c_imm:$CRn, c_imm:$CRm, imm0_7:$opc2),
                NoItinerary, "cdp2\t$cop, $opc1, $CRd, $CRn, $CRm, $opc2",
                [(int_arm_cdp2 imm:$cop, imm:$opc1, imm:$CRd, imm:$CRn,
@@ -4574,7 +4590,7 @@ def : ARMV5TPat<(int_arm_mrc2 imm:$cop, imm:$opc1, imm:$CRn,
 
 class MovRRCopro pattern = []>
   : ABI<0b1100, (outs), (ins p_imm:$cop, imm0_15:$opc1,
-        GPR:$Rt, GPR:$Rt2, c_imm:$CRm),
+        GPRnopc:$Rt, GPRnopc:$Rt2, c_imm:$CRm),
         NoItinerary, opc, "\t$cop, $opc1, $Rt, $Rt2, $CRm", pattern> {
   let Inst{23-21} = 0b010;
   let Inst{20} = direction;
@@ -4593,13 +4609,13 @@ class MovRRCopro pattern = []>
 }
 
 def MCRR : MovRRCopro<"mcrr", 0 /* from ARM core register to coprocessor */,
-                      [(int_arm_mcrr imm:$cop, imm:$opc1, GPR:$Rt, GPR:$Rt2,
+                      [(int_arm_mcrr imm:$cop, imm:$opc1, GPRnopc:$Rt, GPRnopc:$Rt2,
                                      imm:$CRm)]>;
 def MRRC : MovRRCopro<"mrrc", 1 /* from coprocessor to ARM core register */>;
 
 class MovRRCopro2 pattern = []>
   : ABXI<0b1100, (outs), (ins p_imm:$cop, imm0_15:$opc1,
-         GPR:$Rt, GPR:$Rt2, c_imm:$CRm), NoItinerary,
+         GPRnopc:$Rt, GPRnopc:$Rt2, c_imm:$CRm), NoItinerary,
          !strconcat(opc, "\t$cop, $opc1, $Rt, $Rt2, $CRm"), pattern> {
   let Inst{31-28} = 0b1111;
   let Inst{23-21} = 0b010;
@@ -4616,10 +4632,12 @@ class MovRRCopro2 pattern = []>
   let Inst{11-8}  = cop;
   let Inst{7-4}   = opc1;
   let Inst{3-0}   = CRm;
+
+  let DecoderMethod = "DecodeMRRC2";
 }
 
 def MCRR2 : MovRRCopro2<"mcrr2", 0 /* from ARM core register to coprocessor */,
-                        [(int_arm_mcrr2 imm:$cop, imm:$opc1, GPR:$Rt, GPR:$Rt2,
+                        [(int_arm_mcrr2 imm:$cop, imm:$opc1, GPRnopc:$Rt, GPRnopc:$Rt2,
                                         imm:$CRm)]>;
 def MRRC2 : MovRRCopro2<"mrrc2", 1 /* from coprocessor to ARM core register */>;
 
@@ -4628,22 +4646,32 @@ def MRRC2 : MovRRCopro2<"mrrc2", 1 /* from coprocessor to ARM core register */>;
 //
 
 // Move to ARM core register from Special Register
-def MRS : ABI<0b0001, (outs GPR:$Rd), (ins), NoItinerary,
+def MRS : ABI<0b0001, (outs GPRnopc:$Rd), (ins), NoItinerary,
               "mrs", "\t$Rd, apsr", []> {
   bits<4> Rd;
   let Inst{23-16} = 0b00001111;
+  let Unpredictable{19-17} = 0b111;
+
   let Inst{15-12} = Rd;
-  let Inst{7-4} = 0b0000;
+
+  let Inst{11-0} = 0b000000000000;
+  let Unpredictable{11-0} = 0b110100001111;
 }
 
-def : InstAlias<"mrs${p} $Rd, cpsr", (MRS GPR:$Rd, pred:$p)>, Requires<[IsARM]>;
+def : InstAlias<"mrs${p} $Rd, cpsr", (MRS GPRnopc:$Rd, pred:$p)>, Requires<[IsARM]>;
 
-def MRSsys : ABI<0b0001, (outs GPR:$Rd), (ins), NoItinerary,
+// The MRSsys instruction is the MRS instruction from the ARM ARM,
+// section B9.3.9, with the R bit set to 1.
+def MRSsys : ABI<0b0001, (outs GPRnopc:$Rd), (ins), NoItinerary,
                  "mrs", "\t$Rd, spsr", []> {
   bits<4> Rd;
   let Inst{23-16} = 0b01001111;
+  let Unpredictable{19-16} = 0b1111;
+
   let Inst{15-12} = Rd;
-  let Inst{7-4} = 0b0000;
+
+  let Inst{11-0} = 0b000000000000;
+  let Unpredictable{11-0} = 0b110100001111;
 }
 
 // Move from ARM core register to Special Register
diff --git a/lib/Target/ARM/ARMInstrNEON.td b/lib/Target/ARM/ARMInstrNEON.td
index c7219a60f6c..fd8ac0b328e 100644
--- a/lib/Target/ARM/ARMInstrNEON.td
+++ b/lib/Target/ARM/ARMInstrNEON.td
@@ -5634,6 +5634,7 @@ multiclass Lengthen_HalfSingle =
 //     Pat<(v4i32 (extloadvi8 addrmode5:$addr))
 //         (EXTRACT_SUBREG (VMOVLuv4i32 
@@ -5644,28 +5645,63 @@ multiclass Lengthen_HalfSingle;
 multiclass Lengthen_Double {
+                           string Insn2Ty> {
+  def _Any : Pat<(!cast("v" # DestLanes # DestTy)
+                   (!cast("extloadv" # SrcTy) addrmode5:$addr)),
+         (!cast("VMOVLuv" # Insn2Lanes # Insn2Ty)
+           (EXTRACT_SUBREG (!cast("VMOVLuv" # Insn1Lanes # Insn1Ty)
+             (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr),
+              ssub_0)), dsub_0))>;
+  def _Z   : Pat<(!cast("v" # DestLanes # DestTy)
+                   (!cast("zextloadv" # SrcTy) addrmode5:$addr)),
+         (!cast("VMOVLuv" # Insn2Lanes # Insn2Ty)
+           (EXTRACT_SUBREG (!cast("VMOVLuv" # Insn1Lanes # Insn1Ty)
+             (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr),
+              ssub_0)), dsub_0))>;
+  def _S   : Pat<(!cast("v" # DestLanes # DestTy)
+                   (!cast("sextloadv" # SrcTy) addrmode5:$addr)),
+         (!cast("VMOVLsv" # Insn2Lanes # Insn2Ty)
+           (EXTRACT_SUBREG (!cast("VMOVLsv" # Insn1Lanes # Insn1Ty)
+             (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr),
+              ssub_0)), dsub_0))>;
+}
+
+// extload, zextload and sextload for a lengthening load followed by another
+// lengthening load, to quadruple the initial length, but which ends up only
+// requiring half the available lanes (a 64-bit outcome instead of a 128-bit).
+//
+// Lengthen_HalfDouble<"2", "i32", "i8", "8", "i16", "4", "i32"> =
+//     Pat<(v4i32 (extloadvi8 addrmode5:$addr))
+//         (EXTRACT_SUBREG (VMOVLuv4i32 
+//           (EXTRACT_SUBREG (VMOVLuv8i16 (INSERT_SUBREG (f64 (IMPLICIT_DEF)),
+//                                                       (VLDRS addrmode5:$addr),
+//                                                       ssub_0)),
+//                           dsub_0)),
+//           dsub_0)>;
+multiclass Lengthen_HalfDouble {
   def _Any : Pat<(!cast("v" # DestLanes # DestTy)
                    (!cast("extloadv" # SrcTy) addrmode5:$addr)),
          (EXTRACT_SUBREG (!cast("VMOVLuv" # Insn2Lanes # Insn2Ty)
            (EXTRACT_SUBREG (!cast("VMOVLuv" # Insn1Lanes # Insn1Ty)
              (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr),
               ssub_0)), dsub_0)),
-          RegType)>;
+          dsub_0)>;
   def _Z   : Pat<(!cast("v" # DestLanes # DestTy)
                    (!cast("zextloadv" # SrcTy) addrmode5:$addr)),
          (EXTRACT_SUBREG (!cast("VMOVLuv" # Insn2Lanes # Insn2Ty)
            (EXTRACT_SUBREG (!cast("VMOVLuv" # Insn1Lanes # Insn1Ty)
              (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr),
               ssub_0)), dsub_0)),
-          RegType)>;
+          dsub_0)>;
   def _S   : Pat<(!cast("v" # DestLanes # DestTy)
                    (!cast("sextloadv" # SrcTy) addrmode5:$addr)),
          (EXTRACT_SUBREG (!cast("VMOVLsv" # Insn2Lanes # Insn2Ty)
            (EXTRACT_SUBREG (!cast("VMOVLsv" # Insn1Lanes # Insn1Ty)
              (INSERT_SUBREG (f64 (IMPLICIT_DEF)), (VLDRS addrmode5:$addr),
               ssub_0)), dsub_0)),
-          RegType)>;
+          dsub_0)>;
 }
 
 defm : Lengthen_Single<"8", "i16", "i8">; // v8i8 -> v8i16
@@ -5676,12 +5712,12 @@ defm : Lengthen_HalfSingle<"4", "i16", "i8", "8", "i16">; // v4i8 -> v4i16
 defm : Lengthen_HalfSingle<"2", "i16", "i8", "8", "i16">; // v2i8 -> v2i16
 defm : Lengthen_HalfSingle<"2", "i32", "i16", "4", "i32">; // v2i16 -> v2i32
 
-// Double lengthening - v4i8 -> v4i16 -> v4i32 
-defm : Lengthen_Double<"4", "i32", "i8", "8", "i16", "4", "i32", qsub_0>;
+// Double lengthening - v4i8 -> v4i16 -> v4i32
+defm : Lengthen_Double<"4", "i32", "i8", "8", "i16", "4", "i32">;
 // v2i8 -> v2i16 -> v2i32
-defm : Lengthen_Double<"2", "i32", "i8", "8", "i16", "4", "i32", dsub_0>;
+defm : Lengthen_HalfDouble<"2", "i32", "i8", "8", "i16", "4", "i32">;
 // v2i16 -> v2i32 -> v2i64
-defm : Lengthen_Double<"2", "i64", "i16", "4", "i32", "2", "i64", qsub_0>;
+defm : Lengthen_Double<"2", "i64", "i16", "4", "i32", "2", "i64">;
 
 // Triple lengthening - v2i8 -> v2i16 -> v2i32 -> v2i64
 def : Pat<(v2i64 (extloadvi8 addrmode5:$addr)),
@@ -5951,7 +5987,7 @@ def : NEONInstAlias<"vshl${p}.u32 $Vdn, $Vm",
 def : NEONInstAlias<"vshl${p}.u64 $Vdn, $Vm",
                     (VSHLuv2i64 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
 
-// VSHL (immediate) two-operand aliases.
+// VSHR (immediate) two-operand aliases.
 def : NEONInstAlias<"vshr${p}.s8 $Vdn, $imm",
                     (VSHRsv8i8 DPR:$Vdn, DPR:$Vdn, shr_imm8:$imm, pred:$p)>;
 def : NEONInstAlias<"vshr${p}.s16 $Vdn, $imm",
@@ -5988,6 +6024,41 @@ def : NEONInstAlias<"vshr${p}.u32 $Vdn, $imm",
 def : NEONInstAlias<"vshr${p}.u64 $Vdn, $imm",
                     (VSHRuv2i64 QPR:$Vdn, QPR:$Vdn, shr_imm64:$imm, pred:$p)>;
 
+// VRSHL two-operand aliases.
+def : NEONInstAlias<"vrshl${p}.s8 $Vdn, $Vm",
+                    (VRSHLsv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.s16 $Vdn, $Vm",
+                    (VRSHLsv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.s32 $Vdn, $Vm",
+                    (VRSHLsv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.s64 $Vdn, $Vm",
+                    (VRSHLsv1i64 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.u8 $Vdn, $Vm",
+                    (VRSHLuv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.u16 $Vdn, $Vm",
+                    (VRSHLuv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.u32 $Vdn, $Vm",
+                    (VRSHLuv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.u64 $Vdn, $Vm",
+                    (VRSHLuv1i64 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vrshl${p}.s8 $Vdn, $Vm",
+                    (VRSHLsv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.s16 $Vdn, $Vm",
+                    (VRSHLsv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.s32 $Vdn, $Vm",
+                    (VRSHLsv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.s64 $Vdn, $Vm",
+                    (VRSHLsv2i64 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.u8 $Vdn, $Vm",
+                    (VRSHLuv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.u16 $Vdn, $Vm",
+                    (VRSHLuv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.u32 $Vdn, $Vm",
+                    (VRSHLuv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vrshl${p}.u64 $Vdn, $Vm",
+                    (VRSHLuv2i64 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
 // VLD1 single-lane pseudo-instructions. These need special handling for
 // the lane index that an InstAlias can't handle, so we use these instead.
 def VLD1LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vld1${p}", ".8", "$list, $addr",
@@ -6951,6 +7022,100 @@ def : NEONInstAlias<"vsli${p}.32 $Vdm, $imm",
 def : NEONInstAlias<"vsli${p}.64 $Vdm, $imm",
                     (VSLIv2i64 QPR:$Vdm, QPR:$Vdm, shr_imm64:$imm, pred:$p)>;
 
+// Two-operand variants for VHSUB.
+    // Signed.
+def : NEONInstAlias<"vhsub${p}.s8 $Vdn, $Vm",
+                    (VHSUBsv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhsub${p}.s16 $Vdn, $Vm",
+                    (VHSUBsv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhsub${p}.s32 $Vdn, $Vm",
+                    (VHSUBsv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vhsub${p}.s8 $Vdn, $Vm",
+                    (VHSUBsv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhsub${p}.s16 $Vdn, $Vm",
+                    (VHSUBsv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhsub${p}.s32 $Vdn, $Vm",
+                    (VHSUBsv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+    // Unsigned.
+def : NEONInstAlias<"vhsub${p}.u8 $Vdn, $Vm",
+                    (VHSUBuv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhsub${p}.u16 $Vdn, $Vm",
+                    (VHSUBuv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhsub${p}.u32 $Vdn, $Vm",
+                    (VHSUBuv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vhsub${p}.u8 $Vdn, $Vm",
+                    (VHSUBuv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhsub${p}.u16 $Vdn, $Vm",
+                    (VHSUBuv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhsub${p}.u32 $Vdn, $Vm",
+                    (VHSUBuv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+
+// Two-operand variants for VHADD.
+    // Signed.
+def : NEONInstAlias<"vhadd${p}.s8 $Vdn, $Vm",
+                    (VHADDsv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhadd${p}.s16 $Vdn, $Vm",
+                    (VHADDsv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhadd${p}.s32 $Vdn, $Vm",
+                    (VHADDsv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vhadd${p}.s8 $Vdn, $Vm",
+                    (VHADDsv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhadd${p}.s16 $Vdn, $Vm",
+                    (VHADDsv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhadd${p}.s32 $Vdn, $Vm",
+                    (VHADDsv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+    // Unsigned.
+def : NEONInstAlias<"vhadd${p}.u8 $Vdn, $Vm",
+                    (VHADDuv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhadd${p}.u16 $Vdn, $Vm",
+                    (VHADDuv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhadd${p}.u32 $Vdn, $Vm",
+                    (VHADDuv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vhadd${p}.u8 $Vdn, $Vm",
+                    (VHADDuv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhadd${p}.u16 $Vdn, $Vm",
+                    (VHADDuv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vhadd${p}.u32 $Vdn, $Vm",
+                    (VHADDuv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+// Two-operand variants for VRHADD.
+    // Signed.
+def : NEONInstAlias<"vrhadd${p}.s8 $Vdn, $Rm",
+                    (VRHADDsv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Rm, pred:$p)>;
+def : NEONInstAlias<"vrhadd${p}.s16 $Vdn, $Rm",
+                    (VRHADDsv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Rm, pred:$p)>;
+def : NEONInstAlias<"vrhadd${p}.s32 $Vdn, $Rm",
+                    (VRHADDsv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Rm, pred:$p)>;
+
+def : NEONInstAlias<"vrhadd${p}.s8 $Vdn, $Rm",
+                    (VRHADDsv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Rm, pred:$p)>;
+def : NEONInstAlias<"vrhadd${p}.s16 $Vdn, $Rm",
+                    (VRHADDsv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Rm, pred:$p)>;
+def : NEONInstAlias<"vrhadd${p}.s32 $Vdn, $Rm",
+                    (VRHADDsv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Rm, pred:$p)>;
+
+    // Unsigned.
+def : NEONInstAlias<"vrhadd${p}.u8 $Vdn, $Rm",
+                    (VRHADDuv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Rm, pred:$p)>;
+def : NEONInstAlias<"vrhadd${p}.u16 $Vdn, $Rm",
+                    (VRHADDuv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Rm, pred:$p)>;
+def : NEONInstAlias<"vrhadd${p}.u32 $Vdn, $Rm",
+                    (VRHADDuv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Rm, pred:$p)>;
+
+def : NEONInstAlias<"vrhadd${p}.u8 $Vdn, $Rm",
+                    (VRHADDuv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Rm, pred:$p)>;
+def : NEONInstAlias<"vrhadd${p}.u16 $Vdn, $Rm",
+                    (VRHADDuv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Rm, pred:$p)>;
+def : NEONInstAlias<"vrhadd${p}.u32 $Vdn, $Rm",
+                    (VRHADDuv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Rm, pred:$p)>;
+
 // VSWP allows, but does not require, a type suffix.
 defm : NEONDTAnyInstAlias<"vswp${p}", "$Vd, $Vm",
                          (VSWPd DPR:$Vd, DPR:$Vm, pred:$p)>;
diff --git a/lib/Target/ARM/ARMTargetMachine.cpp b/lib/Target/ARM/ARMTargetMachine.cpp
index 047efc23a4e..9aa83089202 100644
--- a/lib/Target/ARM/ARMTargetMachine.cpp
+++ b/lib/Target/ARM/ARMTargetMachine.cpp
@@ -136,22 +136,22 @@ TargetPassConfig *ARMBaseTargetMachine::createPassConfig(PassManagerBase &PM) {
 
 bool ARMPassConfig::addPreISel() {
   if (TM->getOptLevel() != CodeGenOpt::None && EnableGlobalMerge)
-    PM.add(createGlobalMergePass(TM->getTargetLowering()));
+    PM->add(createGlobalMergePass(TM->getTargetLowering()));
 
   return false;
 }
 
 bool ARMPassConfig::addInstSelector() {
-  PM.add(createARMISelDag(getARMTargetMachine(), getOptLevel()));
+  PM->add(createARMISelDag(getARMTargetMachine(), getOptLevel()));
   return false;
 }
 
 bool ARMPassConfig::addPreRegAlloc() {
   // FIXME: temporarily disabling load / store optimization pass for Thumb1.
   if (getOptLevel() != CodeGenOpt::None && !getARMSubtarget().isThumb1Only())
-    PM.add(createARMLoadStoreOptimizationPass(true));
+    PM->add(createARMLoadStoreOptimizationPass(true));
   if (getOptLevel() != CodeGenOpt::None && getARMSubtarget().isCortexA9())
-    PM.add(createMLxExpansionPass());
+    PM->add(createMLxExpansionPass());
   return true;
 }
 
@@ -159,23 +159,23 @@ bool ARMPassConfig::addPreSched2() {
   // FIXME: temporarily disabling load / store optimization pass for Thumb1.
   if (getOptLevel() != CodeGenOpt::None) {
     if (!getARMSubtarget().isThumb1Only()) {
-      PM.add(createARMLoadStoreOptimizationPass());
+      PM->add(createARMLoadStoreOptimizationPass());
       printAndVerify("After ARM load / store optimizer");
     }
     if (getARMSubtarget().hasNEON())
-      PM.add(createExecutionDependencyFixPass(&ARM::DPRRegClass));
+      PM->add(createExecutionDependencyFixPass(&ARM::DPRRegClass));
   }
 
   // Expand some pseudo instructions into multiple instructions to allow
   // proper scheduling.
-  PM.add(createARMExpandPseudoPass());
+  PM->add(createARMExpandPseudoPass());
 
   if (getOptLevel() != CodeGenOpt::None) {
     if (!getARMSubtarget().isThumb1Only())
       addPass(IfConverterID);
   }
   if (getARMSubtarget().isThumb2())
-    PM.add(createThumb2ITBlockPass());
+    PM->add(createThumb2ITBlockPass());
 
   return true;
 }
@@ -183,13 +183,13 @@ bool ARMPassConfig::addPreSched2() {
 bool ARMPassConfig::addPreEmitPass() {
   if (getARMSubtarget().isThumb2()) {
     if (!getARMSubtarget().prefers32BitThumb())
-      PM.add(createThumb2SizeReductionPass());
+      PM->add(createThumb2SizeReductionPass());
 
     // Constant island pass work on unbundled instructions.
     addPass(UnpackMachineBundlesID);
   }
 
-  PM.add(createARMConstantIslandPass());
+  PM->add(createARMConstantIslandPass());
 
   return true;
 }
diff --git a/lib/Target/ARM/AsmParser/ARMAsmParser.cpp b/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
index e55a7dad45d..2c53e3f8f8c 100644
--- a/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
+++ b/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
@@ -82,8 +82,14 @@ class ARMAsmParser : public MCTargetAsmParser {
   MCAsmParser &getParser() const { return Parser; }
   MCAsmLexer &getLexer() const { return Parser.getLexer(); }
 
-  void Warning(SMLoc L, const Twine &Msg) { Parser.Warning(L, Msg); }
-  bool Error(SMLoc L, const Twine &Msg) { return Parser.Error(L, Msg); }
+  bool Warning(SMLoc L, const Twine &Msg,
+               ArrayRef Ranges = ArrayRef()) {
+    return Parser.Warning(L, Msg, Ranges);
+  }
+  bool Error(SMLoc L, const Twine &Msg,
+             ArrayRef Ranges = ArrayRef()) {
+    return Parser.Error(L, Msg, Ranges);
+  }
 
   int tryParseRegister();
   bool tryParseRegisterWithWriteBack(SmallVectorImpl &);
@@ -478,6 +484,8 @@ public:
   /// getEndLoc - Get the location of the last token of this operand.
   SMLoc getEndLoc() const { return EndLoc; }
 
+  SMRange getLocRange() const { return SMRange(StartLoc, EndLoc); }
+
   ARMCC::CondCodes getCondCode() const {
     assert(Kind == k_CondCode && "Invalid access!");
     return CC.Val;
@@ -4518,22 +4526,26 @@ bool ARMAsmParser::parseOperand(SmallVectorImpl &Operands,
   case AsmToken::Dollar:
   case AsmToken::Hash: {
     // #42 -> immediate.
-    // TODO: ":lower16:" and ":upper16:" modifiers after # before immediate
     S = Parser.getTok().getLoc();
     Parser.Lex();
-    bool isNegative = Parser.getTok().is(AsmToken::Minus);
-    const MCExpr *ImmVal;
-    if (getParser().ParseExpression(ImmVal))
-      return true;
-    const MCConstantExpr *CE = dyn_cast(ImmVal);
-    if (CE) {
-      int32_t Val = CE->getValue();
-      if (isNegative && Val == 0)
-        ImmVal = MCConstantExpr::Create(INT32_MIN, getContext());
+
+    if (Parser.getTok().isNot(AsmToken::Colon)) {
+      bool isNegative = Parser.getTok().is(AsmToken::Minus);
+      const MCExpr *ImmVal;
+      if (getParser().ParseExpression(ImmVal))
+        return true;
+      const MCConstantExpr *CE = dyn_cast(ImmVal);
+      if (CE) {
+        int32_t Val = CE->getValue();
+        if (isNegative && Val == 0)
+          ImmVal = MCConstantExpr::Create(INT32_MIN, getContext());
+      }
+      E = SMLoc::getFromPointer(Parser.getTok().getLoc().getPointer() - 1);
+      Operands.push_back(ARMOperand::CreateImm(ImmVal, S, E));
+      return false;
     }
-    E = SMLoc::getFromPointer(Parser.getTok().getLoc().getPointer() - 1);
-    Operands.push_back(ARMOperand::CreateImm(ImmVal, S, E));
-    return false;
+    // w/ a ':' after the '#', it's just like a plain ':'.
+    // FALLTHROUGH
   }
   case AsmToken::Colon: {
     // ":lower16:" and ":upper16:" expression prefixes
@@ -7321,7 +7333,8 @@ MatchAndEmitInstruction(SMLoc IDLoc,
     return Error(ErrorLoc, "invalid operand for instruction");
   }
   case Match_MnemonicFail:
-    return Error(IDLoc, "invalid instruction");
+    return Error(IDLoc, "invalid instruction",
+                 ((ARMOperand*)Operands[0])->getLocRange());
   case Match_ConversionFail:
     // The converter function will have already emited a diagnostic.
     return true;
diff --git a/lib/Target/ARM/Disassembler/ARMDisassembler.cpp b/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
index 2f504b756b1..912935db17a 100644
--- a/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
+++ b/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
@@ -326,6 +326,8 @@ static DecodeStatus DecodeT2ShifterImmOperand(MCInst &Inst, unsigned Val,
 
 static DecodeStatus DecodeLDR(MCInst &Inst, unsigned Val,
                                 uint64_t Address, const void *Decoder);
+static DecodeStatus DecodeMRRC2(llvm::MCInst &Inst, unsigned Val,
+                                uint64_t Address, const void *Decoder);
 #include "ARMGenDisassemblerTables.inc"
 #include "ARMGenInstrInfo.inc"
 #include "ARMGenEDInfo.inc"
@@ -2690,7 +2692,6 @@ static DecodeStatus DecodeVLD2DupInstruction(MCInst &Inst, unsigned Insn,
   unsigned Rm = fieldFromInstruction32(Insn, 0, 4);
   unsigned align = fieldFromInstruction32(Insn, 4, 1);
   unsigned size = 1 << fieldFromInstruction32(Insn, 6, 2);
-  unsigned pred = fieldFromInstruction32(Insn, 22, 4);
   align *= 2*size;
 
   switch (Inst.getOpcode()) {
@@ -2721,16 +2722,11 @@ static DecodeStatus DecodeVLD2DupInstruction(MCInst &Inst, unsigned Insn,
     return MCDisassembler::Fail;
   Inst.addOperand(MCOperand::CreateImm(align));
 
-  if (Rm == 0xD)
-    Inst.addOperand(MCOperand::CreateReg(0));
-  else if (Rm != 0xF) {
+  if (Rm != 0xD && Rm != 0xF) {
     if (!Check(S, DecodeGPRRegisterClass(Inst, Rm, Address, Decoder)))
       return MCDisassembler::Fail;
   }
 
-  if (!Check(S, DecodePredicateOperand(Inst, pred, Address, Decoder)))
-    return MCDisassembler::Fail;
-
   return S;
 }
 
@@ -4314,6 +4310,10 @@ static DecodeStatus DecodeSwap(MCInst &Inst, unsigned Insn,
     return DecodeCPSInstruction(Inst, Insn, Address, Decoder);
 
   DecodeStatus S = MCDisassembler::Success;
+
+  if (Rt == Rn || Rn == Rt2)
+    S = MCDisassembler::SoftFail;
+
   if (!Check(S, DecodeGPRnopcRegisterClass(Inst, Rt, Address, Decoder)))
     return MCDisassembler::Fail;
   if (!Check(S, DecodeGPRnopcRegisterClass(Inst, Rt2, Address, Decoder)))
@@ -4409,3 +4409,31 @@ static DecodeStatus DecodeLDR(MCInst &Inst, unsigned Val,
   return S;
 }
 
+static DecodeStatus DecodeMRRC2(llvm::MCInst &Inst, unsigned Val,
+                                uint64_t Address, const void *Decoder) {
+
+  DecodeStatus S = MCDisassembler::Success;
+
+  unsigned CRm = fieldFromInstruction32(Val, 0, 4);
+  unsigned opc1 = fieldFromInstruction32(Val, 4, 4);
+  unsigned cop = fieldFromInstruction32(Val, 8, 4);
+  unsigned Rt = fieldFromInstruction32(Val, 12, 4);
+  unsigned Rt2 = fieldFromInstruction32(Val, 16, 4);
+
+  if ((cop & ~0x1) == 0xa)
+    return MCDisassembler::Fail;
+
+  if (Rt == Rt2)
+    S = MCDisassembler::SoftFail;
+
+  Inst.addOperand(MCOperand::CreateImm(cop));
+  Inst.addOperand(MCOperand::CreateImm(opc1));
+  if (!Check(S, DecodeGPRnopcRegisterClass(Inst, Rt, Address, Decoder)))
+    return MCDisassembler::Fail;
+  if (!Check(S, DecodeGPRnopcRegisterClass(Inst, Rt2, Address, Decoder)))
+    return MCDisassembler::Fail;
+  Inst.addOperand(MCOperand::CreateImm(CRm));
+
+  return S;
+}
+
diff --git a/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp b/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp
index b3eeafe0831..cbd81c11a45 100644
--- a/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp
+++ b/lib/Target/ARM/InstPrinter/ARMInstPrinter.cpp
@@ -209,12 +209,12 @@ void ARMInstPrinter::printOperand(const MCInst *MI, unsigned OpNo,
   } else {
     assert(Op.isExpr() && "unknown operand kind in printOperand");
     // If a symbolic branch target was added as a constant expression then print
-    // that address in hex.
+    // that address in hex. And only print 32 unsigned bits for the address.
     const MCConstantExpr *BranchTarget = dyn_cast(Op.getExpr());
     int64_t Address;
     if (BranchTarget && BranchTarget->EvaluateAsAbsolute(Address)) {
       O << "0x";
-      O.write_hex(Address);
+      O.write_hex((uint32_t)Address);
     }
     else {
       // Otherwise, just print the expression.
diff --git a/lib/Target/CellSPU/SPUCallingConv.td b/lib/Target/CellSPU/SPUCallingConv.td
index 9f9692bf67f..9bc6be79860 100644
--- a/lib/Target/CellSPU/SPUCallingConv.td
+++ b/lib/Target/CellSPU/SPUCallingConv.td
@@ -11,10 +11,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-/// CCIfSubtarget - Match if the current subtarget has a feature F.
-class CCIfSubtarget
- : CCIf().", F), A>;
-
 //===----------------------------------------------------------------------===//
 // Return Value Calling Convention
 //===----------------------------------------------------------------------===//
diff --git a/lib/Target/CellSPU/SPUTargetMachine.cpp b/lib/Target/CellSPU/SPUTargetMachine.cpp
index 21f6b25bf25..3b90261fe69 100644
--- a/lib/Target/CellSPU/SPUTargetMachine.cpp
+++ b/lib/Target/CellSPU/SPUTargetMachine.cpp
@@ -72,7 +72,7 @@ TargetPassConfig *SPUTargetMachine::createPassConfig(PassManagerBase &PM) {
 
 bool SPUPassConfig::addInstSelector() {
   // Install an instruction selector.
-  PM.add(createSPUISelDag(getSPUTargetMachine()));
+  PM->add(createSPUISelDag(getSPUTargetMachine()));
   return false;
 }
 
@@ -85,9 +85,9 @@ bool SPUPassConfig::addPreEmitPass() {
     (BuilderFunc)(intptr_t)sys::DynamicLibrary::SearchForAddressOfSymbol(
           "createTCESchedulerPass");
   if (schedulerCreator != NULL)
-      PM.add(schedulerCreator("cellspu"));
+      PM->add(schedulerCreator("cellspu"));
 
   //align instructions with nops/lnops for dual issue
-  PM.add(createSPUNopFillerPass(getSPUTargetMachine()));
+  PM->add(createSPUNopFillerPass(getSPUTargetMachine()));
   return true;
 }
diff --git a/lib/Target/Hexagon/CMakeLists.txt b/lib/Target/Hexagon/CMakeLists.txt
index 8a49cd8105a..af9e8136bf1 100644
--- a/lib/Target/Hexagon/CMakeLists.txt
+++ b/lib/Target/Hexagon/CMakeLists.txt
@@ -11,15 +11,15 @@ add_public_tablegen_target(HexagonCommonTableGen)
 
 add_llvm_target(HexagonCodeGen
   HexagonAsmPrinter.cpp
-  HexagonCFGOptimizer.cpp
   HexagonCallingConvLower.cpp
+  HexagonCFGOptimizer.cpp
   HexagonExpandPredSpillCode.cpp
   HexagonFrameLowering.cpp
   HexagonHardwareLoops.cpp
+  HexagonMCInstLower.cpp
+  HexagonInstrInfo.cpp
   HexagonISelDAGToDAG.cpp
   HexagonISelLowering.cpp
-  HexagonInstrInfo.cpp
-  HexagonMCInstLower.cpp
   HexagonPeephole.cpp
   HexagonRegisterInfo.cpp
   HexagonRemoveSZExtArgs.cpp
@@ -28,7 +28,6 @@ add_llvm_target(HexagonCodeGen
   HexagonSubtarget.cpp
   HexagonTargetMachine.cpp
   HexagonTargetObjectFile.cpp
-  HexagonVLIWPacketizer.cpp
 )
 
 add_subdirectory(TargetInfo)
diff --git a/lib/Target/Hexagon/Hexagon.h b/lib/Target/Hexagon/Hexagon.h
index 43858b9624f..08083233367 100644
--- a/lib/Target/Hexagon/Hexagon.h
+++ b/lib/Target/Hexagon/Hexagon.h
@@ -40,7 +40,6 @@ namespace llvm {
   FunctionPass *createHexagonHardwareLoops();
   FunctionPass *createHexagonPeephole();
   FunctionPass *createHexagonFixupHwLoops();
-  FunctionPass *createHexagonPacketizer();
 
 /* TODO: object output.
   MCCodeEmitter *createHexagonMCCodeEmitter(const Target &,
diff --git a/lib/Target/Hexagon/HexagonAsmPrinter.cpp b/lib/Target/Hexagon/HexagonAsmPrinter.cpp
index 2cc8b814a02..39bf45d2d77 100644
--- a/lib/Target/Hexagon/HexagonAsmPrinter.cpp
+++ b/lib/Target/Hexagon/HexagonAsmPrinter.cpp
@@ -13,11 +13,11 @@
 //
 //===----------------------------------------------------------------------===//
 
+
 #define DEBUG_TYPE "asm-printer"
 #include "Hexagon.h"
 #include "HexagonAsmPrinter.h"
 #include "HexagonMachineFunctionInfo.h"
-#include "HexagonMCInst.h"
 #include "HexagonTargetMachine.h"
 #include "HexagonSubtarget.h"
 #include "InstPrinter/HexagonInstPrinter.h"
@@ -54,7 +54,6 @@
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringExtras.h"
-#include 
 
 using namespace llvm;
 
@@ -78,7 +77,8 @@ void HexagonAsmPrinter::printOperand(const MachineInstr *MI, unsigned OpNo,
   const MachineOperand &MO = MI->getOperand(OpNo);
 
   switch (MO.getType()) {
-  default: llvm_unreachable("");
+  default:
+    assert(0 && "");
   case MachineOperand::MO_Register:
     O << HexagonInstPrinter::getRegisterName(MO.getReg());
     return;
@@ -196,45 +196,10 @@ void HexagonAsmPrinter::printPredicateOperand(const MachineInstr *MI,
 /// the current output stream.
 ///
 void HexagonAsmPrinter::EmitInstruction(const MachineInstr *MI) {
-  if (MI->isBundle()) {
-    std::vector BundleMIs;
+  MCInst MCI;
 
-    const MachineBasicBlock *MBB = MI->getParent();
-    MachineBasicBlock::const_instr_iterator MII = MI;
-    ++MII;
-    unsigned int IgnoreCount = 0;
-    while (MII != MBB->end() && MII->isInsideBundle()) {
-      const MachineInstr *MInst = MII;
-      if (MInst->getOpcode() == TargetOpcode::DBG_VALUE ||
-          MInst->getOpcode() == TargetOpcode::IMPLICIT_DEF) {
-          IgnoreCount++;
-          ++MII;
-          continue;
-      }
-      //BundleMIs.push_back(&*MII);
-      BundleMIs.push_back(MInst);
-      ++MII;
-    }
-    unsigned Size = BundleMIs.size();
-    assert((Size+IgnoreCount) == MI->getBundleSize() && "Corrupt Bundle!");
-    for (unsigned Index = 0; Index < Size; Index++) {
-      HexagonMCInst MCI;
-      MCI.setStartPacket(Index == 0);
-      MCI.setEndPacket(Index == (Size-1));
-
-      HexagonLowerToMC(BundleMIs[Index], MCI, *this);
-      OutStreamer.EmitInstruction(MCI);
-    }
-  }
-  else {
-    HexagonMCInst MCI;
-    if (MI->getOpcode() == Hexagon::ENDLOOP0) {
-      MCI.setStartPacket(true);
-      MCI.setEndPacket(true);
-    }
-    HexagonLowerToMC(MI, MCI, *this);
-    OutStreamer.EmitInstruction(MCI);
-  }
+  HexagonLowerToMC(MI, MCI, *this);
+  OutStreamer.EmitInstruction(MCI);
 
   return;
 }
@@ -277,17 +242,17 @@ void HexagonAsmPrinter::printJumpTable(const MachineInstr *MI, int OpNo,
                                        raw_ostream &O) {
   const MachineOperand &MO = MI->getOperand(OpNo);
   assert( (MO.getType() == MachineOperand::MO_JumpTableIndex) &&
-    "Expecting jump table index");
+	  "Expecting jump table index");
 
   // Hexagon_TODO: Do we need name mangling?
   O << *GetJTISymbol(MO.getIndex());
 }
 
 void HexagonAsmPrinter::printConstantPool(const MachineInstr *MI, int OpNo,
-                                       raw_ostream &O) {
+                                          raw_ostream &O) {
   const MachineOperand &MO = MI->getOperand(OpNo);
   assert( (MO.getType() == MachineOperand::MO_ConstantPoolIndex) &&
-   "Expecting constant pool index");
+          "Expecting constant pool index");
 
   // Hexagon_TODO: Do we need name mangling?
   O << *GetCPISymbol(MO.getIndex());
diff --git a/lib/Target/Hexagon/HexagonISelLowering.cpp b/lib/Target/Hexagon/HexagonISelLowering.cpp
index d6da0d0911b..8c4350d1c50 100644
--- a/lib/Target/Hexagon/HexagonISelLowering.cpp
+++ b/lib/Target/Hexagon/HexagonISelLowering.cpp
@@ -32,9 +32,11 @@
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/CodeGen/SelectionDAGISel.h"
 #include "llvm/CodeGen/ValueTypes.h"
+#include "llvm/Support/CommandLine.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
-#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/raw_ostream.h"
+
 using namespace llvm;
 
 const unsigned Hexagon_MAX_RET_SIZE = 64;
diff --git a/lib/Target/Hexagon/HexagonInstrFormats.td b/lib/Target/Hexagon/HexagonInstrFormats.td
index 48f0f01bb4c..c9f16fb538a 100644
--- a/lib/Target/Hexagon/HexagonInstrFormats.td
+++ b/lib/Target/Hexagon/HexagonInstrFormats.td
@@ -13,26 +13,13 @@
 //                    *** Must match HexagonBaseInfo.h ***
 //===----------------------------------------------------------------------===//
 
-class Type t> {
-  bits<5> Value = t;
-}
-def TypePSEUDO : Type<0>;
-def TypeALU32  : Type<1>;
-def TypeCR     : Type<2>;
-def TypeJR     : Type<3>;
-def TypeJ      : Type<4>;
-def TypeLD     : Type<5>;
-def TypeST     : Type<6>;
-def TypeSYSTEM : Type<7>;
-def TypeXTYPE  : Type<8>;
-def TypeMARKER : Type<31>;
 
 //===----------------------------------------------------------------------===//
 //                         Intruction Class Declaration +
 //===----------------------------------------------------------------------===//
 
 class InstHexagon pattern,
-                  string cstr, InstrItinClass itin, Type type> : Instruction {
+                  string cstr, InstrItinClass itin> : Instruction {
   field bits<32> Inst;
 
   let Namespace = "Hexagon";
@@ -44,15 +31,11 @@ class InstHexagon pattern,
   let Constraints = cstr;
   let Itinerary   = itin;
 
-  // *** Must match HexagonBaseInfo.h ***
-  Type HexagonType = type;
-  let TSFlags{4-0} = HexagonType.Value;
-  bits<1> isHexagonSolo = 0;
-  let TSFlags{5} = isHexagonSolo;
+  // *** The code below must match HexagonBaseInfo.h ***
 
   // Predicated instructions.
   bits<1> isPredicated = 0;
-  let TSFlags{6} = isPredicated;
+  let TSFlags{1} = isPredicated;
 
   // *** The code above must match HexagonBaseInfo.h ***
 }
@@ -64,40 +47,28 @@ class InstHexagon pattern,
 // LD Instruction Class in V2/V3/V4.
 // Definition of the instruction class NOT CHANGED.
 class LDInst pattern>
-  : InstHexagon {
+  : InstHexagon {
   bits<5> rd;
   bits<5> rs;
   bits<13> imm13;
-  let mayLoad = 1;
 }
 
 // LD Instruction Class in V2/V3/V4.
 // Definition of the instruction class NOT CHANGED.
 class LDInstPost pattern,
                  string cstr>
-  : InstHexagon {
+  : InstHexagon {
   bits<5> rd;
   bits<5> rs;
   bits<5> rt;
   bits<13> imm13;
-  let mayLoad = 1;
 }
 
 // ST Instruction Class in V2/V3 can take SLOT0 only.
 // ST Instruction Class in V4    can take SLOT0 & SLOT1.
 // Definition of the instruction class CHANGED from V2/V3 to V4.
 class STInst pattern>
-  : InstHexagon {
-  bits<5> rd;
-  bits<5> rs;
-  bits<13> imm13;
-  let mayStore = 1;
-}
-
-// SYSTEM Instruction Class in V4 can take SLOT0 only
-// In V2/V3 we used ST for this but in v4 ST can take SLOT0 or SLOT1.
-class SYSInst pattern>
-  : InstHexagon {
+  : InstHexagon {
   bits<5> rd;
   bits<5> rs;
   bits<13> imm13;
@@ -108,18 +79,17 @@ class SYSInst pattern>
 // Definition of the instruction class CHANGED from V2/V3 to V4.
 class STInstPost pattern,
                  string cstr>
-  : InstHexagon {
+  : InstHexagon {
   bits<5> rd;
   bits<5> rs;
   bits<5> rt;
   bits<13> imm13;
-  let mayStore = 1;
 }
 
 // ALU32 Instruction Class in V2/V3/V4.
 // Definition of the instruction class NOT CHANGED.
 class ALU32Type pattern>
-   : InstHexagon {
+   : InstHexagon {
   bits<5>  rd;
   bits<5>  rs;
   bits<5>  rt;
@@ -132,17 +102,7 @@ class ALU32Type pattern>
 // Definition of the instruction class NOT CHANGED.
 // Name of the Instruction Class changed from ALU64 to XTYPE from V2/V3 to V4.
 class ALU64Type pattern>
-   : InstHexagon {
-  bits<5>  rd;
-  bits<5>  rs;
-  bits<5>  rt;
-  bits<16> imm16;
-  bits<16> imm16_2;
-}
-
-class ALU64_acc pattern,
-   string cstr>
-   : InstHexagon {
+   : InstHexagon {
   bits<5>  rd;
   bits<5>  rs;
   bits<5>  rt;
@@ -155,7 +115,7 @@ class ALU64_acc pattern,
 // Definition of the instruction class NOT CHANGED.
 // Name of the Instruction Class changed from M to XTYPE from V2/V3 to V4.
 class MInst pattern>
-  : InstHexagon {
+  : InstHexagon {
   bits<5> rd;
   bits<5> rs;
   bits<5> rt;
@@ -166,8 +126,8 @@ class MInst pattern>
 // Definition of the instruction class NOT CHANGED.
 // Name of the Instruction Class changed from M to XTYPE from V2/V3 to V4.
 class MInst_acc pattern,
-		string cstr>
-    : InstHexagon {
+                string cstr>
+    : InstHexagon {
   bits<5> rd;
   bits<5> rs;
   bits<5> rt;
@@ -178,7 +138,9 @@ class MInst_acc pattern,
 // Definition of the instruction class NOT CHANGED.
 // Name of the Instruction Class changed from S to XTYPE from V2/V3 to V4.
 class SInst pattern>
-  : InstHexagon {
+//: InstHexagon {
+  : InstHexagon {
+//  : InstHexagon {
   bits<5> rd;
   bits<5> rs;
   bits<5> rt;
@@ -189,8 +151,8 @@ class SInst pattern>
 // Definition of the instruction class NOT CHANGED.
 // Name of the Instruction Class changed from S to XTYPE from V2/V3 to V4.
 class SInst_acc pattern,
-		string cstr>
-  : InstHexagon {
+                string cstr>
+  : InstHexagon {
 //  : InstHexagon {
 //  : InstHexagon {
   bits<5> rd;
@@ -201,14 +163,14 @@ class SInst_acc pattern,
 // J Instruction Class in V2/V3/V4.
 // Definition of the instruction class NOT CHANGED.
 class JType pattern>
-  : InstHexagon {
+  : InstHexagon {
   bits<16> imm16;
 }
 
 // JR Instruction Class in V2/V3/V4.
 // Definition of the instruction class NOT CHANGED.
 class JRType pattern>
-  : InstHexagon {
+  : InstHexagon {
   bits<5>  rs;
   bits<5>  pu; // Predicate register
 }
@@ -216,22 +178,15 @@ class JRType pattern>
 // CR Instruction Class in V2/V3/V4.
 // Definition of the instruction class NOT CHANGED.
 class CRInst pattern>
-  : InstHexagon {
+  : InstHexagon {
   bits<5> rs;
   bits<10> imm10;
 }
 
-class Marker pattern>
-  : InstHexagon {
-  let isCodeGenOnly = 1;
-  let isPseudo = 1;
-}
 
 class Pseudo pattern>
-  : InstHexagon {
-  let isCodeGenOnly = 1;
-  let isPseudo = 1;
-}
+ : InstHexagon;
+
 
 //===----------------------------------------------------------------------===//
 //                         Intruction Classes Definitions -
@@ -267,11 +222,6 @@ class ALU64_rr pattern>
    : ALU64Type {
 }
 
-class ALU64_ri pattern>
-   : ALU64Type {
-  let rt{0-4} = 0;
-}
-
 // J Type Instructions.
 class JInst pattern>
   : JType {
@@ -287,14 +237,12 @@ class JRInst pattern>
 class STInstPI pattern, string cstr>
   : STInstPost {
   let rt{0-4} = 0;
-  let mayStore = 1;
 }
 
 // Post increment LD Instruction.
 class LDInstPI pattern, string cstr>
   : LDInstPost {
   let rt{0-4} = 0;
-  let mayLoad = 1;
 }
 
 //===----------------------------------------------------------------------===//
diff --git a/lib/Target/Hexagon/HexagonInstrFormatsV4.td b/lib/Target/Hexagon/HexagonInstrFormatsV4.td
index 49741a3d1b2..bd5e4493d7c 100644
--- a/lib/Target/Hexagon/HexagonInstrFormatsV4.td
+++ b/lib/Target/Hexagon/HexagonInstrFormatsV4.td
@@ -11,25 +11,11 @@
 //
 //===----------------------------------------------------------------------===//
 
-//----------------------------------------------------------------------------//
-//                         Hexagon Intruction Flags +
-//
-//                        *** Must match BaseInfo.h ***
-//----------------------------------------------------------------------------//
-
-def TypeMEMOP  : Type<9>;
-def TypeNV     : Type<10>;
-def TypePREFIX : Type<30>;
-
-//----------------------------------------------------------------------------//
-//                         Intruction Classes Definitions +
-//----------------------------------------------------------------------------//
-
 //
 // NV type instructions.
 //
 class NVInst_V4 pattern>
-  : InstHexagon {
+  : InstHexagon {
   bits<5> rd;
   bits<5> rs;
   bits<13> imm13;
@@ -38,7 +24,7 @@ class NVInst_V4 pattern>
 // Definition of Post increment new value store.
 class NVInstPost_V4 pattern,
                     string cstr>
-  : InstHexagon {
+  : InstHexagon {
   bits<5> rd;
   bits<5> rs;
   bits<5> rt;
@@ -53,15 +39,8 @@ class NVInstPI_V4 pattern,
 }
 
 class MEMInst_V4 pattern>
-  : InstHexagon {
+  : InstHexagon {
   bits<5> rd;
   bits<5> rs;
   bits<6> imm6;
 }
-
-class Immext pattern>
-  : InstHexagon {
-  let isCodeGenOnly = 1;
-
-  bits<26> imm26;
-}
diff --git a/lib/Target/Hexagon/HexagonInstrInfo.cpp b/lib/Target/Hexagon/HexagonInstrInfo.cpp
index 8685ec192c7..77b366372cf 100644
--- a/lib/Target/Hexagon/HexagonInstrInfo.cpp
+++ b/lib/Target/Hexagon/HexagonInstrInfo.cpp
@@ -11,10 +11,10 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "Hexagon.h"
 #include "HexagonInstrInfo.h"
 #include "HexagonRegisterInfo.h"
 #include "HexagonSubtarget.h"
-#include "Hexagon.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/CodeGen/DFAPacketizer.h"
@@ -466,865 +466,7 @@ unsigned HexagonInstrInfo::createVR(MachineFunction* MF, MVT VT) const {
   return NewReg;
 }
 
-bool HexagonInstrInfo::isExtendable(const MachineInstr *MI) const {
-  switch(MI->getOpcode()) {
-    // JMP_EQri
-    case Hexagon::JMP_EQriPt_nv_V4:
-    case Hexagon::JMP_EQriPnt_nv_V4:
-    case Hexagon::JMP_EQriNotPt_nv_V4:
-    case Hexagon::JMP_EQriNotPnt_nv_V4:
 
-    // JMP_EQri - with -1
-    case Hexagon::JMP_EQriPtneg_nv_V4:
-    case Hexagon::JMP_EQriPntneg_nv_V4:
-    case Hexagon::JMP_EQriNotPtneg_nv_V4:
-    case Hexagon::JMP_EQriNotPntneg_nv_V4:
-
-    // JMP_EQrr
-    case Hexagon::JMP_EQrrPt_nv_V4:
-    case Hexagon::JMP_EQrrPnt_nv_V4:
-    case Hexagon::JMP_EQrrNotPt_nv_V4:
-    case Hexagon::JMP_EQrrNotPnt_nv_V4:
-
-    // JMP_GTri
-    case Hexagon::JMP_GTriPt_nv_V4:
-    case Hexagon::JMP_GTriPnt_nv_V4:
-    case Hexagon::JMP_GTriNotPt_nv_V4:
-    case Hexagon::JMP_GTriNotPnt_nv_V4:
-
-    // JMP_GTri - with -1
-    case Hexagon::JMP_GTriPtneg_nv_V4:
-    case Hexagon::JMP_GTriPntneg_nv_V4:
-    case Hexagon::JMP_GTriNotPtneg_nv_V4:
-    case Hexagon::JMP_GTriNotPntneg_nv_V4:
-
-    // JMP_GTrr
-    case Hexagon::JMP_GTrrPt_nv_V4:
-    case Hexagon::JMP_GTrrPnt_nv_V4:
-    case Hexagon::JMP_GTrrNotPt_nv_V4:
-    case Hexagon::JMP_GTrrNotPnt_nv_V4:
-
-    // JMP_GTrrdn
-    case Hexagon::JMP_GTrrdnPt_nv_V4:
-    case Hexagon::JMP_GTrrdnPnt_nv_V4:
-    case Hexagon::JMP_GTrrdnNotPt_nv_V4:
-    case Hexagon::JMP_GTrrdnNotPnt_nv_V4:
-
-    // JMP_GTUri
-    case Hexagon::JMP_GTUriPt_nv_V4:
-    case Hexagon::JMP_GTUriPnt_nv_V4:
-    case Hexagon::JMP_GTUriNotPt_nv_V4:
-    case Hexagon::JMP_GTUriNotPnt_nv_V4:
-
-    // JMP_GTUrr
-    case Hexagon::JMP_GTUrrPt_nv_V4:
-    case Hexagon::JMP_GTUrrPnt_nv_V4:
-    case Hexagon::JMP_GTUrrNotPt_nv_V4:
-    case Hexagon::JMP_GTUrrNotPnt_nv_V4:
-
-    // JMP_GTUrrdn
-    case Hexagon::JMP_GTUrrdnPt_nv_V4:
-    case Hexagon::JMP_GTUrrdnPnt_nv_V4:
-    case Hexagon::JMP_GTUrrdnNotPt_nv_V4:
-    case Hexagon::JMP_GTUrrdnNotPnt_nv_V4:
-      return true;
-
-    // TFR_FI
-    case Hexagon::TFR_FI:
-      return true;
-
-
-    default:
-      return false;
-  }
-  return  false;
-}
-
-bool HexagonInstrInfo::isExtended(const MachineInstr *MI) const {
-  switch(MI->getOpcode()) {
-    // JMP_EQri
-    case Hexagon::JMP_EQriPt_ie_nv_V4:
-    case Hexagon::JMP_EQriPnt_ie_nv_V4:
-    case Hexagon::JMP_EQriNotPt_ie_nv_V4:
-    case Hexagon::JMP_EQriNotPnt_ie_nv_V4:
-
-    // JMP_EQri - with -1
-    case Hexagon::JMP_EQriPtneg_ie_nv_V4:
-    case Hexagon::JMP_EQriPntneg_ie_nv_V4:
-    case Hexagon::JMP_EQriNotPtneg_ie_nv_V4:
-    case Hexagon::JMP_EQriNotPntneg_ie_nv_V4:
-
-    // JMP_EQrr
-    case Hexagon::JMP_EQrrPt_ie_nv_V4:
-    case Hexagon::JMP_EQrrPnt_ie_nv_V4:
-    case Hexagon::JMP_EQrrNotPt_ie_nv_V4:
-    case Hexagon::JMP_EQrrNotPnt_ie_nv_V4:
-
-    // JMP_GTri
-    case Hexagon::JMP_GTriPt_ie_nv_V4:
-    case Hexagon::JMP_GTriPnt_ie_nv_V4:
-    case Hexagon::JMP_GTriNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTriNotPnt_ie_nv_V4:
-
-    // JMP_GTri - with -1
-    case Hexagon::JMP_GTriPtneg_ie_nv_V4:
-    case Hexagon::JMP_GTriPntneg_ie_nv_V4:
-    case Hexagon::JMP_GTriNotPtneg_ie_nv_V4:
-    case Hexagon::JMP_GTriNotPntneg_ie_nv_V4:
-
-    // JMP_GTrr
-    case Hexagon::JMP_GTrrPt_ie_nv_V4:
-    case Hexagon::JMP_GTrrPnt_ie_nv_V4:
-    case Hexagon::JMP_GTrrNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTrrNotPnt_ie_nv_V4:
-
-    // JMP_GTrrdn
-    case Hexagon::JMP_GTrrdnPt_ie_nv_V4:
-    case Hexagon::JMP_GTrrdnPnt_ie_nv_V4:
-    case Hexagon::JMP_GTrrdnNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTrrdnNotPnt_ie_nv_V4:
-
-    // JMP_GTUri
-    case Hexagon::JMP_GTUriPt_ie_nv_V4:
-    case Hexagon::JMP_GTUriPnt_ie_nv_V4:
-    case Hexagon::JMP_GTUriNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTUriNotPnt_ie_nv_V4:
-
-    // JMP_GTUrr
-    case Hexagon::JMP_GTUrrPt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrPnt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrNotPnt_ie_nv_V4:
-
-    // JMP_GTUrrdn
-    case Hexagon::JMP_GTUrrdnPt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrdnPnt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrdnNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrdnNotPnt_ie_nv_V4:
-
-    // V4 absolute set addressing.
-    case Hexagon::LDrid_abs_setimm_V4:
-    case Hexagon::LDriw_abs_setimm_V4:
-    case Hexagon::LDrih_abs_setimm_V4:
-    case Hexagon::LDrib_abs_setimm_V4:
-    case Hexagon::LDriuh_abs_setimm_V4:
-    case Hexagon::LDriub_abs_setimm_V4:
-
-    case Hexagon::STrid_abs_setimm_V4:
-    case Hexagon::STrib_abs_setimm_V4:
-    case Hexagon::STrih_abs_setimm_V4:
-    case Hexagon::STriw_abs_setimm_V4:
-
-    // V4 global address load.
-    case Hexagon::LDrid_GP_cPt_V4 :
-    case Hexagon::LDrid_GP_cNotPt_V4 :
-    case Hexagon::LDrid_GP_cdnPt_V4 :
-    case Hexagon::LDrid_GP_cdnNotPt_V4 :
-    case Hexagon::LDrib_GP_cPt_V4 :
-    case Hexagon::LDrib_GP_cNotPt_V4 :
-    case Hexagon::LDrib_GP_cdnPt_V4 :
-    case Hexagon::LDrib_GP_cdnNotPt_V4 :
-    case Hexagon::LDriub_GP_cPt_V4 :
-    case Hexagon::LDriub_GP_cNotPt_V4 :
-    case Hexagon::LDriub_GP_cdnPt_V4 :
-    case Hexagon::LDriub_GP_cdnNotPt_V4 :
-    case Hexagon::LDrih_GP_cPt_V4 :
-    case Hexagon::LDrih_GP_cNotPt_V4 :
-    case Hexagon::LDrih_GP_cdnPt_V4 :
-    case Hexagon::LDrih_GP_cdnNotPt_V4 :
-    case Hexagon::LDriuh_GP_cPt_V4 :
-    case Hexagon::LDriuh_GP_cNotPt_V4 :
-    case Hexagon::LDriuh_GP_cdnPt_V4 :
-    case Hexagon::LDriuh_GP_cdnNotPt_V4 :
-    case Hexagon::LDriw_GP_cPt_V4 :
-    case Hexagon::LDriw_GP_cNotPt_V4 :
-    case Hexagon::LDriw_GP_cdnPt_V4 :
-    case Hexagon::LDriw_GP_cdnNotPt_V4 :
-    case Hexagon::LDd_GP_cPt_V4 :
-    case Hexagon::LDd_GP_cNotPt_V4 :
-    case Hexagon::LDd_GP_cdnPt_V4 :
-    case Hexagon::LDd_GP_cdnNotPt_V4 :
-    case Hexagon::LDb_GP_cPt_V4 :
-    case Hexagon::LDb_GP_cNotPt_V4 :
-    case Hexagon::LDb_GP_cdnPt_V4 :
-    case Hexagon::LDb_GP_cdnNotPt_V4 :
-    case Hexagon::LDub_GP_cPt_V4 :
-    case Hexagon::LDub_GP_cNotPt_V4 :
-    case Hexagon::LDub_GP_cdnPt_V4 :
-    case Hexagon::LDub_GP_cdnNotPt_V4 :
-    case Hexagon::LDh_GP_cPt_V4 :
-    case Hexagon::LDh_GP_cNotPt_V4 :
-    case Hexagon::LDh_GP_cdnPt_V4 :
-    case Hexagon::LDh_GP_cdnNotPt_V4 :
-    case Hexagon::LDuh_GP_cPt_V4 :
-    case Hexagon::LDuh_GP_cNotPt_V4 :
-    case Hexagon::LDuh_GP_cdnPt_V4 :
-    case Hexagon::LDuh_GP_cdnNotPt_V4 :
-    case Hexagon::LDw_GP_cPt_V4 :
-    case Hexagon::LDw_GP_cNotPt_V4 :
-    case Hexagon::LDw_GP_cdnPt_V4 :
-    case Hexagon::LDw_GP_cdnNotPt_V4 :
-
-    // V4 global address store.
-    case Hexagon::STrid_GP_cPt_V4 :
-    case Hexagon::STrid_GP_cNotPt_V4 :
-    case Hexagon::STrid_GP_cdnPt_V4 :
-    case Hexagon::STrid_GP_cdnNotPt_V4 :
-    case Hexagon::STrib_GP_cPt_V4 :
-    case Hexagon::STrib_GP_cNotPt_V4 :
-    case Hexagon::STrib_GP_cdnPt_V4 :
-    case Hexagon::STrib_GP_cdnNotPt_V4 :
-    case Hexagon::STrih_GP_cPt_V4 :
-    case Hexagon::STrih_GP_cNotPt_V4 :
-    case Hexagon::STrih_GP_cdnPt_V4 :
-    case Hexagon::STrih_GP_cdnNotPt_V4 :
-    case Hexagon::STriw_GP_cPt_V4 :
-    case Hexagon::STriw_GP_cNotPt_V4 :
-    case Hexagon::STriw_GP_cdnPt_V4 :
-    case Hexagon::STriw_GP_cdnNotPt_V4 :
-    case Hexagon::STd_GP_cPt_V4 :
-    case Hexagon::STd_GP_cNotPt_V4 :
-    case Hexagon::STd_GP_cdnPt_V4 :
-    case Hexagon::STd_GP_cdnNotPt_V4 :
-    case Hexagon::STb_GP_cPt_V4 :
-    case Hexagon::STb_GP_cNotPt_V4 :
-    case Hexagon::STb_GP_cdnPt_V4 :
-    case Hexagon::STb_GP_cdnNotPt_V4 :
-    case Hexagon::STh_GP_cPt_V4 :
-    case Hexagon::STh_GP_cNotPt_V4 :
-    case Hexagon::STh_GP_cdnPt_V4 :
-    case Hexagon::STh_GP_cdnNotPt_V4 :
-    case Hexagon::STw_GP_cPt_V4 :
-    case Hexagon::STw_GP_cNotPt_V4 :
-    case Hexagon::STw_GP_cdnPt_V4 :
-    case Hexagon::STw_GP_cdnNotPt_V4 :
-
-    // V4 predicated global address new value store.
-    case Hexagon::STrib_GP_cPt_nv_V4 :
-    case Hexagon::STrib_GP_cNotPt_nv_V4 :
-    case Hexagon::STrib_GP_cdnPt_nv_V4 :
-    case Hexagon::STrib_GP_cdnNotPt_nv_V4 :
-    case Hexagon::STrih_GP_cPt_nv_V4 :
-    case Hexagon::STrih_GP_cNotPt_nv_V4 :
-    case Hexagon::STrih_GP_cdnPt_nv_V4 :
-    case Hexagon::STrih_GP_cdnNotPt_nv_V4 :
-    case Hexagon::STriw_GP_cPt_nv_V4 :
-    case Hexagon::STriw_GP_cNotPt_nv_V4 :
-    case Hexagon::STriw_GP_cdnPt_nv_V4 :
-    case Hexagon::STriw_GP_cdnNotPt_nv_V4 :
-    case Hexagon::STb_GP_cPt_nv_V4 :
-    case Hexagon::STb_GP_cNotPt_nv_V4 :
-    case Hexagon::STb_GP_cdnPt_nv_V4 :
-    case Hexagon::STb_GP_cdnNotPt_nv_V4 :
-    case Hexagon::STh_GP_cPt_nv_V4 :
-    case Hexagon::STh_GP_cNotPt_nv_V4 :
-    case Hexagon::STh_GP_cdnPt_nv_V4 :
-    case Hexagon::STh_GP_cdnNotPt_nv_V4 :
-    case Hexagon::STw_GP_cPt_nv_V4 :
-    case Hexagon::STw_GP_cNotPt_nv_V4 :
-    case Hexagon::STw_GP_cdnPt_nv_V4 :
-    case Hexagon::STw_GP_cdnNotPt_nv_V4 :
-
-    // TFR_FI
-    case Hexagon::TFR_FI_immext_V4:
-      return true;
-
-    default:
-      return false;
-  }
-  return  false;
-}
-
-bool HexagonInstrInfo::isNewValueJump(const MachineInstr *MI) const {
-  switch (MI->getOpcode()) {
-    // JMP_EQri
-    case Hexagon::JMP_EQriPt_nv_V4:
-    case Hexagon::JMP_EQriPnt_nv_V4:
-    case Hexagon::JMP_EQriNotPt_nv_V4:
-    case Hexagon::JMP_EQriNotPnt_nv_V4:
-    case Hexagon::JMP_EQriPt_ie_nv_V4:
-    case Hexagon::JMP_EQriPnt_ie_nv_V4:
-    case Hexagon::JMP_EQriNotPt_ie_nv_V4:
-    case Hexagon::JMP_EQriNotPnt_ie_nv_V4:
-
-    // JMP_EQri - with -1
-    case Hexagon::JMP_EQriPtneg_nv_V4:
-    case Hexagon::JMP_EQriPntneg_nv_V4:
-    case Hexagon::JMP_EQriNotPtneg_nv_V4:
-    case Hexagon::JMP_EQriNotPntneg_nv_V4:
-    case Hexagon::JMP_EQriPtneg_ie_nv_V4:
-    case Hexagon::JMP_EQriPntneg_ie_nv_V4:
-    case Hexagon::JMP_EQriNotPtneg_ie_nv_V4:
-    case Hexagon::JMP_EQriNotPntneg_ie_nv_V4:
-
-    // JMP_EQrr
-    case Hexagon::JMP_EQrrPt_nv_V4:
-    case Hexagon::JMP_EQrrPnt_nv_V4:
-    case Hexagon::JMP_EQrrNotPt_nv_V4:
-    case Hexagon::JMP_EQrrNotPnt_nv_V4:
-    case Hexagon::JMP_EQrrPt_ie_nv_V4:
-    case Hexagon::JMP_EQrrPnt_ie_nv_V4:
-    case Hexagon::JMP_EQrrNotPt_ie_nv_V4:
-    case Hexagon::JMP_EQrrNotPnt_ie_nv_V4:
-
-    // JMP_GTri
-    case Hexagon::JMP_GTriPt_nv_V4:
-    case Hexagon::JMP_GTriPnt_nv_V4:
-    case Hexagon::JMP_GTriNotPt_nv_V4:
-    case Hexagon::JMP_GTriNotPnt_nv_V4:
-    case Hexagon::JMP_GTriPt_ie_nv_V4:
-    case Hexagon::JMP_GTriPnt_ie_nv_V4:
-    case Hexagon::JMP_GTriNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTriNotPnt_ie_nv_V4:
-
-    // JMP_GTri - with -1
-    case Hexagon::JMP_GTriPtneg_nv_V4:
-    case Hexagon::JMP_GTriPntneg_nv_V4:
-    case Hexagon::JMP_GTriNotPtneg_nv_V4:
-    case Hexagon::JMP_GTriNotPntneg_nv_V4:
-    case Hexagon::JMP_GTriPtneg_ie_nv_V4:
-    case Hexagon::JMP_GTriPntneg_ie_nv_V4:
-    case Hexagon::JMP_GTriNotPtneg_ie_nv_V4:
-    case Hexagon::JMP_GTriNotPntneg_ie_nv_V4:
-
-    // JMP_GTrr
-    case Hexagon::JMP_GTrrPt_nv_V4:
-    case Hexagon::JMP_GTrrPnt_nv_V4:
-    case Hexagon::JMP_GTrrNotPt_nv_V4:
-    case Hexagon::JMP_GTrrNotPnt_nv_V4:
-    case Hexagon::JMP_GTrrPt_ie_nv_V4:
-    case Hexagon::JMP_GTrrPnt_ie_nv_V4:
-    case Hexagon::JMP_GTrrNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTrrNotPnt_ie_nv_V4:
-
-    // JMP_GTrrdn
-    case Hexagon::JMP_GTrrdnPt_nv_V4:
-    case Hexagon::JMP_GTrrdnPnt_nv_V4:
-    case Hexagon::JMP_GTrrdnNotPt_nv_V4:
-    case Hexagon::JMP_GTrrdnNotPnt_nv_V4:
-    case Hexagon::JMP_GTrrdnPt_ie_nv_V4:
-    case Hexagon::JMP_GTrrdnPnt_ie_nv_V4:
-    case Hexagon::JMP_GTrrdnNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTrrdnNotPnt_ie_nv_V4:
-
-    // JMP_GTUri
-    case Hexagon::JMP_GTUriPt_nv_V4:
-    case Hexagon::JMP_GTUriPnt_nv_V4:
-    case Hexagon::JMP_GTUriNotPt_nv_V4:
-    case Hexagon::JMP_GTUriNotPnt_nv_V4:
-    case Hexagon::JMP_GTUriPt_ie_nv_V4:
-    case Hexagon::JMP_GTUriPnt_ie_nv_V4:
-    case Hexagon::JMP_GTUriNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTUriNotPnt_ie_nv_V4:
-
-    // JMP_GTUrr
-    case Hexagon::JMP_GTUrrPt_nv_V4:
-    case Hexagon::JMP_GTUrrPnt_nv_V4:
-    case Hexagon::JMP_GTUrrNotPt_nv_V4:
-    case Hexagon::JMP_GTUrrNotPnt_nv_V4:
-    case Hexagon::JMP_GTUrrPt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrPnt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrNotPnt_ie_nv_V4:
-
-    // JMP_GTUrrdn
-    case Hexagon::JMP_GTUrrdnPt_nv_V4:
-    case Hexagon::JMP_GTUrrdnPnt_nv_V4:
-    case Hexagon::JMP_GTUrrdnNotPt_nv_V4:
-    case Hexagon::JMP_GTUrrdnNotPnt_nv_V4:
-    case Hexagon::JMP_GTUrrdnPt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrdnPnt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrdnNotPt_ie_nv_V4:
-    case Hexagon::JMP_GTUrrdnNotPnt_ie_nv_V4:
-      return true;
-
-    default:
-      return false;
-  }
-  return false;
-}
-
-unsigned HexagonInstrInfo::getImmExtForm(const MachineInstr* MI) const {
-  switch(MI->getOpcode()) {
-  default: llvm_unreachable("Unknown type of instruction");
-
-  // JMP_EQri
-  case Hexagon::JMP_EQriPt_nv_V4:
-    return Hexagon::JMP_EQriPt_ie_nv_V4;
-  case Hexagon::JMP_EQriNotPt_nv_V4:
-    return Hexagon::JMP_EQriNotPt_ie_nv_V4;
-  case Hexagon::JMP_EQriPnt_nv_V4:
-    return Hexagon::JMP_EQriPnt_ie_nv_V4;
-  case Hexagon::JMP_EQriNotPnt_nv_V4:
-    return Hexagon::JMP_EQriNotPnt_ie_nv_V4;
-
-  // JMP_EQri -- with -1
-  case Hexagon::JMP_EQriPtneg_nv_V4:
-    return Hexagon::JMP_EQriPtneg_ie_nv_V4;
-  case Hexagon::JMP_EQriNotPtneg_nv_V4:
-    return Hexagon::JMP_EQriNotPtneg_ie_nv_V4;
-  case Hexagon::JMP_EQriPntneg_nv_V4:
-    return Hexagon::JMP_EQriPntneg_ie_nv_V4;
-  case Hexagon::JMP_EQriNotPntneg_nv_V4:
-    return Hexagon::JMP_EQriNotPntneg_ie_nv_V4;
-
-  // JMP_EQrr
-  case Hexagon::JMP_EQrrPt_nv_V4:
-    return Hexagon::JMP_EQrrPt_ie_nv_V4;
-  case Hexagon::JMP_EQrrNotPt_nv_V4:
-    return Hexagon::JMP_EQrrNotPt_ie_nv_V4;
-  case Hexagon::JMP_EQrrPnt_nv_V4:
-    return Hexagon::JMP_EQrrPnt_ie_nv_V4;
-  case Hexagon::JMP_EQrrNotPnt_nv_V4:
-    return Hexagon::JMP_EQrrNotPnt_ie_nv_V4;
-
-  // JMP_GTri
-  case Hexagon::JMP_GTriPt_nv_V4:
-    return Hexagon::JMP_GTriPt_ie_nv_V4;
-  case Hexagon::JMP_GTriNotPt_nv_V4:
-    return Hexagon::JMP_GTriNotPt_ie_nv_V4;
-  case Hexagon::JMP_GTriPnt_nv_V4:
-    return Hexagon::JMP_GTriPnt_ie_nv_V4;
-  case Hexagon::JMP_GTriNotPnt_nv_V4:
-    return Hexagon::JMP_GTriNotPnt_ie_nv_V4;
-
-  // JMP_GTri -- with -1
-  case Hexagon::JMP_GTriPtneg_nv_V4:
-    return Hexagon::JMP_GTriPtneg_ie_nv_V4;
-  case Hexagon::JMP_GTriNotPtneg_nv_V4:
-    return Hexagon::JMP_GTriNotPtneg_ie_nv_V4;
-  case Hexagon::JMP_GTriPntneg_nv_V4:
-    return Hexagon::JMP_GTriPntneg_ie_nv_V4;
-  case Hexagon::JMP_GTriNotPntneg_nv_V4:
-    return Hexagon::JMP_GTriNotPntneg_ie_nv_V4;
-
-  // JMP_GTrr
-  case Hexagon::JMP_GTrrPt_nv_V4:
-    return Hexagon::JMP_GTrrPt_ie_nv_V4;
-  case Hexagon::JMP_GTrrNotPt_nv_V4:
-    return Hexagon::JMP_GTrrNotPt_ie_nv_V4;
-  case Hexagon::JMP_GTrrPnt_nv_V4:
-    return Hexagon::JMP_GTrrPnt_ie_nv_V4;
-  case Hexagon::JMP_GTrrNotPnt_nv_V4:
-    return Hexagon::JMP_GTrrNotPnt_ie_nv_V4;
-
-  // JMP_GTrrdn
-  case Hexagon::JMP_GTrrdnPt_nv_V4:
-    return Hexagon::JMP_GTrrdnPt_ie_nv_V4;
-  case Hexagon::JMP_GTrrdnNotPt_nv_V4:
-    return Hexagon::JMP_GTrrdnNotPt_ie_nv_V4;
-  case Hexagon::JMP_GTrrdnPnt_nv_V4:
-    return Hexagon::JMP_GTrrdnPnt_ie_nv_V4;
-  case Hexagon::JMP_GTrrdnNotPnt_nv_V4:
-    return Hexagon::JMP_GTrrdnNotPnt_ie_nv_V4;
-
-  // JMP_GTUri
-  case Hexagon::JMP_GTUriPt_nv_V4:
-    return Hexagon::JMP_GTUriPt_ie_nv_V4;
-  case Hexagon::JMP_GTUriNotPt_nv_V4:
-    return Hexagon::JMP_GTUriNotPt_ie_nv_V4;
-  case Hexagon::JMP_GTUriPnt_nv_V4:
-    return Hexagon::JMP_GTUriPnt_ie_nv_V4;
-  case Hexagon::JMP_GTUriNotPnt_nv_V4:
-    return Hexagon::JMP_GTUriNotPnt_ie_nv_V4;
-
-  // JMP_GTUrr
-  case Hexagon::JMP_GTUrrPt_nv_V4:
-    return Hexagon::JMP_GTUrrPt_ie_nv_V4;
-  case Hexagon::JMP_GTUrrNotPt_nv_V4:
-    return Hexagon::JMP_GTUrrNotPt_ie_nv_V4;
-  case Hexagon::JMP_GTUrrPnt_nv_V4:
-    return Hexagon::JMP_GTUrrPnt_ie_nv_V4;
-  case Hexagon::JMP_GTUrrNotPnt_nv_V4:
-    return Hexagon::JMP_GTUrrNotPnt_ie_nv_V4;
-
-  // JMP_GTUrrdn
-  case Hexagon::JMP_GTUrrdnPt_nv_V4:
-    return Hexagon::JMP_GTUrrdnPt_ie_nv_V4;
-  case Hexagon::JMP_GTUrrdnNotPt_nv_V4:
-    return Hexagon::JMP_GTUrrdnNotPt_ie_nv_V4;
-  case Hexagon::JMP_GTUrrdnPnt_nv_V4:
-    return Hexagon::JMP_GTUrrdnPnt_ie_nv_V4;
-  case Hexagon::JMP_GTUrrdnNotPnt_nv_V4:
-    return Hexagon::JMP_GTUrrdnNotPnt_ie_nv_V4;
-
-  case Hexagon::TFR_FI:
-      return Hexagon::TFR_FI_immext_V4;
-
-  case Hexagon::MEMw_ADDSUBi_indexed_MEM_V4 :
-  case Hexagon::MEMw_ADDi_indexed_MEM_V4 :
-  case Hexagon::MEMw_SUBi_indexed_MEM_V4 :
-  case Hexagon::MEMw_ADDr_indexed_MEM_V4 :
-  case Hexagon::MEMw_SUBr_indexed_MEM_V4 :
-  case Hexagon::MEMw_ANDr_indexed_MEM_V4 :
-  case Hexagon::MEMw_ORr_indexed_MEM_V4 :
-  case Hexagon::MEMw_ADDSUBi_MEM_V4 :
-  case Hexagon::MEMw_ADDi_MEM_V4 :
-  case Hexagon::MEMw_SUBi_MEM_V4 :
-  case Hexagon::MEMw_ADDr_MEM_V4 :
-  case Hexagon::MEMw_SUBr_MEM_V4 :
-  case Hexagon::MEMw_ANDr_MEM_V4 :
-  case Hexagon::MEMw_ORr_MEM_V4 :
-  case Hexagon::MEMh_ADDSUBi_indexed_MEM_V4 :
-  case Hexagon::MEMh_ADDi_indexed_MEM_V4 :
-  case Hexagon::MEMh_SUBi_indexed_MEM_V4 :
-  case Hexagon::MEMh_ADDr_indexed_MEM_V4 :
-  case Hexagon::MEMh_SUBr_indexed_MEM_V4 :
-  case Hexagon::MEMh_ANDr_indexed_MEM_V4 :
-  case Hexagon::MEMh_ORr_indexed_MEM_V4 :
-  case Hexagon::MEMh_ADDSUBi_MEM_V4 :
-  case Hexagon::MEMh_ADDi_MEM_V4 :
-  case Hexagon::MEMh_SUBi_MEM_V4 :
-  case Hexagon::MEMh_ADDr_MEM_V4 :
-  case Hexagon::MEMh_SUBr_MEM_V4 :
-  case Hexagon::MEMh_ANDr_MEM_V4 :
-  case Hexagon::MEMh_ORr_MEM_V4 :
-  case Hexagon::MEMb_ADDSUBi_indexed_MEM_V4 :
-  case Hexagon::MEMb_ADDi_indexed_MEM_V4 :
-  case Hexagon::MEMb_SUBi_indexed_MEM_V4 :
-  case Hexagon::MEMb_ADDr_indexed_MEM_V4 :
-  case Hexagon::MEMb_SUBr_indexed_MEM_V4 :
-  case Hexagon::MEMb_ANDr_indexed_MEM_V4 :
-  case Hexagon::MEMb_ORr_indexed_MEM_V4 :
-  case Hexagon::MEMb_ADDSUBi_MEM_V4 :
-  case Hexagon::MEMb_ADDi_MEM_V4 :
-  case Hexagon::MEMb_SUBi_MEM_V4 :
-  case Hexagon::MEMb_ADDr_MEM_V4 :
-  case Hexagon::MEMb_SUBr_MEM_V4 :
-  case Hexagon::MEMb_ANDr_MEM_V4 :
-  case Hexagon::MEMb_ORr_MEM_V4 :
-    llvm_unreachable("Needs implementing");
-  }
-}
-
-unsigned HexagonInstrInfo::getNormalBranchForm(const MachineInstr* MI) const {
-  switch(MI->getOpcode()) {
-  default: llvm_unreachable("Unknown type of jump instruction");
-
-  // JMP_EQri
-  case Hexagon::JMP_EQriPt_ie_nv_V4:
-    return Hexagon::JMP_EQriPt_nv_V4;
-  case Hexagon::JMP_EQriNotPt_ie_nv_V4:
-    return Hexagon::JMP_EQriNotPt_nv_V4;
-  case Hexagon::JMP_EQriPnt_ie_nv_V4:
-    return Hexagon::JMP_EQriPnt_nv_V4;
-  case Hexagon::JMP_EQriNotPnt_ie_nv_V4:
-    return Hexagon::JMP_EQriNotPnt_nv_V4;
-
-  // JMP_EQri -- with -1
-  case Hexagon::JMP_EQriPtneg_ie_nv_V4:
-    return Hexagon::JMP_EQriPtneg_nv_V4;
-  case Hexagon::JMP_EQriNotPtneg_ie_nv_V4:
-    return Hexagon::JMP_EQriNotPtneg_nv_V4;
-  case Hexagon::JMP_EQriPntneg_ie_nv_V4:
-    return Hexagon::JMP_EQriPntneg_nv_V4;
-  case Hexagon::JMP_EQriNotPntneg_ie_nv_V4:
-    return Hexagon::JMP_EQriNotPntneg_nv_V4;
-
-  // JMP_EQrr
-  case Hexagon::JMP_EQrrPt_ie_nv_V4:
-    return Hexagon::JMP_EQrrPt_nv_V4;
-  case Hexagon::JMP_EQrrNotPt_ie_nv_V4:
-    return Hexagon::JMP_EQrrNotPt_nv_V4;
-  case Hexagon::JMP_EQrrPnt_ie_nv_V4:
-    return Hexagon::JMP_EQrrPnt_nv_V4;
-  case Hexagon::JMP_EQrrNotPnt_ie_nv_V4:
-    return Hexagon::JMP_EQrrNotPnt_nv_V4;
-
-  // JMP_GTri
-  case Hexagon::JMP_GTriPt_ie_nv_V4:
-    return Hexagon::JMP_GTriPt_nv_V4;
-  case Hexagon::JMP_GTriNotPt_ie_nv_V4:
-    return Hexagon::JMP_GTriNotPt_nv_V4;
-  case Hexagon::JMP_GTriPnt_ie_nv_V4:
-    return Hexagon::JMP_GTriPnt_nv_V4;
-  case Hexagon::JMP_GTriNotPnt_ie_nv_V4:
-    return Hexagon::JMP_GTriNotPnt_nv_V4;
-
-  // JMP_GTri -- with -1
-  case Hexagon::JMP_GTriPtneg_ie_nv_V4:
-    return Hexagon::JMP_GTriPtneg_nv_V4;
-  case Hexagon::JMP_GTriNotPtneg_ie_nv_V4:
-    return Hexagon::JMP_GTriNotPtneg_nv_V4;
-  case Hexagon::JMP_GTriPntneg_ie_nv_V4:
-    return Hexagon::JMP_GTriPntneg_nv_V4;
-  case Hexagon::JMP_GTriNotPntneg_ie_nv_V4:
-    return Hexagon::JMP_GTriNotPntneg_nv_V4;
-
-  // JMP_GTrr
-  case Hexagon::JMP_GTrrPt_ie_nv_V4:
-    return Hexagon::JMP_GTrrPt_nv_V4;
-  case Hexagon::JMP_GTrrNotPt_ie_nv_V4:
-    return Hexagon::JMP_GTrrNotPt_nv_V4;
-  case Hexagon::JMP_GTrrPnt_ie_nv_V4:
-    return Hexagon::JMP_GTrrPnt_nv_V4;
-  case Hexagon::JMP_GTrrNotPnt_ie_nv_V4:
-    return Hexagon::JMP_GTrrNotPnt_nv_V4;
-
-  // JMP_GTrrdn
-  case Hexagon::JMP_GTrrdnPt_ie_nv_V4:
-    return Hexagon::JMP_GTrrdnPt_nv_V4;
-  case Hexagon::JMP_GTrrdnNotPt_ie_nv_V4:
-    return Hexagon::JMP_GTrrdnNotPt_nv_V4;
-  case Hexagon::JMP_GTrrdnPnt_ie_nv_V4:
-    return Hexagon::JMP_GTrrdnPnt_nv_V4;
-  case Hexagon::JMP_GTrrdnNotPnt_ie_nv_V4:
-    return Hexagon::JMP_GTrrdnNotPnt_nv_V4;
-
-  // JMP_GTUri
-  case Hexagon::JMP_GTUriPt_ie_nv_V4:
-    return Hexagon::JMP_GTUriPt_nv_V4;
-  case Hexagon::JMP_GTUriNotPt_ie_nv_V4:
-    return Hexagon::JMP_GTUriNotPt_nv_V4;
-  case Hexagon::JMP_GTUriPnt_ie_nv_V4:
-    return Hexagon::JMP_GTUriPnt_nv_V4;
-  case Hexagon::JMP_GTUriNotPnt_ie_nv_V4:
-    return Hexagon::JMP_GTUriNotPnt_nv_V4;
-
-  // JMP_GTUrr
-  case Hexagon::JMP_GTUrrPt_ie_nv_V4:
-    return Hexagon::JMP_GTUrrPt_nv_V4;
-  case Hexagon::JMP_GTUrrNotPt_ie_nv_V4:
-    return Hexagon::JMP_GTUrrNotPt_nv_V4;
-  case Hexagon::JMP_GTUrrPnt_ie_nv_V4:
-    return Hexagon::JMP_GTUrrPnt_nv_V4;
-  case Hexagon::JMP_GTUrrNotPnt_ie_nv_V4:
-    return Hexagon::JMP_GTUrrNotPnt_nv_V4;
-
-  // JMP_GTUrrdn
-  case Hexagon::JMP_GTUrrdnPt_ie_nv_V4:
-    return Hexagon::JMP_GTUrrdnPt_nv_V4;
-  case Hexagon::JMP_GTUrrdnNotPt_ie_nv_V4:
-    return Hexagon::JMP_GTUrrdnNotPt_nv_V4;
-  case Hexagon::JMP_GTUrrdnPnt_ie_nv_V4:
-    return Hexagon::JMP_GTUrrdnPnt_nv_V4;
-  case Hexagon::JMP_GTUrrdnNotPnt_ie_nv_V4:
-    return Hexagon::JMP_GTUrrdnNotPnt_nv_V4;
-  }
-}
-
-
-bool HexagonInstrInfo::isNewValueStore(const MachineInstr *MI) const {
-  switch (MI->getOpcode()) {
-
-    // Store Byte
-    case Hexagon::STrib_nv_V4:
-    case Hexagon::STrib_indexed_nv_V4:
-    case Hexagon::STrib_indexed_shl_nv_V4:
-    case Hexagon::STrib_shl_nv_V4:
-    case Hexagon::STrib_GP_nv_V4:
-    case Hexagon::STb_GP_nv_V4:
-    case Hexagon::POST_STbri_nv_V4:
-    case Hexagon::STrib_cPt_nv_V4:
-    case Hexagon::STrib_cdnPt_nv_V4:
-    case Hexagon::STrib_cNotPt_nv_V4:
-    case Hexagon::STrib_cdnNotPt_nv_V4:
-    case Hexagon::STrib_indexed_cPt_nv_V4:
-    case Hexagon::STrib_indexed_cdnPt_nv_V4:
-    case Hexagon::STrib_indexed_cNotPt_nv_V4:
-    case Hexagon::STrib_indexed_cdnNotPt_nv_V4:
-    case Hexagon::STrib_indexed_shl_cPt_nv_V4:
-    case Hexagon::STrib_indexed_shl_cdnPt_nv_V4:
-    case Hexagon::STrib_indexed_shl_cNotPt_nv_V4:
-    case Hexagon::STrib_indexed_shl_cdnNotPt_nv_V4:
-    case Hexagon::POST_STbri_cPt_nv_V4:
-    case Hexagon::POST_STbri_cdnPt_nv_V4:
-    case Hexagon::POST_STbri_cNotPt_nv_V4:
-    case Hexagon::POST_STbri_cdnNotPt_nv_V4:
-    case Hexagon::STb_GP_cPt_nv_V4:
-    case Hexagon::STb_GP_cNotPt_nv_V4:
-    case Hexagon::STb_GP_cdnPt_nv_V4:
-    case Hexagon::STb_GP_cdnNotPt_nv_V4:
-    case Hexagon::STrib_GP_cPt_nv_V4:
-    case Hexagon::STrib_GP_cNotPt_nv_V4:
-    case Hexagon::STrib_GP_cdnPt_nv_V4:
-    case Hexagon::STrib_GP_cdnNotPt_nv_V4:
-    case Hexagon::STrib_abs_nv_V4:
-    case Hexagon::STrib_abs_cPt_nv_V4:
-    case Hexagon::STrib_abs_cdnPt_nv_V4:
-    case Hexagon::STrib_abs_cNotPt_nv_V4:
-    case Hexagon::STrib_abs_cdnNotPt_nv_V4:
-    case Hexagon::STrib_imm_abs_nv_V4:
-    case Hexagon::STrib_imm_abs_cPt_nv_V4:
-    case Hexagon::STrib_imm_abs_cdnPt_nv_V4:
-    case Hexagon::STrib_imm_abs_cNotPt_nv_V4:
-    case Hexagon::STrib_imm_abs_cdnNotPt_nv_V4:
-
-    // Store Halfword
-    case Hexagon::STrih_nv_V4:
-    case Hexagon::STrih_indexed_nv_V4:
-    case Hexagon::STrih_indexed_shl_nv_V4:
-    case Hexagon::STrih_shl_nv_V4:
-    case Hexagon::STrih_GP_nv_V4:
-    case Hexagon::STh_GP_nv_V4:
-    case Hexagon::POST_SThri_nv_V4:
-    case Hexagon::STrih_cPt_nv_V4:
-    case Hexagon::STrih_cdnPt_nv_V4:
-    case Hexagon::STrih_cNotPt_nv_V4:
-    case Hexagon::STrih_cdnNotPt_nv_V4:
-    case Hexagon::STrih_indexed_cPt_nv_V4:
-    case Hexagon::STrih_indexed_cdnPt_nv_V4:
-    case Hexagon::STrih_indexed_cNotPt_nv_V4:
-    case Hexagon::STrih_indexed_cdnNotPt_nv_V4:
-    case Hexagon::STrih_indexed_shl_cPt_nv_V4:
-    case Hexagon::STrih_indexed_shl_cdnPt_nv_V4:
-    case Hexagon::STrih_indexed_shl_cNotPt_nv_V4:
-    case Hexagon::STrih_indexed_shl_cdnNotPt_nv_V4:
-    case Hexagon::POST_SThri_cPt_nv_V4:
-    case Hexagon::POST_SThri_cdnPt_nv_V4:
-    case Hexagon::POST_SThri_cNotPt_nv_V4:
-    case Hexagon::POST_SThri_cdnNotPt_nv_V4:
-    case Hexagon::STh_GP_cPt_nv_V4:
-    case Hexagon::STh_GP_cNotPt_nv_V4:
-    case Hexagon::STh_GP_cdnPt_nv_V4:
-    case Hexagon::STh_GP_cdnNotPt_nv_V4:
-    case Hexagon::STrih_GP_cPt_nv_V4:
-    case Hexagon::STrih_GP_cNotPt_nv_V4:
-    case Hexagon::STrih_GP_cdnPt_nv_V4:
-    case Hexagon::STrih_GP_cdnNotPt_nv_V4:
-    case Hexagon::STrih_abs_nv_V4:
-    case Hexagon::STrih_abs_cPt_nv_V4:
-    case Hexagon::STrih_abs_cdnPt_nv_V4:
-    case Hexagon::STrih_abs_cNotPt_nv_V4:
-    case Hexagon::STrih_abs_cdnNotPt_nv_V4:
-    case Hexagon::STrih_imm_abs_nv_V4:
-    case Hexagon::STrih_imm_abs_cPt_nv_V4:
-    case Hexagon::STrih_imm_abs_cdnPt_nv_V4:
-    case Hexagon::STrih_imm_abs_cNotPt_nv_V4:
-    case Hexagon::STrih_imm_abs_cdnNotPt_nv_V4:
-
-    // Store Word
-    case Hexagon::STriw_nv_V4:
-    case Hexagon::STriw_indexed_nv_V4:
-    case Hexagon::STriw_indexed_shl_nv_V4:
-    case Hexagon::STriw_shl_nv_V4:
-    case Hexagon::STriw_GP_nv_V4:
-    case Hexagon::STw_GP_nv_V4:
-    case Hexagon::POST_STwri_nv_V4:
-    case Hexagon::STriw_cPt_nv_V4:
-    case Hexagon::STriw_cdnPt_nv_V4:
-    case Hexagon::STriw_cNotPt_nv_V4:
-    case Hexagon::STriw_cdnNotPt_nv_V4:
-    case Hexagon::STriw_indexed_cPt_nv_V4:
-    case Hexagon::STriw_indexed_cdnPt_nv_V4:
-    case Hexagon::STriw_indexed_cNotPt_nv_V4:
-    case Hexagon::STriw_indexed_cdnNotPt_nv_V4:
-    case Hexagon::STriw_indexed_shl_cPt_nv_V4:
-    case Hexagon::STriw_indexed_shl_cdnPt_nv_V4:
-    case Hexagon::STriw_indexed_shl_cNotPt_nv_V4:
-    case Hexagon::STriw_indexed_shl_cdnNotPt_nv_V4:
-    case Hexagon::POST_STwri_cPt_nv_V4:
-    case Hexagon::POST_STwri_cdnPt_nv_V4:
-    case Hexagon::POST_STwri_cNotPt_nv_V4:
-    case Hexagon::POST_STwri_cdnNotPt_nv_V4:
-    case Hexagon::STw_GP_cPt_nv_V4:
-    case Hexagon::STw_GP_cNotPt_nv_V4:
-    case Hexagon::STw_GP_cdnPt_nv_V4:
-    case Hexagon::STw_GP_cdnNotPt_nv_V4:
-    case Hexagon::STriw_GP_cPt_nv_V4:
-    case Hexagon::STriw_GP_cNotPt_nv_V4:
-    case Hexagon::STriw_GP_cdnPt_nv_V4:
-    case Hexagon::STriw_GP_cdnNotPt_nv_V4:
-    case Hexagon::STriw_abs_nv_V4:
-    case Hexagon::STriw_abs_cPt_nv_V4:
-    case Hexagon::STriw_abs_cdnPt_nv_V4:
-    case Hexagon::STriw_abs_cNotPt_nv_V4:
-    case Hexagon::STriw_abs_cdnNotPt_nv_V4:
-    case Hexagon::STriw_imm_abs_nv_V4:
-    case Hexagon::STriw_imm_abs_cPt_nv_V4:
-    case Hexagon::STriw_imm_abs_cdnPt_nv_V4:
-    case Hexagon::STriw_imm_abs_cNotPt_nv_V4:
-    case Hexagon::STriw_imm_abs_cdnNotPt_nv_V4:
-      return true;
-
-    default:
-      return false;
-  }
-  return false;
-}
-
-bool HexagonInstrInfo::isPostIncrement (const MachineInstr* MI) const {
-  switch (MI->getOpcode())
-  {
-    // Load Byte
-    case Hexagon::POST_LDrib:
-    case Hexagon::POST_LDrib_cPt:
-    case Hexagon::POST_LDrib_cNotPt:
-    case Hexagon::POST_LDrib_cdnPt_V4:
-    case Hexagon::POST_LDrib_cdnNotPt_V4:
-
-    // Load unsigned byte
-    case Hexagon::POST_LDriub:
-    case Hexagon::POST_LDriub_cPt:
-    case Hexagon::POST_LDriub_cNotPt:
-    case Hexagon::POST_LDriub_cdnPt_V4:
-    case Hexagon::POST_LDriub_cdnNotPt_V4:
-
-    // Load halfword
-    case Hexagon::POST_LDrih:
-    case Hexagon::POST_LDrih_cPt:
-    case Hexagon::POST_LDrih_cNotPt:
-    case Hexagon::POST_LDrih_cdnPt_V4:
-    case Hexagon::POST_LDrih_cdnNotPt_V4:
-
-    // Load unsigned halfword
-    case Hexagon::POST_LDriuh:
-    case Hexagon::POST_LDriuh_cPt:
-    case Hexagon::POST_LDriuh_cNotPt:
-    case Hexagon::POST_LDriuh_cdnPt_V4:
-    case Hexagon::POST_LDriuh_cdnNotPt_V4:
-
-    // Load word
-    case Hexagon::POST_LDriw:
-    case Hexagon::POST_LDriw_cPt:
-    case Hexagon::POST_LDriw_cNotPt:
-    case Hexagon::POST_LDriw_cdnPt_V4:
-    case Hexagon::POST_LDriw_cdnNotPt_V4:
-
-    // Load double word
-    case Hexagon::POST_LDrid:
-    case Hexagon::POST_LDrid_cPt:
-    case Hexagon::POST_LDrid_cNotPt:
-    case Hexagon::POST_LDrid_cdnPt_V4:
-    case Hexagon::POST_LDrid_cdnNotPt_V4:
-
-    // Store byte
-    case Hexagon::POST_STbri:
-    case Hexagon::POST_STbri_cPt:
-    case Hexagon::POST_STbri_cNotPt:
-    case Hexagon::POST_STbri_cdnPt_V4:
-    case Hexagon::POST_STbri_cdnNotPt_V4:
-
-    // Store halfword
-    case Hexagon::POST_SThri:
-    case Hexagon::POST_SThri_cPt:
-    case Hexagon::POST_SThri_cNotPt:
-    case Hexagon::POST_SThri_cdnPt_V4:
-    case Hexagon::POST_SThri_cdnNotPt_V4:
-
-    // Store word
-    case Hexagon::POST_STwri:
-    case Hexagon::POST_STwri_cPt:
-    case Hexagon::POST_STwri_cNotPt:
-    case Hexagon::POST_STwri_cdnPt_V4:
-    case Hexagon::POST_STwri_cdnNotPt_V4:
-
-    // Store double word
-    case Hexagon::POST_STdri:
-    case Hexagon::POST_STdri_cPt:
-    case Hexagon::POST_STdri_cNotPt:
-    case Hexagon::POST_STdri_cdnPt_V4:
-    case Hexagon::POST_STdri_cdnNotPt_V4:
-      return true;
-
-    default:
-      return false;
-  }
-}
-
-bool HexagonInstrInfo::isSaveCalleeSavedRegsCall(const MachineInstr *MI) const {
-  return MI->getOpcode() == Hexagon::SAVE_REGISTERS_CALL_V4;
-}
 
 bool HexagonInstrInfo::isPredicable(MachineInstr *MI) const {
   bool isPred = MI->getDesc().isPredicable();
@@ -2445,24 +1587,6 @@ isSpillPredRegOp(const MachineInstr *MI) const {
   return false;
 }
 
-bool HexagonInstrInfo::
-isConditionalTransfer (const MachineInstr *MI) const {
-  switch (MI->getOpcode()) {
-    case Hexagon::TFR_cPt:
-    case Hexagon::TFR_cNotPt:
-    case Hexagon::TFRI_cPt:
-    case Hexagon::TFRI_cNotPt:
-    case Hexagon::TFR_cdnPt:
-    case Hexagon::TFR_cdnNotPt:
-    case Hexagon::TFRI_cdnPt:
-    case Hexagon::TFRI_cdnNotPt:
-      return true;
-
-    default:
-      return false;
-  }
-  return false;
-}
 
 bool HexagonInstrInfo::isConditionalALU32 (const MachineInstr* MI) const {
   const HexagonRegisterInfo& QRI = getRegisterInfo();
@@ -2502,6 +1626,7 @@ bool HexagonInstrInfo::isConditionalALU32 (const MachineInstr* MI) const {
   }
 }
 
+
 bool HexagonInstrInfo::
 isConditionalLoad (const MachineInstr* MI) const {
   const HexagonRegisterInfo& QRI = getRegisterInfo();
@@ -2575,136 +1700,6 @@ isConditionalLoad (const MachineInstr* MI) const {
   }
 }
 
-// Returns true if an instruction is a conditional store.
-//
-// Note: It doesn't include conditional new-value stores as they can't be
-// converted to .new predicate.
-//
-//               p.new NV store [ if(p0.new)memw(R0+#0)=R2.new ]
-//                ^           ^
-//               /             \ (not OK. it will cause new-value store to be
-//              /               X conditional on p0.new while R2 producer is
-//             /                 \ on p0)
-//            /                   \.
-//     p.new store                 p.old NV store
-// [if(p0.new)memw(R0+#0)=R2]    [if(p0)memw(R0+#0)=R2.new]
-//            ^                  ^
-//             \                /
-//              \              /
-//               \            /
-//                 p.old store
-//             [if (p0)memw(R0+#0)=R2]
-//
-// The above diagram shows the steps involoved in the conversion of a predicated
-// store instruction to its .new predicated new-value form.
-//
-// The following set of instructions further explains the scenario where
-// conditional new-value store becomes invalid when promoted to .new predicate
-// form.
-//
-// { 1) if (p0) r0 = add(r1, r2)
-//   2) p0 = cmp.eq(r3, #0) }
-//
-//   3) if (p0) memb(r1+#0) = r0  --> this instruction can't be grouped with
-// the first two instructions because in instr 1, r0 is conditional on old value
-// of p0 but its use in instr 3 is conditional on p0 modified by instr 2 which
-// is not valid for new-value stores.
-bool HexagonInstrInfo::
-isConditionalStore (const MachineInstr* MI) const {
-  const HexagonRegisterInfo& QRI = getRegisterInfo();
-  switch (MI->getOpcode())
-  {
-    case Hexagon::STrib_imm_cPt_V4 :
-    case Hexagon::STrib_imm_cNotPt_V4 :
-    case Hexagon::STrib_indexed_shl_cPt_V4 :
-    case Hexagon::STrib_indexed_shl_cNotPt_V4 :
-    case Hexagon::STrib_cPt :
-    case Hexagon::STrib_cNotPt :
-    case Hexagon::POST_STbri_cPt :
-    case Hexagon::POST_STbri_cNotPt :
-    case Hexagon::STrid_indexed_cPt :
-    case Hexagon::STrid_indexed_cNotPt :
-    case Hexagon::STrid_indexed_shl_cPt_V4 :
-    case Hexagon::POST_STdri_cPt :
-    case Hexagon::POST_STdri_cNotPt :
-    case Hexagon::STrih_cPt :
-    case Hexagon::STrih_cNotPt :
-    case Hexagon::STrih_indexed_cPt :
-    case Hexagon::STrih_indexed_cNotPt :
-    case Hexagon::STrih_imm_cPt_V4 :
-    case Hexagon::STrih_imm_cNotPt_V4 :
-    case Hexagon::STrih_indexed_shl_cPt_V4 :
-    case Hexagon::STrih_indexed_shl_cNotPt_V4 :
-    case Hexagon::POST_SThri_cPt :
-    case Hexagon::POST_SThri_cNotPt :
-    case Hexagon::STriw_cPt :
-    case Hexagon::STriw_cNotPt :
-    case Hexagon::STriw_indexed_cPt :
-    case Hexagon::STriw_indexed_cNotPt :
-    case Hexagon::STriw_imm_cPt_V4 :
-    case Hexagon::STriw_imm_cNotPt_V4 :
-    case Hexagon::STriw_indexed_shl_cPt_V4 :
-    case Hexagon::STriw_indexed_shl_cNotPt_V4 :
-    case Hexagon::POST_STwri_cPt :
-    case Hexagon::POST_STwri_cNotPt :
-      return QRI.Subtarget.hasV4TOps();
-
-    // V4 global address store before promoting to dot new.
-    case Hexagon::STrid_GP_cPt_V4 :
-    case Hexagon::STrid_GP_cNotPt_V4 :
-    case Hexagon::STrib_GP_cPt_V4 :
-    case Hexagon::STrib_GP_cNotPt_V4 :
-    case Hexagon::STrih_GP_cPt_V4 :
-    case Hexagon::STrih_GP_cNotPt_V4 :
-    case Hexagon::STriw_GP_cPt_V4 :
-    case Hexagon::STriw_GP_cNotPt_V4 :
-    case Hexagon::STd_GP_cPt_V4 :
-    case Hexagon::STd_GP_cNotPt_V4 :
-    case Hexagon::STb_GP_cPt_V4 :
-    case Hexagon::STb_GP_cNotPt_V4 :
-    case Hexagon::STh_GP_cPt_V4 :
-    case Hexagon::STh_GP_cNotPt_V4 :
-    case Hexagon::STw_GP_cPt_V4 :
-    case Hexagon::STw_GP_cNotPt_V4 :
-      return QRI.Subtarget.hasV4TOps();
-
-    // Predicated new value stores (i.e. if (p0) memw(..)=r0.new) are excluded
-    // from the "Conditional Store" list. Because a predicated new value store
-    // would NOT be promoted to a double dot new store. See diagram below:
-    // This function returns yes for those stores that are predicated but not
-    // yet promoted to predicate dot new instructions.
-    //
-    //                          +---------------------+
-    //                    /-----| if (p0) memw(..)=r0 |---------\~
-    //                   ||     +---------------------+         ||
-    //          promote  ||       /\       /\                   ||  promote
-    //                   ||      /||\     /||\                  ||
-    //                  \||/    demote     ||                  \||/
-    //                   \/       ||       ||                   \/
-    //       +-------------------------+   ||   +-------------------------+
-    //       | if (p0.new) memw(..)=r0 |   ||   | if (p0) memw(..)=r0.new |
-    //       +-------------------------+   ||   +-------------------------+
-    //                        ||           ||         ||
-    //                        ||         demote      \||/
-    //                      promote        ||         \/ NOT possible
-    //                        ||           ||         /\~
-    //                       \||/          ||        /||\~
-    //                        \/           ||         ||
-    //                      +-----------------------------+
-    //                      | if (p0.new) memw(..)=r0.new |
-    //                      +-----------------------------+
-    //                           Double Dot New Store
-    //
-
-    default:
-      return false;
-
-  }
-  return false;
-}
-
-
-
 DFAPacketizer *HexagonInstrInfo::
 CreateTargetScheduleState(const TargetMachine *TM,
                            const ScheduleDAG *DAG) const {
diff --git a/lib/Target/Hexagon/HexagonInstrInfo.h b/lib/Target/Hexagon/HexagonInstrInfo.h
index 6a45871b67e..730687036c8 100644
--- a/lib/Target/Hexagon/HexagonInstrInfo.h
+++ b/lib/Target/Hexagon/HexagonInstrInfo.h
@@ -160,20 +160,10 @@ public:
   bool isS8_Immediate(const int value) const;
   bool isS6_Immediate(const int value) const;
 
-  bool isSaveCalleeSavedRegsCall(const MachineInstr* MI) const;
-  bool isConditionalTransfer(const MachineInstr* MI) const;
   bool isConditionalALU32 (const MachineInstr* MI) const;
   bool isConditionalLoad (const MachineInstr* MI) const;
-  bool isConditionalStore(const MachineInstr* MI) const;
   bool isDeallocRet(const MachineInstr *MI) const;
   unsigned getInvertedPredicatedOpcode(const int Opc) const;
-  bool isExtendable(const MachineInstr* MI) const;
-  bool isExtended(const MachineInstr* MI) const;
-  bool isPostIncrement(const MachineInstr* MI) const;
-  bool isNewValueStore(const MachineInstr* MI) const;
-  bool isNewValueJump(const MachineInstr* MI) const;
-  unsigned getImmExtForm(const MachineInstr* MI) const;
-  unsigned getNormalBranchForm(const MachineInstr* MI) const;
 
 private:
   int getMatchingCondBranchOpcode(int Opc, bool sense) const;
diff --git a/lib/Target/Hexagon/HexagonInstrInfo.td b/lib/Target/Hexagon/HexagonInstrInfo.td
index fd5adef0f63..b563ac3c613 100644
--- a/lib/Target/Hexagon/HexagonInstrInfo.td
+++ b/lib/Target/Hexagon/HexagonInstrInfo.td
@@ -3046,7 +3046,3 @@ include "HexagonInstrInfoV3.td"
 //===----------------------------------------------------------------------===//
 
 include "HexagonInstrInfoV4.td"
-
-//===----------------------------------------------------------------------===//
-// V4 Instructions -
-//===----------------------------------------------------------------------===//
diff --git a/lib/Target/Hexagon/HexagonInstrInfoV3.td b/lib/Target/Hexagon/HexagonInstrInfoV3.td
index 2bd6770efd7..a73897ee345 100644
--- a/lib/Target/Hexagon/HexagonInstrInfoV3.td
+++ b/lib/Target/Hexagon/HexagonInstrInfoV3.td
@@ -41,11 +41,10 @@ let isCall = 1, neverHasSideEffects = 1,
  }
 
 
-// Jump to address from register
 // if(p?.new) jumpr:t r?
 let isReturn = 1, isTerminator = 1, isBarrier = 1,
   Defs = [PC], Uses = [R31] in {
-  def JMPR_cdnPt_V3: JRInst<(outs), (ins PredRegs:$src1, IntRegs:$src2),
+  def JMPR_cPnewt: JRInst<(outs), (ins PredRegs:$src1, IntRegs:$src2),
                        "if ($src1.new) jumpr:t $src2",
                        []>, Requires<[HasV3T]>;
 }
@@ -53,7 +52,7 @@ let isReturn = 1, isTerminator = 1, isBarrier = 1,
 // if (!p?.new) jumpr:t r?
 let isReturn = 1, isTerminator = 1, isBarrier = 1,
   Defs = [PC], Uses = [R31] in {
-  def JMPR_cdnNotPt_V3: JRInst<(outs), (ins PredRegs:$src1, IntRegs:$src2),
+  def JMPR_cNotPnewt: JRInst<(outs), (ins PredRegs:$src1, IntRegs:$src2),
                        "if (!$src1.new) jumpr:t $src2",
                        []>, Requires<[HasV3T]>;
 }
@@ -62,7 +61,7 @@ let isReturn = 1, isTerminator = 1, isBarrier = 1,
 // if(p?.new) jumpr:nt r?
 let isReturn = 1, isTerminator = 1, isBarrier = 1,
   Defs = [PC], Uses = [R31] in {
-  def JMPR_cdnPnt: JRInst<(outs), (ins PredRegs:$src1, IntRegs:$src2),
+  def JMPR_cPnewNt: JRInst<(outs), (ins PredRegs:$src1, IntRegs:$src2),
                        "if ($src1.new) jumpr:nt $src2",
                        []>, Requires<[HasV3T]>;
 }
@@ -70,7 +69,7 @@ let isReturn = 1, isTerminator = 1, isBarrier = 1,
 // if (!p?.new) jumpr:nt r?
 let isReturn = 1, isTerminator = 1, isBarrier = 1,
   Defs = [PC], Uses = [R31] in {
-  def JMPR_cdnNotPnt: JRInst<(outs), (ins PredRegs:$src1, IntRegs:$src2),
+  def JMPR_cNotPnewNt: JRInst<(outs), (ins PredRegs:$src1, IntRegs:$src2),
                        "if (!$src1.new) jumpr:nt $src2",
                        []>, Requires<[HasV3T]>;
 }
@@ -87,22 +86,20 @@ let AddedComplexity = 200 in
 def MAXw_dd : ALU64_rr<(outs DoubleRegs:$dst), (ins DoubleRegs:$src1,
                                                     DoubleRegs:$src2),
               "$dst = max($src2, $src1)",
-              [(set (i64 DoubleRegs:$dst),
-                    (i64 (select (i1 (setlt (i64 DoubleRegs:$src2),
-                                            (i64 DoubleRegs:$src1))),
-                                 (i64 DoubleRegs:$src1),
-                                 (i64 DoubleRegs:$src2))))]>,
+              [(set DoubleRegs:$dst, (select (i1 (setlt DoubleRegs:$src2,
+                                                        DoubleRegs:$src1)),
+                                             DoubleRegs:$src1,
+                                             DoubleRegs:$src2))]>,
 Requires<[HasV3T]>;
 
 let AddedComplexity = 200 in
 def MINw_dd : ALU64_rr<(outs DoubleRegs:$dst), (ins DoubleRegs:$src1,
                                                     DoubleRegs:$src2),
               "$dst = min($src2, $src1)",
-              [(set (i64 DoubleRegs:$dst),
-                    (i64 (select (i1 (setgt (i64 DoubleRegs:$src2),
-                                            (i64 DoubleRegs:$src1))),
-                                 (i64 DoubleRegs:$src1),
-                                 (i64 DoubleRegs:$src2))))]>,
+              [(set DoubleRegs:$dst, (select (i1 (setgt DoubleRegs:$src2,
+                                                        DoubleRegs:$src1)),
+                                             DoubleRegs:$src1,
+                                             DoubleRegs:$src2))]>,
 Requires<[HasV3T]>;
 
 //===----------------------------------------------------------------------===//
@@ -112,25 +109,25 @@ Requires<[HasV3T]>;
 
 
 
-//def : Pat <(brcond (i1 (seteq (i32 IntRegs:$src1), 0)), bb:$offset),
-//      (JMP_RegEzt (i32 IntRegs:$src1), bb:$offset)>, Requires<[HasV3T]>;
+//def : Pat <(brcond (i1 (seteq IntRegs:$src1, 0)), bb:$offset),
+//      (JMP_RegEzt IntRegs:$src1, bb:$offset)>, Requires<[HasV3T]>;
 
-//def : Pat <(brcond (i1 (setne (i32 IntRegs:$src1), 0)), bb:$offset),
-//      (JMP_RegNzt (i32 IntRegs:$src1), bb:$offset)>, Requires<[HasV3T]>;
+//def : Pat <(brcond (i1 (setne IntRegs:$src1, 0)), bb:$offset),
+//      (JMP_RegNzt IntRegs:$src1, bb:$offset)>, Requires<[HasV3T]>;
 
-//def : Pat <(brcond (i1 (setle (i32 IntRegs:$src1), 0)), bb:$offset),
-//      (JMP_RegLezt (i32 IntRegs:$src1), bb:$offset)>, Requires<[HasV3T]>;
+//def : Pat <(brcond (i1 (setle IntRegs:$src1, 0)), bb:$offset),
+//      (JMP_RegLezt IntRegs:$src1, bb:$offset)>, Requires<[HasV3T]>;
 
-//def : Pat <(brcond (i1 (setge (i32 IntRegs:$src1), 0)), bb:$offset),
-//      (JMP_RegGezt (i32 IntRegs:$src1), bb:$offset)>, Requires<[HasV3T]>;
+//def : Pat <(brcond (i1 (setge IntRegs:$src1, 0)), bb:$offset),
+//      (JMP_RegGezt IntRegs:$src1, bb:$offset)>, Requires<[HasV3T]>;
 
-//def : Pat <(brcond (i1 (setgt (i32 IntRegs:$src1), -1)), bb:$offset),
-//      (JMP_RegGezt (i32 IntRegs:$src1), bb:$offset)>, Requires<[HasV3T]>;
+//def : Pat <(brcond (i1 (setgt IntRegs:$src1, -1)), bb:$offset),
+//      (JMP_RegGezt IntRegs:$src1, bb:$offset)>, Requires<[HasV3T]>;
 
 
 // Map call instruction
-def : Pat<(call (i32 IntRegs:$dst)),
-      (CALLRv3 (i32 IntRegs:$dst))>, Requires<[HasV3T]>;
+def : Pat<(call IntRegs:$dst),
+      (CALLRv3 IntRegs:$dst)>, Requires<[HasV3T]>;
 def : Pat<(call tglobaladdr:$dst),
       (CALLv3 tglobaladdr:$dst)>, Requires<[HasV3T]>;
 def : Pat<(call texternalsym:$dst),
diff --git a/lib/Target/Hexagon/HexagonInstrInfoV4.td b/lib/Target/Hexagon/HexagonInstrInfoV4.td
index f507e4f37c1..9e60cf26d08 100644
--- a/lib/Target/Hexagon/HexagonInstrInfoV4.td
+++ b/lib/Target/Hexagon/HexagonInstrInfoV4.td
@@ -11,11 +11,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-def IMMEXT : Immext<(outs), (ins),
-                    "##immext //should never emit this",
-                    []>,
-                    Requires<[HasV4T]>;
-
 // Hexagon V4 Architecture spec defines 8 instruction classes:
 // LD ST ALU32 XTYPE J JR MEMOP NV CR SYSTEM(system is not implemented in the
 // compiler)
@@ -255,151 +250,23 @@ def ZXTH_cdnNotPt_V4 : ALU32_rr<(outs IntRegs:$dst),
             []>,
             Requires<[HasV4T]>;
 
-// Generate frame index addresses.
-let neverHasSideEffects = 1, isReMaterializable = 1 in
-def TFR_FI_immext_V4 : ALU32_ri<(outs IntRegs:$dst),
-            (ins IntRegs:$src1, s32Imm:$offset),
-            "$dst = add($src1, ##$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
 
 //===----------------------------------------------------------------------===//
 // ALU32 -
 //===----------------------------------------------------------------------===//
 
 
-//===----------------------------------------------------------------------===//
-// ALU32/PERM +
-//===----------------------------------------------------------------------===//
-
-// Combine
-// Rdd=combine(Rs, #s8)
-let neverHasSideEffects = 1 in
-def COMBINE_ri_V4 : ALU32_ri<(outs DoubleRegs:$dst),
-            (ins IntRegs:$src1, s8Imm:$src2),
-            "$dst = combine($src1, #$src2)",
-            []>,
-            Requires<[HasV4T]>;
-// Rdd=combine(#s8, Rs)
-let neverHasSideEffects = 1 in
-def COMBINE_ir_V4 : ALU32_ir<(outs DoubleRegs:$dst),
-            (ins s8Imm:$src1, IntRegs:$src2),
-            "$dst = combine(#$src1, $src2)",
-            []>,
-            Requires<[HasV4T]>;
-//===----------------------------------------------------------------------===//
-// ALU32/PERM +
-//===----------------------------------------------------------------------===//
 
 //===----------------------------------------------------------------------===//
 // LD +
 //===----------------------------------------------------------------------===//
-//
-// These absolute set addressing mode instructions accept immediate as
-// an operand. We have duplicated these patterns to take global address.
+///
+/// Make sure that in post increment load, the first operand is always the post
+/// increment operand.
+///
+//// Load doubleword.
+// Rdd=memd(Re=#U6)
 
-let neverHasSideEffects = 1 in
-def LDrid_abs_setimm_V4 : LDInst<(outs DoubleRegs:$dst1, IntRegs:$dst2),
-            (ins u6Imm:$addr),
-            "$dst1 = memd($dst2=#$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Rd=memb(Re=#U6)
-let neverHasSideEffects = 1 in
-def LDrib_abs_setimm_V4 : LDInst<(outs IntRegs:$dst1, IntRegs:$dst2),
-            (ins u6Imm:$addr),
-            "$dst1 = memb($dst2=#$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Rd=memh(Re=#U6)
-let neverHasSideEffects = 1 in
-def LDrih_abs_setimm_V4 : LDInst<(outs IntRegs:$dst1, IntRegs:$dst2),
-            (ins u6Imm:$addr),
-            "$dst1 = memh($dst2=#$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Rd=memub(Re=#U6)
-let neverHasSideEffects = 1 in
-def LDriub_abs_setimm_V4 : LDInst<(outs IntRegs:$dst1, IntRegs:$dst2),
-            (ins u6Imm:$addr),
-            "$dst1 = memub($dst2=#$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Rd=memuh(Re=#U6)
-let neverHasSideEffects = 1 in
-def LDriuh_abs_setimm_V4 : LDInst<(outs IntRegs:$dst1, IntRegs:$dst2),
-            (ins u6Imm:$addr),
-            "$dst1 = memuh($dst2=#$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Rd=memw(Re=#U6)
-let neverHasSideEffects = 1 in
-def LDriw_abs_setimm_V4 : LDInst<(outs IntRegs:$dst1, IntRegs:$dst2),
-            (ins u6Imm:$addr),
-            "$dst1 = memw($dst2=#$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Following patterns are defined for absolute set addressing mode
-// instruction which take global address as operand.
-let mayLoad = 1, neverHasSideEffects = 1 in
-def LDrid_abs_set_V4 : LDInst<(outs DoubleRegs:$dst1, IntRegs:$dst2),
-            (ins globaladdress:$addr),
-            "$dst1 = memd($dst2=##$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Rd=memb(Re=#U6)
-let mayLoad = 1, neverHasSideEffects = 1 in
-def LDrib_abs_set_V4 : LDInst<(outs IntRegs:$dst1, IntRegs:$dst2),
-            (ins globaladdress:$addr),
-            "$dst1 = memb($dst2=##$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Rd=memh(Re=#U6)
-let mayLoad = 1, neverHasSideEffects = 1 in
-def LDrih_abs_set_V4 : LDInst<(outs IntRegs:$dst1, IntRegs:$dst2),
-            (ins globaladdress:$addr),
-            "$dst1 = memh($dst2=##$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Rd=memub(Re=#U6)
-let mayLoad = 1, neverHasSideEffects = 1 in
-def LDriub_abs_set_V4 : LDInst<(outs IntRegs:$dst1, IntRegs:$dst2),
-            (ins globaladdress:$addr),
-            "$dst1 = memub($dst2=##$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Rd=memuh(Re=#U6)
-let mayLoad = 1, neverHasSideEffects = 1 in
-def LDriuh_abs_set_V4 : LDInst<(outs IntRegs:$dst1, IntRegs:$dst2),
-            (ins globaladdress:$addr),
-            "$dst1 = memuh($dst2=##$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Rd=memw(Re=#U6)
-let mayLoad = 1, neverHasSideEffects = 1 in
-def LDriw_abs_set_V4 : LDInst<(outs IntRegs:$dst1, IntRegs:$dst2),
-            (ins globaladdress:$addr),
-            "$dst1 = memw($dst2=##$addr)",
-            []>,
-            Requires<[HasV4T]>;
-
-// Load doubleword.
-//
-// Make sure that in post increment load, the first operand is always the post
-// increment operand.
-//
 // Rdd=memd(Rs+Rt<<#u2)
 // Special case pattern for indexed load without offset which is easier to
 // match. AddedComplexity of this pattern should be lower than base+offset load
@@ -409,19 +276,17 @@ let AddedComplexity = 10, isPredicable = 1 in
 def LDrid_indexed_V4 : LDInst<(outs DoubleRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2),
                     "$dst=memd($src1+$src2<<#0)",
-                    [(set (i64 DoubleRegs:$dst),
-                          (i64 (load (add (i32 IntRegs:$src1),
-                                          (i32 IntRegs:$src2)))))]>,
+                    [(set DoubleRegs:$dst, (load (add IntRegs:$src1,
+                                                      IntRegs:$src2)))]>,
                     Requires<[HasV4T]>;
 
 let AddedComplexity = 40, isPredicable = 1 in
 def LDrid_indexed_shl_V4 : LDInst<(outs DoubleRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$offset),
                     "$dst=memd($src1+$src2<<#$offset)",
-                    [(set (i64 DoubleRegs:$dst),
-                          (i64 (load (add (i32 IntRegs:$src1),
-                                          (shl (i32 IntRegs:$src2),
-                                               u2ImmPred:$offset)))))]>,
+                    [(set DoubleRegs:$dst, (load (add IntRegs:$src1,
+                                                 (shl IntRegs:$src2,
+                                                      u2ImmPred:$offset))))]>,
                     Requires<[HasV4T]>;
 
 //// Load doubleword conditionally.
@@ -497,62 +362,60 @@ def LDrid_indexed_shl_cdnNotPt_V4 : LDInst<(outs DoubleRegs:$dst),
 // Rdd=memd(Rt<<#u2+#U6)
 
 //// Load byte.
+// Rd=memb(Re=#U6)
+
 // Rd=memb(Rs+Rt<<#u2)
 let AddedComplexity = 10, isPredicable = 1 in
 def LDrib_indexed_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2),
                     "$dst=memb($src1+$src2<<#0)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (sextloadi8 (add (i32 IntRegs:$src1),
-                                                (i32 IntRegs:$src2)))))]>,
+                    [(set IntRegs:$dst, (sextloadi8 (add IntRegs:$src1,
+                                                         IntRegs:$src2)))]>,
                     Requires<[HasV4T]>;
 
 let AddedComplexity = 10, isPredicable = 1 in
 def LDriub_indexed_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2),
                     "$dst=memub($src1+$src2<<#0)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (zextloadi8 (add (i32 IntRegs:$src1),
-                                                (i32 IntRegs:$src2)))))]>,
+                    [(set IntRegs:$dst, (zextloadi8 (add IntRegs:$src1,
+                                                         IntRegs:$src2)))]>,
                     Requires<[HasV4T]>;
 
 let AddedComplexity = 10, isPredicable = 1 in
 def LDriub_ae_indexed_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2),
                     "$dst=memub($src1+$src2<<#0)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (extloadi8 (add (i32 IntRegs:$src1),
-                                               (i32 IntRegs:$src2)))))]>,
+                    [(set IntRegs:$dst, (extloadi8 (add IntRegs:$src1,
+                                                        IntRegs:$src2)))]>,
                     Requires<[HasV4T]>;
 
 let AddedComplexity = 40, isPredicable = 1 in
 def LDrib_indexed_shl_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$offset),
                     "$dst=memb($src1+$src2<<#$offset)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (sextloadi8 (add (i32 IntRegs:$src1),
-                                                (shl (i32 IntRegs:$src2),
-                                                     u2ImmPred:$offset)))))]>,
+                    [(set IntRegs:$dst,
+                          (sextloadi8 (add IntRegs:$src1,
+                                           (shl IntRegs:$src2,
+                                                u2ImmPred:$offset))))]>,
                     Requires<[HasV4T]>;
 
 let AddedComplexity = 40, isPredicable = 1 in
 def LDriub_indexed_shl_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$offset),
                     "$dst=memub($src1+$src2<<#$offset)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (zextloadi8 (add (i32 IntRegs:$src1),
-                                                (shl (i32 IntRegs:$src2),
-                                                     u2ImmPred:$offset)))))]>,
+                    [(set IntRegs:$dst,
+                          (zextloadi8 (add IntRegs:$src1,
+                                           (shl IntRegs:$src2,
+                                                u2ImmPred:$offset))))]>,
                     Requires<[HasV4T]>;
 
 let AddedComplexity = 40, isPredicable = 1 in
 def LDriub_ae_indexed_shl_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$offset),
                     "$dst=memub($src1+$src2<<#$offset)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (extloadi8 (add (i32 IntRegs:$src1),
-                                               (shl (i32 IntRegs:$src2),
-                                                    u2ImmPred:$offset)))))]>,
+                    [(set IntRegs:$dst, (extloadi8 (add IntRegs:$src1,
+                                                   (shl IntRegs:$src2,
+                                                        u2ImmPred:$offset))))]>,
                     Requires<[HasV4T]>;
 
 //// Load byte conditionally.
@@ -698,32 +561,31 @@ def LDriub_indexed_shl_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
 // Rd=memb(Rt<<#u2+#U6)
 
 //// Load halfword
+// Rd=memh(Re=#U6)
+
 // Rd=memh(Rs+Rt<<#u2)
 let AddedComplexity = 10, isPredicable = 1 in
 def LDrih_indexed_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2),
                     "$dst=memh($src1+$src2<<#0)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (sextloadi16 (add (i32 IntRegs:$src1),
-                                                 (i32 IntRegs:$src2)))))]>,
+                    [(set IntRegs:$dst, (sextloadi16 (add IntRegs:$src1,
+                                                          IntRegs:$src2)))]>,
                     Requires<[HasV4T]>;
 
 let AddedComplexity = 10, isPredicable = 1 in
 def LDriuh_indexed_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2),
                     "$dst=memuh($src1+$src2<<#0)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (zextloadi16 (add (i32 IntRegs:$src1),
-                                                 (i32 IntRegs:$src2)))))]>,
+                    [(set IntRegs:$dst, (zextloadi16 (add IntRegs:$src1,
+                                                          IntRegs:$src2)))]>,
                     Requires<[HasV4T]>;
 
 let AddedComplexity = 10, isPredicable = 1 in
 def LDriuh_ae_indexed_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2),
                     "$dst=memuh($src1+$src2<<#0)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (extloadi16 (add (i32 IntRegs:$src1),
-                                                (i32 IntRegs:$src2)))))]>,
+                    [(set IntRegs:$dst, (extloadi16 (add IntRegs:$src1,
+                                                         IntRegs:$src2)))]>,
                     Requires<[HasV4T]>;
 
 // Rd=memh(Rs+Rt<<#u2)
@@ -731,30 +593,30 @@ let AddedComplexity = 40, isPredicable = 1 in
 def LDrih_indexed_shl_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$offset),
                     "$dst=memh($src1+$src2<<#$offset)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (sextloadi16 (add (i32 IntRegs:$src1),
-                                                 (shl (i32 IntRegs:$src2),
-                                                      u2ImmPred:$offset)))))]>,
+                    [(set IntRegs:$dst,
+                          (sextloadi16 (add IntRegs:$src1,
+                                            (shl IntRegs:$src2,
+                                                 u2ImmPred:$offset))))]>,
                     Requires<[HasV4T]>;
 
 let AddedComplexity = 40, isPredicable = 1 in
 def LDriuh_indexed_shl_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$offset),
                     "$dst=memuh($src1+$src2<<#$offset)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (zextloadi16 (add (i32 IntRegs:$src1),
-                                                 (shl (i32 IntRegs:$src2),
-                                                      u2ImmPred:$offset)))))]>,
+                    [(set IntRegs:$dst,
+                          (zextloadi16 (add IntRegs:$src1,
+                                            (shl IntRegs:$src2,
+                                                 u2ImmPred:$offset))))]>,
                     Requires<[HasV4T]>;
 
 let AddedComplexity = 40, isPredicable = 1 in
 def LDriuh_ae_indexed_shl_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$offset),
                     "$dst=memuh($src1+$src2<<#$offset)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (extloadi16 (add (i32 IntRegs:$src1),
-                                                (shl (i32 IntRegs:$src2),
-                                                     u2ImmPred:$offset)))))]>,
+                    [(set IntRegs:$dst,
+                          (extloadi16 (add IntRegs:$src1,
+                                           (shl IntRegs:$src2,
+                                                u2ImmPred:$offset))))]>,
                     Requires<[HasV4T]>;
 
 //// Load halfword conditionally.
@@ -900,14 +762,6 @@ def LDriuh_indexed_shl_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
 // Rd=memh(Rt<<#u2+#U6)
 
 //// Load word.
-// Load predicate: Fix for bug 5279.
-let mayLoad = 1, neverHasSideEffects = 1 in
-def LDriw_pred_V4 : LDInst<(outs PredRegs:$dst),
-            (ins MEMri:$addr),
-            "Error; should not emit",
-            []>,
-            Requires<[HasV4T]>;
-
 // Rd=memw(Re=#U6)
 
 // Rd=memw(Rs+Rt<<#u2)
@@ -915,9 +769,8 @@ let AddedComplexity = 10, isPredicable = 1 in
 def LDriw_indexed_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2),
                     "$dst=memw($src1+$src2<<#0)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (load (add (i32 IntRegs:$src1),
-                                          (i32 IntRegs:$src2)))))]>,
+                    [(set IntRegs:$dst, (load (add IntRegs:$src1,
+                                                   IntRegs:$src2)))]>,
                     Requires<[HasV4T]>;
 
 // Rd=memw(Rs+Rt<<#u2)
@@ -925,10 +778,9 @@ let AddedComplexity = 40, isPredicable = 1 in
 def LDriw_indexed_shl_V4 : LDInst<(outs IntRegs:$dst),
                     (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$offset),
                     "$dst=memw($src1+$src2<<#$offset)",
-                    [(set (i32 IntRegs:$dst),
-                          (i32 (load (add (i32 IntRegs:$src1),
-                                          (shl (i32 IntRegs:$src2),
-                                               u2ImmPred:$offset)))))]>,
+                    [(set IntRegs:$dst, (load (add IntRegs:$src1,
+                                              (shl IntRegs:$src2,
+                                                   u2ImmPred:$offset))))]>,
                     Requires<[HasV4T]>;
 
 //// Load word conditionally.
@@ -1103,633 +955,6 @@ def POST_LDriw_cdnNotPt_V4 : LDInstPI<(outs IntRegs:$dst1, IntRegs:$dst2),
             "$src2 = $dst2">,
             Requires<[HasV4T]>;
 
-/// Load from global offset
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDrid_GP_V4 : LDInst<(outs DoubleRegs:$dst),
-            (ins globaladdress:$global, u16Imm:$offset),
-            "$dst=memd(#$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrid_GP_cPt_V4 : LDInst<(outs DoubleRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1) $dst=memd(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrid_GP_cNotPt_V4 : LDInst<(outs DoubleRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1) $dst=memd(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrid_GP_cdnPt_V4 : LDInst<(outs DoubleRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1.new) $dst=memd(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrid_GP_cdnNotPt_V4 : LDInst<(outs DoubleRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1.new) $dst=memd(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDrib_GP_V4 : LDInst<(outs IntRegs:$dst),
-            (ins globaladdress:$global, u16Imm:$offset),
-            "$dst=memb(#$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrib_GP_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1) $dst=memb(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrib_GP_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1) $dst=memb(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrib_GP_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1.new) $dst=memb(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrib_GP_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1.new) $dst=memb(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDriub_GP_V4 : LDInst<(outs IntRegs:$dst),
-            (ins globaladdress:$global, u16Imm:$offset),
-            "$dst=memub(#$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriub_GP_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1) $dst=memub(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriub_GP_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1) $dst=memub(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriub_GP_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1.new) $dst=memub(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriub_GP_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1.new) $dst=memub(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDrih_GP_V4 : LDInst<(outs IntRegs:$dst),
-            (ins globaladdress:$global, u16Imm:$offset),
-            "$dst=memh(#$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrih_GP_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1) $dst=memh(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrih_GP_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1) $dst=memh(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrih_GP_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1.new) $dst=memh(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDrih_GP_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1.new) $dst=memh(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDriuh_GP_V4 : LDInst<(outs IntRegs:$dst),
-            (ins globaladdress:$global, u16Imm:$offset),
-            "$dst=memuh(#$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriuh_GP_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1) $dst=memuh(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriuh_GP_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1) $dst=memuh(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriuh_GP_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1.new) $dst=memuh(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriuh_GP_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1.new) $dst=memuh(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDriw_GP_V4 : LDInst<(outs IntRegs:$dst),
-            (ins globaladdress:$global, u16Imm:$offset),
-            "$dst=memw(#$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriw_GP_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1) $dst=memw(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriw_GP_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1) $dst=memw(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriw_GP_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if ($src1.new) $dst=memw(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDriw_GP_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset),
-            "if (!$src1.new) $dst=memw(##$global+$offset)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDd_GP_V4 : LDInst<(outs DoubleRegs:$dst),
-            (ins globaladdress:$global),
-            "$dst=memd(#$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rtt=memd(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDd_GP_cPt_V4 : LDInst<(outs DoubleRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1) $dst=memd(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-// if (!Pv) Rtt=memd(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDd_GP_cNotPt_V4 : LDInst<(outs DoubleRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1) $dst=memd(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rtt=memd(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDd_GP_cdnPt_V4 : LDInst<(outs DoubleRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1.new) $dst=memd(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-// if (!Pv) Rtt=memd(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDd_GP_cdnNotPt_V4 : LDInst<(outs DoubleRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1.new) $dst=memd(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDb_GP_V4 : LDInst<(outs IntRegs:$dst),
-            (ins globaladdress:$global),
-            "$dst=memb(#$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rt=memb(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDb_GP_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1) $dst=memb(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) Rt=memb(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDb_GP_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1) $dst=memb(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rt=memb(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDb_GP_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1.new) $dst=memb(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) Rt=memb(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDb_GP_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1.new) $dst=memb(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDub_GP_V4 : LDInst<(outs IntRegs:$dst),
-            (ins globaladdress:$global),
-            "$dst=memub(#$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rt=memub(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDub_GP_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1) $dst=memub(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-// if (!Pv) Rt=memub(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDub_GP_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1) $dst=memub(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rt=memub(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDub_GP_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1.new) $dst=memub(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-// if (!Pv) Rt=memub(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDub_GP_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1.new) $dst=memub(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDh_GP_V4 : LDInst<(outs IntRegs:$dst),
-            (ins globaladdress:$global),
-            "$dst=memh(#$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rt=memh(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDh_GP_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1) $dst=memh(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) Rt=memh(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDh_GP_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1) $dst=memh(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rt=memh(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDh_GP_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1.new) $dst=memh(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) Rt=memh(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDh_GP_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1.new) $dst=memh(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDuh_GP_V4 : LDInst<(outs IntRegs:$dst),
-            (ins globaladdress:$global),
-            "$dst=memuh(#$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rt=memuh(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDuh_GP_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1) $dst=memuh(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) Rt=memuh(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDuh_GP_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1) $dst=memuh(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rt=memuh(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDuh_GP_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1.new) $dst=memuh(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) Rt=memuh(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDuh_GP_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1.new) $dst=memuh(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-let isPredicable = 1, mayLoad = 1, neverHasSideEffects = 1 in
-def LDw_GP_V4 : LDInst<(outs IntRegs:$dst),
-            (ins globaladdress:$global),
-            "$dst=memw(#$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rt=memw(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDw_GP_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1) $dst=memw(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-// if (!Pv) Rt=memw(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDw_GP_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1) $dst=memw(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) Rt=memw(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDw_GP_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if ($src1.new) $dst=memw(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-// if (!Pv) Rt=memw(##global)
-let mayLoad = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def LDw_GP_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$global),
-            "if (!$src1.new) $dst=memw(##$global)",
-            []>,
-            Requires<[HasV4T]>;
-
-
-
-def : Pat <(atomic_load_64 (HexagonCONST32_GP tglobaladdr:$global)),
-           (i64 (LDd_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-def : Pat <(atomic_load_32 (HexagonCONST32_GP tglobaladdr:$global)),
-           (i32 (LDw_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-def : Pat <(atomic_load_16 (HexagonCONST32_GP tglobaladdr:$global)),
-           (i32 (LDuh_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-def : Pat <(atomic_load_8 (HexagonCONST32_GP tglobaladdr:$global)),
-           (i32 (LDub_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-// Map from load(globaladdress) -> memw(#foo + 0)
-let AddedComplexity = 100 in
-def : Pat <(i64 (load (HexagonCONST32_GP tglobaladdr:$global))),
-           (i64 (LDd_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-// Map from Pd = load(globaladdress) -> Rd = memb(globaladdress), Pd = Rd
-let AddedComplexity = 100 in
-def : Pat <(i1 (load (HexagonCONST32_GP tglobaladdr:$global))),
-           (i1 (TFR_PdRs (i32 (LDb_GP_V4 tglobaladdr:$global))))>,
-           Requires<[HasV4T]>;
-
-// When the Interprocedural Global Variable optimizer realizes that a certain
-// global variable takes only two constant values, it shrinks the global to
-// a boolean. Catch those loads here in the following 3 patterns.
-let AddedComplexity = 100 in
-def : Pat <(i32 (extloadi1 (HexagonCONST32_GP tglobaladdr:$global))),
-           (i32 (LDb_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-let AddedComplexity = 100 in
-def : Pat <(i32 (sextloadi1 (HexagonCONST32_GP tglobaladdr:$global))),
-           (i32 (LDb_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-// Map from load(globaladdress) -> memb(#foo)
-let AddedComplexity = 100 in
-def : Pat <(i32 (extloadi8 (HexagonCONST32_GP tglobaladdr:$global))),
-           (i32 (LDb_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-// Map from load(globaladdress) -> memb(#foo)
-let AddedComplexity = 100 in
-def : Pat <(i32 (sextloadi8 (HexagonCONST32_GP tglobaladdr:$global))),
-           (i32 (LDb_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-let AddedComplexity = 100 in
-def : Pat <(i32 (zextloadi1 (HexagonCONST32_GP tglobaladdr:$global))),
-           (i32 (LDub_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-// Map from load(globaladdress) -> memub(#foo)
-let AddedComplexity = 100 in
-def : Pat <(i32 (zextloadi8 (HexagonCONST32_GP tglobaladdr:$global))),
-           (i32 (LDub_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-// Map from load(globaladdress) -> memh(#foo)
-let AddedComplexity = 100 in
-def : Pat <(i32 (extloadi16 (HexagonCONST32_GP tglobaladdr:$global))),
-           (i32 (LDh_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-// Map from load(globaladdress) -> memh(#foo)
-let AddedComplexity = 100 in
-def : Pat <(i32 (sextloadi16 (HexagonCONST32_GP tglobaladdr:$global))),
-           (i32 (LDh_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-// Map from load(globaladdress) -> memuh(#foo)
-let AddedComplexity = 100 in
-def : Pat <(i32 (zextloadi16 (HexagonCONST32_GP tglobaladdr:$global))),
-           (i32 (LDuh_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-// Map from load(globaladdress) -> memw(#foo)
-let AddedComplexity = 100 in
-def : Pat <(i32 (load (HexagonCONST32_GP tglobaladdr:$global))),
-           (i32 (LDw_GP_V4 tglobaladdr:$global))>,
-            Requires<[HasV4T]>;
-
-def : Pat <(atomic_load_64 (add (HexagonCONST32_GP tglobaladdr:$global),
-                                u16ImmPred:$offset)),
-           (i64 (LDrid_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-           Requires<[HasV4T]>;
-
-def : Pat <(atomic_load_32 (add (HexagonCONST32_GP tglobaladdr:$global),
-                                u16ImmPred:$offset)),
-           (i32 (LDriw_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-            Requires<[HasV4T]>;
-
-def : Pat <(atomic_load_16 (add (HexagonCONST32_GP tglobaladdr:$global),
-                                u16ImmPred:$offset)),
-           (i32 (LDriuh_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-            Requires<[HasV4T]>;
-
-def : Pat <(atomic_load_8 (add (HexagonCONST32_GP tglobaladdr:$global),
-                               u16ImmPred:$offset)),
-           (i32 (LDriub_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-           Requires<[HasV4T]>;
-
-// Map from load(globaladdress + x) -> memd(#foo + x)
-let AddedComplexity = 100 in
-def : Pat <(i64 (load (add (HexagonCONST32_GP tglobaladdr:$global),
-                           u16ImmPred:$offset))),
-           (i64 (LDrid_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-           Requires<[HasV4T]>;
-
-// Map from load(globaladdress + x) -> memb(#foo + x)
-let AddedComplexity = 100 in
-def : Pat <(i32 (extloadi8 (add (HexagonCONST32_GP tglobaladdr:$global),
-                           u16ImmPred:$offset))),
-           (i32 (LDrib_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-           Requires<[HasV4T]>;
-
-// Map from load(globaladdress + x) -> memb(#foo + x)
-let AddedComplexity = 100 in
-def : Pat <(i32 (sextloadi8 (add (HexagonCONST32_GP tglobaladdr:$global),
-                            u16ImmPred:$offset))),
-           (i32 (LDrib_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-           Requires<[HasV4T]>;
-
-// Map from load(globaladdress + x) -> memub(#foo + x)
-let AddedComplexity = 100 in
-def : Pat <(i32 (zextloadi8 (add (HexagonCONST32_GP tglobaladdr:$global),
-                            u16ImmPred:$offset))),
-           (i32 (LDriub_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-           Requires<[HasV4T]>;
-
-// Map from load(globaladdress + x) -> memuh(#foo + x)
-let AddedComplexity = 100 in
-def : Pat <(i32 (extloadi16 (add (HexagonCONST32_GP tglobaladdr:$global),
-                            u16ImmPred:$offset))),
-           (i32 (LDrih_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-            Requires<[HasV4T]>;
-
-// Map from load(globaladdress + x) -> memh(#foo + x)
-let AddedComplexity = 100 in
-def : Pat <(i32 (sextloadi16 (add (HexagonCONST32_GP tglobaladdr:$global),
-                             u16ImmPred:$offset))),
-           (i32 (LDrih_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-           Requires<[HasV4T]>;
-
-
-// Map from load(globaladdress + x) -> memuh(#foo + x)
-let AddedComplexity = 100 in
-def : Pat <(i32 (zextloadi16 (add (HexagonCONST32_GP tglobaladdr:$global),
-                             u16ImmPred:$offset))),
-           (i32 (LDriuh_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-            Requires<[HasV4T]>;
-
-// Map from load(globaladdress + x) -> memw(#foo + x)
-let AddedComplexity = 100 in
-def : Pat <(i32 (load (add (HexagonCONST32_GP tglobaladdr:$global),
-                      u16ImmPred:$offset))),
-           (i32 (LDriw_GP_V4 tglobaladdr:$global, u16ImmPred:$offset))>,
-            Requires<[HasV4T]>;
-
 
 //===----------------------------------------------------------------------===//
 // LD -
@@ -1746,70 +971,18 @@ def : Pat <(i32 (load (add (HexagonCONST32_GP tglobaladdr:$global),
 ///    last operand.
 ///
 
+// Store doubleword.
 // memd(Re=#U6)=Rtt
-def STrid_abs_setimm_V4 : STInst<(outs IntRegs:$dst1),
-            (ins DoubleRegs:$src1, u6Imm:$src2),
-            "memd($dst1=#$src2) = $src1",
-            []>,
-            Requires<[HasV4T]>;
-
-// memb(Re=#U6)=Rs
-def STrib_abs_setimm_V4 : STInst<(outs IntRegs:$dst1),
-            (ins IntRegs:$src1, u6Imm:$src2),
-            "memb($dst1=#$src2) = $src1",
-            []>,
-            Requires<[HasV4T]>;
-
-// memh(Re=#U6)=Rs
-def STrih_abs_setimm_V4 : STInst<(outs IntRegs:$dst1),
-            (ins IntRegs:$src1, u6Imm:$src2),
-            "memh($dst1=#$src2) = $src1",
-            []>,
-            Requires<[HasV4T]>;
-
-// memw(Re=#U6)=Rs
-def STriw_abs_setimm_V4 : STInst<(outs IntRegs:$dst1),
-            (ins IntRegs:$src1, u6Imm:$src2),
-            "memw($dst1=#$src2) = $src1",
-            []>,
-            Requires<[HasV4T]>;
-
-// memd(Re=#U6)=Rtt
-def STrid_abs_set_V4 : STInst<(outs IntRegs:$dst1),
-            (ins DoubleRegs:$src1, globaladdress:$src2),
-            "memd($dst1=##$src2) = $src1",
-            []>,
-            Requires<[HasV4T]>;
-
-// memb(Re=#U6)=Rs
-def STrib_abs_set_V4 : STInst<(outs IntRegs:$dst1),
-            (ins IntRegs:$src1, globaladdress:$src2),
-            "memb($dst1=##$src2) = $src1",
-            []>,
-            Requires<[HasV4T]>;
-
-// memh(Re=#U6)=Rs
-def STrih_abs_set_V4 : STInst<(outs IntRegs:$dst1),
-            (ins IntRegs:$src1, globaladdress:$src2),
-            "memh($dst1=##$src2) = $src1",
-            []>,
-            Requires<[HasV4T]>;
-
-// memw(Re=#U6)=Rs
-def STriw_abs_set_V4 : STInst<(outs IntRegs:$dst1),
-            (ins IntRegs:$src1, globaladdress:$src2),
-            "memw($dst1=##$src2) = $src1",
-            []>,
-            Requires<[HasV4T]>;
+// TODO: needs to be implemented
 
+// memd(Rs+#s11:3)=Rtt
 // memd(Rs+Ru<<#u2)=Rtt
 let AddedComplexity = 10, isPredicable = 1 in
 def STrid_indexed_shl_V4 : STInst<(outs),
             (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$src3, DoubleRegs:$src4),
             "memd($src1+$src2<<#$src3) = $src4",
-            [(store (i64 DoubleRegs:$src4),
-                    (add (i32 IntRegs:$src1),
-                         (shl (i32 IntRegs:$src2), u2ImmPred:$src3)))]>,
+            [(store DoubleRegs:$src4, (add IntRegs:$src1,
+                                      (shl IntRegs:$src2, u2ImmPred:$src3)))]>,
             Requires<[HasV4T]>;
 
 // memd(Ru<<#u2+#U6)=Rtt
@@ -1817,9 +990,9 @@ let AddedComplexity = 10 in
 def STrid_shl_V4 : STInst<(outs),
             (ins IntRegs:$src1, u2Imm:$src2, u6Imm:$src3, DoubleRegs:$src4),
             "memd($src1<<#$src2+#$src3) = $src4",
-            [(store (i64 DoubleRegs:$src4),
-                    (add (shl (i32 IntRegs:$src1), u2ImmPred:$src2),
-                         u6ImmPred:$src3))]>,
+            [(store DoubleRegs:$src4, (shl IntRegs:$src1,
+                                      (add u2ImmPred:$src2,
+                                           u6ImmPred:$src3)))]>,
             Requires<[HasV4T]>;
 
 // memd(Rx++#s4:3)=Rtt
@@ -1836,8 +1009,7 @@ def STrid_shl_V4 : STInst<(outs),
 // if ([!]Pv[.new]) memd(Rs+#u6:3)=Rtt
 // if (Pv) memd(Rs+#u6:3)=Rtt
 // if (Pv.new) memd(Rs+#u6:3)=Rtt
-let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1 in
 def STrid_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, MEMri:$addr, DoubleRegs:$src2),
             "if ($src1.new) memd($addr) = $src2",
@@ -1846,8 +1018,7 @@ def STrid_cdnPt_V4 : STInst<(outs),
 
 // if (!Pv) memd(Rs+#u6:3)=Rtt
 // if (!Pv.new) memd(Rs+#u6:3)=Rtt
-let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1 in
 def STrid_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, MEMri:$addr, DoubleRegs:$src2),
             "if (!$src1.new) memd($addr) = $src2",
@@ -1856,8 +1027,7 @@ def STrid_cdnNotPt_V4 : STInst<(outs),
 
 // if (Pv) memd(Rs+#u6:3)=Rtt
 // if (Pv.new) memd(Rs+#u6:3)=Rtt
-let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1 in
 def STrid_indexed_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_3Imm:$src3,
                  DoubleRegs:$src4),
@@ -1867,8 +1037,7 @@ def STrid_indexed_cdnPt_V4 : STInst<(outs),
 
 // if (!Pv) memd(Rs+#u6:3)=Rtt
 // if (!Pv.new) memd(Rs+#u6:3)=Rtt
-let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1 in
 def STrid_indexed_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_3Imm:$src3,
                  DoubleRegs:$src4),
@@ -1878,8 +1047,7 @@ def STrid_indexed_cdnNotPt_V4 : STInst<(outs),
 
 // if ([!]Pv[.new]) memd(Rs+Ru<<#u2)=Rtt
 // if (Pv) memd(Rs+Ru<<#u2)=Rtt
-let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1 in
 def STrid_indexed_shl_cPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  DoubleRegs:$src5),
@@ -1888,17 +1056,15 @@ def STrid_indexed_shl_cPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memd(Rs+Ru<<#u2)=Rtt
-let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1 in
 def STrid_indexed_shl_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  DoubleRegs:$src5),
-            "if ($src1.new) memd($src2+$src3<<#$src4) = $src5",
+            "if ($src1) memd($src2+$src3<<#$src4) = $src5",
             []>,
             Requires<[HasV4T]>;
 // if (!Pv) memd(Rs+Ru<<#u2)=Rtt
-let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1 in
 def STrid_indexed_shl_cNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  DoubleRegs:$src5),
@@ -1906,8 +1072,7 @@ def STrid_indexed_shl_cNotPt_V4 : STInst<(outs),
             []>,
             Requires<[HasV4T]>;
 // if (!Pv.new) memd(Rs+Ru<<#u2)=Rtt
-let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1 in
 def STrid_indexed_shl_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  DoubleRegs:$src5),
@@ -1918,8 +1083,7 @@ def STrid_indexed_shl_cdnNotPt_V4 : STInst<(outs),
 // if ([!]Pv[.new]) memd(Rx++#s4:3)=Rtt
 // if (Pv) memd(Rx++#s4:3)=Rtt
 // if (Pv.new) memd(Rx++#s4:3)=Rtt
-let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1 in
 def POST_STdri_cdnPt_V4 : STInstPI<(outs IntRegs:$dst),
             (ins PredRegs:$src1, DoubleRegs:$src2, IntRegs:$src3,
                  s4_3Imm:$offset),
@@ -1930,8 +1094,7 @@ def POST_STdri_cdnPt_V4 : STInstPI<(outs IntRegs:$dst),
 
 // if (!Pv) memd(Rx++#s4:3)=Rtt
 // if (!Pv.new) memd(Rx++#s4:3)=Rtt
-let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let AddedComplexity = 10, mayStore = 1, neverHasSideEffects = 1 in
 def POST_STdri_cdnNotPt_V4 : STInstPI<(outs IntRegs:$dst),
             (ins PredRegs:$src1, DoubleRegs:$src2, IntRegs:$src3,
                  s4_3Imm:$offset),
@@ -1942,12 +1105,15 @@ def POST_STdri_cdnNotPt_V4 : STInstPI<(outs IntRegs:$dst),
 
 
 // Store byte.
+// memb(Re=#U6)=Rt
+// TODO: needs to be implemented.
+// memb(Rs+#s11:0)=Rt
 // memb(Rs+#u6:0)=#S8
 let AddedComplexity = 10, isPredicable = 1 in
 def STrib_imm_V4 : STInst<(outs),
             (ins IntRegs:$src1, u6_0Imm:$src2, s8Imm:$src3),
             "memb($src1+#$src2) = #$src3",
-            [(truncstorei8 s8ImmPred:$src3, (add (i32 IntRegs:$src1),
+            [(truncstorei8 s8ImmPred:$src3, (add IntRegs:$src1,
                                                  u6_0ImmPred:$src2))]>,
             Requires<[HasV4T]>;
 
@@ -1956,10 +1122,9 @@ let AddedComplexity = 10, isPredicable = 1 in
 def STrib_indexed_shl_V4 : STInst<(outs),
             (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$src3, IntRegs:$src4),
             "memb($src1+$src2<<#$src3) = $src4",
-            [(truncstorei8 (i32 IntRegs:$src4),
-                           (add (i32 IntRegs:$src1),
-                                (shl (i32 IntRegs:$src2),
-                                          u2ImmPred:$src3)))]>,
+            [(truncstorei8 IntRegs:$src4, (add IntRegs:$src1,
+                                          (shl IntRegs:$src2,
+                                               u2ImmPred:$src3)))]>,
             Requires<[HasV4T]>;
 
 // memb(Ru<<#u2+#U6)=Rt
@@ -1967,9 +1132,9 @@ let AddedComplexity = 10 in
 def STrib_shl_V4 : STInst<(outs),
             (ins IntRegs:$src1, u2Imm:$src2, u6Imm:$src3, IntRegs:$src4),
             "memb($src1<<#$src2+#$src3) = $src4",
-            [(truncstorei8 (i32 IntRegs:$src4),
-                           (add (shl (i32 IntRegs:$src1), u2ImmPred:$src2),
-                                u6ImmPred:$src3))]>,
+            [(truncstorei8 IntRegs:$src4, (shl IntRegs:$src1,
+                                          (add u2ImmPred:$src2,
+                                               u6ImmPred:$src3)))]>,
             Requires<[HasV4T]>;
 
 // memb(Rx++#s4:0:circ(Mu))=Rt
@@ -1983,8 +1148,7 @@ def STrib_shl_V4 : STInst<(outs),
 // if ([!]Pv[.new]) memb(#u6)=Rt
 // if ([!]Pv[.new]) memb(Rs+#u6:0)=#S6
 // if (Pv) memb(Rs+#u6:0)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_imm_cPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_0Imm:$src3, s6Imm:$src4),
             "if ($src1) memb($src2+#$src3) = #$src4",
@@ -1992,8 +1156,7 @@ def STrib_imm_cPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memb(Rs+#u6:0)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_imm_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_0Imm:$src3, s6Imm:$src4),
             "if ($src1.new) memb($src2+#$src3) = #$src4",
@@ -2001,8 +1164,7 @@ def STrib_imm_cdnPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memb(Rs+#u6:0)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_imm_cNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_0Imm:$src3, s6Imm:$src4),
             "if (!$src1) memb($src2+#$src3) = #$src4",
@@ -2010,8 +1172,7 @@ def STrib_imm_cNotPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memb(Rs+#u6:0)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_imm_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_0Imm:$src3, s6Imm:$src4),
             "if (!$src1.new) memb($src2+#$src3) = #$src4",
@@ -2021,8 +1182,7 @@ def STrib_imm_cdnNotPt_V4 : STInst<(outs),
 // if ([!]Pv[.new]) memb(Rs+#u6:0)=Rt
 // if (Pv) memb(Rs+#u6:0)=Rt
 // if (Pv.new) memb(Rs+#u6:0)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if ($src1.new) memb($addr) = $src2",
@@ -2031,8 +1191,7 @@ def STrib_cdnPt_V4 : STInst<(outs),
 
 // if (!Pv) memb(Rs+#u6:0)=Rt
 // if (!Pv.new) memb(Rs+#u6:0)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if (!$src1.new) memb($addr) = $src2",
@@ -2042,8 +1201,7 @@ def STrib_cdnNotPt_V4 : STInst<(outs),
 // if (Pv) memb(Rs+#u6:0)=Rt
 // if (!Pv) memb(Rs+#u6:0)=Rt
 // if (Pv.new) memb(Rs+#u6:0)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_indexed_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_0Imm:$src3, IntRegs:$src4),
             "if ($src1.new) memb($src2+#$src3) = $src4",
@@ -2051,8 +1209,7 @@ def STrib_indexed_cdnPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memb(Rs+#u6:0)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_indexed_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_0Imm:$src3, IntRegs:$src4),
             "if (!$src1.new) memb($src2+#$src3) = $src4",
@@ -2061,8 +1218,7 @@ def STrib_indexed_cdnNotPt_V4 : STInst<(outs),
 
 // if ([!]Pv[.new]) memb(Rs+Ru<<#u2)=Rt
 // if (Pv) memb(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrib_indexed_shl_cPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2071,8 +1227,7 @@ def STrib_indexed_shl_cPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memb(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrib_indexed_shl_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2081,8 +1236,7 @@ def STrib_indexed_shl_cdnPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memb(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrib_indexed_shl_cNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2091,8 +1245,7 @@ def STrib_indexed_shl_cNotPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memb(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrib_indexed_shl_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2103,8 +1256,7 @@ def STrib_indexed_shl_cdnNotPt_V4 : STInst<(outs),
 // if ([!]Pv[.new]) memb(Rx++#s4:0)=Rt
 // if (Pv) memb(Rx++#s4:0)=Rt
 // if (Pv.new) memb(Rx++#s4:0)=Rt
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STbri_cdnPt_V4 : STInstPI<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_0Imm:$offset),
             "if ($src1.new) memb($src3++#$offset) = $src2",
@@ -2113,8 +1265,7 @@ def POST_STbri_cdnPt_V4 : STInstPI<(outs IntRegs:$dst),
 
 // if (!Pv) memb(Rx++#s4:0)=Rt
 // if (!Pv.new) memb(Rx++#s4:0)=Rt
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STbri_cdnNotPt_V4 : STInstPI<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_0Imm:$offset),
             "if (!$src1.new) memb($src3++#$offset) = $src2",
@@ -2123,15 +1274,20 @@ def POST_STbri_cdnNotPt_V4 : STInstPI<(outs IntRegs:$dst),
 
 
 // Store halfword.
-// TODO: needs to be implemented
 // memh(Re=#U6)=Rt.H
+// TODO: needs to be implemented
+
+// memh(Re=#U6)=Rt
+// TODO: needs to be implemented
+
 // memh(Rs+#s11:1)=Rt.H
+// memh(Rs+#s11:1)=Rt
 // memh(Rs+#u6:1)=#S8
 let AddedComplexity = 10, isPredicable = 1 in
 def STrih_imm_V4 : STInst<(outs),
             (ins IntRegs:$src1, u6_1Imm:$src2, s8Imm:$src3),
             "memh($src1+#$src2) = #$src3",
-            [(truncstorei16 s8ImmPred:$src3, (add (i32 IntRegs:$src1),
+            [(truncstorei16 s8ImmPred:$src3, (add IntRegs:$src1,
                                                   u6_1ImmPred:$src2))]>,
             Requires<[HasV4T]>;
 
@@ -2143,10 +1299,9 @@ let AddedComplexity = 10, isPredicable = 1 in
 def STrih_indexed_shl_V4 : STInst<(outs),
             (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$src3, IntRegs:$src4),
             "memh($src1+$src2<<#$src3) = $src4",
-            [(truncstorei16 (i32 IntRegs:$src4),
-                            (add (i32 IntRegs:$src1),
-                                 (shl (i32 IntRegs:$src2),
-                                      u2ImmPred:$src3)))]>,
+            [(truncstorei16 IntRegs:$src4, (add IntRegs:$src1,
+                                          (shl IntRegs:$src2,
+                                               u2ImmPred:$src3)))]>,
             Requires<[HasV4T]>;
 
 // memh(Ru<<#u2+#U6)=Rt.H
@@ -2155,9 +1310,9 @@ let AddedComplexity = 10 in
 def STrih_shl_V4 : STInst<(outs),
             (ins IntRegs:$src1, u2Imm:$src2, u6Imm:$src3, IntRegs:$src4),
             "memh($src1<<#$src2+#$src3) = $src4",
-            [(truncstorei16 (i32 IntRegs:$src4),
-                            (add (shl (i32 IntRegs:$src1), u2ImmPred:$src2),
-                                 u6ImmPred:$src3))]>,
+            [(truncstorei16 IntRegs:$src4, (shl IntRegs:$src1,
+                                          (add u2ImmPred:$src2,
+                                               u6ImmPred:$src3)))]>,
             Requires<[HasV4T]>;
 
 // memh(Rx++#s4:1:circ(Mu))=Rt.H
@@ -2168,14 +1323,17 @@ def STrih_shl_V4 : STInst<(outs),
 // memh(Rx++Mu)=Rt
 // memh(Rx++Mu:brev)=Rt.H
 // memh(Rx++Mu:brev)=Rt
+// memh(gp+#u16:1)=Rt.H
 // memh(gp+#u16:1)=Rt
+
+
+// Store halfword conditionally.
 // if ([!]Pv[.new]) memh(#u6)=Rt.H
 // if ([!]Pv[.new]) memh(#u6)=Rt
 
 // if ([!]Pv[.new]) memh(Rs+#u6:1)=#S6
 // if (Pv) memh(Rs+#u6:1)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_imm_cPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_1Imm:$src3, s6Imm:$src4),
             "if ($src1) memh($src2+#$src3) = #$src4",
@@ -2183,8 +1341,7 @@ def STrih_imm_cPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memh(Rs+#u6:1)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_imm_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_1Imm:$src3, s6Imm:$src4),
             "if ($src1.new) memh($src2+#$src3) = #$src4",
@@ -2192,8 +1349,7 @@ def STrih_imm_cdnPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memh(Rs+#u6:1)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_imm_cNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_1Imm:$src3, s6Imm:$src4),
             "if (!$src1) memh($src2+#$src3) = #$src4",
@@ -2201,8 +1357,7 @@ def STrih_imm_cNotPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memh(Rs+#u6:1)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_imm_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_1Imm:$src3, s6Imm:$src4),
             "if (!$src1.new) memh($src2+#$src3) = #$src4",
@@ -2215,8 +1370,7 @@ def STrih_imm_cdnNotPt_V4 : STInst<(outs),
 // if ([!]Pv[.new]) memh(Rs+#u6:1)=Rt
 // if (Pv) memh(Rs+#u6:1)=Rt
 // if (Pv.new) memh(Rs+#u6:1)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if ($src1.new) memh($addr) = $src2",
@@ -2225,8 +1379,7 @@ def STrih_cdnPt_V4 : STInst<(outs),
 
 // if (!Pv) memh(Rs+#u6:1)=Rt
 // if (!Pv.new) memh(Rs+#u6:1)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if (!$src1.new) memh($addr) = $src2",
@@ -2234,8 +1387,7 @@ def STrih_cdnNotPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memh(Rs+#u6:1)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_indexed_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_1Imm:$src3, IntRegs:$src4),
             "if ($src1.new) memh($src2+#$src3) = $src4",
@@ -2243,8 +1395,7 @@ def STrih_indexed_cdnPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memh(Rs+#u6:1)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_indexed_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_1Imm:$src3, IntRegs:$src4),
             "if (!$src1.new) memh($src2+#$src3) = $src4",
@@ -2254,8 +1405,7 @@ def STrih_indexed_cdnNotPt_V4 : STInst<(outs),
 // if ([!]Pv[.new]) memh(Rs+Ru<<#u2)=Rt.H
 // if ([!]Pv[.new]) memh(Rs+Ru<<#u2)=Rt
 // if (Pv) memh(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrih_indexed_shl_cPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2264,8 +1414,6 @@ def STrih_indexed_shl_cPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memh(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
 def STrih_indexed_shl_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2274,8 +1422,7 @@ def STrih_indexed_shl_cdnPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memh(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrih_indexed_shl_cNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2284,8 +1431,7 @@ def STrih_indexed_shl_cNotPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memh(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrih_indexed_shl_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2299,8 +1445,7 @@ def STrih_indexed_shl_cdnNotPt_V4 : STInst<(outs),
 // if ([!]Pv[.new]) memh(Rx++#s4:1)=Rt
 // if (Pv) memh(Rx++#s4:1)=Rt
 // if (Pv.new) memh(Rx++#s4:1)=Rt
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_SThri_cdnPt_V4 : STInstPI<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_1Imm:$offset),
             "if ($src1.new) memh($src3++#$offset) = $src2",
@@ -2309,8 +1454,7 @@ def POST_SThri_cdnPt_V4 : STInstPI<(outs IntRegs:$dst),
 
 // if (!Pv) memh(Rx++#s4:1)=Rt
 // if (!Pv.new) memh(Rx++#s4:1)=Rt
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_SThri_cdnNotPt_V4 : STInstPI<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_1Imm:$offset),
             "if (!$src1.new) memh($src3++#$offset) = $src2",
@@ -2322,22 +1466,13 @@ def POST_SThri_cdnNotPt_V4 : STInstPI<(outs IntRegs:$dst),
 // memw(Re=#U6)=Rt
 // TODO: Needs to be implemented.
 
-// Store predicate:
-let mayStore = 1, neverHasSideEffects = 1 in
-def STriw_pred_V4 : STInst<(outs),
-            (ins MEMri:$addr, PredRegs:$src1),
-            "Error; should not emit",
-            []>,
-            Requires<[HasV4T]>;
-
-
+// memw(Rs+#s11:2)=Rt
 // memw(Rs+#u6:2)=#S8
 let AddedComplexity = 10, isPredicable = 1 in
 def STriw_imm_V4 : STInst<(outs),
             (ins IntRegs:$src1, u6_2Imm:$src2, s8Imm:$src3),
             "memw($src1+#$src2) = #$src3",
-            [(store s8ImmPred:$src3, (add (i32 IntRegs:$src1),
-                                          u6_2ImmPred:$src2))]>,
+            [(store s8ImmPred:$src3, (add IntRegs:$src1, u6_2ImmPred:$src2))]>,
             Requires<[HasV4T]>;
 
 // memw(Rs+Ru<<#u2)=Rt
@@ -2345,9 +1480,8 @@ let AddedComplexity = 10, isPredicable = 1 in
 def STriw_indexed_shl_V4 : STInst<(outs),
             (ins IntRegs:$src1, IntRegs:$src2, u2Imm:$src3, IntRegs:$src4),
             "memw($src1+$src2<<#$src3) = $src4",
-            [(store (i32 IntRegs:$src4), (add (i32 IntRegs:$src1),
-                                    (shl (i32 IntRegs:$src2),
-                                         u2ImmPred:$src3)))]>,
+            [(store IntRegs:$src4, (add IntRegs:$src1,
+                                    (shl IntRegs:$src2, u2ImmPred:$src3)))]>,
             Requires<[HasV4T]>;
 
 // memw(Ru<<#u2+#U6)=Rt
@@ -2355,9 +1489,8 @@ let AddedComplexity = 10 in
 def STriw_shl_V4 : STInst<(outs),
             (ins IntRegs:$src1, u2Imm:$src2, u6Imm:$src3, IntRegs:$src4),
             "memw($src1<<#$src2+#$src3) = $src4",
-            [(store (i32 IntRegs:$src4),
-                    (add (shl (i32 IntRegs:$src1), u2ImmPred:$src2),
-                              u6ImmPred:$src3))]>,
+            [(store IntRegs:$src4, (shl IntRegs:$src1,
+                                   (add u2ImmPred:$src2, u6ImmPred:$src3)))]>,
             Requires<[HasV4T]>;
 
 // memw(Rx++#s4:2)=Rt
@@ -2369,11 +1502,12 @@ def STriw_shl_V4 : STInst<(outs),
 
 
 // Store word conditionally.
+// if ([!]Pv[.new]) memw(#u6)=Rt
+// TODO: Needs to be implemented.
 
 // if ([!]Pv[.new]) memw(Rs+#u6:2)=#S6
 // if (Pv) memw(Rs+#u6:2)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_imm_cPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_2Imm:$src3, s6Imm:$src4),
             "if ($src1) memw($src2+#$src3) = #$src4",
@@ -2381,8 +1515,7 @@ def STriw_imm_cPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memw(Rs+#u6:2)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_imm_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_2Imm:$src3, s6Imm:$src4),
             "if ($src1.new) memw($src2+#$src3) = #$src4",
@@ -2390,8 +1523,7 @@ def STriw_imm_cdnPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memw(Rs+#u6:2)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_imm_cNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_2Imm:$src3, s6Imm:$src4),
             "if (!$src1) memw($src2+#$src3) = #$src4",
@@ -2399,8 +1531,7 @@ def STriw_imm_cNotPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memw(Rs+#u6:2)=#S6
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_imm_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_2Imm:$src3, s6Imm:$src4),
             "if (!$src1.new) memw($src2+#$src3) = #$src4",
@@ -2410,8 +1541,7 @@ def STriw_imm_cdnNotPt_V4 : STInst<(outs),
 // if ([!]Pv[.new]) memw(Rs+#u6:2)=Rt
 // if (Pv) memw(Rs+#u6:2)=Rt
 // if (Pv.new) memw(Rs+#u6:2)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if ($src1.new) memw($addr) = $src2",
@@ -2420,8 +1550,7 @@ def STriw_cdnPt_V4 : STInst<(outs),
 
 // if (!Pv) memw(Rs+#u6:2)=Rt
 // if (!Pv.new) memw(Rs+#u6:2)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if (!$src1.new) memw($addr) = $src2",
@@ -2431,8 +1560,7 @@ def STriw_cdnNotPt_V4 : STInst<(outs),
 // if (Pv) memw(Rs+#u6:2)=Rt
 // if (!Pv) memw(Rs+#u6:2)=Rt
 // if (Pv.new) memw(Rs+#u6:2)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_indexed_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_2Imm:$src3, IntRegs:$src4),
             "if ($src1.new) memw($src2+#$src3) = $src4",
@@ -2440,8 +1568,7 @@ def STriw_indexed_cdnPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memw(Rs+#u6:2)=Rt
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_indexed_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_2Imm:$src3, IntRegs:$src4),
             "if (!$src1.new) memw($src2+#$src3) = $src4",
@@ -2450,8 +1577,7 @@ def STriw_indexed_cdnNotPt_V4 : STInst<(outs),
 
 // if ([!]Pv[.new]) memw(Rs+Ru<<#u2)=Rt
 // if (Pv) memw(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STriw_indexed_shl_cPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2460,8 +1586,7 @@ def STriw_indexed_shl_cPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memw(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STriw_indexed_shl_cdnPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2470,8 +1595,7 @@ def STriw_indexed_shl_cdnPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memw(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STriw_indexed_shl_cNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2480,8 +1604,7 @@ def STriw_indexed_shl_cNotPt_V4 : STInst<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memw(Rs+Ru<<#u2)=Rt
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STriw_indexed_shl_cdnNotPt_V4 : STInst<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -2492,8 +1615,7 @@ def STriw_indexed_shl_cdnNotPt_V4 : STInst<(outs),
 // if ([!]Pv[.new]) memw(Rx++#s4:2)=Rt
 // if (Pv) memw(Rx++#s4:2)=Rt
 // if (Pv.new) memw(Rx++#s4:2)=Rt
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STwri_cdnPt_V4 : STInstPI<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_2Imm:$offset),
             "if ($src1.new) memw($src3++#$offset) = $src2",
@@ -2502,8 +1624,7 @@ def POST_STwri_cdnPt_V4 : STInstPI<(outs IntRegs:$dst),
 
 // if (!Pv) memw(Rx++#s4:2)=Rt
 // if (!Pv.new) memw(Rx++#s4:2)=Rt
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STwri_cdnNotPt_V4 : STInstPI<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_2Imm:$offset),
             "if (!$src1.new) memw($src3++#$offset) = $src2",
@@ -2511,439 +1632,6 @@ def POST_STwri_cdnNotPt_V4 : STInstPI<(outs IntRegs:$dst),
             Requires<[HasV4T]>;
 
 
-/// store to global address
-
-let isPredicable = 1, mayStore = 1, neverHasSideEffects = 1 in
-def STrid_GP_V4 : STInst<(outs),
-            (ins globaladdress:$global, u16Imm:$offset, DoubleRegs:$src),
-            "memd(#$global+$offset) = $src",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrid_GP_cPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        DoubleRegs:$src2),
-            "if ($src1) memd(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrid_GP_cNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        DoubleRegs:$src2),
-            "if (!$src1) memd(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrid_GP_cdnPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        DoubleRegs:$src2),
-            "if ($src1.new) memd(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrid_GP_cdnNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        DoubleRegs:$src2),
-            "if (!$src1.new) memd(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let isPredicable = 1, mayStore = 1, neverHasSideEffects = 1 in
-def STrib_GP_V4 : STInst<(outs),
-            (ins globaladdress:$global, u16Imm:$offset, IntRegs:$src),
-            "memb(#$global+$offset) = $src",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrib_GP_cPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1) memb(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrib_GP_cNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1) memb(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrib_GP_cdnPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1.new) memb(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrib_GP_cdnNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1.new) memb(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let isPredicable = 1, mayStore = 1, neverHasSideEffects = 1 in
-def STrih_GP_V4 : STInst<(outs),
-            (ins globaladdress:$global, u16Imm:$offset, IntRegs:$src),
-            "memh(#$global+$offset) = $src",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrih_GP_cPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1) memh(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrih_GP_cNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1) memh(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrih_GP_cdnPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1.new) memh(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STrih_GP_cdnNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1.new) memh(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let isPredicable = 1, mayStore = 1, neverHasSideEffects = 1 in
-def STriw_GP_V4 : STInst<(outs),
-            (ins globaladdress:$global, u16Imm:$offset, IntRegs:$src),
-            "memw(#$global+$offset) = $src",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STriw_GP_cPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1) memw(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STriw_GP_cNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1) memw(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STriw_GP_cdnPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1.new) memw(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STriw_GP_cdnNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1.new) memw(##$global+$offset) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-// memd(#global)=Rtt
-let isPredicable = 1, mayStore = 1, neverHasSideEffects = 1 in
-def STd_GP_V4 : STInst<(outs),
-            (ins globaladdress:$global, DoubleRegs:$src),
-            "memd(#$global) = $src",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) memd(##global) = Rtt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STd_GP_cPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, DoubleRegs:$src2),
-            "if ($src1) memd(##$global) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) memd(##global) = Rtt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STd_GP_cNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, DoubleRegs:$src2),
-            "if (!$src1) memd(##$global) = $src2",
-            []>,
-              Requires<[HasV4T]>;
-
-// if (Pv) memd(##global) = Rtt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STd_GP_cdnPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, DoubleRegs:$src2),
-            "if ($src1.new) memd(##$global) = $src2",
-            []>,
-              Requires<[HasV4T]>;
-
-// if (!Pv) memd(##global) = Rtt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STd_GP_cdnNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, DoubleRegs:$src2),
-            "if (!$src1.new) memd(##$global) = $src2",
-            []>,
-            Requires<[HasV4T]>;
-
-// memb(#global)=Rt
-let isPredicable = 1, mayStore = 1, neverHasSideEffects = 1 in
-def STb_GP_V4 : STInst<(outs),
-            (ins globaladdress:$global, IntRegs:$src),
-            "memb(#$global) = $src",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) memb(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STb_GP_cPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1) memb(##$global) = $src2",
-              []>,
-              Requires<[HasV4T]>;
-
-// if (!Pv) memb(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STb_GP_cNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1) memb(##$global) = $src2",
-              []>,
-              Requires<[HasV4T]>;
-
-// if (Pv) memb(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STb_GP_cdnPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1.new) memb(##$global) = $src2",
-              []>,
-              Requires<[HasV4T]>;
-
-// if (!Pv) memb(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STb_GP_cdnNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1.new) memb(##$global) = $src2",
-              []>,
-              Requires<[HasV4T]>;
-
-// memh(#global)=Rt
-let isPredicable = 1, mayStore = 1, neverHasSideEffects = 1 in
-def STh_GP_V4 : STInst<(outs),
-            (ins globaladdress:$global, IntRegs:$src),
-            "memh(#$global) = $src",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) memh(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STh_GP_cPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1) memh(##$global) = $src2",
-              []>,
-              Requires<[HasV4T]>;
-
-// if (!Pv) memh(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STh_GP_cNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1) memh(##$global) = $src2",
-              []>,
-              Requires<[HasV4T]>;
-
-// if (Pv) memh(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STh_GP_cdnPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1.new) memh(##$global) = $src2",
-              []>,
-              Requires<[HasV4T]>;
-
-// if (!Pv) memh(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STh_GP_cdnNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1.new) memh(##$global) = $src2",
-              []>,
-              Requires<[HasV4T]>;
-
-// memw(#global)=Rt
-let isPredicable = 1, mayStore = 1, neverHasSideEffects = 1 in
-def STw_GP_V4 : STInst<(outs),
-            (ins globaladdress:$global, IntRegs:$src),
-            "memw(#$global) = $src",
-              []>,
-              Requires<[HasV4T]>;
-
-// if (Pv) memw(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STw_GP_cPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1) memw(##$global) = $src2",
-              []>,
-              Requires<[HasV4T]>;
-
-// if (!Pv) memw(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STw_GP_cNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1) memw(##$global) = $src2",
-              []>,
-              Requires<[HasV4T]>;
-
-// if (Pv) memw(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STw_GP_cdnPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1.new) memw(##$global) = $src2",
-              []>,
-              Requires<[HasV4T]>;
-
-// if (!Pv) memw(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1, isPredicated = 1 in
-def STw_GP_cdnNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1.new) memw(##$global) = $src2",
-            []>,
-              Requires<[HasV4T]>;
-
-// 64 bit atomic store
-def : Pat <(atomic_store_64 (HexagonCONST32_GP tglobaladdr:$global),
-                            (i64 DoubleRegs:$src1)),
-           (STd_GP_V4 tglobaladdr:$global, (i64 DoubleRegs:$src1))>,
-           Requires<[HasV4T]>;
-
-// Map from store(globaladdress) -> memd(#foo)
-let AddedComplexity = 100 in
-def : Pat <(store (i64 DoubleRegs:$src1), (HexagonCONST32_GP tglobaladdr:$global)),
-           (STd_GP_V4 tglobaladdr:$global, (i64 DoubleRegs:$src1))>,
-           Requires<[HasV4T]>;
-
-// 8 bit atomic store
-def : Pat < (atomic_store_8 (HexagonCONST32_GP tglobaladdr:$global),
-                            (i32 IntRegs:$src1)),
-            (STb_GP_V4 tglobaladdr:$global, (i32 IntRegs:$src1))>,
-              Requires<[HasV4T]>;
-
-// Map from store(globaladdress) -> memb(#foo)
-let AddedComplexity = 100 in
-def : Pat<(truncstorei8 (i32 IntRegs:$src1),
-          (HexagonCONST32_GP tglobaladdr:$global)),
-          (STb_GP_V4 tglobaladdr:$global, (i32 IntRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-// Map from "i1 = constant<-1>; memw(CONST32(#foo)) = i1"
-//       to "r0 = 1; memw(#foo) = r0"
-let AddedComplexity = 100 in
-def : Pat<(store (i1 -1), (HexagonCONST32_GP tglobaladdr:$global)),
-          (STb_GP_V4 tglobaladdr:$global, (TFRI 1))>,
-          Requires<[HasV4T]>;
-
-def : Pat<(atomic_store_16 (HexagonCONST32_GP tglobaladdr:$global),
-                           (i32 IntRegs:$src1)),
-          (STh_GP_V4 tglobaladdr:$global, (i32 IntRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-// Map from store(globaladdress) -> memh(#foo)
-let AddedComplexity = 100 in
-def : Pat<(truncstorei16 (i32 IntRegs:$src1),
-                         (HexagonCONST32_GP tglobaladdr:$global)),
-          (STh_GP_V4 tglobaladdr:$global, (i32 IntRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-// 32 bit atomic store
-def : Pat<(atomic_store_32 (HexagonCONST32_GP tglobaladdr:$global),
-                           (i32 IntRegs:$src1)),
-          (STw_GP_V4 tglobaladdr:$global, (i32 IntRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-// Map from store(globaladdress) -> memw(#foo)
-let AddedComplexity = 100 in
-def : Pat<(store (i32 IntRegs:$src1), (HexagonCONST32_GP tglobaladdr:$global)),
-          (STw_GP_V4 tglobaladdr:$global, (i32 IntRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-def : Pat<(atomic_store_64 (add (HexagonCONST32_GP tglobaladdr:$global),
-                                u16ImmPred:$offset),
-                           (i64 DoubleRegs:$src1)),
-          (STrid_GP_V4 tglobaladdr:$global, u16ImmPred:$offset,
-                                            (i64 DoubleRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-def : Pat<(atomic_store_32 (add (HexagonCONST32_GP tglobaladdr:$global),
-                                u16ImmPred:$offset),
-                           (i32 IntRegs:$src1)),
-          (STriw_GP_V4 tglobaladdr:$global, u16ImmPred:$offset, (i32 IntRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-def : Pat<(atomic_store_16 (add (HexagonCONST32_GP tglobaladdr:$global),
-                                u16ImmPred:$offset),
-                           (i32 IntRegs:$src1)),
-          (STrih_GP_V4 tglobaladdr:$global, u16ImmPred:$offset, (i32 IntRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-def : Pat<(atomic_store_8 (add (HexagonCONST32_GP tglobaladdr:$global),
-                               u16ImmPred:$offset),
-                          (i32 IntRegs:$src1)),
-          (STrib_GP_V4 tglobaladdr:$global, u16ImmPred:$offset, (i32 IntRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-// Map from store(globaladdress + x) -> memd(#foo + x)
-let AddedComplexity = 100 in
-def : Pat<(store (i64 DoubleRegs:$src1), (add (HexagonCONST32_GP tglobaladdr:$global),
-                                        u16ImmPred:$offset)),
-          (STrid_GP_V4 tglobaladdr:$global, u16ImmPred:$offset,
-                                            (i64 DoubleRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-// Map from store(globaladdress + x) -> memb(#foo + x)
-let AddedComplexity = 100 in
-def : Pat<(truncstorei8 (i32 IntRegs:$src1),
-                        (add (HexagonCONST32_GP tglobaladdr:$global),
-                             u16ImmPred:$offset)),
-          (STrib_GP_V4 tglobaladdr:$global, u16ImmPred:$offset, (i32 IntRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-// Map from store(globaladdress + x) -> memh(#foo + x)
-let AddedComplexity = 100 in
-def : Pat<(truncstorei16 (i32 IntRegs:$src1),
-                         (add (HexagonCONST32_GP tglobaladdr:$global),
-                              u16ImmPred:$offset)),
-          (STrih_GP_V4 tglobaladdr:$global, u16ImmPred:$offset, (i32 IntRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-// Map from store(globaladdress + x) -> memw(#foo + x)
-let AddedComplexity = 100 in
-def : Pat<(store (i32 IntRegs:$src1),
-                 (add (HexagonCONST32_GP tglobaladdr:$global),
-                                u16ImmPred:$offset)),
-          (STriw_GP_V4 tglobaladdr:$global, u16ImmPred:$offset, (i32 IntRegs:$src1))>,
-          Requires<[HasV4T]>;
-
-
-
 //===----------------------------------------------------------------------===
 // ST -
 //===----------------------------------------------------------------------===
@@ -3008,19 +1696,11 @@ def STrib_GP_nv_V4 : NVInst_V4<(outs),
             []>,
             Requires<[HasV4T]>;
 
-// memb(#global)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1 in
-def STb_GP_nv_V4 : NVInst_V4<(outs),
-            (ins globaladdress:$global, IntRegs:$src),
-            "memb(#$global) = $src.new",
-            []>,
-            Requires<[HasV4T]>;
 
 // Store new-value byte conditionally.
 // if ([!]Pv[.new]) memb(#u6)=Nt.new
 // if (Pv) memb(Rs+#u6:0)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_cPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if ($src1) memb($addr) = $src2.new",
@@ -3028,8 +1708,7 @@ def STrib_cPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memb(Rs+#u6:0)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_cdnPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if ($src1.new) memb($addr) = $src2.new",
@@ -3037,8 +1716,7 @@ def STrib_cdnPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memb(Rs+#u6:0)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_cNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if (!$src1) memb($addr) = $src2.new",
@@ -3046,8 +1724,7 @@ def STrib_cNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memb(Rs+#u6:0)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if (!$src1.new) memb($addr) = $src2.new",
@@ -3055,8 +1732,7 @@ def STrib_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv) memb(Rs+#u6:0)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_indexed_cPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_0Imm:$src3, IntRegs:$src4),
             "if ($src1) memb($src2+#$src3) = $src4.new",
@@ -3064,8 +1740,7 @@ def STrib_indexed_cPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memb(Rs+#u6:0)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_indexed_cdnPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_0Imm:$src3, IntRegs:$src4),
             "if ($src1.new) memb($src2+#$src3) = $src4.new",
@@ -3073,8 +1748,7 @@ def STrib_indexed_cdnPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memb(Rs+#u6:0)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_indexed_cNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_0Imm:$src3, IntRegs:$src4),
             "if (!$src1) memb($src2+#$src3) = $src4.new",
@@ -3082,8 +1756,7 @@ def STrib_indexed_cNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memb(Rs+#u6:0)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrib_indexed_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_0Imm:$src3, IntRegs:$src4),
             "if (!$src1.new) memb($src2+#$src3) = $src4.new",
@@ -3093,8 +1766,7 @@ def STrib_indexed_cdnNotPt_nv_V4 : NVInst_V4<(outs),
 
 // if ([!]Pv[.new]) memb(Rs+Ru<<#u2)=Nt.new
 // if (Pv) memb(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrib_indexed_shl_cPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3103,8 +1775,7 @@ def STrib_indexed_shl_cPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memb(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrib_indexed_shl_cdnPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3113,8 +1784,7 @@ def STrib_indexed_shl_cdnPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memb(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrib_indexed_shl_cNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3123,8 +1793,7 @@ def STrib_indexed_shl_cNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memb(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrib_indexed_shl_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3134,8 +1803,7 @@ def STrib_indexed_shl_cdnNotPt_nv_V4 : NVInst_V4<(outs),
 
 // if ([!]Pv[.new]) memb(Rx++#s4:0)=Nt.new
 // if (Pv) memb(Rx++#s4:0)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STbri_cPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_0Imm:$offset),
             "if ($src1) memb($src3++#$offset) = $src2.new",
@@ -3143,8 +1811,7 @@ def POST_STbri_cPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memb(Rx++#s4:0)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STbri_cdnPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_0Imm:$offset),
             "if ($src1.new) memb($src3++#$offset) = $src2.new",
@@ -3152,8 +1819,7 @@ def POST_STbri_cdnPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             Requires<[HasV4T]>;
 
 // if (!Pv) memb(Rx++#s4:0)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STbri_cNotPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_0Imm:$offset),
             "if (!$src1) memb($src3++#$offset) = $src2.new",
@@ -3161,8 +1827,7 @@ def POST_STbri_cNotPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memb(Rx++#s4:0)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STbri_cdnNotPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_0Imm:$offset),
             "if (!$src1.new) memb($src3++#$offset) = $src2.new",
@@ -3224,14 +1889,6 @@ def STrih_GP_nv_V4 : NVInst_V4<(outs),
             []>,
             Requires<[HasV4T]>;
 
-// memh(#global)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1 in
-def STh_GP_nv_V4 : NVInst_V4<(outs),
-            (ins globaladdress:$global, IntRegs:$src),
-            "memh(#$global) = $src.new",
-            []>,
-            Requires<[HasV4T]>;
-
 
 // Store new-value halfword conditionally.
 
@@ -3239,8 +1896,7 @@ def STh_GP_nv_V4 : NVInst_V4<(outs),
 
 // if ([!]Pv[.new]) memh(Rs+#u6:1)=Nt.new
 // if (Pv) memh(Rs+#u6:1)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_cPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if ($src1) memh($addr) = $src2.new",
@@ -3248,8 +1904,7 @@ def STrih_cPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memh(Rs+#u6:1)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_cdnPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if ($src1.new) memh($addr) = $src2.new",
@@ -3257,8 +1912,7 @@ def STrih_cdnPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memh(Rs+#u6:1)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_cNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if (!$src1) memh($addr) = $src2.new",
@@ -3266,8 +1920,7 @@ def STrih_cNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memh(Rs+#u6:1)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if (!$src1.new) memh($addr) = $src2.new",
@@ -3275,8 +1928,7 @@ def STrih_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv) memh(Rs+#u6:1)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_indexed_cPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_1Imm:$src3, IntRegs:$src4),
             "if ($src1) memh($src2+#$src3) = $src4.new",
@@ -3284,8 +1936,7 @@ def STrih_indexed_cPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memh(Rs+#u6:1)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_indexed_cdnPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_1Imm:$src3, IntRegs:$src4),
             "if ($src1.new) memh($src2+#$src3) = $src4.new",
@@ -3293,8 +1944,7 @@ def STrih_indexed_cdnPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memh(Rs+#u6:1)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_indexed_cNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_1Imm:$src3, IntRegs:$src4),
             "if (!$src1) memh($src2+#$src3) = $src4.new",
@@ -3302,8 +1952,7 @@ def STrih_indexed_cNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memh(Rs+#u6:1)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STrih_indexed_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_1Imm:$src3, IntRegs:$src4),
             "if (!$src1.new) memh($src2+#$src3) = $src4.new",
@@ -3312,8 +1961,7 @@ def STrih_indexed_cdnNotPt_nv_V4 : NVInst_V4<(outs),
 
 // if ([!]Pv[.new]) memh(Rs+Ru<<#u2)=Nt.new
 // if (Pv) memh(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrih_indexed_shl_cPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3322,8 +1970,7 @@ def STrih_indexed_shl_cPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memh(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrih_indexed_shl_cdnPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3332,8 +1979,7 @@ def STrih_indexed_shl_cdnPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memh(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrih_indexed_shl_cNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3342,8 +1988,7 @@ def STrih_indexed_shl_cNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memh(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STrih_indexed_shl_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3353,8 +1998,7 @@ def STrih_indexed_shl_cdnNotPt_nv_V4 : NVInst_V4<(outs),
 
 // if ([!]Pv[]) memh(Rx++#s4:1)=Nt.new
 // if (Pv) memh(Rx++#s4:1)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_SThri_cPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_1Imm:$offset),
             "if ($src1) memh($src3++#$offset) = $src2.new",
@@ -3362,8 +2006,7 @@ def POST_SThri_cPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memh(Rx++#s4:1)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_SThri_cdnPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_1Imm:$offset),
             "if ($src1.new) memh($src3++#$offset) = $src2.new",
@@ -3371,8 +2014,7 @@ def POST_SThri_cdnPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             Requires<[HasV4T]>;
 
 // if (!Pv) memh(Rx++#s4:1)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_SThri_cNotPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_1Imm:$offset),
             "if (!$src1) memh($src3++#$offset) = $src2.new",
@@ -3380,8 +2022,7 @@ def POST_SThri_cNotPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memh(Rx++#s4:1)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_SThri_cdnNotPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_1Imm:$offset),
             "if (!$src1.new) memh($src3++#$offset) = $src2.new",
@@ -3444,12 +2085,6 @@ def STriw_GP_nv_V4 : NVInst_V4<(outs),
             []>,
             Requires<[HasV4T]>;
 
-let mayStore = 1, neverHasSideEffects = 1 in
-def STw_GP_nv_V4 : NVInst_V4<(outs),
-            (ins globaladdress:$global, IntRegs:$src),
-            "memw(#$global) = $src.new",
-            []>,
-            Requires<[HasV4T]>;
 
 // Store new-value word conditionally.
 
@@ -3457,8 +2092,7 @@ def STw_GP_nv_V4 : NVInst_V4<(outs),
 
 // if ([!]Pv[.new]) memw(Rs+#u6:2)=Nt.new
 // if (Pv) memw(Rs+#u6:2)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_cPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if ($src1) memw($addr) = $src2.new",
@@ -3466,8 +2100,7 @@ def STriw_cPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memw(Rs+#u6:2)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_cdnPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if ($src1.new) memw($addr) = $src2.new",
@@ -3475,8 +2108,7 @@ def STriw_cdnPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memw(Rs+#u6:2)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_cNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if (!$src1) memw($addr) = $src2.new",
@@ -3484,8 +2116,7 @@ def STriw_cNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memw(Rs+#u6:2)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, MEMri:$addr, IntRegs:$src2),
             "if (!$src1.new) memw($addr) = $src2.new",
@@ -3493,8 +2124,7 @@ def STriw_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv) memw(Rs+#u6:2)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_indexed_cPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_2Imm:$src3, IntRegs:$src4),
             "if ($src1) memw($src2+#$src3) = $src4.new",
@@ -3502,8 +2132,7 @@ def STriw_indexed_cPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memw(Rs+#u6:2)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_indexed_cdnPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_2Imm:$src3, IntRegs:$src4),
             "if ($src1.new) memw($src2+#$src3) = $src4.new",
@@ -3511,8 +2140,7 @@ def STriw_indexed_cdnPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memw(Rs+#u6:2)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_indexed_cNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_2Imm:$src3, IntRegs:$src4),
             "if (!$src1) memw($src2+#$src3) = $src4.new",
@@ -3520,8 +2148,7 @@ def STriw_indexed_cNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memw(Rs+#u6:2)=Nt.new
-let mayStore = 1, neverHasSideEffects = 1,
-    isPredicated = 1 in
+let mayStore = 1, neverHasSideEffects = 1 in
 def STriw_indexed_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, u6_2Imm:$src3, IntRegs:$src4),
             "if (!$src1.new) memw($src2+#$src3) = $src4.new",
@@ -3531,8 +2158,7 @@ def STriw_indexed_cdnNotPt_nv_V4 : NVInst_V4<(outs),
 
 // if ([!]Pv[.new]) memw(Rs+Ru<<#u2)=Nt.new
 // if (Pv) memw(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STriw_indexed_shl_cPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3541,8 +2167,7 @@ def STriw_indexed_shl_cPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memw(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STriw_indexed_shl_cdnPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3551,8 +2176,7 @@ def STriw_indexed_shl_cdnPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv) memw(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STriw_indexed_shl_cNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3561,8 +2185,7 @@ def STriw_indexed_shl_cNotPt_nv_V4 : NVInst_V4<(outs),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memw(Rs+Ru<<#u2)=Nt.new
-let mayStore = 1, AddedComplexity = 10,
-    isPredicated = 1 in
+let mayStore = 1, AddedComplexity = 10 in
 def STriw_indexed_shl_cdnNotPt_nv_V4 : NVInst_V4<(outs),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, u2Imm:$src4,
                  IntRegs:$src5),
@@ -3572,8 +2195,7 @@ def STriw_indexed_shl_cdnNotPt_nv_V4 : NVInst_V4<(outs),
 
 // if ([!]Pv[.new]) memw(Rx++#s4:2)=Nt.new
 // if (Pv) memw(Rx++#s4:2)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STwri_cPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_2Imm:$offset),
             "if ($src1) memw($src3++#$offset) = $src2.new",
@@ -3581,8 +2203,7 @@ def POST_STwri_cPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             Requires<[HasV4T]>;
 
 // if (Pv.new) memw(Rx++#s4:2)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STwri_cdnPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_2Imm:$offset),
             "if ($src1.new) memw($src3++#$offset) = $src2.new",
@@ -3590,8 +2211,7 @@ def POST_STwri_cdnPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             Requires<[HasV4T]>;
 
 // if (!Pv) memw(Rx++#s4:2)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STwri_cNotPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_2Imm:$offset),
             "if (!$src1) memw($src3++#$offset) = $src2.new",
@@ -3599,8 +2219,7 @@ def POST_STwri_cNotPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             Requires<[HasV4T]>;
 
 // if (!Pv.new) memw(Rx++#s4:2)=Nt.new
-let mayStore = 1, hasCtrlDep = 1,
-    isPredicated = 1 in
+let mayStore = 1, hasCtrlDep = 1 in
 def POST_STwri_cdnNotPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             (ins PredRegs:$src1, IntRegs:$src2, IntRegs:$src3, s4_2Imm:$offset),
             "if (!$src1.new) memw($src3++#$offset) = $src2.new",
@@ -3608,199 +2227,6 @@ def POST_STwri_cdnNotPt_nv_V4 : NVInstPI_V4<(outs IntRegs:$dst),
             Requires<[HasV4T]>;
 
 
-
-// if (Pv) memb(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STb_GP_cPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1) memb(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) memb(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STb_GP_cNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1) memb(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) memb(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STb_GP_cdnPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1.new) memb(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) memb(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STb_GP_cdnNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1.new) memb(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) memh(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STh_GP_cPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1) memh(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) memh(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STh_GP_cNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1) memh(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) memh(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STh_GP_cdnPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1.new) memh(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) memh(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STh_GP_cdnNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1.new) memh(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) memw(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STw_GP_cPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1) memw(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) memw(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STw_GP_cNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1) memw(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (Pv) memw(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STw_GP_cdnPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if ($src1.new) memw(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-// if (!Pv) memw(##global) = Rt
-let mayStore = 1, neverHasSideEffects = 1 in
-def STw_GP_cdnNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, IntRegs:$src2),
-            "if (!$src1.new) memw(##$global) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STrib_GP_cPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1) memb(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STrib_GP_cNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1) memb(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STrib_GP_cdnPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1.new) memb(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STrib_GP_cdnNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1.new) memb(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STrih_GP_cPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1) memh(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STrih_GP_cNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1) memh(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STrih_GP_cdnPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1.new) memh(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STrih_GP_cdnNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1.new) memh(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STriw_GP_cPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1) memw(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STriw_GP_cNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1) memw(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STriw_GP_cdnPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if ($src1.new) memw(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
-let mayStore = 1, neverHasSideEffects = 1 in
-def STriw_GP_cdnNotPt_nv_V4 : NVInst_V4<(outs),
-            (ins PredRegs:$src1, globaladdress:$global, u16Imm:$offset,
-                                                        IntRegs:$src2),
-            "if (!$src1.new) memw(##$global+$offset) = $src2.new",
-            []>,
-            Requires<[HasV4T]>;
-
 //===----------------------------------------------------------------------===//
 // NV/ST -
 //===----------------------------------------------------------------------===//
@@ -3990,18 +2416,16 @@ let isBranch = 1, isTerminator=1, neverHasSideEffects = 1, Defs = [PC] in {
 def ADDr_ADDri_V4 : MInst<(outs IntRegs:$dst),
           (ins IntRegs:$src1, IntRegs:$src2, s6Imm:$src3),
           "$dst = add($src1, add($src2, #$src3))",
-          [(set (i32 IntRegs:$dst),
-           (add (i32 IntRegs:$src1), (add (i32 IntRegs:$src2),
-                                          s6ImmPred:$src3)))]>,
+          [(set IntRegs:$dst,
+           (add IntRegs:$src1, (add IntRegs:$src2, s6ImmPred:$src3)))]>,
           Requires<[HasV4T]>;
 
 //  Rd=add(Rs,sub(#s6,Ru))
 def ADDr_SUBri_V4 : MInst<(outs IntRegs:$dst),
           (ins IntRegs:$src1, s6Imm:$src2, IntRegs:$src3),
           "$dst = add($src1, sub(#$src2, $src3))",
-          [(set (i32 IntRegs:$dst),
-           (add (i32 IntRegs:$src1), (sub s6ImmPred:$src2,
-                                          (i32 IntRegs:$src3))))]>,
+          [(set IntRegs:$dst,
+           (add IntRegs:$src1, (sub s6ImmPred:$src2, IntRegs:$src3)))]>,
           Requires<[HasV4T]>;
 
 // Generates the same instruction as ADDr_SUBri_V4 but matches different
@@ -4010,9 +2434,8 @@ def ADDr_SUBri_V4 : MInst<(outs IntRegs:$dst),
 def ADDri_SUBr_V4 : MInst<(outs IntRegs:$dst),
           (ins IntRegs:$src1, s6Imm:$src2, IntRegs:$src3),
           "$dst = add($src1, sub(#$src2, $src3))",
-          [(set (i32 IntRegs:$dst),
-                (sub (add (i32 IntRegs:$src1), s6ImmPred:$src2),
-                     (i32 IntRegs:$src3)))]>,
+          [(set IntRegs:$dst,
+           (sub (add IntRegs:$src1, s6ImmPred:$src2), IntRegs:$src3))]>,
           Requires<[HasV4T]>;
 
 
@@ -4028,16 +2451,16 @@ def ADDri_SUBr_V4 : MInst<(outs IntRegs:$dst),
 def ANDd_NOTd_V4 : MInst<(outs DoubleRegs:$dst),
           (ins DoubleRegs:$src1, DoubleRegs:$src2),
           "$dst = and($src1, ~$src2)",
-          [(set (i64 DoubleRegs:$dst), (and (i64 DoubleRegs:$src1),
-                                      (not (i64 DoubleRegs:$src2))))]>,
+          [(set DoubleRegs:$dst, (and DoubleRegs:$src1,
+                                      (not DoubleRegs:$src2)))]>,
           Requires<[HasV4T]>;
 
 //  Rdd=or(Rtt,~Rss)
 def ORd_NOTd_V4 : MInst<(outs DoubleRegs:$dst),
           (ins DoubleRegs:$src1, DoubleRegs:$src2),
           "$dst = or($src1, ~$src2)",
-          [(set (i64 DoubleRegs:$dst),
-           (or (i64 DoubleRegs:$src1), (not (i64 DoubleRegs:$src2))))]>,
+          [(set DoubleRegs:$dst,
+           (or DoubleRegs:$src1, (not DoubleRegs:$src2)))]>,
           Requires<[HasV4T]>;
 
 
@@ -4046,9 +2469,8 @@ def ORd_NOTd_V4 : MInst<(outs DoubleRegs:$dst),
 def XORd_XORdd: MInst_acc<(outs DoubleRegs:$dst),
           (ins DoubleRegs:$src1, DoubleRegs:$src2, DoubleRegs:$src3),
           "$dst ^= xor($src2, $src3)",
-          [(set (i64 DoubleRegs:$dst),
-           (xor (i64 DoubleRegs:$src1), (xor (i64 DoubleRegs:$src2),
-                                             (i64 DoubleRegs:$src3))))],
+          [(set DoubleRegs:$dst,
+           (xor DoubleRegs:$src1, (xor DoubleRegs:$src2, DoubleRegs:$src3)))],
           "$src1 = $dst">,
           Requires<[HasV4T]>;
 
@@ -4058,9 +2480,8 @@ def XORd_XORdd: MInst_acc<(outs DoubleRegs:$dst),
 def ORr_ANDri_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, s10Imm:$src3),
             "$dst = or($src1, and($src2, #$src3))",
-            [(set (i32 IntRegs:$dst),
-                  (or (i32 IntRegs:$src1), (and (i32 IntRegs:$src2),
-                                                s10ImmPred:$src3)))],
+            [(set IntRegs:$dst,
+             (or IntRegs:$src1, (and IntRegs:$src2, s10ImmPred:$src3)))],
             "$src2 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4069,9 +2490,8 @@ def ORr_ANDri_V4 : MInst_acc<(outs IntRegs:$dst),
 def ANDr_ANDrr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst &= and($src2, $src3)",
-            [(set (i32 IntRegs:$dst),
-                  (and (i32 IntRegs:$src1), (and (i32 IntRegs:$src2),
-                                                 (i32 IntRegs:$src3))))],
+            [(set IntRegs:$dst,
+             (and IntRegs:$src1, (and IntRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4079,9 +2499,8 @@ def ANDr_ANDrr_V4 : MInst_acc<(outs IntRegs:$dst),
 def ORr_ANDrr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst |= and($src2, $src3)",
-            [(set (i32 IntRegs:$dst),
-                  (or (i32 IntRegs:$src1), (and (i32 IntRegs:$src2),
-                                                (i32 IntRegs:$src3))))],
+            [(set IntRegs:$dst,
+             (or IntRegs:$src1, (and IntRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4089,9 +2508,8 @@ def ORr_ANDrr_V4 : MInst_acc<(outs IntRegs:$dst),
 def XORr_ANDrr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst ^= and($src2, $src3)",
-            [(set (i32 IntRegs:$dst),
-             (xor (i32 IntRegs:$src1), (and (i32 IntRegs:$src2),
-                                            (i32 IntRegs:$src3))))],
+            [(set IntRegs:$dst,
+             (xor IntRegs:$src1, (and IntRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4100,9 +2518,8 @@ def XORr_ANDrr_V4 : MInst_acc<(outs IntRegs:$dst),
 def ANDr_ANDr_NOTr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst &= and($src2, ~$src3)",
-            [(set (i32 IntRegs:$dst),
-                  (and (i32 IntRegs:$src1), (and (i32 IntRegs:$src2),
-                                                 (not (i32 IntRegs:$src3)))))],
+            [(set IntRegs:$dst,
+             (and IntRegs:$src1, (and IntRegs:$src2, (not IntRegs:$src3))))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4110,9 +2527,8 @@ def ANDr_ANDr_NOTr_V4 : MInst_acc<(outs IntRegs:$dst),
 def ORr_ANDr_NOTr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst |= and($src2, ~$src3)",
-            [(set (i32 IntRegs:$dst),
-             (or (i32 IntRegs:$src1), (and (i32 IntRegs:$src2),
-                                           (not (i32 IntRegs:$src3)))))],
+            [(set IntRegs:$dst,
+             (or IntRegs:$src1, (and IntRegs:$src2, (not IntRegs:$src3))))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4120,9 +2536,8 @@ def ORr_ANDr_NOTr_V4 : MInst_acc<(outs IntRegs:$dst),
 def XORr_ANDr_NOTr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst ^= and($src2, ~$src3)",
-            [(set (i32 IntRegs:$dst),
-             (xor (i32 IntRegs:$src1), (and (i32 IntRegs:$src2),
-                                            (not (i32 IntRegs:$src3)))))],
+            [(set IntRegs:$dst,
+             (xor IntRegs:$src1, (and IntRegs:$src2, (not IntRegs:$src3))))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4131,9 +2546,8 @@ def XORr_ANDr_NOTr_V4 : MInst_acc<(outs IntRegs:$dst),
 def ANDr_ORrr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst &= or($src2, $src3)",
-            [(set (i32 IntRegs:$dst),
-                  (and (i32 IntRegs:$src1), (or (i32 IntRegs:$src2),
-                                                (i32 IntRegs:$src3))))],
+            [(set IntRegs:$dst,
+             (and IntRegs:$src1, (or IntRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4141,9 +2555,8 @@ def ANDr_ORrr_V4 : MInst_acc<(outs IntRegs:$dst),
 def ORr_ORrr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst |= or($src2, $src3)",
-            [(set (i32 IntRegs:$dst),
-                  (or (i32 IntRegs:$src1), (or (i32 IntRegs:$src2),
-                                               (i32 IntRegs:$src3))))],
+            [(set IntRegs:$dst,
+             (or IntRegs:$src1, (or IntRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4151,9 +2564,8 @@ def ORr_ORrr_V4 : MInst_acc<(outs IntRegs:$dst),
 def XORr_ORrr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst ^= or($src2, $src3)",
-            [(set (i32 IntRegs:$dst),
-             (xor (i32 IntRegs:$src1), (or (i32 IntRegs:$src2),
-                                           (i32 IntRegs:$src3))))],
+            [(set IntRegs:$dst,
+             (xor IntRegs:$src1, (or IntRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4162,9 +2574,8 @@ def XORr_ORrr_V4 : MInst_acc<(outs IntRegs:$dst),
 def ANDr_XORrr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst &= xor($src2, $src3)",
-            [(set (i32 IntRegs:$dst),
-                  (and (i32 IntRegs:$src1), (xor (i32 IntRegs:$src2),
-                                                 (i32 IntRegs:$src3))))],
+            [(set IntRegs:$dst,
+             (and IntRegs:$src1, (xor IntRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4172,9 +2583,8 @@ def ANDr_XORrr_V4 : MInst_acc<(outs IntRegs:$dst),
 def ORr_XORrr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst |= xor($src2, $src3)",
-            [(set (i32 IntRegs:$dst),
-                  (and (i32 IntRegs:$src1), (xor (i32 IntRegs:$src2),
-                                                 (i32 IntRegs:$src3))))],
+            [(set IntRegs:$dst,
+             (and IntRegs:$src1, (xor IntRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4182,9 +2592,8 @@ def ORr_XORrr_V4 : MInst_acc<(outs IntRegs:$dst),
 def XORr_XORrr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, IntRegs:$src3),
             "$dst ^= xor($src2, $src3)",
-            [(set (i32 IntRegs:$dst),
-             (and (i32 IntRegs:$src1), (xor (i32 IntRegs:$src2),
-                                            (i32 IntRegs:$src3))))],
+            [(set IntRegs:$dst,
+             (and IntRegs:$src1, (xor IntRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4192,9 +2601,8 @@ def XORr_XORrr_V4 : MInst_acc<(outs IntRegs:$dst),
 def ORr_ANDri2_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, s10Imm:$src3),
             "$dst |= and($src2, #$src3)",
-            [(set (i32 IntRegs:$dst),
-                  (or (i32 IntRegs:$src1), (and (i32 IntRegs:$src2),
-                                                s10ImmPred:$src3)))],
+            [(set IntRegs:$dst,
+             (or IntRegs:$src1, (and IntRegs:$src2, s10ImmPred:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4202,9 +2610,8 @@ def ORr_ANDri2_V4 : MInst_acc<(outs IntRegs:$dst),
 def ORr_ORri_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs: $src2, s10Imm:$src3),
             "$dst |= or($src2, #$src3)",
-            [(set (i32 IntRegs:$dst),
-                  (or (i32 IntRegs:$src1), (and (i32 IntRegs:$src2),
-                                                s10ImmPred:$src3)))],
+            [(set IntRegs:$dst,
+             (or IntRegs:$src1, (and IntRegs:$src2, s10ImmPred:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4256,9 +2663,8 @@ def ORr_ORri_V4 : MInst_acc<(outs IntRegs:$dst),
 def ADDi_MPYri_V4 : MInst<(outs IntRegs:$dst),
             (ins u6Imm:$src1, IntRegs:$src2, u6Imm:$src3),
             "$dst = add(#$src1, mpyi($src2, #$src3))",
-            [(set (i32 IntRegs:$dst),
-                  (add (mul (i32 IntRegs:$src2), u6ImmPred:$src3),
-                       u6ImmPred:$src1))]>,
+            [(set IntRegs:$dst,
+             (add (mul IntRegs:$src2, u6ImmPred:$src3), u6ImmPred:$src1))]>,
             Requires<[HasV4T]>;
 
 // Rd=add(#u6,mpyi(Rs,Rt))
@@ -4266,36 +2672,32 @@ def ADDi_MPYri_V4 : MInst<(outs IntRegs:$dst),
 def ADDi_MPYrr_V4 : MInst<(outs IntRegs:$dst),
             (ins u6Imm:$src1, IntRegs:$src2, IntRegs:$src3),
             "$dst = add(#$src1, mpyi($src2, $src3))",
-            [(set (i32 IntRegs:$dst),
-                  (add (mul (i32 IntRegs:$src2), (i32 IntRegs:$src3)),
-                       u6ImmPred:$src1))]>,
+            [(set IntRegs:$dst,
+             (add (mul IntRegs:$src2, IntRegs:$src3), u6ImmPred:$src1))]>,
             Requires<[HasV4T]>;
 
 // Rd=add(Ru,mpyi(#u6:2,Rs))
 def ADDr_MPYir_V4 : MInst<(outs IntRegs:$dst),
             (ins IntRegs:$src1, u6Imm:$src2, IntRegs:$src3),
             "$dst = add($src1, mpyi(#$src2, $src3))",
-            [(set (i32 IntRegs:$dst),
-             (add (i32 IntRegs:$src1), (mul (i32 IntRegs:$src3),
-                                            u6_2ImmPred:$src2)))]>,
+            [(set IntRegs:$dst,
+             (add IntRegs:$src1, (mul IntRegs:$src3, u6_2ImmPred:$src2)))]>,
             Requires<[HasV4T]>;
 
 // Rd=add(Ru,mpyi(Rs,#u6))
 def ADDr_MPYri_V4 : MInst<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs:$src2, u6Imm:$src3),
             "$dst = add($src1, mpyi($src2, #$src3))",
-            [(set (i32 IntRegs:$dst),
-                  (add (i32 IntRegs:$src1), (mul (i32 IntRegs:$src2),
-                                                 u6ImmPred:$src3)))]>,
+            [(set IntRegs:$dst,
+             (add IntRegs:$src1, (mul IntRegs:$src2, u6ImmPred:$src3)))]>,
             Requires<[HasV4T]>;
 
 // Rx=add(Ru,mpyi(Rx,Rs))
 def ADDr_MPYrr_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins IntRegs:$src1, IntRegs:$src2, IntRegs:$src3),
             "$dst = add($src1, mpyi($src2, $src3))",
-            [(set (i32 IntRegs:$dst),
-             (add (i32 IntRegs:$src1), (mul (i32 IntRegs:$src2),
-                                            (i32 IntRegs:$src3))))],
+            [(set IntRegs:$dst,
+             (add IntRegs:$src1, (mul IntRegs:$src2, IntRegs:$src3)))],
             "$src2 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4343,9 +2745,8 @@ def ADDr_MPYrr_V4 : MInst_acc<(outs IntRegs:$dst),
 def ADDi_ASLri_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins u8Imm:$src1, IntRegs:$src2, u5Imm:$src3),
             "$dst = add(#$src1, asl($src2, #$src3))",
-            [(set (i32 IntRegs:$dst),
-                  (add (shl (i32 IntRegs:$src2), u5ImmPred:$src3),
-                       u8ImmPred:$src1))],
+            [(set IntRegs:$dst,
+              (add (shl IntRegs:$src2, u5ImmPred:$src3), u8ImmPred:$src1))],
             "$src2 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4353,9 +2754,8 @@ def ADDi_ASLri_V4 : MInst_acc<(outs IntRegs:$dst),
 def ADDi_LSRri_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins u8Imm:$src1, IntRegs:$src2, u5Imm:$src3),
             "$dst = add(#$src1, lsr($src2, #$src3))",
-            [(set (i32 IntRegs:$dst),
-                  (add (srl (i32 IntRegs:$src2), u5ImmPred:$src3),
-                       u8ImmPred:$src1))],
+            [(set IntRegs:$dst,
+             (add (srl IntRegs:$src2, u5ImmPred:$src3), u8ImmPred:$src1))],
             "$src2 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4363,9 +2763,8 @@ def ADDi_LSRri_V4 : MInst_acc<(outs IntRegs:$dst),
 def SUBi_ASLri_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins u8Imm:$src1, IntRegs:$src2, u5Imm:$src3),
             "$dst = sub(#$src1, asl($src2, #$src3))",
-            [(set (i32 IntRegs:$dst),
-                  (sub (shl (i32 IntRegs:$src2), u5ImmPred:$src3),
-                       u8ImmPred:$src1))],
+            [(set IntRegs:$dst,
+             (sub (shl IntRegs:$src2, u5ImmPred:$src3), u8ImmPred:$src1))],
             "$src2 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4373,9 +2772,8 @@ def SUBi_ASLri_V4 : MInst_acc<(outs IntRegs:$dst),
 def SUBi_LSRri_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins u8Imm:$src1, IntRegs:$src2, u5Imm:$src3),
             "$dst = sub(#$src1, lsr($src2, #$src3))",
-            [(set (i32 IntRegs:$dst),
-                  (sub (srl (i32 IntRegs:$src2), u5ImmPred:$src3),
-                       u8ImmPred:$src1))],
+            [(set IntRegs:$dst,
+             (sub (srl IntRegs:$src2, u5ImmPred:$src3), u8ImmPred:$src1))],
             "$src2 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4385,9 +2783,8 @@ def SUBi_LSRri_V4 : MInst_acc<(outs IntRegs:$dst),
 def ANDi_ASLri_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins u8Imm:$src1, IntRegs:$src2, u5Imm:$src3),
             "$dst = and(#$src1, asl($src2, #$src3))",
-            [(set (i32 IntRegs:$dst),
-                  (and (shl (i32 IntRegs:$src2), u5ImmPred:$src3),
-                       u8ImmPred:$src1))],
+            [(set IntRegs:$dst,
+             (and (shl IntRegs:$src2, u5ImmPred:$src3), u8ImmPred:$src1))],
             "$src2 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4395,31 +2792,26 @@ def ANDi_ASLri_V4 : MInst_acc<(outs IntRegs:$dst),
 def ANDi_LSRri_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins u8Imm:$src1, IntRegs:$src2, u5Imm:$src3),
             "$dst = and(#$src1, lsr($src2, #$src3))",
-            [(set (i32 IntRegs:$dst),
-                  (and (srl (i32 IntRegs:$src2), u5ImmPred:$src3),
-                       u8ImmPred:$src1))],
+            [(set IntRegs:$dst,
+             (and (srl IntRegs:$src2, u5ImmPred:$src3), u8ImmPred:$src1))],
             "$src2 = $dst">,
             Requires<[HasV4T]>;
 
 //Rx=or(#u8,asl(Rx,#U5))
-let AddedComplexity = 30 in
 def ORi_ASLri_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins u8Imm:$src1, IntRegs:$src2, u5Imm:$src3),
             "$dst = or(#$src1, asl($src2, #$src3))",
-            [(set (i32 IntRegs:$dst),
-                  (or (shl (i32 IntRegs:$src2), u5ImmPred:$src3),
-                      u8ImmPred:$src1))],
+            [(set IntRegs:$dst,
+             (or (shl IntRegs:$src2, u5ImmPred:$src3), u8ImmPred:$src1))],
             "$src2 = $dst">,
             Requires<[HasV4T]>;
 
 //Rx=or(#u8,lsr(Rx,#U5))
-let AddedComplexity = 30 in
 def ORi_LSRri_V4 : MInst_acc<(outs IntRegs:$dst),
             (ins u8Imm:$src1, IntRegs:$src2, u5Imm:$src3),
             "$dst = or(#$src1, lsr($src2, #$src3))",
-            [(set (i32 IntRegs:$dst),
-                  (or (srl (i32 IntRegs:$src2), u5ImmPred:$src3),
-                      u8ImmPred:$src1))],
+            [(set IntRegs:$dst,
+             (or (srl IntRegs:$src2, u5ImmPred:$src3), u8ImmPred:$src1))],
             "$src2 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4428,8 +2820,7 @@ def ORi_LSRri_V4 : MInst_acc<(outs IntRegs:$dst),
 //Rd=lsl(#s6,Rt)
 def LSLi_V4 : MInst<(outs IntRegs:$dst), (ins s6Imm:$src1, IntRegs:$src2),
             "$dst = lsl(#$src1, $src2)",
-            [(set (i32 IntRegs:$dst), (shl s6ImmPred:$src1,
-                                           (i32 IntRegs:$src2)))]>,
+            [(set IntRegs:$dst, (shl s6ImmPred:$src1, IntRegs:$src2))]>,
             Requires<[HasV4T]>;
 
 
@@ -4438,9 +2829,8 @@ def LSLi_V4 : MInst<(outs IntRegs:$dst), (ins s6Imm:$src1, IntRegs:$src2),
 def ASLd_rr_xor_V4 : MInst_acc<(outs DoubleRegs:$dst),
             (ins DoubleRegs:$src1, DoubleRegs:$src2, IntRegs:$src3),
             "$dst ^= asl($src2, $src3)",
-            [(set (i64 DoubleRegs:$dst),
-                  (xor (i64 DoubleRegs:$src1), (shl (i64 DoubleRegs:$src2),
-                                                    (i32 IntRegs:$src3))))],
+            [(set DoubleRegs:$dst,
+             (xor DoubleRegs:$src1, (shl DoubleRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4448,9 +2838,8 @@ def ASLd_rr_xor_V4 : MInst_acc<(outs DoubleRegs:$dst),
 def ASRd_rr_xor_V4 : MInst_acc<(outs DoubleRegs:$dst),
             (ins DoubleRegs:$src1, DoubleRegs:$src2, IntRegs:$src3),
             "$dst ^= asr($src2, $src3)",
-            [(set (i64 DoubleRegs:$dst),
-                  (xor (i64 DoubleRegs:$src1), (sra (i64 DoubleRegs:$src2),
-                                                    (i32 IntRegs:$src3))))],
+            [(set DoubleRegs:$dst,
+             (xor DoubleRegs:$src1, (sra DoubleRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4458,9 +2847,8 @@ def ASRd_rr_xor_V4 : MInst_acc<(outs DoubleRegs:$dst),
 def LSLd_rr_xor_V4 : MInst_acc<(outs DoubleRegs:$dst),
             (ins DoubleRegs:$src1, DoubleRegs:$src2, IntRegs:$src3),
             "$dst ^= lsl($src2, $src3)",
-            [(set (i64 DoubleRegs:$dst), (xor (i64 DoubleRegs:$src1),
-                                              (shl (i64 DoubleRegs:$src2),
-                                                   (i32 IntRegs:$src3))))],
+            [(set DoubleRegs:$dst,
+             (xor DoubleRegs:$src1, (shl DoubleRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4468,9 +2856,8 @@ def LSLd_rr_xor_V4 : MInst_acc<(outs DoubleRegs:$dst),
 def LSRd_rr_xor_V4 : MInst_acc<(outs DoubleRegs:$dst),
             (ins DoubleRegs:$src1, DoubleRegs:$src2, IntRegs:$src3),
             "$dst ^= lsr($src2, $src3)",
-            [(set (i64 DoubleRegs:$dst),
-                  (xor (i64 DoubleRegs:$src1), (srl (i64 DoubleRegs:$src2),
-                                                    (i32 IntRegs:$src3))))],
+            [(set DoubleRegs:$dst,
+             (xor DoubleRegs:$src1, (srl DoubleRegs:$src2, IntRegs:$src3)))],
             "$src1 = $dst">,
             Requires<[HasV4T]>;
 
@@ -4516,16 +2903,16 @@ let AddedComplexity = 30 in
 def MEMw_ADDSUBi_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_2Imm:$offset, m6Imm:$addend),
             "Error; should not emit",
-            [(store (add (load (add (i32 IntRegs:$base), u6_2ImmPred:$offset)),
-                         m6ImmPred:$addend),
-                    (add (i32 IntRegs:$base), u6_2ImmPred:$offset))]>,
+            [(store (add (load (add IntRegs:$base, u6_2ImmPred:$offset)),
+m6ImmPred:$addend),
+                    (add IntRegs:$base, u6_2ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memw(Rs+#u6:2) += #U5
 let AddedComplexity = 30 in
 def MEMw_ADDi_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_2Imm:$offset, u5Imm:$addend),
-            "memw($base+#$offset) += #$addend",
+            "memw($base+#$offset) += $addend",
             []>,
             Requires<[HasV4T, UseMEMOP]>;
 
@@ -4533,7 +2920,7 @@ def MEMw_ADDi_indexed_MEM_V4 : MEMInst_V4<(outs),
 let AddedComplexity = 30 in
 def MEMw_SUBi_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_2Imm:$offset, u5Imm:$subend),
-            "memw($base+#$offset) -= #$subend",
+            "memw($base+#$offset) -= $subend",
             []>,
             Requires<[HasV4T, UseMEMOP]>;
 
@@ -4542,9 +2929,9 @@ let AddedComplexity = 30 in
 def MEMw_ADDr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_2Imm:$offset, IntRegs:$addend),
             "memw($base+#$offset) += $addend",
-            [(store (add (load (add (i32 IntRegs:$base), u6_2ImmPred:$offset)),
-                         (i32 IntRegs:$addend)),
-                    (add (i32 IntRegs:$base), u6_2ImmPred:$offset))]>,
+            [(store (add (load (add IntRegs:$base, u6_2ImmPred:$offset)),
+IntRegs:$addend),
+                    (add IntRegs:$base, u6_2ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memw(Rs+#u6:2) -= Rt
@@ -4552,19 +2939,19 @@ let AddedComplexity = 30 in
 def MEMw_SUBr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_2Imm:$offset, IntRegs:$subend),
             "memw($base+#$offset) -= $subend",
-            [(store (sub (load (add (i32 IntRegs:$base), u6_2ImmPred:$offset)),
-                         (i32 IntRegs:$subend)),
-                    (add (i32 IntRegs:$base), u6_2ImmPred:$offset))]>,
+            [(store (sub (load (add IntRegs:$base, u6_2ImmPred:$offset)),
+IntRegs:$subend),
+                    (add IntRegs:$base, u6_2ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memw(Rs+#u6:2) &= Rt
 let AddedComplexity = 30 in
 def MEMw_ANDr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_2Imm:$offset, IntRegs:$andend),
-            "memw($base+#$offset) &= $andend",
-            [(store (and (load (add (i32 IntRegs:$base), u6_2ImmPred:$offset)),
-                         (i32 IntRegs:$andend)),
-                    (add (i32 IntRegs:$base), u6_2ImmPred:$offset))]>,
+            "memw($base+#$offset) += $andend",
+            [(store (and (load (add IntRegs:$base, u6_2ImmPred:$offset)),
+IntRegs:$andend),
+                    (add IntRegs:$base, u6_2ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memw(Rs+#u6:2) |= Rt
@@ -4572,9 +2959,9 @@ let AddedComplexity = 30 in
 def MEMw_ORr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_2Imm:$offset, IntRegs:$orend),
             "memw($base+#$offset) |= $orend",
-            [(store (or (load (add (i32 IntRegs:$base), u6_2ImmPred:$offset)),
-                        (i32 IntRegs:$orend)),
-                    (add (i32 IntRegs:$base), u6_2ImmPred:$offset))]>,
+            [(store (or (load (add IntRegs:$base, u6_2ImmPred:$offset)),
+                        IntRegs:$orend),
+                    (add IntRegs:$base, u6_2ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // MEMw_ADDSUBi_V4:
@@ -4609,7 +2996,7 @@ let AddedComplexity = 30 in
 def MEMw_ADDr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$addend),
             "memw($addr) += $addend",
-            [(store (add (load ADDRriU6_2:$addr), (i32 IntRegs:$addend)),
+            [(store (add (load ADDRriU6_2:$addr), IntRegs:$addend),
                     ADDRriU6_2:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
@@ -4618,7 +3005,7 @@ let AddedComplexity = 30 in
 def MEMw_SUBr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$subend),
             "memw($addr) -= $subend",
-            [(store (sub (load ADDRriU6_2:$addr), (i32 IntRegs:$subend)),
+            [(store (sub (load ADDRriU6_2:$addr), IntRegs:$subend),
                     ADDRriU6_2:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
@@ -4627,7 +3014,7 @@ let AddedComplexity = 30 in
 def MEMw_ANDr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$andend),
             "memw($addr) &= $andend",
-            [(store (and (load ADDRriU6_2:$addr), (i32 IntRegs:$andend)),
+            [(store (and (load ADDRriU6_2:$addr), IntRegs:$andend),
                     ADDRriU6_2:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
@@ -4636,8 +3023,8 @@ let AddedComplexity = 30 in
 def MEMw_ORr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$orend),
             "memw($addr) |= $orend",
-            [(store (or (load ADDRriU6_2:$addr), (i32 IntRegs:$orend)),
-                    ADDRriU6_2:$addr)]>,
+            [(store (or (load ADDRriU6_2:$addr), IntRegs:$orend),
+ADDRriU6_2:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 //===----------------------------------------------------------------------===//
@@ -4673,10 +3060,10 @@ let AddedComplexity = 30 in
 def MEMh_ADDSUBi_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_1Imm:$offset, m6Imm:$addend),
             "Error; should not emit",
-            [(truncstorei16 (add (sextloadi16 (add (i32 IntRegs:$base),
+            [(truncstorei16 (add (sextloadi16 (add IntRegs:$base,
                                                    u6_1ImmPred:$offset)),
                                  m6ImmPred:$addend),
-                            (add (i32 IntRegs:$base), u6_1ImmPred:$offset))]>,
+                            (add IntRegs:$base, u6_1ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memh(Rs+#u6:1) += #U5
@@ -4700,10 +3087,10 @@ let AddedComplexity = 30 in
 def MEMh_ADDr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_1Imm:$offset, IntRegs:$addend),
             "memh($base+#$offset) += $addend",
-            [(truncstorei16 (add (sextloadi16 (add (i32 IntRegs:$base),
+            [(truncstorei16 (add (sextloadi16 (add IntRegs:$base,
                                                    u6_1ImmPred:$offset)),
-                                 (i32 IntRegs:$addend)),
-                            (add (i32 IntRegs:$base), u6_1ImmPred:$offset))]>,
+                                 IntRegs:$addend),
+                            (add IntRegs:$base, u6_1ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memh(Rs+#u6:1) -= Rt
@@ -4711,10 +3098,10 @@ let AddedComplexity = 30 in
 def MEMh_SUBr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_1Imm:$offset, IntRegs:$subend),
             "memh($base+#$offset) -= $subend",
-            [(truncstorei16 (sub (sextloadi16 (add (i32 IntRegs:$base),
+            [(truncstorei16 (sub (sextloadi16 (add IntRegs:$base,
                                                    u6_1ImmPred:$offset)),
-                                 (i32 IntRegs:$subend)),
-                            (add (i32 IntRegs:$base), u6_1ImmPred:$offset))]>,
+                                 IntRegs:$subend),
+                            (add IntRegs:$base, u6_1ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memh(Rs+#u6:1) &= Rt
@@ -4722,10 +3109,10 @@ let AddedComplexity = 30 in
 def MEMh_ANDr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_1Imm:$offset, IntRegs:$andend),
             "memh($base+#$offset) += $andend",
-            [(truncstorei16 (and (sextloadi16 (add (i32 IntRegs:$base),
+            [(truncstorei16 (and (sextloadi16 (add IntRegs:$base,
                                                    u6_1ImmPred:$offset)),
-                                 (i32 IntRegs:$andend)),
-                            (add (i32 IntRegs:$base), u6_1ImmPred:$offset))]>,
+                                 IntRegs:$andend),
+                            (add IntRegs:$base, u6_1ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memh(Rs+#u6:1) |= Rt
@@ -4733,10 +3120,10 @@ let AddedComplexity = 30 in
 def MEMh_ORr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_1Imm:$offset, IntRegs:$orend),
             "memh($base+#$offset) |= $orend",
-            [(truncstorei16 (or (sextloadi16 (add (i32 IntRegs:$base),
+            [(truncstorei16 (or (sextloadi16 (add IntRegs:$base,
                                               u6_1ImmPred:$offset)),
-                             (i32 IntRegs:$orend)),
-                            (add (i32 IntRegs:$base), u6_1ImmPred:$offset))]>,
+                             IntRegs:$orend),
+                            (add IntRegs:$base, u6_1ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // MEMh_ADDSUBi_V4:
@@ -4772,7 +3159,7 @@ def MEMh_ADDr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$addend),
             "memh($addr) += $addend",
             [(truncstorei16 (add (sextloadi16 ADDRriU6_1:$addr),
-                                 (i32 IntRegs:$addend)), ADDRriU6_1:$addr)]>,
+                                 IntRegs:$addend), ADDRriU6_1:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memh(Rs+#u6:1) -= Rt
@@ -4781,7 +3168,7 @@ def MEMh_SUBr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$subend),
             "memh($addr) -= $subend",
             [(truncstorei16 (sub (sextloadi16 ADDRriU6_1:$addr),
-                                 (i32 IntRegs:$subend)), ADDRriU6_1:$addr)]>,
+                                 IntRegs:$subend), ADDRriU6_1:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memh(Rs+#u6:1) &= Rt
@@ -4790,7 +3177,7 @@ def MEMh_ANDr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$andend),
             "memh($addr) &= $andend",
             [(truncstorei16 (and (sextloadi16 ADDRriU6_1:$addr),
-                                 (i32 IntRegs:$andend)), ADDRriU6_1:$addr)]>,
+                                 IntRegs:$andend), ADDRriU6_1:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memh(Rs+#u6:1) |= Rt
@@ -4799,7 +3186,7 @@ def MEMh_ORr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$orend),
             "memh($addr) |= $orend",
             [(truncstorei16 (or (sextloadi16 ADDRriU6_1:$addr),
-                                (i32 IntRegs:$orend)), ADDRriU6_1:$addr)]>,
+                                IntRegs:$orend), ADDRriU6_1:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 
@@ -4836,10 +3223,10 @@ let AddedComplexity = 30 in
 def MEMb_ADDSUBi_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_0Imm:$offset, m6Imm:$addend),
             "Error; should not emit",
-            [(truncstorei8 (add (sextloadi8 (add (i32 IntRegs:$base),
+            [(truncstorei8 (add (sextloadi8 (add IntRegs:$base,
                                                  u6_0ImmPred:$offset)),
                                 m6ImmPred:$addend),
-                           (add (i32 IntRegs:$base), u6_0ImmPred:$offset))]>,
+                           (add IntRegs:$base, u6_0ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memb(Rs+#u6:0) += #U5
@@ -4863,10 +3250,10 @@ let AddedComplexity = 30 in
 def MEMb_ADDr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_0Imm:$offset, IntRegs:$addend),
             "memb($base+#$offset) += $addend",
-            [(truncstorei8 (add (sextloadi8 (add (i32 IntRegs:$base),
+            [(truncstorei8 (add (sextloadi8 (add IntRegs:$base,
                                                  u6_0ImmPred:$offset)),
-                                (i32 IntRegs:$addend)),
-                           (add (i32 IntRegs:$base), u6_0ImmPred:$offset))]>,
+                                IntRegs:$addend),
+                           (add IntRegs:$base, u6_0ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memb(Rs+#u6:0) -= Rt
@@ -4874,10 +3261,10 @@ let AddedComplexity = 30 in
 def MEMb_SUBr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_0Imm:$offset, IntRegs:$subend),
             "memb($base+#$offset) -= $subend",
-            [(truncstorei8 (sub (sextloadi8 (add (i32 IntRegs:$base),
+            [(truncstorei8 (sub (sextloadi8 (add IntRegs:$base,
                                                  u6_0ImmPred:$offset)),
-                                (i32 IntRegs:$subend)),
-                           (add (i32 IntRegs:$base), u6_0ImmPred:$offset))]>,
+                                IntRegs:$subend),
+                           (add IntRegs:$base, u6_0ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memb(Rs+#u6:0) &= Rt
@@ -4885,10 +3272,10 @@ let AddedComplexity = 30 in
 def MEMb_ANDr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_0Imm:$offset, IntRegs:$andend),
             "memb($base+#$offset) += $andend",
-            [(truncstorei8 (and (sextloadi8 (add (i32 IntRegs:$base),
+            [(truncstorei8 (and (sextloadi8 (add IntRegs:$base,
                                                  u6_0ImmPred:$offset)),
-                                (i32 IntRegs:$andend)),
-                           (add (i32 IntRegs:$base), u6_0ImmPred:$offset))]>,
+                                IntRegs:$andend),
+                           (add IntRegs:$base, u6_0ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memb(Rs+#u6:0) |= Rt
@@ -4896,10 +3283,10 @@ let AddedComplexity = 30 in
 def MEMb_ORr_indexed_MEM_V4 : MEMInst_V4<(outs),
             (ins IntRegs:$base, u6_0Imm:$offset, IntRegs:$orend),
             "memb($base+#$offset) |= $orend",
-            [(truncstorei8 (or (sextloadi8 (add (i32 IntRegs:$base),
+            [(truncstorei8 (or (sextloadi8 (add IntRegs:$base,
                                                 u6_0ImmPred:$offset)),
-                               (i32 IntRegs:$orend)),
-                           (add (i32 IntRegs:$base), u6_0ImmPred:$offset))]>,
+                                IntRegs:$orend),
+                           (add IntRegs:$base, u6_0ImmPred:$offset))]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // MEMb_ADDSUBi_V4:
@@ -4935,7 +3322,7 @@ def MEMb_ADDr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$addend),
             "memb($addr) += $addend",
             [(truncstorei8 (add (sextloadi8 ADDRriU6_0:$addr),
-                                (i32 IntRegs:$addend)), ADDRriU6_0:$addr)]>,
+                                IntRegs:$addend), ADDRriU6_0:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memb(Rs+#u6:0) -= Rt
@@ -4944,7 +3331,7 @@ def MEMb_SUBr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$subend),
             "memb($addr) -= $subend",
             [(truncstorei8 (sub (sextloadi8 ADDRriU6_0:$addr),
-                                (i32 IntRegs:$subend)), ADDRriU6_0:$addr)]>,
+                                IntRegs:$subend), ADDRriU6_0:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memb(Rs+#u6:0) &= Rt
@@ -4953,7 +3340,7 @@ def MEMb_ANDr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$andend),
             "memb($addr) &= $andend",
             [(truncstorei8 (and (sextloadi8 ADDRriU6_0:$addr),
-                                (i32 IntRegs:$andend)), ADDRriU6_0:$addr)]>,
+                                IntRegs:$andend), ADDRriU6_0:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 // memb(Rs+#u6:0) |= Rt
@@ -4962,7 +3349,7 @@ def MEMb_ORr_MEM_V4 : MEMInst_V4<(outs),
             (ins MEMri:$addr, IntRegs:$orend),
             "memb($addr) |= $orend",
             [(truncstorei8 (or (sextloadi8 ADDRriU6_0:$addr),
-                               (i32 IntRegs:$orend)), ADDRriU6_0:$addr)]>,
+                                IntRegs:$orend), ADDRriU6_0:$addr)]>,
             Requires<[HasV4T, UseMEMOP]>;
 
 
@@ -4977,16 +3364,13 @@ def MEMb_ORr_MEM_V4 : MEMInst_V4<(outs),
 // The implemented patterns are: EQ/GT/GTU.
 // Missing patterns are: GE/GEU/LT/LTU/LE/LEU.
 
-// Following instruction is not being extended as it results into the
-// incorrect code for negative numbers.
 // Pd=cmpb.eq(Rs,#u8)
-
 let isCompare = 1 in
 def CMPbEQri_V4 : MInst<(outs PredRegs:$dst),
             (ins IntRegs:$src1, u8Imm:$src2),
             "$dst = cmpb.eq($src1, #$src2)",
-            [(set (i1 PredRegs:$dst),
-                  (seteq (and (i32 IntRegs:$src1), 255), u8ImmPred:$src2))]>,
+            [(set PredRegs:$dst, (seteq (and IntRegs:$src1, 255),
+                                        u8ImmPred:$src2))]>,
             Requires<[HasV4T]>;
 
 // Pd=cmpb.eq(Rs,Rt)
@@ -4994,9 +3378,10 @@ let isCompare = 1 in
 def CMPbEQrr_ubub_V4 : MInst<(outs PredRegs:$dst),
             (ins IntRegs:$src1, IntRegs:$src2),
             "$dst = cmpb.eq($src1, $src2)",
-            [(set (i1 PredRegs:$dst),
-                  (seteq (and (xor (i32 IntRegs:$src1),
-                                   (i32 IntRegs:$src2)), 255), 0))]>,
+            [(set PredRegs:$dst, (seteq (and (xor IntRegs:$src1,
+                                                  IntRegs:$src2),
+                                             255),
+                                        0))]>,
             Requires<[HasV4T]>;
 
 // Pd=cmpb.eq(Rs,Rt)
@@ -5004,31 +3389,26 @@ let isCompare = 1 in
 def CMPbEQrr_sbsb_V4 : MInst<(outs PredRegs:$dst),
             (ins IntRegs:$src1, IntRegs:$src2),
             "$dst = cmpb.eq($src1, $src2)",
-            [(set (i1 PredRegs:$dst),
-                  (seteq (shl (i32 IntRegs:$src1), (i32 24)),
-                         (shl (i32 IntRegs:$src2), (i32 24))))]>,
+            [(set PredRegs:$dst, (seteq (shl IntRegs:$src1, (i32 24)),
+                                        (shl IntRegs:$src2, (i32 24))))]>,
             Requires<[HasV4T]>;
 
-/* Incorrect Pattern -- immediate should be right shifted before being
-used in the cmpb.gt instruction.
 // Pd=cmpb.gt(Rs,#s8)
 let isCompare = 1 in
 def CMPbGTri_V4 : MInst<(outs PredRegs:$dst),
-            (ins IntRegs:$src1, s8Imm:$src2),
+            (ins IntRegs:$src1, s32Imm:$src2),
             "$dst = cmpb.gt($src1, #$src2)",
-            [(set (i1 PredRegs:$dst), (setgt (shl (i32 IntRegs:$src1), (i32 24)),
-                                             s8ImmPred:$src2))]>,
+            [(set PredRegs:$dst, (setgt (shl IntRegs:$src1, (i32 24)),
+                                        s32_24ImmPred:$src2))]>,
             Requires<[HasV4T]>;
-*/
 
 // Pd=cmpb.gt(Rs,Rt)
 let isCompare = 1 in
 def CMPbGTrr_V4 : MInst<(outs PredRegs:$dst),
             (ins IntRegs:$src1, IntRegs:$src2),
             "$dst = cmpb.gt($src1, $src2)",
-            [(set (i1 PredRegs:$dst),
-                  (setgt (shl (i32 IntRegs:$src1), (i32 24)),
-                         (shl (i32 IntRegs:$src2), (i32 24))))]>,
+            [(set PredRegs:$dst, (setgt (shl IntRegs:$src1, (i32 24)),
+                                        (shl IntRegs:$src2, (i32 24))))]>,
             Requires<[HasV4T]>;
 
 // Pd=cmpb.gtu(Rs,#u7)
@@ -5036,8 +3416,8 @@ let isCompare = 1 in
 def CMPbGTUri_V4 : MInst<(outs PredRegs:$dst),
             (ins IntRegs:$src1, u7Imm:$src2),
             "$dst = cmpb.gtu($src1, #$src2)",
-            [(set (i1 PredRegs:$dst), (setugt (and (i32 IntRegs:$src1), 255),
-                                              u7ImmPred:$src2))]>,
+            [(set PredRegs:$dst, (setugt (and IntRegs:$src1, 255),
+                                         u7ImmPred:$src2))]>,
             Requires<[HasV4T]>;
 
 // Pd=cmpb.gtu(Rs,Rt)
@@ -5045,21 +3425,18 @@ let isCompare = 1 in
 def CMPbGTUrr_V4 : MInst<(outs PredRegs:$dst),
             (ins IntRegs:$src1, IntRegs:$src2),
             "$dst = cmpb.gtu($src1, $src2)",
-            [(set (i1 PredRegs:$dst), (setugt (and (i32 IntRegs:$src1), 255),
-                                              (and (i32 IntRegs:$src2), 255)))]>,
+            [(set PredRegs:$dst, (setugt (and IntRegs:$src1, 255),
+                                         (and IntRegs:$src2, 255)))]>,
             Requires<[HasV4T]>;
 
-// Following instruction is not being extended as it results into the incorrect
-// code for negative numbers.
-
 // Signed half compare(.eq) ri.
 // Pd=cmph.eq(Rs,#s8)
 let isCompare = 1 in
 def CMPhEQri_V4 : MInst<(outs PredRegs:$dst),
-            (ins IntRegs:$src1, s8Imm:$src2),
+            (ins IntRegs:$src1, u16Imm:$src2),
             "$dst = cmph.eq($src1, #$src2)",
-            [(set (i1 PredRegs:$dst), (seteq (and (i32 IntRegs:$src1), 65535),
-                                             s8ImmPred:$src2))]>,
+            [(set PredRegs:$dst, (seteq (and IntRegs:$src1, 65535),
+                                        u16_s8ImmPred:$src2))]>,
             Requires<[HasV4T]>;
 
 // Signed half compare(.eq) rr.
@@ -5072,9 +3449,10 @@ let isCompare = 1 in
 def CMPhEQrr_xor_V4 : MInst<(outs PredRegs:$dst),
             (ins IntRegs:$src1, IntRegs:$src2),
             "$dst = cmph.eq($src1, $src2)",
-            [(set (i1 PredRegs:$dst), (seteq (and (xor (i32 IntRegs:$src1),
-                                                       (i32 IntRegs:$src2)),
-                                                  65535), 0))]>,
+            [(set PredRegs:$dst, (seteq (and (xor IntRegs:$src1,
+                                                  IntRegs:$src2),
+                                             65535),
+                                        0))]>,
             Requires<[HasV4T]>;
 
 // Signed half compare(.eq) rr.
@@ -5087,25 +3465,19 @@ let isCompare = 1 in
 def CMPhEQrr_shl_V4 : MInst<(outs PredRegs:$dst),
             (ins IntRegs:$src1, IntRegs:$src2),
             "$dst = cmph.eq($src1, $src2)",
-            [(set (i1 PredRegs:$dst),
-                  (seteq (shl (i32 IntRegs:$src1), (i32 16)),
-                         (shl (i32 IntRegs:$src2), (i32 16))))]>,
+            [(set PredRegs:$dst, (seteq (shl IntRegs:$src1, (i32 16)),
+                                        (shl IntRegs:$src2, (i32 16))))]>,
             Requires<[HasV4T]>;
 
-/* Incorrect Pattern -- immediate should be right shifted before being
-used in the cmph.gt instruction.
 // Signed half compare(.gt) ri.
 // Pd=cmph.gt(Rs,#s8)
-
 let isCompare = 1 in
 def CMPhGTri_V4 : MInst<(outs PredRegs:$dst),
-            (ins IntRegs:$src1, s8Imm:$src2),
+            (ins IntRegs:$src1, s32Imm:$src2),
             "$dst = cmph.gt($src1, #$src2)",
-            [(set (i1 PredRegs:$dst),
-                  (setgt (shl (i32 IntRegs:$src1), (i32 16)),
-                         s8ImmPred:$src2))]>,
+            [(set PredRegs:$dst, (setgt (shl IntRegs:$src1, (i32 16)),
+                                        s32_16s8ImmPred:$src2))]>,
             Requires<[HasV4T]>;
-*/
 
 // Signed half compare(.gt) rr.
 // Pd=cmph.gt(Rs,Rt)
@@ -5113,9 +3485,8 @@ let isCompare = 1 in
 def CMPhGTrr_shl_V4 : MInst<(outs PredRegs:$dst),
             (ins IntRegs:$src1, IntRegs:$src2),
             "$dst = cmph.gt($src1, $src2)",
-            [(set (i1 PredRegs:$dst),
-                  (setgt (shl (i32 IntRegs:$src1), (i32 16)),
-                         (shl (i32 IntRegs:$src2), (i32 16))))]>,
+            [(set PredRegs:$dst, (setgt (shl IntRegs:$src1, (i32 16)),
+                                        (shl IntRegs:$src2, (i32 16))))]>,
             Requires<[HasV4T]>;
 
 // Unsigned half compare rr (.gtu).
@@ -5124,9 +3495,8 @@ let isCompare = 1 in
 def CMPhGTUrr_V4 : MInst<(outs PredRegs:$dst),
             (ins IntRegs:$src1, IntRegs:$src2),
             "$dst = cmph.gtu($src1, $src2)",
-            [(set (i1 PredRegs:$dst),
-                  (setugt (and (i32 IntRegs:$src1), 65535),
-                          (and (i32 IntRegs:$src2), 65535)))]>,
+            [(set PredRegs:$dst, (setugt (and IntRegs:$src1, 65535),
+                                         (and IntRegs:$src2, 65535)))]>,
             Requires<[HasV4T]>;
 
 // Unsigned half compare ri (.gtu).
@@ -5135,8 +3505,8 @@ let isCompare = 1 in
 def CMPhGTUri_V4 : MInst<(outs PredRegs:$dst),
             (ins IntRegs:$src1, u7Imm:$src2),
             "$dst = cmph.gtu($src1, #$src2)",
-            [(set (i1 PredRegs:$dst), (setugt (and (i32 IntRegs:$src1), 65535),
-                                              u7ImmPred:$src2))]>,
+            [(set PredRegs:$dst, (setugt (and IntRegs:$src1, 65535),
+                                         u7ImmPred:$src2))]>,
             Requires<[HasV4T]>;
 
 //===----------------------------------------------------------------------===//
@@ -5153,37 +3523,9 @@ let isReturn = 1, isTerminator = 1, isBarrier = 1, isPredicable = 1,
             Requires<[HasV4T]>;
 }
 
-// Restore registers and dealloc return function call.
-let isCall = 1, isBarrier = 1, isReturn = 1, isTerminator = 1,
-  Defs = [R29, R30, R31, PC] in {
-  def RESTORE_DEALLOC_RET_JMP_V4 : JInst<(outs), (ins calltarget:$dst, variable_ops),
-             "jump $dst // Restore_and_dealloc_return",
-             []>,
-             Requires<[HasV4T]>;
-}
-
-// Restore registers and dealloc frame before a tail call.
-let isCall = 1, isBarrier = 1,
-  Defs = [R29, R30, R31, PC] in {
-  def RESTORE_DEALLOC_BEFORE_TAILCALL_V4 : JInst<(outs), (ins calltarget:$dst, variable_ops),
-             "call $dst // Restore_and_dealloc_before_tailcall",
-             []>,
-             Requires<[HasV4T]>;
-}
-
-// Save registers function call.
-let isCall = 1, isBarrier = 1,
-  Uses = [R29, R31] in {
-  def SAVE_REGISTERS_CALL_V4 : JInst<(outs), (ins calltarget:$dst, variable_ops),
-             "call $dst // Save_calle_saved_registers",
-             []>,
-             Requires<[HasV4T]>;
-}
-
 //    if (Ps) dealloc_return
 let isReturn = 1, isTerminator = 1,
-    Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1,
-    isPredicated = 1 in {
+  Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1 in {
   def DEALLOC_RET_cPt_V4 : NVInst_V4<(outs), (ins PredRegs:$src1, i32imm:$amt1),
             "if ($src1) dealloc_return",
             []>,
@@ -5192,8 +3534,7 @@ let isReturn = 1, isTerminator = 1,
 
 //    if (!Ps) dealloc_return
 let isReturn = 1, isTerminator = 1,
-    Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1,
-    isPredicated = 1 in {
+  Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1 in {
   def DEALLOC_RET_cNotPt_V4 : NVInst_V4<(outs), (ins PredRegs:$src1,
                                                      i32imm:$amt1),
             "if (!$src1) dealloc_return",
@@ -5203,8 +3544,7 @@ let isReturn = 1, isTerminator = 1,
 
 //    if (Ps.new) dealloc_return:nt
 let isReturn = 1, isTerminator = 1,
-    Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1,
-    isPredicated = 1 in {
+  Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1 in {
   def DEALLOC_RET_cdnPnt_V4 : NVInst_V4<(outs), (ins PredRegs:$src1,
                                                      i32imm:$amt1),
             "if ($src1.new) dealloc_return:nt",
@@ -5214,8 +3554,7 @@ let isReturn = 1, isTerminator = 1,
 
 //    if (!Ps.new) dealloc_return:nt
 let isReturn = 1, isTerminator = 1,
-    Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1,
-    isPredicated = 1 in {
+  Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1 in {
   def DEALLOC_RET_cNotdnPnt_V4 : NVInst_V4<(outs), (ins PredRegs:$src1,
                                                         i32imm:$amt1),
             "if (!$src1.new) dealloc_return:nt",
@@ -5225,8 +3564,7 @@ let isReturn = 1, isTerminator = 1,
 
 //    if (Ps.new) dealloc_return:t
 let isReturn = 1, isTerminator = 1,
-    Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1,
-    isPredicated = 1 in {
+  Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1 in {
   def DEALLOC_RET_cdnPt_V4 : NVInst_V4<(outs), (ins PredRegs:$src1,
                                                     i32imm:$amt1),
             "if ($src1.new) dealloc_return:t",
@@ -5236,511 +3574,10 @@ let isReturn = 1, isTerminator = 1,
 
 //    if (!Ps.new) dealloc_return:nt
 let isReturn = 1, isTerminator = 1,
-    Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1,
-    isPredicated = 1 in {
+  Defs = [R29, R30, R31, PC], Uses = [R29, R31], neverHasSideEffects = 1 in {
   def DEALLOC_RET_cNotdnPt_V4 : NVInst_V4<(outs), (ins PredRegs:$src1,
                                                        i32imm:$amt1),
             "if (!$src1.new) dealloc_return:t",
             []>,
             Requires<[HasV4T]>;
 }
-
-
-// Load/Store with absolute addressing mode
-// memw(#u6)=Rt
-
-multiclass ST_abs {
-  let isPredicable = 1 in
-  def _abs_V4 : STInst<(outs),
-            (ins globaladdress:$absaddr, IntRegs:$src),
-            !strconcat(OpcStr, "(##$absaddr) = $src"),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$absaddr, IntRegs:$src2),
-            !strconcat("if ($src1)", !strconcat(OpcStr, "(##$absaddr) = $src2")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$absaddr, IntRegs:$src2),
-            !strconcat("if (!$src1)", !strconcat(OpcStr, "(##$absaddr) = $src2")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$absaddr, IntRegs:$src2),
-            !strconcat("if ($src1.new)", !strconcat(OpcStr, "(##$absaddr) = $src2")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$absaddr, IntRegs:$src2),
-            !strconcat("if (!$src1.new)", !strconcat(OpcStr, "(##$absaddr) = $src2")),
-            []>,
-            Requires<[HasV4T]>;
-
-  def _abs_nv_V4 : STInst<(outs),
-            (ins globaladdress:$absaddr, IntRegs:$src),
-            !strconcat(OpcStr, "(##$absaddr) = $src.new"),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cPt_nv_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$absaddr, IntRegs:$src2),
-            !strconcat("if ($src1)", !strconcat(OpcStr, "(##$absaddr) = $src2.new")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cNotPt_nv_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$absaddr, IntRegs:$src2),
-            !strconcat("if (!$src1)", !strconcat(OpcStr, "(##$absaddr) = $src2.new")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnPt_nv_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$absaddr, IntRegs:$src2),
-            !strconcat("if ($src1.new)", !strconcat(OpcStr, "(##$absaddr) = $src2.new")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnNotPt_nv_V4 : STInst<(outs),
-            (ins PredRegs:$src1, globaladdress:$absaddr, IntRegs:$src2),
-            !strconcat("if (!$src1.new)", !strconcat(OpcStr, "(##$absaddr) = $src2.new")),
-            []>,
-            Requires<[HasV4T]>;
-}
-
-let AddedComplexity = 30, isPredicable = 1 in
-def STrid_abs_V4 : STInst<(outs),
-          (ins globaladdress:$absaddr, DoubleRegs:$src),
-           "memd(##$absaddr) = $src",
-          [(store (i64 DoubleRegs:$src), (HexagonCONST32 tglobaladdr:$absaddr))]>,
-          Requires<[HasV4T]>;
-
-let AddedComplexity = 30, isPredicated = 1 in
-def STrid_abs_cPt_V4 : STInst<(outs),
-          (ins PredRegs:$src1, globaladdress:$absaddr, DoubleRegs:$src2),
-          "if ($src1) memd(##$absaddr) = $src2",
-          []>,
-          Requires<[HasV4T]>;
-
-let AddedComplexity = 30, isPredicated = 1 in
-def STrid_abs_cNotPt_V4 : STInst<(outs),
-          (ins PredRegs:$src1, globaladdress:$absaddr, DoubleRegs:$src2),
-          "if (!$src1) memd(##$absaddr) = $src2",
-          []>,
-          Requires<[HasV4T]>;
-
-let AddedComplexity = 30, isPredicated = 1 in
-def STrid_abs_cdnPt_V4 : STInst<(outs),
-          (ins PredRegs:$src1, globaladdress:$absaddr, DoubleRegs:$src2),
-          "if ($src1.new) memd(##$absaddr) = $src2",
-          []>,
-          Requires<[HasV4T]>;
-
-let AddedComplexity = 30, isPredicated = 1 in
-def STrid_abs_cdnNotPt_V4 : STInst<(outs),
-          (ins PredRegs:$src1, globaladdress:$absaddr, DoubleRegs:$src2),
-          "if (!$src1.new) memd(##$absaddr) = $src2",
-          []>,
-          Requires<[HasV4T]>;
-
-defm STrib : ST_abs<"memb">;
-defm STrih : ST_abs<"memh">;
-defm STriw : ST_abs<"memw">;
-
-let Predicates = [HasV4T], AddedComplexity  = 30 in
-def : Pat<(truncstorei8 (i32 IntRegs:$src1), (HexagonCONST32 tglobaladdr:$absaddr)),
-          (STrib_abs_V4 tglobaladdr: $absaddr, IntRegs: $src1)>;
-
-let Predicates = [HasV4T], AddedComplexity  = 30 in
-def : Pat<(truncstorei16 (i32 IntRegs:$src1), (HexagonCONST32 tglobaladdr:$absaddr)),
-          (STrih_abs_V4 tglobaladdr: $absaddr, IntRegs: $src1)>;
-
-let Predicates = [HasV4T], AddedComplexity  = 30 in
-def : Pat<(store (i32 IntRegs:$src1), (HexagonCONST32 tglobaladdr:$absaddr)),
-          (STriw_abs_V4 tglobaladdr: $absaddr, IntRegs: $src1)>;
-
-
-multiclass LD_abs {
-  let isPredicable = 1 in
-  def _abs_V4 : LDInst<(outs IntRegs:$dst),
-            (ins globaladdress:$absaddr),
-            !strconcat("$dst = ", !strconcat(OpcStr, "(##$absaddr)")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$absaddr),
-            !strconcat("if ($src1) $dst = ", !strconcat(OpcStr, "(##$absaddr)")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$absaddr),
-            !strconcat("if (!$src1) $dst = ", !strconcat(OpcStr, "(##$absaddr)")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$absaddr),
-            !strconcat("if ($src1.new) $dst = ", !strconcat(OpcStr, "(##$absaddr)")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, globaladdress:$absaddr),
-            !strconcat("if (!$src1.new) $dst = ", !strconcat(OpcStr, "(##$absaddr)")),
-            []>,
-            Requires<[HasV4T]>;
-}
-
-let AddedComplexity = 30 in
-def LDrid_abs_V4 : LDInst<(outs DoubleRegs:$dst),
-          (ins globaladdress:$absaddr),
-          "$dst = memd(##$absaddr)",
-          [(set (i64 DoubleRegs:$dst), (load (HexagonCONST32 tglobaladdr:$absaddr)))]>,
-          Requires<[HasV4T]>;
-
-let AddedComplexity = 30, isPredicated = 1 in
-def LDrid_abs_cPt_V4 : LDInst<(outs DoubleRegs:$dst),
-          (ins PredRegs:$src1, globaladdress:$absaddr),
-          "if ($src1) $dst = memd(##$absaddr)",
-          []>,
-          Requires<[HasV4T]>;
-
-let AddedComplexity = 30, isPredicated = 1 in
-def LDrid_abs_cNotPt_V4 : LDInst<(outs DoubleRegs:$dst),
-          (ins PredRegs:$src1, globaladdress:$absaddr),
-          "if (!$src1) $dst = memd(##$absaddr)",
-          []>,
-          Requires<[HasV4T]>;
-
-let AddedComplexity = 30, isPredicated = 1 in
-def LDrid_abs_cdnPt_V4 : LDInst<(outs DoubleRegs:$dst),
-          (ins PredRegs:$src1, globaladdress:$absaddr),
-          "if ($src1.new) $dst = memd(##$absaddr)",
-          []>,
-          Requires<[HasV4T]>;
-
-let AddedComplexity = 30, isPredicated = 1 in
-def LDrid_abs_cdnNotPt_V4 : LDInst<(outs DoubleRegs:$dst),
-          (ins PredRegs:$src1, globaladdress:$absaddr),
-          "if (!$src1.new) $dst = memd(##$absaddr)",
-          []>,
-          Requires<[HasV4T]>;
-
-defm LDrib : LD_abs<"memb">;
-defm LDriub : LD_abs<"memub">;
-defm LDrih : LD_abs<"memh">;
-defm LDriuh : LD_abs<"memuh">;
-defm LDriw : LD_abs<"memw">;
-
-
-let Predicates = [HasV4T], AddedComplexity  = 30 in
-def : Pat<(i32 (load (HexagonCONST32 tglobaladdr:$absaddr))),
-          (LDriw_abs_V4 tglobaladdr: $absaddr)>;
-
-let Predicates = [HasV4T], AddedComplexity=30 in
-def : Pat<(i32 (sextloadi8 (HexagonCONST32 tglobaladdr:$absaddr))),
-          (LDrib_abs_V4 tglobaladdr:$absaddr)>;
-
-let Predicates = [HasV4T], AddedComplexity=30 in
-def : Pat<(i32 (zextloadi8 (HexagonCONST32 tglobaladdr:$absaddr))),
-          (LDriub_abs_V4 tglobaladdr:$absaddr)>;
-
-let Predicates = [HasV4T], AddedComplexity=30 in
-def : Pat<(i32 (sextloadi16 (HexagonCONST32 tglobaladdr:$absaddr))),
-          (LDrih_abs_V4 tglobaladdr:$absaddr)>;
-
-let Predicates = [HasV4T], AddedComplexity=30 in
-def : Pat<(i32 (zextloadi16 (HexagonCONST32 tglobaladdr:$absaddr))),
-          (LDriuh_abs_V4 tglobaladdr:$absaddr)>;
-
-// Transfer global address into a register
-let AddedComplexity=50, isMoveImm = 1, isReMaterializable = 1 in
-def TFRI_V4 : ALU32_ri<(outs IntRegs:$dst), (ins globaladdress:$src1),
-           "$dst = ##$src1",
-           [(set IntRegs:$dst, (HexagonCONST32 tglobaladdr:$src1))]>,
-           Requires<[HasV4T]>;
-
-let AddedComplexity=50, neverHasSideEffects = 1, isPredicated = 1 in
-def TFRI_cPt_V4 : ALU32_ri<(outs IntRegs:$dst),
-                           (ins PredRegs:$src1, globaladdress:$src2),
-           "if($src1) $dst = ##$src2",
-           []>,
-           Requires<[HasV4T]>;
-
-let AddedComplexity=50, neverHasSideEffects = 1, isPredicated = 1 in
-def TFRI_cNotPt_V4 : ALU32_ri<(outs IntRegs:$dst),
-                              (ins PredRegs:$src1, globaladdress:$src2),
-           "if(!$src1) $dst = ##$src2",
-           []>,
-           Requires<[HasV4T]>;
-
-let AddedComplexity=50, neverHasSideEffects = 1, isPredicated = 1 in
-def TFRI_cdnPt_V4 : ALU32_ri<(outs IntRegs:$dst),
-                             (ins PredRegs:$src1, globaladdress:$src2),
-           "if($src1.new) $dst = ##$src2",
-           []>,
-           Requires<[HasV4T]>;
-
-let AddedComplexity=50, neverHasSideEffects = 1, isPredicated = 1 in
-def TFRI_cdnNotPt_V4 : ALU32_ri<(outs IntRegs:$dst),
-                                (ins PredRegs:$src1, globaladdress:$src2),
-           "if(!$src1.new) $dst = ##$src2",
-           []>,
-           Requires<[HasV4T]>;
-
-let AddedComplexity = 50, Predicates = [HasV4T] in
-def : Pat<(HexagonCONST32_GP tglobaladdr:$src1),
-           (TFRI_V4 tglobaladdr:$src1)>;
-
-
-// Load - Indirect with long offset: These instructions take global address
-// as an operand
-let AddedComplexity = 10 in
-def LDrid_ind_lo_V4 : LDInst<(outs DoubleRegs:$dst),
-            (ins IntRegs:$src1, u2Imm:$src2, globaladdress:$offset),
-            "$dst=memd($src1<<#$src2+##$offset)",
-            [(set (i64 DoubleRegs:$dst),
-                  (load (add (shl IntRegs:$src1, u2ImmPred:$src2),
-                        (HexagonCONST32 tglobaladdr:$offset))))]>,
-            Requires<[HasV4T]>;
-
-let AddedComplexity = 10 in
-multiclass LD_indirect_lo {
-  def _lo_V4 : LDInst<(outs IntRegs:$dst),
-            (ins IntRegs:$src1, u2Imm:$src2, globaladdress:$offset),
-            !strconcat("$dst = ", !strconcat(OpcStr, "($src1<<#$src2+##$offset)")),
-            [(set IntRegs:$dst,
-                  (i32 (OpNode (add (shl IntRegs:$src1, u2ImmPred:$src2),
-                          (HexagonCONST32 tglobaladdr:$offset)))))]>,
-            Requires<[HasV4T]>;
-}
-
-defm LDrib_ind : LD_indirect_lo<"memb", sextloadi8>;
-defm LDriub_ind : LD_indirect_lo<"memub", zextloadi8>;
-defm LDrih_ind : LD_indirect_lo<"memh", sextloadi16>;
-defm LDriuh_ind : LD_indirect_lo<"memuh", zextloadi16>;
-defm LDriw_ind : LD_indirect_lo<"memw", load>;
-
-// Store - Indirect with long offset: These instructions take global address
-// as an operand
-let AddedComplexity = 10 in
-def STrid_ind_lo_V4 : STInst<(outs),
-            (ins IntRegs:$src1, u2Imm:$src2, globaladdress:$src3,
-                 DoubleRegs:$src4),
-            "memd($src1<<#$src2+#$src3) = $src4",
-            [(store (i64 DoubleRegs:$src4),
-                 (add (shl IntRegs:$src1, u2ImmPred:$src2),
-                      (HexagonCONST32 tglobaladdr:$src3)))]>,
-             Requires<[HasV4T]>;
-
-let AddedComplexity = 10 in
-multiclass ST_indirect_lo {
-  def _lo_V4 : STInst<(outs),
-            (ins IntRegs:$src1, u2Imm:$src2, globaladdress:$src3,
-                 IntRegs:$src4),
-            !strconcat(OpcStr, "($src1<<#$src2+##$src3) = $src4"),
-            [(OpNode (i32 IntRegs:$src4),
-                 (add (shl IntRegs:$src1, u2ImmPred:$src2),
-                      (HexagonCONST32 tglobaladdr:$src3)))]>,
-             Requires<[HasV4T]>;
-}
-
-defm STrib_ind : ST_indirect_lo<"memb", truncstorei8>;
-defm STrih_ind : ST_indirect_lo<"memh", truncstorei16>;
-defm STriw_ind : ST_indirect_lo<"memw", store>;
-
-// Store - absolute addressing mode: These instruction take constant
-// value as the extended operand
-multiclass ST_absimm {
-  let isPredicable = 1 in
-  def _abs_V4 : STInst<(outs),
-            (ins u6Imm:$src1, IntRegs:$src2),
-            !strconcat(OpcStr, "(#$src1) = $src2"),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, u6Imm:$src2, IntRegs:$src3),
-            !strconcat("if ($src1)", !strconcat(OpcStr, "(#$src2) = $src3")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, u6Imm:$src2, IntRegs:$src3),
-            !strconcat("if (!$src1)", !strconcat(OpcStr, "(#$src2) = $src3")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, u6Imm:$src2, IntRegs:$src3),
-            !strconcat("if ($src1.new)", !strconcat(OpcStr, "(#$src2) = $src3")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnNotPt_V4 : STInst<(outs),
-            (ins PredRegs:$src1, u6Imm:$src2, IntRegs:$src3),
-            !strconcat("if (!$src1.new)", !strconcat(OpcStr, "(#$src2) = $src3")),
-            []>,
-            Requires<[HasV4T]>;
-
-  def _abs_nv_V4 : STInst<(outs),
-            (ins u6Imm:$src1, IntRegs:$src2),
-            !strconcat(OpcStr, "(#$src1) = $src2.new"),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cPt_nv_V4 : STInst<(outs),
-            (ins PredRegs:$src1, u6Imm:$src2, IntRegs:$src3),
-            !strconcat("if ($src1)", !strconcat(OpcStr, "(#$src2) = $src3.new")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cNotPt_nv_V4 : STInst<(outs),
-            (ins PredRegs:$src1, u6Imm:$src2, IntRegs:$src3),
-            !strconcat("if (!$src1)", !strconcat(OpcStr, "(#$src2) = $src3.new")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnPt_nv_V4 : STInst<(outs),
-            (ins PredRegs:$src1, u6Imm:$src2, IntRegs:$src3),
-            !strconcat("if ($src1.new)", !strconcat(OpcStr, "(#$src2) = $src3.new")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnNotPt_nv_V4 : STInst<(outs),
-            (ins PredRegs:$src1, u6Imm:$src2, IntRegs:$src3),
-            !strconcat("if (!$src1.new)", !strconcat(OpcStr, "(#$src2) = $src3.new")),
-            []>,
-            Requires<[HasV4T]>;
-}
-
-defm STrib_imm : ST_absimm<"memb">;
-defm STrih_imm : ST_absimm<"memh">;
-defm STriw_imm : ST_absimm<"memw">;
-
-let Predicates = [HasV4T], AddedComplexity  = 30 in
-def : Pat<(truncstorei8 (i32 IntRegs:$src1), u6ImmPred:$src2),
-          (STrib_imm_abs_V4 u6ImmPred:$src2, IntRegs: $src1)>;
-
-let Predicates = [HasV4T], AddedComplexity  = 30 in
-def : Pat<(truncstorei16 (i32 IntRegs:$src1), u6ImmPred:$src2),
-          (STrih_imm_abs_V4 u6ImmPred:$src2, IntRegs: $src1)>;
-
-let Predicates = [HasV4T], AddedComplexity  = 30 in
-def : Pat<(store (i32 IntRegs:$src1), u6ImmPred:$src2),
-          (STriw_imm_abs_V4 u6ImmPred:$src2, IntRegs: $src1)>;
-
-
-// Load - absolute addressing mode: These instruction take constant
-// value as the extended operand
-
-multiclass LD_absimm {
-  let isPredicable = 1 in
-  def _abs_V4 : LDInst<(outs IntRegs:$dst),
-            (ins u6Imm:$src),
-            !strconcat("$dst = ", !strconcat(OpcStr, "(#$src)")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, u6Imm:$src2),
-            !strconcat("if ($src1) $dst = ", !strconcat(OpcStr, "(#$src2)")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, u6Imm:$src2),
-            !strconcat("if (!$src1) $dst = ", !strconcat(OpcStr, "(#$src2)")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, u6Imm:$src2),
-            !strconcat("if ($src1.new) $dst = ", !strconcat(OpcStr, "(#$src2)")),
-            []>,
-            Requires<[HasV4T]>;
-
-  let isPredicated = 1 in
-  def _abs_cdnNotPt_V4 : LDInst<(outs IntRegs:$dst),
-            (ins PredRegs:$src1, u6Imm:$src2),
-            !strconcat("if (!$src1.new) $dst = ", !strconcat(OpcStr, "(#$src2)")),
-            []>,
-            Requires<[HasV4T]>;
-}
-
-defm LDrib_imm : LD_absimm<"memb">;
-defm LDriub_imm : LD_absimm<"memub">;
-defm LDrih_imm : LD_absimm<"memh">;
-defm LDriuh_imm : LD_absimm<"memuh">;
-defm LDriw_imm : LD_absimm<"memw">;
-
-let Predicates = [HasV4T], AddedComplexity  = 30 in
-def : Pat<(i32 (load u6ImmPred:$src)),
-          (LDriw_imm_abs_V4 u6ImmPred:$src)>;
-
-let Predicates = [HasV4T], AddedComplexity=30 in
-def : Pat<(i32 (sextloadi8 u6ImmPred:$src)),
-          (LDrib_imm_abs_V4 u6ImmPred:$src)>;
-
-let Predicates = [HasV4T], AddedComplexity=30 in
-def : Pat<(i32 (zextloadi8 u6ImmPred:$src)),
-          (LDriub_imm_abs_V4 u6ImmPred:$src)>;
-
-let Predicates = [HasV4T], AddedComplexity=30 in
-def : Pat<(i32 (sextloadi16 u6ImmPred:$src)),
-          (LDrih_imm_abs_V4 u6ImmPred:$src)>;
-
-let Predicates = [HasV4T], AddedComplexity=30 in
-def : Pat<(i32 (zextloadi16 u6ImmPred:$src)),
-          (LDriuh_imm_abs_V4 u6ImmPred:$src)>;
-
-
-// Indexed store double word - global address.
-// memw(Rs+#u6:2)=#S8
-let AddedComplexity = 10 in
-def STriw_offset_ext_V4 : STInst<(outs),
-            (ins IntRegs:$src1, u6_2Imm:$src2, globaladdress:$src3),
-            "memw($src1+#$src2) = ##$src3",
-            [(store (HexagonCONST32 tglobaladdr:$src3),
-                    (add IntRegs:$src1, u6_2ImmPred:$src2))]>,
-            Requires<[HasV4T]>;
-
-
-// Indexed store double word - global address.
-// memw(Rs+#u6:2)=#S8
-let AddedComplexity = 10 in
-def STrih_offset_ext_V4 : STInst<(outs),
-            (ins IntRegs:$src1, u6_1Imm:$src2, globaladdress:$src3),
-            "memh($src1+#$src2) = ##$src3",
-            [(truncstorei16 (HexagonCONST32 tglobaladdr:$src3),
-                    (add IntRegs:$src1, u6_1ImmPred:$src2))]>,
-            Requires<[HasV4T]>;
diff --git a/lib/Target/Hexagon/HexagonMCInst.h b/lib/Target/Hexagon/HexagonMCInst.h
deleted file mode 100644
index 16ea7cf6ed7..00000000000
--- a/lib/Target/Hexagon/HexagonMCInst.h
+++ /dev/null
@@ -1,41 +0,0 @@
-//===- HexagonMCInst.h - Hexagon sub-class of MCInst ----------------------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This class extends MCInst to allow some VLIW annotation.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef HEXAGONMCINST_H
-#define HEXAGONMCINST_H
-
-#include "llvm/MC/MCInst.h"
-#include "llvm/CodeGen/MachineInstr.h"
-
-namespace llvm {
-  class HexagonMCInst: public MCInst {
-    // Packet start and end markers
-    unsigned startPacket: 1, endPacket: 1;
-    const MachineInstr *MachineI;
-  public:
-    explicit HexagonMCInst(): MCInst(),
-                              startPacket(0), endPacket(0) {}
-
-    const MachineInstr* getMI() const { return MachineI; };
-
-    void setMI(const MachineInstr *MI) { MachineI = MI; };
-
-    bool isStartPacket() const { return (startPacket); };
-    bool isEndPacket() const { return (endPacket); };
-
-    void setStartPacket(bool yes) { startPacket = yes; };
-    void setEndPacket(bool yes) { endPacket = yes; };
-  };
-}
-
-#endif
diff --git a/lib/Target/Hexagon/HexagonMCInstLower.cpp b/lib/Target/Hexagon/HexagonMCInstLower.cpp
index 70bddcc76a5..fbb331bdd8b 100644
--- a/lib/Target/Hexagon/HexagonMCInstLower.cpp
+++ b/lib/Target/Hexagon/HexagonMCInstLower.cpp
@@ -49,7 +49,7 @@ void llvm::HexagonLowerToMC(const MachineInstr* MI, MCInst& MCI,
     switch (MO.getType()) {
     default:
       MI->dump();
-      llvm_unreachable("unknown operand type");
+      assert(0 && "unknown operand type");
     case MachineOperand::MO_Register:
       // Ignore all implicit register operands.
       if (MO.isImplicit()) continue;
diff --git a/lib/Target/Hexagon/HexagonSchedule.td b/lib/Target/Hexagon/HexagonSchedule.td
index c4887963895..fbea4452ec6 100644
--- a/lib/Target/Hexagon/HexagonSchedule.td
+++ b/lib/Target/Hexagon/HexagonSchedule.td
@@ -13,6 +13,7 @@ def LSUNIT    : FuncUnit;
 def MUNIT     : FuncUnit;
 def SUNIT     : FuncUnit;
 
+
 // Itinerary classes
 def ALU32     : InstrItinClass;
 def ALU64     : InstrItinClass;
@@ -23,25 +24,23 @@ def LD        : InstrItinClass;
 def M         : InstrItinClass;
 def ST        : InstrItinClass;
 def S         : InstrItinClass;
-def SYS       : InstrItinClass;
-def MARKER    : InstrItinClass;
 def PSEUDO    : InstrItinClass;
 
+
 def HexagonItineraries :
-      ProcessorItineraries<[LUNIT, LSUNIT, MUNIT, SUNIT], [], [
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>
-      ]>;
+ ProcessorItineraries<[LUNIT, LSUNIT, MUNIT, SUNIT], [], [
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>
+]>;
+
 
 //===----------------------------------------------------------------------===//
 // V4 Machine Info +
diff --git a/lib/Target/Hexagon/HexagonScheduleV4.td b/lib/Target/Hexagon/HexagonScheduleV4.td
index 1d82dbb90e9..4cf66fe7434 100644
--- a/lib/Target/Hexagon/HexagonScheduleV4.td
+++ b/lib/Target/Hexagon/HexagonScheduleV4.td
@@ -23,6 +23,7 @@
 //    | SLOT3     |  XTYPE          ALU32     J         CR           |
 //    |===========|==================================================|
 
+
 // Functional Units.
 def SLOT0       : FuncUnit;
 def SLOT1       : FuncUnit;
@@ -33,26 +34,22 @@ def SLOT3       : FuncUnit;
 def NV_V4       : InstrItinClass;
 def MEM_V4      : InstrItinClass;
 // ALU64/M/S Instruction classes of V2 are collectively knownn as XTYPE in V4.
-def PREFIX      : InstrItinClass;
 
-def HexagonItinerariesV4 :
-      ProcessorItineraries<[SLOT0, SLOT1, SLOT2, SLOT3], [], [
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>,
-        InstrItinData]>
-      ]>;
+def HexagonItinerariesV4 : ProcessorItineraries<
+                    [SLOT0, SLOT1, SLOT2, SLOT3], [], [
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>,
+  InstrItinData]>
+]>;
 
 //===----------------------------------------------------------------------===//
 // Hexagon V4 Resource Definitions -
diff --git a/lib/Target/Hexagon/HexagonTargetMachine.cpp b/lib/Target/Hexagon/HexagonTargetMachine.cpp
index 411325bf963..55bbba7251a 100644
--- a/lib/Target/Hexagon/HexagonTargetMachine.cpp
+++ b/lib/Target/Hexagon/HexagonTargetMachine.cpp
@@ -100,23 +100,23 @@ TargetPassConfig *HexagonTargetMachine::createPassConfig(PassManagerBase &PM) {
 }
 
 bool HexagonPassConfig::addInstSelector() {
-  PM.add(createHexagonRemoveExtendOps(getHexagonTargetMachine()));
-  PM.add(createHexagonISelDag(getHexagonTargetMachine()));
-  PM.add(createHexagonPeephole());
+  PM->add(createHexagonRemoveExtendOps(getHexagonTargetMachine()));
+  PM->add(createHexagonISelDag(getHexagonTargetMachine()));
+  PM->add(createHexagonPeephole());
   return false;
 }
 
 
 bool HexagonPassConfig::addPreRegAlloc() {
   if (!DisableHardwareLoops) {
-    PM.add(createHexagonHardwareLoops());
+    PM->add(createHexagonHardwareLoops());
   }
 
   return false;
 }
 
 bool HexagonPassConfig::addPostRegAlloc() {
-  PM.add(createHexagonCFGOptimizer(getHexagonTargetMachine()));
+  PM->add(createHexagonCFGOptimizer(getHexagonTargetMachine()));
   return true;
 }
 
@@ -129,17 +129,14 @@ bool HexagonPassConfig::addPreSched2() {
 bool HexagonPassConfig::addPreEmitPass() {
 
   if (!DisableHardwareLoops) {
-    PM.add(createHexagonFixupHwLoops());
+    PM->add(createHexagonFixupHwLoops());
   }
 
   // Expand Spill code for predicate registers.
-  PM.add(createHexagonExpandPredSpillCode(getHexagonTargetMachine()));
+  PM->add(createHexagonExpandPredSpillCode(getHexagonTargetMachine()));
 
   // Split up TFRcondsets into conditional transfers.
-  PM.add(createHexagonSplitTFRCondSets(getHexagonTargetMachine()));
-
-  // Create Packets.
-  PM.add(createHexagonPacketizer());
+  PM->add(createHexagonSplitTFRCondSets(getHexagonTargetMachine()));
 
   return false;
 }
diff --git a/lib/Target/Hexagon/HexagonVLIWPacketizer.cpp b/lib/Target/Hexagon/HexagonVLIWPacketizer.cpp
deleted file mode 100644
index c6e7bd1f53d..00000000000
--- a/lib/Target/Hexagon/HexagonVLIWPacketizer.cpp
+++ /dev/null
@@ -1,3642 +0,0 @@
-//===----- HexagonPacketizer.cpp - vliw packetizer ---------------------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This implements a simple VLIW packetizer using DFA. The packetizer works on
-// machine basic blocks. For each instruction I in BB, the packetizer consults
-// the DFA to see if machine resources are available to execute I. If so, the
-// packetizer checks if I depends on any instruction J in the current packet.
-// If no dependency is found, I is added to current packet and machine resource
-// is marked as taken. If any dependency is found, a target API call is made to
-// prune the dependence.
-//
-//===----------------------------------------------------------------------===//
-#define DEBUG_TYPE "packets"
-#include "llvm/CodeGen/DFAPacketizer.h"
-#include "llvm/CodeGen/Passes.h"
-#include "llvm/CodeGen/MachineDominators.h"
-#include "llvm/CodeGen/MachineFunctionPass.h"
-#include "llvm/CodeGen/MachineLoopInfo.h"
-#include "llvm/CodeGen/ScheduleDAG.h"
-#include "llvm/CodeGen/ScheduleDAGInstrs.h"
-#include "llvm/CodeGen/LatencyPriorityQueue.h"
-#include "llvm/CodeGen/SchedulerRegistry.h"
-#include "llvm/CodeGen/MachineFrameInfo.h"
-#include "llvm/CodeGen/MachineInstrBuilder.h"
-#include "llvm/CodeGen/MachineRegisterInfo.h"
-#include "llvm/CodeGen/MachineFunctionAnalysis.h"
-#include "llvm/CodeGen/ScheduleHazardRecognizer.h"
-#include "llvm/Target/TargetMachine.h"
-#include "llvm/Target/TargetInstrInfo.h"
-#include "llvm/Target/TargetRegisterInfo.h"
-#include "llvm/ADT/DenseMap.h"
-#include "llvm/ADT/Statistic.h"
-#include "llvm/Support/MathExtras.h"
-#include "llvm/MC/MCInstrItineraries.h"
-#include "llvm/Support/Compiler.h"
-#include "llvm/Support/CommandLine.h"
-#include "llvm/Support/Debug.h"
-#include "Hexagon.h"
-#include "HexagonTargetMachine.h"
-#include "HexagonRegisterInfo.h"
-#include "HexagonSubtarget.h"
-#include "HexagonMachineFunctionInfo.h"
-
-#include 
-
-using namespace llvm;
-
-namespace {
-  class HexagonPacketizer : public MachineFunctionPass {
-
-  public:
-    static char ID;
-    HexagonPacketizer() : MachineFunctionPass(ID) {}
-
-    void getAnalysisUsage(AnalysisUsage &AU) const {
-      AU.setPreservesCFG();
-      AU.addRequired();
-      AU.addPreserved();
-      AU.addRequired();
-      AU.addPreserved();
-      MachineFunctionPass::getAnalysisUsage(AU);
-    }
-
-    const char *getPassName() const {
-      return "Hexagon Packetizer";
-    }
-
-    bool runOnMachineFunction(MachineFunction &Fn);
-  };
-  char HexagonPacketizer::ID = 0;
-
-  class HexagonPacketizerList : public VLIWPacketizerList {
-
-  private:
-
-    // Has the instruction been promoted to a dot-new instruction.
-    bool PromotedToDotNew;
-
-    // Has the instruction been glued to allocframe.
-    bool GlueAllocframeStore;
-
-    // Has the feeder instruction been glued to new value jump.
-    bool GlueToNewValueJump;
-
-    // Check if there is a dependence between some instruction already in this
-    // packet and this instruction.
-    bool Dependence;
-
-    // Only check for dependence if there are resources available to
-    // schedule this instruction.
-    bool FoundSequentialDependence;
-
-  public:
-    // Ctor.
-    HexagonPacketizerList(MachineFunction &MF, MachineLoopInfo &MLI,
-                          MachineDominatorTree &MDT);
-
-    // initPacketizerState - initialize some internal flags.
-    void initPacketizerState();
-
-    // ignorePseudoInstruction - Ignore bundling of pseudo instructions.
-    bool ignorePseudoInstruction(MachineInstr *MI, MachineBasicBlock *MBB);
-
-    // isSoloInstruction - return true if instruction MI can not be packetized
-    // with any other instruction, which means that MI itself is a packet.
-    bool isSoloInstruction(MachineInstr *MI);
-
-    // isLegalToPacketizeTogether - Is it legal to packetize SUI and SUJ
-    // together.
-    bool isLegalToPacketizeTogether(SUnit *SUI, SUnit *SUJ);
-
-    // isLegalToPruneDependencies - Is it legal to prune dependece between SUI
-    // and SUJ.
-    bool isLegalToPruneDependencies(SUnit *SUI, SUnit *SUJ);
-
-    MachineBasicBlock::iterator addToPacket(MachineInstr *MI);
-  private:
-    bool IsCallDependent(MachineInstr* MI, SDep::Kind DepType, unsigned DepReg);
-    bool PromoteToDotNew(MachineInstr* MI, SDep::Kind DepType,
-                    MachineBasicBlock::iterator &MII,
-                    const TargetRegisterClass* RC);
-    bool CanPromoteToDotNew(MachineInstr* MI, SUnit* PacketSU,
-                    unsigned DepReg,
-                    std::map  MIToSUnit,
-                    MachineBasicBlock::iterator &MII,
-                    const TargetRegisterClass* RC);
-    bool CanPromoteToNewValue(MachineInstr* MI, SUnit* PacketSU,
-                    unsigned DepReg,
-                    std::map  MIToSUnit,
-                    MachineBasicBlock::iterator &MII);
-    bool CanPromoteToNewValueStore(MachineInstr* MI, MachineInstr* PacketMI,
-                    unsigned DepReg,
-                    std::map  MIToSUnit);
-    bool DemoteToDotOld(MachineInstr* MI);
-    bool ArePredicatesComplements(MachineInstr* MI1, MachineInstr* MI2,
-                    std::map  MIToSUnit);
-    bool RestrictingDepExistInPacket(MachineInstr*,
-                    unsigned, std::map );
-    bool isNewifiable(MachineInstr* MI);
-    bool isCondInst(MachineInstr* MI);
-    bool IsNewifyStore (MachineInstr* MI);
-    bool tryAllocateResourcesForConstExt(MachineInstr* MI);
-    bool canReserveResourcesForConstExt(MachineInstr *MI);
-    void reserveResourcesForConstExt(MachineInstr* MI);
-    bool isNewValueInst(MachineInstr* MI);
-    bool isDotNewInst(MachineInstr* MI);
-  };
-}
-
-// HexagonPacketizerList Ctor.
-HexagonPacketizerList::HexagonPacketizerList(
-  MachineFunction &MF, MachineLoopInfo &MLI,MachineDominatorTree &MDT)
-  : VLIWPacketizerList(MF, MLI, MDT, true){
-}
-
-bool HexagonPacketizer::runOnMachineFunction(MachineFunction &Fn) {
-  const TargetInstrInfo *TII = Fn.getTarget().getInstrInfo();
-  MachineLoopInfo &MLI = getAnalysis();
-  MachineDominatorTree &MDT = getAnalysis();
-
-  // Instantiate the packetizer.
-  HexagonPacketizerList Packetizer(Fn, MLI, MDT);
-
-  // DFA state table should not be empty.
-  assert(Packetizer.getResourceTracker() && "Empty DFA table!");
-
-  //
-  // Loop over all basic blocks and remove KILL pseudo-instructions
-  // These instructions confuse the dependence analysis. Consider:
-  // D0 = ...   (Insn 0)
-  // R0 = KILL R0, D0 (Insn 1)
-  // R0 = ... (Insn 2)
-  // Here, Insn 1 will result in the dependence graph not emitting an output
-  // dependence between Insn 0 and Insn 2. This can lead to incorrect
-  // packetization
-  //
-  for (MachineFunction::iterator MBB = Fn.begin(), MBBe = Fn.end();
-       MBB != MBBe; ++MBB) {
-    MachineBasicBlock::iterator End = MBB->end();
-    MachineBasicBlock::iterator MI = MBB->begin();
-    while (MI != End) {
-      if (MI->isKill()) {
-        MachineBasicBlock::iterator DeleteMI = MI;
-        ++MI;
-        MBB->erase(DeleteMI);
-        End = MBB->end();
-        continue;
-      }
-      ++MI;
-    }
-  }
-
-  // Loop over all of the basic blocks.
-  for (MachineFunction::iterator MBB = Fn.begin(), MBBe = Fn.end();
-       MBB != MBBe; ++MBB) {
-    // Find scheduling regions and schedule / packetize each region.
-    unsigned RemainingCount = MBB->size();
-    for(MachineBasicBlock::iterator RegionEnd = MBB->end();
-        RegionEnd != MBB->begin();) {
-      // The next region starts above the previous region. Look backward in the
-      // instruction stream until we find the nearest boundary.
-      MachineBasicBlock::iterator I = RegionEnd;
-      for(;I != MBB->begin(); --I, --RemainingCount) {
-        if (TII->isSchedulingBoundary(llvm::prior(I), MBB, Fn))
-          break;
-      }
-      I = MBB->begin();
-
-      // Skip empty scheduling regions.
-      if (I == RegionEnd) {
-        RegionEnd = llvm::prior(RegionEnd);
-        --RemainingCount;
-        continue;
-      }
-      // Skip regions with one instruction.
-      if (I == llvm::prior(RegionEnd)) {
-        RegionEnd = llvm::prior(RegionEnd);
-        continue;
-      }
-
-      Packetizer.PacketizeMIs(MBB, I, RegionEnd);
-      RegionEnd = I;
-    }
-  }
-
-  return true;
-}
-
-
-static bool IsIndirectCall(MachineInstr* MI) {
-  return ((MI->getOpcode() == Hexagon::CALLR) ||
-          (MI->getOpcode() == Hexagon::CALLRv3));
-}
-
-// Reserve resources for constant extender. Trigure an assertion if
-// reservation fail.
-void HexagonPacketizerList::reserveResourcesForConstExt(MachineInstr* MI) {
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-  MachineInstr *PseudoMI = MI->getParent()->getParent()->CreateMachineInstr(
-                                  QII->get(Hexagon::IMMEXT), MI->getDebugLoc());
-
-  if (ResourceTracker->canReserveResources(PseudoMI)) {
-    ResourceTracker->reserveResources(PseudoMI);
-    MI->getParent()->getParent()->DeleteMachineInstr(PseudoMI);
-  } else {
-    MI->getParent()->getParent()->DeleteMachineInstr(PseudoMI);
-    llvm_unreachable("can not reserve resources for constant extender.");
-  }
-  return;
-}
-
-bool HexagonPacketizerList::canReserveResourcesForConstExt(MachineInstr *MI) {
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-  assert(QII->isExtended(MI) &&
-         "Should only be called for constant extended instructions");
-  MachineFunction *MF = MI->getParent()->getParent();
-  MachineInstr *PseudoMI = MF->CreateMachineInstr(QII->get(Hexagon::IMMEXT),
-                                                  MI->getDebugLoc());
-  bool CanReserve = ResourceTracker->canReserveResources(PseudoMI);
-  MF->DeleteMachineInstr(PseudoMI);
-  return CanReserve;
-}
-
-// Allocate resources (i.e. 4 bytes) for constant extender. If succeed, return
-// true, otherwise, return false.
-bool HexagonPacketizerList::tryAllocateResourcesForConstExt(MachineInstr* MI) {
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-  MachineInstr *PseudoMI = MI->getParent()->getParent()->CreateMachineInstr(
-                                  QII->get(Hexagon::IMMEXT), MI->getDebugLoc());
-
-  if (ResourceTracker->canReserveResources(PseudoMI)) {
-    ResourceTracker->reserveResources(PseudoMI);
-    MI->getParent()->getParent()->DeleteMachineInstr(PseudoMI);
-    return true;
-  } else {
-    MI->getParent()->getParent()->DeleteMachineInstr(PseudoMI);
-    return false;
-  }
-}
-
-
-bool HexagonPacketizerList::IsCallDependent(MachineInstr* MI,
-                                          SDep::Kind DepType,
-                                          unsigned DepReg) {
-
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-  const HexagonRegisterInfo* QRI = (const HexagonRegisterInfo *) TM.getRegisterInfo();
-
-  // Check for lr dependence
-  if (DepReg == QRI->getRARegister()) {
-    return true;
-  }
-
-  if (QII->isDeallocRet(MI)) {
-    if (DepReg == QRI->getFrameRegister() ||
-        DepReg == QRI->getStackRegister())
-      return true;
-  }
-
-  // Check if this is a predicate dependence
-  const TargetRegisterClass* RC = QRI->getMinimalPhysRegClass(DepReg);
-  if (RC == Hexagon::PredRegsRegisterClass) {
-    return true;
-  }
-
-  //
-  // Lastly check for an operand used in an indirect call
-  // If we had an attribute for checking if an instruction is an indirect call,
-  // then we could have avoided this relatively brittle implementation of
-  // IsIndirectCall()
-  //
-  // Assumes that the first operand of the CALLr is the function address
-  //
-  if (IsIndirectCall(MI) && (DepType == SDep::Data)) {
-    MachineOperand MO = MI->getOperand(0);
-    if (MO.isReg() && MO.isUse() && (MO.getReg() == DepReg)) {
-      return true;
-    }
-  }
-
-  return false;
-}
-
-static bool IsRegDependence(const SDep::Kind DepType) {
-  return (DepType == SDep::Data || DepType == SDep::Anti ||
-          DepType == SDep::Output);
-}
-
-static bool IsDirectJump(MachineInstr* MI) {
-  return (MI->getOpcode() == Hexagon::JMP);
-}
-
-static bool IsSchedBarrier(MachineInstr* MI) {
-  switch (MI->getOpcode()) {
-  case Hexagon::BARRIER:
-    return true;
-  }
-  return false;
-}
-
-static bool IsControlFlow(MachineInstr* MI) {
-  return (MI->getDesc().isTerminator() || MI->getDesc().isCall());
-}
-
-bool HexagonPacketizerList::isNewValueInst(MachineInstr* MI) {
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-  if (QII->isNewValueJump(MI))
-    return true;
-
-  if (QII->isNewValueStore(MI))
-    return true;
-
-  return false;
-}
-
-// Function returns true if an instruction can be promoted to the new-value
-// store. It will always return false for v2 and v3.
-// It lists all the conditional and unconditional stores that can be promoted
-// to the new-value stores.
-
-bool HexagonPacketizerList::IsNewifyStore (MachineInstr* MI) {
-  const HexagonRegisterInfo* QRI = (const HexagonRegisterInfo *) TM.getRegisterInfo();
-  switch (MI->getOpcode())
-  {
-    // store byte
-    case Hexagon::STrib:
-    case Hexagon::STrib_indexed:
-    case Hexagon::STrib_indexed_shl_V4:
-    case Hexagon::STrib_shl_V4:
-    case Hexagon::STrib_GP_V4:
-    case Hexagon::STb_GP_V4:
-    case Hexagon::POST_STbri:
-    case Hexagon::STrib_cPt:
-    case Hexagon::STrib_cdnPt_V4:
-    case Hexagon::STrib_cNotPt:
-    case Hexagon::STrib_cdnNotPt_V4:
-    case Hexagon::STrib_indexed_cPt:
-    case Hexagon::STrib_indexed_cdnPt_V4:
-    case Hexagon::STrib_indexed_cNotPt:
-    case Hexagon::STrib_indexed_cdnNotPt_V4:
-    case Hexagon::STrib_indexed_shl_cPt_V4:
-    case Hexagon::STrib_indexed_shl_cdnPt_V4:
-    case Hexagon::STrib_indexed_shl_cNotPt_V4:
-    case Hexagon::STrib_indexed_shl_cdnNotPt_V4:
-    case Hexagon::POST_STbri_cPt:
-    case Hexagon::POST_STbri_cdnPt_V4:
-    case Hexagon::POST_STbri_cNotPt:
-    case Hexagon::POST_STbri_cdnNotPt_V4:
-    case Hexagon::STb_GP_cPt_V4:
-    case Hexagon::STb_GP_cNotPt_V4:
-    case Hexagon::STb_GP_cdnPt_V4:
-    case Hexagon::STb_GP_cdnNotPt_V4:
-    case Hexagon::STrib_GP_cPt_V4:
-    case Hexagon::STrib_GP_cNotPt_V4:
-    case Hexagon::STrib_GP_cdnPt_V4:
-    case Hexagon::STrib_GP_cdnNotPt_V4:
-
-    // store halfword
-    case Hexagon::STrih:
-    case Hexagon::STrih_indexed:
-    case Hexagon::STrih_indexed_shl_V4:
-    case Hexagon::STrih_shl_V4:
-    case Hexagon::STrih_GP_V4:
-    case Hexagon::STh_GP_V4:
-    case Hexagon::POST_SThri:
-    case Hexagon::STrih_cPt:
-    case Hexagon::STrih_cdnPt_V4:
-    case Hexagon::STrih_cNotPt:
-    case Hexagon::STrih_cdnNotPt_V4:
-    case Hexagon::STrih_indexed_cPt:
-    case Hexagon::STrih_indexed_cdnPt_V4:
-    case Hexagon::STrih_indexed_cNotPt:
-    case Hexagon::STrih_indexed_cdnNotPt_V4:
-    case Hexagon::STrih_indexed_shl_cPt_V4:
-    case Hexagon::STrih_indexed_shl_cdnPt_V4:
-    case Hexagon::STrih_indexed_shl_cNotPt_V4:
-    case Hexagon::STrih_indexed_shl_cdnNotPt_V4:
-    case Hexagon::POST_SThri_cPt:
-    case Hexagon::POST_SThri_cdnPt_V4:
-    case Hexagon::POST_SThri_cNotPt:
-    case Hexagon::POST_SThri_cdnNotPt_V4:
-    case Hexagon::STh_GP_cPt_V4:
-    case Hexagon::STh_GP_cNotPt_V4:
-    case Hexagon::STh_GP_cdnPt_V4:
-    case Hexagon::STh_GP_cdnNotPt_V4:
-    case Hexagon::STrih_GP_cPt_V4:
-    case Hexagon::STrih_GP_cNotPt_V4:
-    case Hexagon::STrih_GP_cdnPt_V4:
-    case Hexagon::STrih_GP_cdnNotPt_V4:
-
-    // store word
-    case Hexagon::STriw:
-    case Hexagon::STriw_indexed:
-    case Hexagon::STriw_indexed_shl_V4:
-    case Hexagon::STriw_shl_V4:
-    case Hexagon::STriw_GP_V4:
-    case Hexagon::STw_GP_V4:
-    case Hexagon::POST_STwri:
-    case Hexagon::STriw_cPt:
-    case Hexagon::STriw_cdnPt_V4:
-    case Hexagon::STriw_cNotPt:
-    case Hexagon::STriw_cdnNotPt_V4:
-    case Hexagon::STriw_indexed_cPt:
-    case Hexagon::STriw_indexed_cdnPt_V4:
-    case Hexagon::STriw_indexed_cNotPt:
-    case Hexagon::STriw_indexed_cdnNotPt_V4:
-    case Hexagon::STriw_indexed_shl_cPt_V4:
-    case Hexagon::STriw_indexed_shl_cdnPt_V4:
-    case Hexagon::STriw_indexed_shl_cNotPt_V4:
-    case Hexagon::STriw_indexed_shl_cdnNotPt_V4:
-    case Hexagon::POST_STwri_cPt:
-    case Hexagon::POST_STwri_cdnPt_V4:
-    case Hexagon::POST_STwri_cNotPt:
-    case Hexagon::POST_STwri_cdnNotPt_V4:
-    case Hexagon::STw_GP_cPt_V4:
-    case Hexagon::STw_GP_cNotPt_V4:
-    case Hexagon::STw_GP_cdnPt_V4:
-    case Hexagon::STw_GP_cdnNotPt_V4:
-    case Hexagon::STriw_GP_cPt_V4:
-    case Hexagon::STriw_GP_cNotPt_V4:
-    case Hexagon::STriw_GP_cdnPt_V4:
-    case Hexagon::STriw_GP_cdnNotPt_V4:
-        return QRI->Subtarget.hasV4TOps();
-  }
-  return false;
-}
-
-static bool IsLoopN(MachineInstr *MI) {
-  return (MI->getOpcode() == Hexagon::LOOP0_i ||
-          MI->getOpcode() == Hexagon::LOOP0_r);
-}
-
-/// DoesModifyCalleeSavedReg - Returns true if the instruction modifies a
-/// callee-saved register.
-static bool DoesModifyCalleeSavedReg(MachineInstr *MI,
-                                     const TargetRegisterInfo *TRI) {
-  for (const uint16_t *CSR = TRI->getCalleeSavedRegs(); *CSR; ++CSR) {
-    unsigned CalleeSavedReg = *CSR;
-    if (MI->modifiesRegister(CalleeSavedReg, TRI))
-      return true;
-  }
-  return false;
-}
-
-// Return the new value instruction for a given store.
-static int GetDotNewOp(const int opc) {
-  switch (opc) {
-  default: llvm_unreachable("Unknown .new type");
-
-  // store new value byte
-  case Hexagon::STrib:
-    return Hexagon::STrib_nv_V4;
-
-  case Hexagon::STrib_indexed:
-    return Hexagon::STrib_indexed_nv_V4;
-
-  case Hexagon::STrib_indexed_shl_V4:
-    return Hexagon::STrib_indexed_shl_nv_V4;
-
-  case Hexagon::STrib_shl_V4:
-    return Hexagon::STrib_shl_nv_V4;
-
-  case Hexagon::STrib_GP_V4:
-    return Hexagon::STrib_GP_nv_V4;
-
-  case Hexagon::STb_GP_V4:
-    return Hexagon::STb_GP_nv_V4;
-
-  case Hexagon::POST_STbri:
-    return Hexagon::POST_STbri_nv_V4;
-
-  case Hexagon::STrib_cPt:
-    return Hexagon::STrib_cPt_nv_V4;
-
-  case Hexagon::STrib_cdnPt_V4:
-    return Hexagon::STrib_cdnPt_nv_V4;
-
-  case Hexagon::STrib_cNotPt:
-    return Hexagon::STrib_cNotPt_nv_V4;
-
-  case Hexagon::STrib_cdnNotPt_V4:
-    return Hexagon::STrib_cdnNotPt_nv_V4;
-
-  case Hexagon::STrib_indexed_cPt:
-    return Hexagon::STrib_indexed_cPt_nv_V4;
-
-  case Hexagon::STrib_indexed_cdnPt_V4:
-    return Hexagon::STrib_indexed_cdnPt_nv_V4;
-
-  case Hexagon::STrib_indexed_cNotPt:
-    return Hexagon::STrib_indexed_cNotPt_nv_V4;
-
-  case Hexagon::STrib_indexed_cdnNotPt_V4:
-    return Hexagon::STrib_indexed_cdnNotPt_nv_V4;
-
-  case Hexagon::STrib_indexed_shl_cPt_V4:
-    return Hexagon::STrib_indexed_shl_cPt_nv_V4;
-
-  case Hexagon::STrib_indexed_shl_cdnPt_V4:
-    return Hexagon::STrib_indexed_shl_cdnPt_nv_V4;
-
-  case Hexagon::STrib_indexed_shl_cNotPt_V4:
-    return Hexagon::STrib_indexed_shl_cNotPt_nv_V4;
-
-  case Hexagon::STrib_indexed_shl_cdnNotPt_V4:
-    return Hexagon::STrib_indexed_shl_cdnNotPt_nv_V4;
-
-  case Hexagon::POST_STbri_cPt:
-    return Hexagon::POST_STbri_cPt_nv_V4;
-
-  case Hexagon::POST_STbri_cdnPt_V4:
-    return Hexagon::POST_STbri_cdnPt_nv_V4;
-
-  case Hexagon::POST_STbri_cNotPt:
-    return Hexagon::POST_STbri_cNotPt_nv_V4;
-
-  case Hexagon::POST_STbri_cdnNotPt_V4:
-    return Hexagon::POST_STbri_cdnNotPt_nv_V4;
-
-  case Hexagon::STb_GP_cPt_V4:
-    return Hexagon::STb_GP_cPt_nv_V4;
-
-  case Hexagon::STb_GP_cNotPt_V4:
-    return Hexagon::STb_GP_cNotPt_nv_V4;
-
-  case Hexagon::STb_GP_cdnPt_V4:
-    return Hexagon::STb_GP_cdnPt_nv_V4;
-
-  case Hexagon::STb_GP_cdnNotPt_V4:
-    return Hexagon::STb_GP_cdnNotPt_nv_V4;
-
-  case Hexagon::STrib_GP_cPt_V4:
-    return Hexagon::STrib_GP_cPt_nv_V4;
-
-  case Hexagon::STrib_GP_cNotPt_V4:
-    return Hexagon::STrib_GP_cNotPt_nv_V4;
-
-  case Hexagon::STrib_GP_cdnPt_V4:
-    return Hexagon::STrib_GP_cdnPt_nv_V4;
-
-  case Hexagon::STrib_GP_cdnNotPt_V4:
-    return Hexagon::STrib_GP_cdnNotPt_nv_V4;
-
-  // store new value halfword
-  case Hexagon::STrih:
-    return Hexagon::STrih_nv_V4;
-
-  case Hexagon::STrih_indexed:
-    return Hexagon::STrih_indexed_nv_V4;
-
-  case Hexagon::STrih_indexed_shl_V4:
-    return Hexagon::STrih_indexed_shl_nv_V4;
-
-  case Hexagon::STrih_shl_V4:
-    return Hexagon::STrih_shl_nv_V4;
-
-  case Hexagon::STrih_GP_V4:
-    return Hexagon::STrih_GP_nv_V4;
-
-  case Hexagon::STh_GP_V4:
-    return Hexagon::STh_GP_nv_V4;
-
-  case Hexagon::POST_SThri:
-    return Hexagon::POST_SThri_nv_V4;
-
-  case Hexagon::STrih_cPt:
-    return Hexagon::STrih_cPt_nv_V4;
-
-  case Hexagon::STrih_cdnPt_V4:
-    return Hexagon::STrih_cdnPt_nv_V4;
-
-  case Hexagon::STrih_cNotPt:
-    return Hexagon::STrih_cNotPt_nv_V4;
-
-  case Hexagon::STrih_cdnNotPt_V4:
-    return Hexagon::STrih_cdnNotPt_nv_V4;
-
-  case Hexagon::STrih_indexed_cPt:
-    return Hexagon::STrih_indexed_cPt_nv_V4;
-
-  case Hexagon::STrih_indexed_cdnPt_V4:
-    return Hexagon::STrih_indexed_cdnPt_nv_V4;
-
-  case Hexagon::STrih_indexed_cNotPt:
-    return Hexagon::STrih_indexed_cNotPt_nv_V4;
-
-  case Hexagon::STrih_indexed_cdnNotPt_V4:
-    return Hexagon::STrih_indexed_cdnNotPt_nv_V4;
-
-  case Hexagon::STrih_indexed_shl_cPt_V4:
-    return Hexagon::STrih_indexed_shl_cPt_nv_V4;
-
-  case Hexagon::STrih_indexed_shl_cdnPt_V4:
-    return Hexagon::STrih_indexed_shl_cdnPt_nv_V4;
-
-  case Hexagon::STrih_indexed_shl_cNotPt_V4:
-    return Hexagon::STrih_indexed_shl_cNotPt_nv_V4;
-
-  case Hexagon::STrih_indexed_shl_cdnNotPt_V4:
-    return Hexagon::STrih_indexed_shl_cdnNotPt_nv_V4;
-
-  case Hexagon::POST_SThri_cPt:
-    return Hexagon::POST_SThri_cPt_nv_V4;
-
-  case Hexagon::POST_SThri_cdnPt_V4:
-    return Hexagon::POST_SThri_cdnPt_nv_V4;
-
-  case Hexagon::POST_SThri_cNotPt:
-    return Hexagon::POST_SThri_cNotPt_nv_V4;
-
-  case Hexagon::POST_SThri_cdnNotPt_V4:
-    return Hexagon::POST_SThri_cdnNotPt_nv_V4;
-
-  case Hexagon::STh_GP_cPt_V4:
-    return Hexagon::STh_GP_cPt_nv_V4;
-
-  case Hexagon::STh_GP_cNotPt_V4:
-    return Hexagon::STh_GP_cNotPt_nv_V4;
-
-  case Hexagon::STh_GP_cdnPt_V4:
-    return Hexagon::STh_GP_cdnPt_nv_V4;
-
-  case Hexagon::STh_GP_cdnNotPt_V4:
-    return Hexagon::STh_GP_cdnNotPt_nv_V4;
-
-  case Hexagon::STrih_GP_cPt_V4:
-    return Hexagon::STrih_GP_cPt_nv_V4;
-
-  case Hexagon::STrih_GP_cNotPt_V4:
-    return Hexagon::STrih_GP_cNotPt_nv_V4;
-
-  case Hexagon::STrih_GP_cdnPt_V4:
-    return Hexagon::STrih_GP_cdnPt_nv_V4;
-
-  case Hexagon::STrih_GP_cdnNotPt_V4:
-    return Hexagon::STrih_GP_cdnNotPt_nv_V4;
-
-  // store new value word
-  case Hexagon::STriw:
-    return Hexagon::STriw_nv_V4;
-
-  case Hexagon::STriw_indexed:
-    return Hexagon::STriw_indexed_nv_V4;
-
-  case Hexagon::STriw_indexed_shl_V4:
-    return Hexagon::STriw_indexed_shl_nv_V4;
-
-  case Hexagon::STriw_shl_V4:
-    return Hexagon::STriw_shl_nv_V4;
-
-  case Hexagon::STriw_GP_V4:
-    return Hexagon::STriw_GP_nv_V4;
-
-  case Hexagon::STw_GP_V4:
-    return Hexagon::STw_GP_nv_V4;
-
-  case Hexagon::POST_STwri:
-    return Hexagon::POST_STwri_nv_V4;
-
-  case Hexagon::STriw_cPt:
-    return Hexagon::STriw_cPt_nv_V4;
-
-  case Hexagon::STriw_cdnPt_V4:
-    return Hexagon::STriw_cdnPt_nv_V4;
-
-  case Hexagon::STriw_cNotPt:
-    return Hexagon::STriw_cNotPt_nv_V4;
-
-  case Hexagon::STriw_cdnNotPt_V4:
-    return Hexagon::STriw_cdnNotPt_nv_V4;
-
-  case Hexagon::STriw_indexed_cPt:
-    return Hexagon::STriw_indexed_cPt_nv_V4;
-
-  case Hexagon::STriw_indexed_cdnPt_V4:
-    return Hexagon::STriw_indexed_cdnPt_nv_V4;
-
-  case Hexagon::STriw_indexed_cNotPt:
-    return Hexagon::STriw_indexed_cNotPt_nv_V4;
-
-  case Hexagon::STriw_indexed_cdnNotPt_V4:
-    return Hexagon::STriw_indexed_cdnNotPt_nv_V4;
-
-  case Hexagon::STriw_indexed_shl_cPt_V4:
-    return Hexagon::STriw_indexed_shl_cPt_nv_V4;
-
-  case Hexagon::STriw_indexed_shl_cdnPt_V4:
-    return Hexagon::STriw_indexed_shl_cdnPt_nv_V4;
-
-  case Hexagon::STriw_indexed_shl_cNotPt_V4:
-    return Hexagon::STriw_indexed_shl_cNotPt_nv_V4;
-
-  case Hexagon::STriw_indexed_shl_cdnNotPt_V4:
-    return Hexagon::STriw_indexed_shl_cdnNotPt_nv_V4;
-
-  case Hexagon::POST_STwri_cPt:
-    return Hexagon::POST_STwri_cPt_nv_V4;
-
-  case Hexagon::POST_STwri_cdnPt_V4:
-    return Hexagon::POST_STwri_cdnPt_nv_V4;
-
-  case Hexagon::POST_STwri_cNotPt:
-    return Hexagon::POST_STwri_cNotPt_nv_V4;
-
-  case Hexagon::POST_STwri_cdnNotPt_V4:
-    return Hexagon::POST_STwri_cdnNotPt_nv_V4;
-
-  case Hexagon::STw_GP_cPt_V4:
-    return Hexagon::STw_GP_cPt_nv_V4;
-
-  case Hexagon::STw_GP_cNotPt_V4:
-    return Hexagon::STw_GP_cNotPt_nv_V4;
-
-  case Hexagon::STw_GP_cdnPt_V4:
-    return Hexagon::STw_GP_cdnPt_nv_V4;
-
-  case Hexagon::STw_GP_cdnNotPt_V4:
-    return Hexagon::STw_GP_cdnNotPt_nv_V4;
-
-  case Hexagon::STriw_GP_cPt_V4:
-    return Hexagon::STriw_GP_cPt_nv_V4;
-
-  case Hexagon::STriw_GP_cNotPt_V4:
-    return Hexagon::STriw_GP_cNotPt_nv_V4;
-
-  case Hexagon::STriw_GP_cdnPt_V4:
-    return Hexagon::STriw_GP_cdnPt_nv_V4;
-
-  case Hexagon::STriw_GP_cdnNotPt_V4:
-    return Hexagon::STriw_GP_cdnNotPt_nv_V4;
-  }
-}
-
-// Return .new predicate version for an instruction
-static int GetDotNewPredOp(const int opc) {
-  switch (opc) {
-  default: llvm_unreachable("Unknown .new type");
-
-  // Conditional stores
-  // Store byte conditionally
-  case Hexagon::STrib_cPt :
-    return Hexagon::STrib_cdnPt_V4;
-
-  case Hexagon::STrib_cNotPt :
-    return Hexagon::STrib_cdnNotPt_V4;
-
-  case Hexagon::STrib_indexed_cPt :
-    return Hexagon::STrib_indexed_cdnPt_V4;
-
-  case Hexagon::STrib_indexed_cNotPt :
-    return Hexagon::STrib_indexed_cdnNotPt_V4;
-
-  case Hexagon::STrib_imm_cPt_V4 :
-    return Hexagon::STrib_imm_cdnPt_V4;
-
-  case Hexagon::STrib_imm_cNotPt_V4 :
-    return Hexagon::STrib_imm_cdnNotPt_V4;
-
-  case Hexagon::POST_STbri_cPt :
-    return Hexagon::POST_STbri_cdnPt_V4;
-
-  case Hexagon::POST_STbri_cNotPt :
-    return Hexagon::POST_STbri_cdnNotPt_V4;
-
-  case Hexagon::STrib_indexed_shl_cPt_V4 :
-    return Hexagon::STrib_indexed_shl_cdnPt_V4;
-
-  case Hexagon::STrib_indexed_shl_cNotPt_V4 :
-    return Hexagon::STrib_indexed_shl_cdnNotPt_V4;
-
-  case Hexagon::STb_GP_cPt_V4 :
-    return Hexagon::STb_GP_cdnPt_V4;
-
-  case Hexagon::STb_GP_cNotPt_V4 :
-    return Hexagon::STb_GP_cdnNotPt_V4;
-
-  case Hexagon::STrib_GP_cPt_V4 :
-    return Hexagon::STrib_GP_cdnPt_V4;
-
-  case Hexagon::STrib_GP_cNotPt_V4 :
-    return Hexagon::STrib_GP_cdnNotPt_V4;
-
-  // Store doubleword conditionally
-  case Hexagon::STrid_cPt :
-    return Hexagon::STrid_cdnPt_V4;
-
-  case Hexagon::STrid_cNotPt :
-    return Hexagon::STrid_cdnNotPt_V4;
-
-  case Hexagon::STrid_indexed_cPt :
-    return Hexagon::STrid_indexed_cdnPt_V4;
-
-  case Hexagon::STrid_indexed_cNotPt :
-    return Hexagon::STrid_indexed_cdnNotPt_V4;
-
-  case Hexagon::STrid_indexed_shl_cPt_V4 :
-    return Hexagon::STrid_indexed_shl_cdnPt_V4;
-
-  case Hexagon::STrid_indexed_shl_cNotPt_V4 :
-    return Hexagon::STrid_indexed_shl_cdnNotPt_V4;
-
-  case Hexagon::POST_STdri_cPt :
-    return Hexagon::POST_STdri_cdnPt_V4;
-
-  case Hexagon::POST_STdri_cNotPt :
-    return Hexagon::POST_STdri_cdnNotPt_V4;
-
-  case Hexagon::STd_GP_cPt_V4 :
-    return Hexagon::STd_GP_cdnPt_V4;
-
-  case Hexagon::STd_GP_cNotPt_V4 :
-    return Hexagon::STd_GP_cdnNotPt_V4;
-
-  case Hexagon::STrid_GP_cPt_V4 :
-    return Hexagon::STrid_GP_cdnPt_V4;
-
-  case Hexagon::STrid_GP_cNotPt_V4 :
-    return Hexagon::STrid_GP_cdnNotPt_V4;
-
-  // Store halfword conditionally
-  case Hexagon::STrih_cPt :
-    return Hexagon::STrih_cdnPt_V4;
-
-  case Hexagon::STrih_cNotPt :
-    return Hexagon::STrih_cdnNotPt_V4;
-
-  case Hexagon::STrih_indexed_cPt :
-    return Hexagon::STrih_indexed_cdnPt_V4;
-
-  case Hexagon::STrih_indexed_cNotPt :
-    return Hexagon::STrih_indexed_cdnNotPt_V4;
-
-  case Hexagon::STrih_imm_cPt_V4 :
-    return Hexagon::STrih_imm_cdnPt_V4;
-
-  case Hexagon::STrih_imm_cNotPt_V4 :
-    return Hexagon::STrih_imm_cdnNotPt_V4;
-
-  case Hexagon::STrih_indexed_shl_cPt_V4 :
-    return Hexagon::STrih_indexed_shl_cdnPt_V4;
-
-  case Hexagon::STrih_indexed_shl_cNotPt_V4 :
-    return Hexagon::STrih_indexed_shl_cdnNotPt_V4;
-
-  case Hexagon::POST_SThri_cPt :
-    return Hexagon::POST_SThri_cdnPt_V4;
-
-  case Hexagon::POST_SThri_cNotPt :
-    return Hexagon::POST_SThri_cdnNotPt_V4;
-
-  case Hexagon::STh_GP_cPt_V4 :
-    return Hexagon::STh_GP_cdnPt_V4;
-
-  case Hexagon::STh_GP_cNotPt_V4 :
-    return Hexagon::STh_GP_cdnNotPt_V4;
-
-  case Hexagon::STrih_GP_cPt_V4 :
-    return Hexagon::STrih_GP_cdnPt_V4;
-
-  case Hexagon::STrih_GP_cNotPt_V4 :
-    return Hexagon::STrih_GP_cdnNotPt_V4;
-
-  // Store word conditionally
-  case Hexagon::STriw_cPt :
-    return Hexagon::STriw_cdnPt_V4;
-
-  case Hexagon::STriw_cNotPt :
-    return Hexagon::STriw_cdnNotPt_V4;
-
-  case Hexagon::STriw_indexed_cPt :
-    return Hexagon::STriw_indexed_cdnPt_V4;
-
-  case Hexagon::STriw_indexed_cNotPt :
-    return Hexagon::STriw_indexed_cdnNotPt_V4;
-
-  case Hexagon::STriw_imm_cPt_V4 :
-    return Hexagon::STriw_imm_cdnPt_V4;
-
-  case Hexagon::STriw_imm_cNotPt_V4 :
-    return Hexagon::STriw_imm_cdnNotPt_V4;
-
-  case Hexagon::STriw_indexed_shl_cPt_V4 :
-    return Hexagon::STriw_indexed_shl_cdnPt_V4;
-
-  case Hexagon::STriw_indexed_shl_cNotPt_V4 :
-    return Hexagon::STriw_indexed_shl_cdnNotPt_V4;
-
-  case Hexagon::POST_STwri_cPt :
-    return Hexagon::POST_STwri_cdnPt_V4;
-
-  case Hexagon::POST_STwri_cNotPt :
-    return Hexagon::POST_STwri_cdnNotPt_V4;
-
-  case Hexagon::STw_GP_cPt_V4 :
-    return Hexagon::STw_GP_cdnPt_V4;
-
-  case Hexagon::STw_GP_cNotPt_V4 :
-    return Hexagon::STw_GP_cdnNotPt_V4;
-
-  case Hexagon::STriw_GP_cPt_V4 :
-    return Hexagon::STriw_GP_cdnPt_V4;
-
-  case Hexagon::STriw_GP_cNotPt_V4 :
-    return Hexagon::STriw_GP_cdnNotPt_V4;
-
-  // Condtional Jumps
-  case Hexagon::JMP_c:
-    return Hexagon::JMP_cdnPt;
-
-  case Hexagon::JMP_cNot:
-    return Hexagon::JMP_cdnNotPt;
-
-  case Hexagon::JMPR_cPt:
-    return Hexagon::JMPR_cdnPt_V3;
-
-  case Hexagon::JMPR_cNotPt:
-    return Hexagon::JMPR_cdnNotPt_V3;
-
-  // Conditional Transfers
-  case Hexagon::TFR_cPt:
-    return Hexagon::TFR_cdnPt;
-
-  case Hexagon::TFR_cNotPt:
-    return Hexagon::TFR_cdnNotPt;
-
-  case Hexagon::TFRI_cPt:
-    return Hexagon::TFRI_cdnPt;
-
-  case Hexagon::TFRI_cNotPt:
-    return Hexagon::TFRI_cdnNotPt;
-
-  // Load double word
-  case Hexagon::LDrid_cPt :
-    return Hexagon::LDrid_cdnPt;
-
-  case Hexagon::LDrid_cNotPt :
-    return Hexagon::LDrid_cdnNotPt;
-
-  case Hexagon::LDrid_indexed_cPt :
-    return Hexagon::LDrid_indexed_cdnPt;
-
-  case Hexagon::LDrid_indexed_cNotPt :
-    return Hexagon::LDrid_indexed_cdnNotPt;
-
-  case Hexagon::POST_LDrid_cPt :
-    return Hexagon::POST_LDrid_cdnPt_V4;
-
-  case Hexagon::POST_LDrid_cNotPt :
-    return Hexagon::POST_LDrid_cdnNotPt_V4;
-
-  // Load word
-  case Hexagon::LDriw_cPt :
-    return Hexagon::LDriw_cdnPt;
-
-  case Hexagon::LDriw_cNotPt :
-    return Hexagon::LDriw_cdnNotPt;
-
-  case Hexagon::LDriw_indexed_cPt :
-    return Hexagon::LDriw_indexed_cdnPt;
-
-  case Hexagon::LDriw_indexed_cNotPt :
-    return Hexagon::LDriw_indexed_cdnNotPt;
-
-  case Hexagon::POST_LDriw_cPt :
-    return Hexagon::POST_LDriw_cdnPt_V4;
-
-  case Hexagon::POST_LDriw_cNotPt :
-    return Hexagon::POST_LDriw_cdnNotPt_V4;
-
-  // Load halfword
-  case Hexagon::LDrih_cPt :
-    return Hexagon::LDrih_cdnPt;
-
-  case Hexagon::LDrih_cNotPt :
-    return Hexagon::LDrih_cdnNotPt;
-
-  case Hexagon::LDrih_indexed_cPt :
-    return Hexagon::LDrih_indexed_cdnPt;
-
-  case Hexagon::LDrih_indexed_cNotPt :
-    return Hexagon::LDrih_indexed_cdnNotPt;
-
-  case Hexagon::POST_LDrih_cPt :
-    return Hexagon::POST_LDrih_cdnPt_V4;
-
-  case Hexagon::POST_LDrih_cNotPt :
-    return Hexagon::POST_LDrih_cdnNotPt_V4;
-
-  // Load byte
-  case Hexagon::LDrib_cPt :
-    return Hexagon::LDrib_cdnPt;
-
-  case Hexagon::LDrib_cNotPt :
-    return Hexagon::LDrib_cdnNotPt;
-
-  case Hexagon::LDrib_indexed_cPt :
-    return Hexagon::LDrib_indexed_cdnPt;
-
-  case Hexagon::LDrib_indexed_cNotPt :
-    return Hexagon::LDrib_indexed_cdnNotPt;
-
-  case Hexagon::POST_LDrib_cPt :
-    return Hexagon::POST_LDrib_cdnPt_V4;
-
-  case Hexagon::POST_LDrib_cNotPt :
-    return Hexagon::POST_LDrib_cdnNotPt_V4;
-
-  // Load unsigned halfword
-  case Hexagon::LDriuh_cPt :
-    return Hexagon::LDriuh_cdnPt;
-
-  case Hexagon::LDriuh_cNotPt :
-    return Hexagon::LDriuh_cdnNotPt;
-
-  case Hexagon::LDriuh_indexed_cPt :
-    return Hexagon::LDriuh_indexed_cdnPt;
-
-  case Hexagon::LDriuh_indexed_cNotPt :
-    return Hexagon::LDriuh_indexed_cdnNotPt;
-
-  case Hexagon::POST_LDriuh_cPt :
-    return Hexagon::POST_LDriuh_cdnPt_V4;
-
-  case Hexagon::POST_LDriuh_cNotPt :
-    return Hexagon::POST_LDriuh_cdnNotPt_V4;
-
-  // Load unsigned byte
-  case Hexagon::LDriub_cPt :
-    return Hexagon::LDriub_cdnPt;
-
-  case Hexagon::LDriub_cNotPt :
-    return Hexagon::LDriub_cdnNotPt;
-
-  case Hexagon::LDriub_indexed_cPt :
-    return Hexagon::LDriub_indexed_cdnPt;
-
-  case Hexagon::LDriub_indexed_cNotPt :
-    return Hexagon::LDriub_indexed_cdnNotPt;
-
-  case Hexagon::POST_LDriub_cPt :
-    return Hexagon::POST_LDriub_cdnPt_V4;
-
-  case Hexagon::POST_LDriub_cNotPt :
-    return Hexagon::POST_LDriub_cdnNotPt_V4;
-
-  // V4 indexed+scaled load
-
-  case Hexagon::LDrid_indexed_cPt_V4 :
-    return Hexagon::LDrid_indexed_cdnPt_V4;
-
-  case Hexagon::LDrid_indexed_cNotPt_V4 :
-    return Hexagon::LDrid_indexed_cdnNotPt_V4;
-
-  case Hexagon::LDrid_indexed_shl_cPt_V4 :
-    return Hexagon::LDrid_indexed_shl_cdnPt_V4;
-
-  case Hexagon::LDrid_indexed_shl_cNotPt_V4 :
-    return Hexagon::LDrid_indexed_shl_cdnNotPt_V4;
-
-  case Hexagon::LDrib_indexed_cPt_V4 :
-    return Hexagon::LDrib_indexed_cdnPt_V4;
-
-  case Hexagon::LDrib_indexed_cNotPt_V4 :
-    return Hexagon::LDrib_indexed_cdnNotPt_V4;
-
-  case Hexagon::LDrib_indexed_shl_cPt_V4 :
-    return Hexagon::LDrib_indexed_shl_cdnPt_V4;
-
-  case Hexagon::LDrib_indexed_shl_cNotPt_V4 :
-    return Hexagon::LDrib_indexed_shl_cdnNotPt_V4;
-
-  case Hexagon::LDriub_indexed_cPt_V4 :
-    return Hexagon::LDriub_indexed_cdnPt_V4;
-
-  case Hexagon::LDriub_indexed_cNotPt_V4 :
-    return Hexagon::LDriub_indexed_cdnNotPt_V4;
-
-  case Hexagon::LDriub_indexed_shl_cPt_V4 :
-    return Hexagon::LDriub_indexed_shl_cdnPt_V4;
-
-  case Hexagon::LDriub_indexed_shl_cNotPt_V4 :
-    return Hexagon::LDriub_indexed_shl_cdnNotPt_V4;
-
-  case Hexagon::LDrih_indexed_cPt_V4 :
-    return Hexagon::LDrih_indexed_cdnPt_V4;
-
-  case Hexagon::LDrih_indexed_cNotPt_V4 :
-    return Hexagon::LDrih_indexed_cdnNotPt_V4;
-
-  case Hexagon::LDrih_indexed_shl_cPt_V4 :
-    return Hexagon::LDrih_indexed_shl_cdnPt_V4;
-
-  case Hexagon::LDrih_indexed_shl_cNotPt_V4 :
-    return Hexagon::LDrih_indexed_shl_cdnNotPt_V4;
-
-  case Hexagon::LDriuh_indexed_cPt_V4 :
-    return Hexagon::LDriuh_indexed_cdnPt_V4;
-
-  case Hexagon::LDriuh_indexed_cNotPt_V4 :
-    return Hexagon::LDriuh_indexed_cdnNotPt_V4;
-
-  case Hexagon::LDriuh_indexed_shl_cPt_V4 :
-    return Hexagon::LDriuh_indexed_shl_cdnPt_V4;
-
-  case Hexagon::LDriuh_indexed_shl_cNotPt_V4 :
-    return Hexagon::LDriuh_indexed_shl_cdnNotPt_V4;
-
-  case Hexagon::LDriw_indexed_cPt_V4 :
-    return Hexagon::LDriw_indexed_cdnPt_V4;
-
-  case Hexagon::LDriw_indexed_cNotPt_V4 :
-    return Hexagon::LDriw_indexed_cdnNotPt_V4;
-
-  case Hexagon::LDriw_indexed_shl_cPt_V4 :
-    return Hexagon::LDriw_indexed_shl_cdnPt_V4;
-
-  case Hexagon::LDriw_indexed_shl_cNotPt_V4 :
-    return Hexagon::LDriw_indexed_shl_cdnNotPt_V4;
-
-  // V4 global address load
-
-  case Hexagon::LDd_GP_cPt_V4:
-    return Hexagon::LDd_GP_cdnPt_V4;
-
-  case Hexagon::LDd_GP_cNotPt_V4:
-    return Hexagon::LDd_GP_cdnNotPt_V4;
-
-  case Hexagon::LDb_GP_cPt_V4:
-    return Hexagon::LDb_GP_cdnPt_V4;
-
-  case Hexagon::LDb_GP_cNotPt_V4:
-    return Hexagon::LDb_GP_cdnNotPt_V4;
-
-  case Hexagon::LDub_GP_cPt_V4:
-    return Hexagon::LDub_GP_cdnPt_V4;
-
-  case Hexagon::LDub_GP_cNotPt_V4:
-    return Hexagon::LDub_GP_cdnNotPt_V4;
-
-  case Hexagon::LDh_GP_cPt_V4:
-    return Hexagon::LDh_GP_cdnPt_V4;
-
-  case Hexagon::LDh_GP_cNotPt_V4:
-    return Hexagon::LDh_GP_cdnNotPt_V4;
-
-  case Hexagon::LDuh_GP_cPt_V4:
-    return Hexagon::LDuh_GP_cdnPt_V4;
-
-  case Hexagon::LDuh_GP_cNotPt_V4:
-    return Hexagon::LDuh_GP_cdnNotPt_V4;
-
-  case Hexagon::LDw_GP_cPt_V4:
-    return Hexagon::LDw_GP_cdnPt_V4;
-
-  case Hexagon::LDw_GP_cNotPt_V4:
-    return Hexagon::LDw_GP_cdnNotPt_V4;
-
-  case Hexagon::LDrid_GP_cPt_V4:
-    return Hexagon::LDrid_GP_cdnPt_V4;
-
-  case Hexagon::LDrid_GP_cNotPt_V4:
-    return Hexagon::LDrid_GP_cdnNotPt_V4;
-
-  case Hexagon::LDrib_GP_cPt_V4:
-    return Hexagon::LDrib_GP_cdnPt_V4;
-
-  case Hexagon::LDrib_GP_cNotPt_V4:
-    return Hexagon::LDrib_GP_cdnNotPt_V4;
-
-  case Hexagon::LDriub_GP_cPt_V4:
-    return Hexagon::LDriub_GP_cdnPt_V4;
-
-  case Hexagon::LDriub_GP_cNotPt_V4:
-    return Hexagon::LDriub_GP_cdnNotPt_V4;
-
-  case Hexagon::LDrih_GP_cPt_V4:
-    return Hexagon::LDrih_GP_cdnPt_V4;
-
-  case Hexagon::LDrih_GP_cNotPt_V4:
-    return Hexagon::LDrih_GP_cdnNotPt_V4;
-
-  case Hexagon::LDriuh_GP_cPt_V4:
-    return Hexagon::LDriuh_GP_cdnPt_V4;
-
-  case Hexagon::LDriuh_GP_cNotPt_V4:
-    return Hexagon::LDriuh_GP_cdnNotPt_V4;
-
-  case Hexagon::LDriw_GP_cPt_V4:
-    return Hexagon::LDriw_GP_cdnPt_V4;
-
-  case Hexagon::LDriw_GP_cNotPt_V4:
-    return Hexagon::LDriw_GP_cdnNotPt_V4;
-
-  // Conditional store new-value byte
-  case Hexagon::STrib_cPt_nv_V4 :
-    return Hexagon::STrib_cdnPt_nv_V4;
-  case Hexagon::STrib_cNotPt_nv_V4 :
-    return Hexagon::STrib_cdnNotPt_nv_V4;
-
-  case Hexagon::STrib_indexed_cPt_nv_V4 :
-    return Hexagon::STrib_indexed_cdnPt_nv_V4;
-  case Hexagon::STrib_indexed_cNotPt_nv_V4 :
-    return Hexagon::STrib_indexed_cdnNotPt_nv_V4;
-
-  case Hexagon::STrib_indexed_shl_cPt_nv_V4 :
-    return Hexagon::STrib_indexed_shl_cdnPt_nv_V4;
-  case Hexagon::STrib_indexed_shl_cNotPt_nv_V4 :
-    return Hexagon::STrib_indexed_shl_cdnNotPt_nv_V4;
-
-  case Hexagon::POST_STbri_cPt_nv_V4 :
-    return Hexagon::POST_STbri_cdnPt_nv_V4;
-  case Hexagon::POST_STbri_cNotPt_nv_V4 :
-    return Hexagon::POST_STbri_cdnNotPt_nv_V4;
-
-  case Hexagon::STb_GP_cPt_nv_V4 :
-    return Hexagon::STb_GP_cdnPt_nv_V4;
-
-  case Hexagon::STb_GP_cNotPt_nv_V4 :
-    return Hexagon::STb_GP_cdnNotPt_nv_V4;
-
-  case Hexagon::STrib_GP_cPt_nv_V4 :
-    return Hexagon::STrib_GP_cdnPt_nv_V4;
-
-  case Hexagon::STrib_GP_cNotPt_nv_V4 :
-    return Hexagon::STrib_GP_cdnNotPt_nv_V4;
-
-  // Conditional store new-value halfword
-  case Hexagon::STrih_cPt_nv_V4 :
-    return Hexagon::STrih_cdnPt_nv_V4;
-  case Hexagon::STrih_cNotPt_nv_V4 :
-    return Hexagon::STrih_cdnNotPt_nv_V4;
-
-  case Hexagon::STrih_indexed_cPt_nv_V4 :
-    return Hexagon::STrih_indexed_cdnPt_nv_V4;
-  case Hexagon::STrih_indexed_cNotPt_nv_V4 :
-    return Hexagon::STrih_indexed_cdnNotPt_nv_V4;
-
-  case Hexagon::STrih_indexed_shl_cPt_nv_V4 :
-    return Hexagon::STrih_indexed_shl_cdnPt_nv_V4;
-  case Hexagon::STrih_indexed_shl_cNotPt_nv_V4 :
-    return Hexagon::STrih_indexed_shl_cdnNotPt_nv_V4;
-
-  case Hexagon::POST_SThri_cPt_nv_V4 :
-    return Hexagon::POST_SThri_cdnPt_nv_V4;
-  case Hexagon::POST_SThri_cNotPt_nv_V4 :
-    return Hexagon::POST_SThri_cdnNotPt_nv_V4;
-
-  case Hexagon::STh_GP_cPt_nv_V4 :
-    return Hexagon::STh_GP_cdnPt_nv_V4;
-
-  case Hexagon::STh_GP_cNotPt_nv_V4 :
-    return Hexagon::STh_GP_cdnNotPt_nv_V4;
-
-  case Hexagon::STrih_GP_cPt_nv_V4 :
-    return Hexagon::STrih_GP_cdnPt_nv_V4;
-
-  case Hexagon::STrih_GP_cNotPt_nv_V4 :
-    return Hexagon::STrih_GP_cdnNotPt_nv_V4;
-
-  // Conditional store new-value word
-  case Hexagon::STriw_cPt_nv_V4 :
-    return  Hexagon::STriw_cdnPt_nv_V4;
-  case Hexagon::STriw_cNotPt_nv_V4 :
-    return Hexagon::STriw_cdnNotPt_nv_V4;
-
-  case Hexagon::STriw_indexed_cPt_nv_V4 :
-    return Hexagon::STriw_indexed_cdnPt_nv_V4;
-  case Hexagon::STriw_indexed_cNotPt_nv_V4 :
-    return Hexagon::STriw_indexed_cdnNotPt_nv_V4;
-
-  case Hexagon::STriw_indexed_shl_cPt_nv_V4 :
-    return Hexagon::STriw_indexed_shl_cdnPt_nv_V4;
-  case Hexagon::STriw_indexed_shl_cNotPt_nv_V4 :
-    return Hexagon::STriw_indexed_shl_cdnNotPt_nv_V4;
-
-  case Hexagon::POST_STwri_cPt_nv_V4 :
-    return Hexagon::POST_STwri_cdnPt_nv_V4;
-  case Hexagon::POST_STwri_cNotPt_nv_V4:
-    return Hexagon::POST_STwri_cdnNotPt_nv_V4;
-
-  case Hexagon::STw_GP_cPt_nv_V4 :
-    return Hexagon::STw_GP_cdnPt_nv_V4;
-
-  case Hexagon::STw_GP_cNotPt_nv_V4 :
-    return Hexagon::STw_GP_cdnNotPt_nv_V4;
-
-  case Hexagon::STriw_GP_cPt_nv_V4 :
-    return Hexagon::STriw_GP_cdnPt_nv_V4;
-
-  case Hexagon::STriw_GP_cNotPt_nv_V4 :
-    return Hexagon::STriw_GP_cdnNotPt_nv_V4;
-
-  // Conditional add
-  case Hexagon::ADD_ri_cPt :
-    return Hexagon::ADD_ri_cdnPt;
-  case Hexagon::ADD_ri_cNotPt :
-    return Hexagon::ADD_ri_cdnNotPt;
-
-  case Hexagon::ADD_rr_cPt :
-    return Hexagon::ADD_rr_cdnPt;
-  case Hexagon::ADD_rr_cNotPt :
-    return Hexagon::ADD_rr_cdnNotPt;
-
-  // Conditional logical Operations
-  case Hexagon::XOR_rr_cPt :
-    return Hexagon::XOR_rr_cdnPt;
-  case Hexagon::XOR_rr_cNotPt :
-    return Hexagon::XOR_rr_cdnNotPt;
-
-  case Hexagon::AND_rr_cPt :
-    return Hexagon::AND_rr_cdnPt;
-  case Hexagon::AND_rr_cNotPt :
-    return Hexagon::AND_rr_cdnNotPt;
-
-  case Hexagon::OR_rr_cPt :
-    return Hexagon::OR_rr_cdnPt;
-  case Hexagon::OR_rr_cNotPt :
-    return Hexagon::OR_rr_cdnNotPt;
-
-  // Conditional Subtract
-  case Hexagon::SUB_rr_cPt :
-    return Hexagon::SUB_rr_cdnPt;
-  case Hexagon::SUB_rr_cNotPt :
-    return Hexagon::SUB_rr_cdnNotPt;
-
-  // Conditional combine
-  case Hexagon::COMBINE_rr_cPt :
-    return Hexagon::COMBINE_rr_cdnPt;
-  case Hexagon::COMBINE_rr_cNotPt :
-    return Hexagon::COMBINE_rr_cdnNotPt;
-
-  case Hexagon::ASLH_cPt_V4 :
-    return Hexagon::ASLH_cdnPt_V4;
-  case Hexagon::ASLH_cNotPt_V4 :
-    return Hexagon::ASLH_cdnNotPt_V4;
-
-  case Hexagon::ASRH_cPt_V4 :
-    return Hexagon::ASRH_cdnPt_V4;
-  case Hexagon::ASRH_cNotPt_V4 :
-    return Hexagon::ASRH_cdnNotPt_V4;
-
-  case Hexagon::SXTB_cPt_V4 :
-    return Hexagon::SXTB_cdnPt_V4;
-  case Hexagon::SXTB_cNotPt_V4 :
-    return Hexagon::SXTB_cdnNotPt_V4;
-
-  case Hexagon::SXTH_cPt_V4 :
-    return Hexagon::SXTH_cdnPt_V4;
-  case Hexagon::SXTH_cNotPt_V4 :
-    return Hexagon::SXTH_cdnNotPt_V4;
-
-  case Hexagon::ZXTB_cPt_V4 :
-    return Hexagon::ZXTB_cdnPt_V4;
-  case Hexagon::ZXTB_cNotPt_V4 :
-    return Hexagon::ZXTB_cdnNotPt_V4;
-
-  case Hexagon::ZXTH_cPt_V4 :
-    return Hexagon::ZXTH_cdnPt_V4;
-  case Hexagon::ZXTH_cNotPt_V4 :
-    return Hexagon::ZXTH_cdnNotPt_V4;
-  }
-}
-
-// Returns true if an instruction can be promoted to .new predicate
-// or new-value store.
-bool HexagonPacketizerList::isNewifiable(MachineInstr* MI) {
-  if ( isCondInst(MI) || IsNewifyStore(MI))
-    return true;
-  else
-    return false;
-}
-
-bool HexagonPacketizerList::isCondInst (MachineInstr* MI) {
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-  const MCInstrDesc& TID = MI->getDesc();
-                                    // bug 5670: until that is fixed,
-                                    // this portion is disabled.
-  if (   TID.isConditionalBranch()  // && !IsRegisterJump(MI)) ||
-      || QII->isConditionalTransfer(MI)
-      || QII->isConditionalALU32(MI)
-      || QII->isConditionalLoad(MI)
-      || QII->isConditionalStore(MI)) {
-    return true;
-  }
-  return false;
-}
-
-
-// Promote an instructiont to its .new form.
-// At this time, we have already made a call to CanPromoteToDotNew
-// and made sure that it can *indeed* be promoted.
-bool HexagonPacketizerList::PromoteToDotNew(MachineInstr* MI,
-                        SDep::Kind DepType, MachineBasicBlock::iterator &MII,
-                        const TargetRegisterClass* RC) {
-
-  assert (DepType == SDep::Data);
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-
-  int NewOpcode;
-  if (RC == Hexagon::PredRegsRegisterClass)
-    NewOpcode = GetDotNewPredOp(MI->getOpcode());
-  else
-    NewOpcode = GetDotNewOp(MI->getOpcode());
-  MI->setDesc(QII->get(NewOpcode));
-
-  return true;
-}
-
-// Returns the most basic instruction for the .new predicated instructions and
-// new-value stores.
-// For example, all of the following instructions will be converted back to the
-// same instruction:
-// 1) if (p0.new) memw(R0+#0) = R1.new  --->
-// 2) if (p0) memw(R0+#0)= R1.new      -------> if (p0) memw(R0+#0) = R1
-// 3) if (p0.new) memw(R0+#0) = R1      --->
-//
-// To understand the translation of instruction 1 to its original form, consider
-// a packet with 3 instructions.
-// { p0 = cmp.eq(R0,R1)
-//   if (p0.new) R2 = add(R3, R4)
-//   R5 = add (R3, R1)
-//   }
-// if (p0) memw(R5+#0) = R2 <--- trying to include it in the previous packet
-//
-// This instruction can be part of the previous packet only if both p0 and R2
-// are promoted to .new values. This promotion happens in steps, first
-// predicate register is promoted to .new and in the next iteration R2 is
-// promoted. Therefore, in case of dependence check failure (due to R5) during
-// next iteration, it should be converted back to its most basic form.
-
-static int GetDotOldOp(const int opc) {
-  switch (opc) {
-  default: llvm_unreachable("Unknown .old type");
-
-  case Hexagon::TFR_cdnPt:
-    return Hexagon::TFR_cPt;
-
-  case Hexagon::TFR_cdnNotPt:
-    return Hexagon::TFR_cNotPt;
-
-  case Hexagon::TFRI_cdnPt:
-    return Hexagon::TFRI_cPt;
-
-  case Hexagon::TFRI_cdnNotPt:
-    return Hexagon::TFRI_cNotPt;
-
-  case Hexagon::JMP_cdnPt:
-    return Hexagon::JMP_c;
-
-  case Hexagon::JMP_cdnNotPt:
-    return Hexagon::JMP_cNot;
-
-  case Hexagon::JMPR_cdnPt_V3:
-    return Hexagon::JMPR_cPt;
-
-  case Hexagon::JMPR_cdnNotPt_V3:
-    return Hexagon::JMPR_cNotPt;
-
-  // Load double word
-
-  case Hexagon::LDrid_cdnPt :
-    return Hexagon::LDrid_cPt;
-
-  case Hexagon::LDrid_cdnNotPt :
-    return Hexagon::LDrid_cNotPt;
-
-  case Hexagon::LDrid_indexed_cdnPt :
-    return Hexagon::LDrid_indexed_cPt;
-
-  case Hexagon::LDrid_indexed_cdnNotPt :
-    return Hexagon::LDrid_indexed_cNotPt;
-
-  case Hexagon::POST_LDrid_cdnPt_V4 :
-    return Hexagon::POST_LDrid_cPt;
-
-  case Hexagon::POST_LDrid_cdnNotPt_V4 :
-    return Hexagon::POST_LDrid_cNotPt;
-
-  // Load word
-
-  case Hexagon::LDriw_cdnPt :
-    return Hexagon::LDriw_cPt;
-
-  case Hexagon::LDriw_cdnNotPt :
-    return Hexagon::LDriw_cNotPt;
-
-  case Hexagon::LDriw_indexed_cdnPt :
-    return Hexagon::LDriw_indexed_cPt;
-
-  case Hexagon::LDriw_indexed_cdnNotPt :
-    return Hexagon::LDriw_indexed_cNotPt;
-
-  case Hexagon::POST_LDriw_cdnPt_V4 :
-    return Hexagon::POST_LDriw_cPt;
-
-  case Hexagon::POST_LDriw_cdnNotPt_V4 :
-    return Hexagon::POST_LDriw_cNotPt;
-
-  // Load half
-
-  case Hexagon::LDrih_cdnPt :
-    return Hexagon::LDrih_cPt;
-
-  case Hexagon::LDrih_cdnNotPt :
-    return Hexagon::LDrih_cNotPt;
-
-  case Hexagon::LDrih_indexed_cdnPt :
-    return Hexagon::LDrih_indexed_cPt;
-
-  case Hexagon::LDrih_indexed_cdnNotPt :
-    return Hexagon::LDrih_indexed_cNotPt;
-
-  case Hexagon::POST_LDrih_cdnPt_V4 :
-    return Hexagon::POST_LDrih_cPt;
-
-  case Hexagon::POST_LDrih_cdnNotPt_V4 :
-    return Hexagon::POST_LDrih_cNotPt;
-
-  // Load byte
-
-  case Hexagon::LDrib_cdnPt :
-    return Hexagon::LDrib_cPt;
-
-  case Hexagon::LDrib_cdnNotPt :
-    return Hexagon::LDrib_cNotPt;
-
-  case Hexagon::LDrib_indexed_cdnPt :
-    return Hexagon::LDrib_indexed_cPt;
-
-  case Hexagon::LDrib_indexed_cdnNotPt :
-    return Hexagon::LDrib_indexed_cNotPt;
-
-  case Hexagon::POST_LDrib_cdnPt_V4 :
-    return Hexagon::POST_LDrib_cPt;
-
-  case Hexagon::POST_LDrib_cdnNotPt_V4 :
-    return Hexagon::POST_LDrib_cNotPt;
-
-  // Load unsigned half
-
-  case Hexagon::LDriuh_cdnPt :
-    return Hexagon::LDriuh_cPt;
-
-  case Hexagon::LDriuh_cdnNotPt :
-    return Hexagon::LDriuh_cNotPt;
-
-  case Hexagon::LDriuh_indexed_cdnPt :
-    return Hexagon::LDriuh_indexed_cPt;
-
-  case Hexagon::LDriuh_indexed_cdnNotPt :
-    return Hexagon::LDriuh_indexed_cNotPt;
-
-  case Hexagon::POST_LDriuh_cdnPt_V4 :
-    return Hexagon::POST_LDriuh_cPt;
-
-  case Hexagon::POST_LDriuh_cdnNotPt_V4 :
-    return Hexagon::POST_LDriuh_cNotPt;
-
-  // Load unsigned byte
-  case Hexagon::LDriub_cdnPt :
-    return Hexagon::LDriub_cPt;
-
-  case Hexagon::LDriub_cdnNotPt :
-    return Hexagon::LDriub_cNotPt;
-
-  case Hexagon::LDriub_indexed_cdnPt :
-    return Hexagon::LDriub_indexed_cPt;
-
-  case Hexagon::LDriub_indexed_cdnNotPt :
-    return Hexagon::LDriub_indexed_cNotPt;
-
-  case Hexagon::POST_LDriub_cdnPt_V4 :
-    return Hexagon::POST_LDriub_cPt;
-
-  case Hexagon::POST_LDriub_cdnNotPt_V4 :
-    return Hexagon::POST_LDriub_cNotPt;
-
-  // V4 indexed+scaled Load
-
-  case Hexagon::LDrid_indexed_cdnPt_V4 :
-    return Hexagon::LDrid_indexed_cPt_V4;
-
-  case Hexagon::LDrid_indexed_cdnNotPt_V4 :
-    return Hexagon::LDrid_indexed_cNotPt_V4;
-
-  case Hexagon::LDrid_indexed_shl_cdnPt_V4 :
-    return Hexagon::LDrid_indexed_shl_cPt_V4;
-
-  case Hexagon::LDrid_indexed_shl_cdnNotPt_V4 :
-    return Hexagon::LDrid_indexed_shl_cNotPt_V4;
-
-  case Hexagon::LDrib_indexed_cdnPt_V4 :
-    return Hexagon::LDrib_indexed_cPt_V4;
-
-  case Hexagon::LDrib_indexed_cdnNotPt_V4 :
-    return Hexagon::LDrib_indexed_cNotPt_V4;
-
-  case Hexagon::LDrib_indexed_shl_cdnPt_V4 :
-    return Hexagon::LDrib_indexed_shl_cPt_V4;
-
-  case Hexagon::LDrib_indexed_shl_cdnNotPt_V4 :
-    return Hexagon::LDrib_indexed_shl_cNotPt_V4;
-
-  case Hexagon::LDriub_indexed_cdnPt_V4 :
-    return Hexagon::LDriub_indexed_cPt_V4;
-
-  case Hexagon::LDriub_indexed_cdnNotPt_V4 :
-    return Hexagon::LDriub_indexed_cNotPt_V4;
-
-  case Hexagon::LDriub_indexed_shl_cdnPt_V4 :
-    return Hexagon::LDriub_indexed_shl_cPt_V4;
-
-  case Hexagon::LDriub_indexed_shl_cdnNotPt_V4 :
-    return Hexagon::LDriub_indexed_shl_cNotPt_V4;
-
-  case Hexagon::LDrih_indexed_cdnPt_V4 :
-    return Hexagon::LDrih_indexed_cPt_V4;
-
-  case Hexagon::LDrih_indexed_cdnNotPt_V4 :
-    return Hexagon::LDrih_indexed_cNotPt_V4;
-
-  case Hexagon::LDrih_indexed_shl_cdnPt_V4 :
-    return Hexagon::LDrih_indexed_shl_cPt_V4;
-
-  case Hexagon::LDrih_indexed_shl_cdnNotPt_V4 :
-    return Hexagon::LDrih_indexed_shl_cNotPt_V4;
-
-  case Hexagon::LDriuh_indexed_cdnPt_V4 :
-    return Hexagon::LDriuh_indexed_cPt_V4;
-
-  case Hexagon::LDriuh_indexed_cdnNotPt_V4 :
-    return Hexagon::LDriuh_indexed_cNotPt_V4;
-
-  case Hexagon::LDriuh_indexed_shl_cdnPt_V4 :
-    return Hexagon::LDriuh_indexed_shl_cPt_V4;
-
-  case Hexagon::LDriuh_indexed_shl_cdnNotPt_V4 :
-    return Hexagon::LDriuh_indexed_shl_cNotPt_V4;
-
-  case Hexagon::LDriw_indexed_cdnPt_V4 :
-    return Hexagon::LDriw_indexed_cPt_V4;
-
-  case Hexagon::LDriw_indexed_cdnNotPt_V4 :
-    return Hexagon::LDriw_indexed_cNotPt_V4;
-
-  case Hexagon::LDriw_indexed_shl_cdnPt_V4 :
-    return Hexagon::LDriw_indexed_shl_cPt_V4;
-
-  case Hexagon::LDriw_indexed_shl_cdnNotPt_V4 :
-    return Hexagon::LDriw_indexed_shl_cNotPt_V4;
-
-  // V4 global address load
-
-  case Hexagon::LDd_GP_cdnPt_V4:
-    return Hexagon::LDd_GP_cPt_V4;
-
-  case Hexagon::LDd_GP_cdnNotPt_V4:
-    return Hexagon::LDd_GP_cNotPt_V4;
-
-  case Hexagon::LDb_GP_cdnPt_V4:
-    return Hexagon::LDb_GP_cPt_V4;
-
-  case Hexagon::LDb_GP_cdnNotPt_V4:
-    return Hexagon::LDb_GP_cNotPt_V4;
-
-  case Hexagon::LDub_GP_cdnPt_V4:
-    return Hexagon::LDub_GP_cPt_V4;
-
-  case Hexagon::LDub_GP_cdnNotPt_V4:
-    return Hexagon::LDub_GP_cNotPt_V4;
-
-  case Hexagon::LDh_GP_cdnPt_V4:
-    return Hexagon::LDh_GP_cPt_V4;
-
-  case Hexagon::LDh_GP_cdnNotPt_V4:
-    return Hexagon::LDh_GP_cNotPt_V4;
-
-  case Hexagon::LDuh_GP_cdnPt_V4:
-    return Hexagon::LDuh_GP_cPt_V4;
-
-  case Hexagon::LDuh_GP_cdnNotPt_V4:
-    return Hexagon::LDuh_GP_cNotPt_V4;
-
-  case Hexagon::LDw_GP_cdnPt_V4:
-    return Hexagon::LDw_GP_cPt_V4;
-
-  case Hexagon::LDw_GP_cdnNotPt_V4:
-    return Hexagon::LDw_GP_cNotPt_V4;
-
-  case Hexagon::LDrid_GP_cdnPt_V4:
-    return Hexagon::LDrid_GP_cPt_V4;
-
-  case Hexagon::LDrid_GP_cdnNotPt_V4:
-    return Hexagon::LDrid_GP_cNotPt_V4;
-
-  case Hexagon::LDrib_GP_cdnPt_V4:
-    return Hexagon::LDrib_GP_cPt_V4;
-
-  case Hexagon::LDrib_GP_cdnNotPt_V4:
-    return Hexagon::LDrib_GP_cNotPt_V4;
-
-  case Hexagon::LDriub_GP_cdnPt_V4:
-    return Hexagon::LDriub_GP_cPt_V4;
-
-  case Hexagon::LDriub_GP_cdnNotPt_V4:
-    return Hexagon::LDriub_GP_cNotPt_V4;
-
-  case Hexagon::LDrih_GP_cdnPt_V4:
-    return Hexagon::LDrih_GP_cPt_V4;
-
-  case Hexagon::LDrih_GP_cdnNotPt_V4:
-    return Hexagon::LDrih_GP_cNotPt_V4;
-
-  case Hexagon::LDriuh_GP_cdnPt_V4:
-    return Hexagon::LDriuh_GP_cPt_V4;
-
-  case Hexagon::LDriuh_GP_cdnNotPt_V4:
-    return Hexagon::LDriuh_GP_cNotPt_V4;
-
-  case Hexagon::LDriw_GP_cdnPt_V4:
-    return Hexagon::LDriw_GP_cPt_V4;
-
-  case Hexagon::LDriw_GP_cdnNotPt_V4:
-    return Hexagon::LDriw_GP_cNotPt_V4;
-
-  // Conditional add
-
-  case Hexagon::ADD_ri_cdnPt :
-    return Hexagon::ADD_ri_cPt;
-  case Hexagon::ADD_ri_cdnNotPt :
-    return Hexagon::ADD_ri_cNotPt;
-
-  case Hexagon::ADD_rr_cdnPt :
-    return Hexagon::ADD_rr_cPt;
-  case Hexagon::ADD_rr_cdnNotPt:
-    return Hexagon::ADD_rr_cNotPt;
-
-  // Conditional logical Operations
-
-  case Hexagon::XOR_rr_cdnPt :
-    return Hexagon::XOR_rr_cPt;
-  case Hexagon::XOR_rr_cdnNotPt :
-    return Hexagon::XOR_rr_cNotPt;
-
-  case Hexagon::AND_rr_cdnPt :
-    return Hexagon::AND_rr_cPt;
-  case Hexagon::AND_rr_cdnNotPt :
-    return Hexagon::AND_rr_cNotPt;
-
-  case Hexagon::OR_rr_cdnPt :
-    return Hexagon::OR_rr_cPt;
-  case Hexagon::OR_rr_cdnNotPt :
-    return Hexagon::OR_rr_cNotPt;
-
-  // Conditional Subtract
-
-  case Hexagon::SUB_rr_cdnPt :
-    return Hexagon::SUB_rr_cPt;
-  case Hexagon::SUB_rr_cdnNotPt :
-    return Hexagon::SUB_rr_cNotPt;
-
-  // Conditional combine
-
-  case Hexagon::COMBINE_rr_cdnPt :
-    return Hexagon::COMBINE_rr_cPt;
-  case Hexagon::COMBINE_rr_cdnNotPt :
-    return Hexagon::COMBINE_rr_cNotPt;
-
-// Conditional shift operations
-
-  case Hexagon::ASLH_cdnPt_V4 :
-    return Hexagon::ASLH_cPt_V4;
-  case Hexagon::ASLH_cdnNotPt_V4 :
-    return Hexagon::ASLH_cNotPt_V4;
-
-  case Hexagon::ASRH_cdnPt_V4 :
-    return Hexagon::ASRH_cPt_V4;
-  case Hexagon::ASRH_cdnNotPt_V4 :
-    return Hexagon::ASRH_cNotPt_V4;
-
-  case Hexagon::SXTB_cdnPt_V4 :
-    return Hexagon::SXTB_cPt_V4;
-  case Hexagon::SXTB_cdnNotPt_V4 :
-    return Hexagon::SXTB_cNotPt_V4;
-
-  case Hexagon::SXTH_cdnPt_V4 :
-    return Hexagon::SXTH_cPt_V4;
-  case Hexagon::SXTH_cdnNotPt_V4 :
-    return Hexagon::SXTH_cNotPt_V4;
-
-  case Hexagon::ZXTB_cdnPt_V4 :
-    return Hexagon::ZXTB_cPt_V4;
-  case Hexagon::ZXTB_cdnNotPt_V4 :
-    return Hexagon::ZXTB_cNotPt_V4;
-
-  case Hexagon::ZXTH_cdnPt_V4 :
-    return Hexagon::ZXTH_cPt_V4;
-  case Hexagon::ZXTH_cdnNotPt_V4 :
-    return Hexagon::ZXTH_cNotPt_V4;
-
-  // Store byte
-
-  case Hexagon::STrib_imm_cdnPt_V4 :
-    return Hexagon::STrib_imm_cPt_V4;
-
-  case Hexagon::STrib_imm_cdnNotPt_V4 :
-    return Hexagon::STrib_imm_cNotPt_V4;
-
-  case Hexagon::STrib_cdnPt_nv_V4 :
-  case Hexagon::STrib_cPt_nv_V4 :
-  case Hexagon::STrib_cdnPt_V4 :
-    return Hexagon::STrib_cPt;
-
-  case Hexagon::STrib_cdnNotPt_nv_V4 :
-  case Hexagon::STrib_cNotPt_nv_V4 :
-  case Hexagon::STrib_cdnNotPt_V4 :
-    return Hexagon::STrib_cNotPt;
-
-  case Hexagon::STrib_indexed_cdnPt_V4 :
-  case Hexagon::STrib_indexed_cPt_nv_V4 :
-  case Hexagon::STrib_indexed_cdnPt_nv_V4 :
-    return Hexagon::STrib_indexed_cPt;
-
-  case Hexagon::STrib_indexed_cdnNotPt_V4 :
-  case Hexagon::STrib_indexed_cNotPt_nv_V4 :
-  case Hexagon::STrib_indexed_cdnNotPt_nv_V4 :
-    return Hexagon::STrib_indexed_cNotPt;
-
-  case Hexagon::STrib_indexed_shl_cdnPt_nv_V4:
-  case Hexagon::STrib_indexed_shl_cPt_nv_V4 :
-  case Hexagon::STrib_indexed_shl_cdnPt_V4 :
-    return Hexagon::STrib_indexed_shl_cPt_V4;
-
-  case Hexagon::STrib_indexed_shl_cdnNotPt_nv_V4:
-  case Hexagon::STrib_indexed_shl_cNotPt_nv_V4 :
-  case Hexagon::STrib_indexed_shl_cdnNotPt_V4 :
-    return Hexagon::STrib_indexed_shl_cNotPt_V4;
-
-  case Hexagon::POST_STbri_cdnPt_nv_V4 :
-  case Hexagon::POST_STbri_cPt_nv_V4 :
-  case Hexagon::POST_STbri_cdnPt_V4 :
-    return Hexagon::POST_STbri_cPt;
-
-  case Hexagon::POST_STbri_cdnNotPt_nv_V4 :
-  case Hexagon::POST_STbri_cNotPt_nv_V4:
-  case Hexagon::POST_STbri_cdnNotPt_V4 :
-    return Hexagon::POST_STbri_cNotPt;
-
-  case Hexagon::STb_GP_cdnPt_nv_V4:
-  case Hexagon::STb_GP_cdnPt_V4:
-  case Hexagon::STb_GP_cPt_nv_V4:
-    return Hexagon::STb_GP_cPt_V4;
-
-  case Hexagon::STb_GP_cdnNotPt_nv_V4:
-  case Hexagon::STb_GP_cdnNotPt_V4:
-  case Hexagon::STb_GP_cNotPt_nv_V4:
-    return Hexagon::STb_GP_cNotPt_V4;
-
-  case Hexagon::STrib_GP_cdnPt_nv_V4:
-  case Hexagon::STrib_GP_cdnPt_V4:
-  case Hexagon::STrib_GP_cPt_nv_V4:
-    return Hexagon::STrib_GP_cPt_V4;
-
-  case Hexagon::STrib_GP_cdnNotPt_nv_V4:
-  case Hexagon::STrib_GP_cdnNotPt_V4:
-  case Hexagon::STrib_GP_cNotPt_nv_V4:
-    return Hexagon::STrib_GP_cNotPt_V4;
-
-  // Store new-value byte - unconditional
-  case Hexagon::STrib_nv_V4:
-    return Hexagon::STrib;
-
-  case Hexagon::STrib_indexed_nv_V4:
-    return Hexagon::STrib_indexed;
-
-  case Hexagon::STrib_indexed_shl_nv_V4:
-    return Hexagon::STrib_indexed_shl_V4;
-
-  case Hexagon::STrib_shl_nv_V4:
-    return Hexagon::STrib_shl_V4;
-
-  case Hexagon::STrib_GP_nv_V4:
-    return Hexagon::STrib_GP_V4;
-
-  case Hexagon::STb_GP_nv_V4:
-    return Hexagon::STb_GP_V4;
-
-  case Hexagon::POST_STbri_nv_V4:
-    return Hexagon::POST_STbri;
-
-  // Store halfword
-  case Hexagon::STrih_imm_cdnPt_V4 :
-    return Hexagon::STrih_imm_cPt_V4;
-
-  case Hexagon::STrih_imm_cdnNotPt_V4 :
-    return Hexagon::STrih_imm_cNotPt_V4;
-
-  case Hexagon::STrih_cdnPt_nv_V4 :
-  case Hexagon::STrih_cPt_nv_V4 :
-  case Hexagon::STrih_cdnPt_V4 :
-    return Hexagon::STrih_cPt;
-
-  case Hexagon::STrih_cdnNotPt_nv_V4 :
-  case Hexagon::STrih_cNotPt_nv_V4 :
-  case Hexagon::STrih_cdnNotPt_V4 :
-    return Hexagon::STrih_cNotPt;
-
-  case Hexagon::STrih_indexed_cdnPt_nv_V4:
-  case Hexagon::STrih_indexed_cPt_nv_V4 :
-  case Hexagon::STrih_indexed_cdnPt_V4 :
-    return Hexagon::STrih_indexed_cPt;
-
-  case Hexagon::STrih_indexed_cdnNotPt_nv_V4:
-  case Hexagon::STrih_indexed_cNotPt_nv_V4 :
-  case Hexagon::STrih_indexed_cdnNotPt_V4 :
-    return Hexagon::STrih_indexed_cNotPt;
-
-  case Hexagon::STrih_indexed_shl_cdnPt_nv_V4 :
-  case Hexagon::STrih_indexed_shl_cPt_nv_V4 :
-  case Hexagon::STrih_indexed_shl_cdnPt_V4 :
-    return Hexagon::STrih_indexed_shl_cPt_V4;
-
-  case Hexagon::STrih_indexed_shl_cdnNotPt_nv_V4 :
-  case Hexagon::STrih_indexed_shl_cNotPt_nv_V4 :
-  case Hexagon::STrih_indexed_shl_cdnNotPt_V4 :
-    return Hexagon::STrih_indexed_shl_cNotPt_V4;
-
-  case Hexagon::POST_SThri_cdnPt_nv_V4 :
-  case Hexagon::POST_SThri_cPt_nv_V4 :
-  case Hexagon::POST_SThri_cdnPt_V4 :
-    return Hexagon::POST_SThri_cPt;
-
-  case Hexagon::POST_SThri_cdnNotPt_nv_V4 :
-  case Hexagon::POST_SThri_cNotPt_nv_V4 :
-  case Hexagon::POST_SThri_cdnNotPt_V4 :
-    return Hexagon::POST_SThri_cNotPt;
-
-  case Hexagon::STh_GP_cdnPt_nv_V4:
-  case Hexagon::STh_GP_cdnPt_V4:
-  case Hexagon::STh_GP_cPt_nv_V4:
-    return Hexagon::STh_GP_cPt_V4;
-
-  case Hexagon::STh_GP_cdnNotPt_nv_V4:
-  case Hexagon::STh_GP_cdnNotPt_V4:
-  case Hexagon::STh_GP_cNotPt_nv_V4:
-    return Hexagon::STh_GP_cNotPt_V4;
-
-  case Hexagon::STrih_GP_cdnPt_nv_V4:
-  case Hexagon::STrih_GP_cdnPt_V4:
-  case Hexagon::STrih_GP_cPt_nv_V4:
-    return Hexagon::STrih_GP_cPt_V4;
-
-  case Hexagon::STrih_GP_cdnNotPt_nv_V4:
-  case Hexagon::STrih_GP_cdnNotPt_V4:
-  case Hexagon::STrih_GP_cNotPt_nv_V4:
-    return Hexagon::STrih_GP_cNotPt_V4;
-
-  // Store new-value halfword - unconditional
-
-  case Hexagon::STrih_nv_V4:
-    return Hexagon::STrih;
-
-  case Hexagon::STrih_indexed_nv_V4:
-    return Hexagon::STrih_indexed;
-
-  case Hexagon::STrih_indexed_shl_nv_V4:
-    return Hexagon::STrih_indexed_shl_V4;
-
-  case Hexagon::STrih_shl_nv_V4:
-    return Hexagon::STrih_shl_V4;
-
-  case Hexagon::STrih_GP_nv_V4:
-    return Hexagon::STrih_GP_V4;
-
-  case Hexagon::STh_GP_nv_V4:
-    return Hexagon::STh_GP_V4;
-
-  case Hexagon::POST_SThri_nv_V4:
-    return Hexagon::POST_SThri;
-
-   // Store word
-
-   case Hexagon::STriw_imm_cdnPt_V4 :
-    return Hexagon::STriw_imm_cPt_V4;
-
-  case Hexagon::STriw_imm_cdnNotPt_V4 :
-    return Hexagon::STriw_imm_cNotPt_V4;
-
-  case Hexagon::STriw_cdnPt_nv_V4 :
-  case Hexagon::STriw_cPt_nv_V4 :
-  case Hexagon::STriw_cdnPt_V4 :
-    return Hexagon::STriw_cPt;
-
-  case Hexagon::STriw_cdnNotPt_nv_V4 :
-  case Hexagon::STriw_cNotPt_nv_V4 :
-  case Hexagon::STriw_cdnNotPt_V4 :
-    return Hexagon::STriw_cNotPt;
-
-  case Hexagon::STriw_indexed_cdnPt_nv_V4 :
-  case Hexagon::STriw_indexed_cPt_nv_V4 :
-  case Hexagon::STriw_indexed_cdnPt_V4 :
-    return Hexagon::STriw_indexed_cPt;
-
-  case Hexagon::STriw_indexed_cdnNotPt_nv_V4 :
-  case Hexagon::STriw_indexed_cNotPt_nv_V4 :
-  case Hexagon::STriw_indexed_cdnNotPt_V4 :
-    return Hexagon::STriw_indexed_cNotPt;
-
-  case Hexagon::STriw_indexed_shl_cdnPt_nv_V4 :
-  case Hexagon::STriw_indexed_shl_cPt_nv_V4 :
-  case Hexagon::STriw_indexed_shl_cdnPt_V4 :
-    return Hexagon::STriw_indexed_shl_cPt_V4;
-
-  case Hexagon::STriw_indexed_shl_cdnNotPt_nv_V4 :
-  case Hexagon::STriw_indexed_shl_cNotPt_nv_V4 :
-  case Hexagon::STriw_indexed_shl_cdnNotPt_V4 :
-    return Hexagon::STriw_indexed_shl_cNotPt_V4;
-
-  case Hexagon::POST_STwri_cdnPt_nv_V4 :
-  case Hexagon::POST_STwri_cPt_nv_V4 :
-  case Hexagon::POST_STwri_cdnPt_V4 :
-    return Hexagon::POST_STwri_cPt;
-
-  case Hexagon::POST_STwri_cdnNotPt_nv_V4 :
-  case Hexagon::POST_STwri_cNotPt_nv_V4 :
-  case Hexagon::POST_STwri_cdnNotPt_V4 :
-    return Hexagon::POST_STwri_cNotPt;
-
-  case Hexagon::STw_GP_cdnPt_nv_V4:
-  case Hexagon::STw_GP_cdnPt_V4:
-  case Hexagon::STw_GP_cPt_nv_V4:
-    return Hexagon::STw_GP_cPt_V4;
-
-  case Hexagon::STw_GP_cdnNotPt_nv_V4:
-  case Hexagon::STw_GP_cdnNotPt_V4:
-  case Hexagon::STw_GP_cNotPt_nv_V4:
-    return Hexagon::STw_GP_cNotPt_V4;
-
-  case Hexagon::STriw_GP_cdnPt_nv_V4:
-  case Hexagon::STriw_GP_cdnPt_V4:
-  case Hexagon::STriw_GP_cPt_nv_V4:
-    return Hexagon::STriw_GP_cPt_V4;
-
-  case Hexagon::STriw_GP_cdnNotPt_nv_V4:
-  case Hexagon::STriw_GP_cdnNotPt_V4:
-  case Hexagon::STriw_GP_cNotPt_nv_V4:
-    return Hexagon::STriw_GP_cNotPt_V4;
-
-  // Store new-value word - unconditional
-
-  case Hexagon::STriw_nv_V4:
-    return Hexagon::STriw;
-
-  case Hexagon::STriw_indexed_nv_V4:
-    return Hexagon::STriw_indexed;
-
-  case Hexagon::STriw_indexed_shl_nv_V4:
-    return Hexagon::STriw_indexed_shl_V4;
-
-  case Hexagon::STriw_shl_nv_V4:
-    return Hexagon::STriw_shl_V4;
-
-  case Hexagon::STriw_GP_nv_V4:
-    return Hexagon::STriw_GP_V4;
-
-  case Hexagon::STw_GP_nv_V4:
-    return Hexagon::STw_GP_V4;
-
-  case Hexagon::POST_STwri_nv_V4:
-    return Hexagon::POST_STwri;
-
- // Store doubleword
-
-  case Hexagon::STrid_cdnPt_V4 :
-    return Hexagon::STrid_cPt;
-
-  case Hexagon::STrid_cdnNotPt_V4 :
-    return Hexagon::STrid_cNotPt;
-
-  case Hexagon::STrid_indexed_cdnPt_V4 :
-    return Hexagon::STrid_indexed_cPt;
-
-  case Hexagon::STrid_indexed_cdnNotPt_V4 :
-    return Hexagon::STrid_indexed_cNotPt;
-
-  case Hexagon::STrid_indexed_shl_cdnPt_V4 :
-    return Hexagon::STrid_indexed_shl_cPt_V4;
-
-  case Hexagon::STrid_indexed_shl_cdnNotPt_V4 :
-    return Hexagon::STrid_indexed_shl_cNotPt_V4;
-
-  case Hexagon::POST_STdri_cdnPt_V4 :
-    return Hexagon::POST_STdri_cPt;
-
-  case Hexagon::POST_STdri_cdnNotPt_V4 :
-    return Hexagon::POST_STdri_cNotPt;
-
-  case Hexagon::STd_GP_cdnPt_V4 :
-    return Hexagon::STd_GP_cPt_V4;
-
-  case Hexagon::STd_GP_cdnNotPt_V4 :
-    return Hexagon::STd_GP_cNotPt_V4;
-
-  case Hexagon::STrid_GP_cdnPt_V4 :
-    return Hexagon::STrid_GP_cPt_V4;
-
-  case Hexagon::STrid_GP_cdnNotPt_V4 :
-    return Hexagon::STrid_GP_cNotPt_V4;
-  }
-}
-
-bool HexagonPacketizerList::DemoteToDotOld(MachineInstr* MI) {
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-  int NewOpcode = GetDotOldOp(MI->getOpcode());
-  MI->setDesc(QII->get(NewOpcode));
-  return true;
-}
-
-// Returns true if an instruction is predicated on p0 and false if it's
-// predicated on !p0.
-
-static bool GetPredicateSense(MachineInstr* MI,
-                              const HexagonInstrInfo *QII) {
-
-  switch (MI->getOpcode()) {
-  case Hexagon::TFR_cPt:
-  case Hexagon::TFR_cdnPt:
-  case Hexagon::TFRI_cPt:
-  case Hexagon::TFRI_cdnPt:
-  case Hexagon::STrib_cPt :
-  case Hexagon::STrib_cdnPt_V4 :
-  case Hexagon::STrib_indexed_cPt :
-  case Hexagon::STrib_indexed_cdnPt_V4 :
-  case Hexagon::STrib_indexed_shl_cPt_V4 :
-  case Hexagon::STrib_indexed_shl_cdnPt_V4 :
-  case Hexagon::POST_STbri_cPt :
-  case Hexagon::POST_STbri_cdnPt_V4 :
-  case Hexagon::STrih_cPt :
-  case Hexagon::STrih_cdnPt_V4 :
-  case Hexagon::STrih_indexed_cPt :
-  case Hexagon::STrih_indexed_cdnPt_V4 :
-  case Hexagon::STrih_indexed_shl_cPt_V4 :
-  case Hexagon::STrih_indexed_shl_cdnPt_V4 :
-  case Hexagon::POST_SThri_cPt :
-  case Hexagon::POST_SThri_cdnPt_V4 :
-  case Hexagon::STriw_cPt :
-  case Hexagon::STriw_cdnPt_V4 :
-  case Hexagon::STriw_indexed_cPt :
-  case Hexagon::STriw_indexed_cdnPt_V4 :
-  case Hexagon::STriw_indexed_shl_cPt_V4 :
-  case Hexagon::STriw_indexed_shl_cdnPt_V4 :
-  case Hexagon::POST_STwri_cPt :
-  case Hexagon::POST_STwri_cdnPt_V4 :
-  case Hexagon::STrib_imm_cPt_V4 :
-  case Hexagon::STrib_imm_cdnPt_V4 :
-  case Hexagon::STrid_cPt :
-  case Hexagon::STrid_cdnPt_V4 :
-  case Hexagon::STrid_indexed_cPt :
-  case Hexagon::STrid_indexed_cdnPt_V4 :
-  case Hexagon::STrid_indexed_shl_cPt_V4 :
-  case Hexagon::STrid_indexed_shl_cdnPt_V4 :
-  case Hexagon::POST_STdri_cPt :
-  case Hexagon::POST_STdri_cdnPt_V4 :
-  case Hexagon::STrih_imm_cPt_V4 :
-  case Hexagon::STrih_imm_cdnPt_V4 :
-  case Hexagon::STriw_imm_cPt_V4 :
-  case Hexagon::STriw_imm_cdnPt_V4 :
-  case Hexagon::JMP_cdnPt :
-  case Hexagon::LDrid_cPt :
-  case Hexagon::LDrid_cdnPt :
-  case Hexagon::LDrid_indexed_cPt :
-  case Hexagon::LDrid_indexed_cdnPt :
-  case Hexagon::POST_LDrid_cPt :
-  case Hexagon::POST_LDrid_cdnPt_V4 :
-  case Hexagon::LDriw_cPt :
-  case Hexagon::LDriw_cdnPt :
-  case Hexagon::LDriw_indexed_cPt :
-  case Hexagon::LDriw_indexed_cdnPt :
-  case Hexagon::POST_LDriw_cPt :
-  case Hexagon::POST_LDriw_cdnPt_V4 :
-  case Hexagon::LDrih_cPt :
-  case Hexagon::LDrih_cdnPt :
-  case Hexagon::LDrih_indexed_cPt :
-  case Hexagon::LDrih_indexed_cdnPt :
-  case Hexagon::POST_LDrih_cPt :
-  case Hexagon::POST_LDrih_cdnPt_V4 :
-  case Hexagon::LDrib_cPt :
-  case Hexagon::LDrib_cdnPt :
-  case Hexagon::LDrib_indexed_cPt :
-  case Hexagon::LDrib_indexed_cdnPt :
-  case Hexagon::POST_LDrib_cPt :
-  case Hexagon::POST_LDrib_cdnPt_V4 :
-  case Hexagon::LDriuh_cPt :
-  case Hexagon::LDriuh_cdnPt :
-  case Hexagon::LDriuh_indexed_cPt :
-  case Hexagon::LDriuh_indexed_cdnPt :
-  case Hexagon::POST_LDriuh_cPt :
-  case Hexagon::POST_LDriuh_cdnPt_V4 :
-  case Hexagon::LDriub_cPt :
-  case Hexagon::LDriub_cdnPt :
-  case Hexagon::LDriub_indexed_cPt :
-  case Hexagon::LDriub_indexed_cdnPt :
-  case Hexagon::POST_LDriub_cPt :
-  case Hexagon::POST_LDriub_cdnPt_V4 :
-  case Hexagon::LDrid_indexed_cPt_V4 :
-  case Hexagon::LDrid_indexed_cdnPt_V4 :
-  case Hexagon::LDrid_indexed_shl_cPt_V4 :
-  case Hexagon::LDrid_indexed_shl_cdnPt_V4 :
-  case Hexagon::LDrib_indexed_cPt_V4 :
-  case Hexagon::LDrib_indexed_cdnPt_V4 :
-  case Hexagon::LDrib_indexed_shl_cPt_V4 :
-  case Hexagon::LDrib_indexed_shl_cdnPt_V4 :
-  case Hexagon::LDriub_indexed_cPt_V4 :
-  case Hexagon::LDriub_indexed_cdnPt_V4 :
-  case Hexagon::LDriub_indexed_shl_cPt_V4 :
-  case Hexagon::LDriub_indexed_shl_cdnPt_V4 :
-  case Hexagon::LDrih_indexed_cPt_V4 :
-  case Hexagon::LDrih_indexed_cdnPt_V4 :
-  case Hexagon::LDrih_indexed_shl_cPt_V4 :
-  case Hexagon::LDrih_indexed_shl_cdnPt_V4 :
-  case Hexagon::LDriuh_indexed_cPt_V4 :
-  case Hexagon::LDriuh_indexed_cdnPt_V4 :
-  case Hexagon::LDriuh_indexed_shl_cPt_V4 :
-  case Hexagon::LDriuh_indexed_shl_cdnPt_V4 :
-  case Hexagon::LDriw_indexed_cPt_V4 :
-  case Hexagon::LDriw_indexed_cdnPt_V4 :
-  case Hexagon::LDriw_indexed_shl_cPt_V4 :
-  case Hexagon::LDriw_indexed_shl_cdnPt_V4 :
-  case Hexagon::ADD_ri_cPt :
-  case Hexagon::ADD_ri_cdnPt :
-  case Hexagon::ADD_rr_cPt :
-  case Hexagon::ADD_rr_cdnPt :
-  case Hexagon::XOR_rr_cPt :
-  case Hexagon::XOR_rr_cdnPt :
-  case Hexagon::AND_rr_cPt :
-  case Hexagon::AND_rr_cdnPt :
-  case Hexagon::OR_rr_cPt :
-  case Hexagon::OR_rr_cdnPt :
-  case Hexagon::SUB_rr_cPt :
-  case Hexagon::SUB_rr_cdnPt :
-  case Hexagon::COMBINE_rr_cPt :
-  case Hexagon::COMBINE_rr_cdnPt :
-  case Hexagon::ASLH_cPt_V4 :
-  case Hexagon::ASLH_cdnPt_V4 :
-  case Hexagon::ASRH_cPt_V4 :
-  case Hexagon::ASRH_cdnPt_V4 :
-  case Hexagon::SXTB_cPt_V4 :
-  case Hexagon::SXTB_cdnPt_V4 :
-  case Hexagon::SXTH_cPt_V4 :
-  case Hexagon::SXTH_cdnPt_V4 :
-  case Hexagon::ZXTB_cPt_V4 :
-  case Hexagon::ZXTB_cdnPt_V4 :
-  case Hexagon::ZXTH_cPt_V4 :
-  case Hexagon::ZXTH_cdnPt_V4 :
-  case Hexagon::LDrid_GP_cPt_V4 :
-  case Hexagon::LDrib_GP_cPt_V4 :
-  case Hexagon::LDriub_GP_cPt_V4 :
-  case Hexagon::LDrih_GP_cPt_V4 :
-  case Hexagon::LDriuh_GP_cPt_V4 :
-  case Hexagon::LDriw_GP_cPt_V4 :
-  case Hexagon::LDd_GP_cPt_V4 :
-  case Hexagon::LDb_GP_cPt_V4 :
-  case Hexagon::LDub_GP_cPt_V4 :
-  case Hexagon::LDh_GP_cPt_V4 :
-  case Hexagon::LDuh_GP_cPt_V4 :
-  case Hexagon::LDw_GP_cPt_V4 :
-  case Hexagon::STrid_GP_cPt_V4 :
-  case Hexagon::STrib_GP_cPt_V4 :
-  case Hexagon::STrih_GP_cPt_V4 :
-  case Hexagon::STriw_GP_cPt_V4 :
-  case Hexagon::STd_GP_cPt_V4 :
-  case Hexagon::STb_GP_cPt_V4 :
-  case Hexagon::STh_GP_cPt_V4 :
-  case Hexagon::STw_GP_cPt_V4 :
-  case Hexagon::LDrid_GP_cdnPt_V4 :
-  case Hexagon::LDrib_GP_cdnPt_V4 :
-  case Hexagon::LDriub_GP_cdnPt_V4 :
-  case Hexagon::LDrih_GP_cdnPt_V4 :
-  case Hexagon::LDriuh_GP_cdnPt_V4 :
-  case Hexagon::LDriw_GP_cdnPt_V4 :
-  case Hexagon::LDd_GP_cdnPt_V4 :
-  case Hexagon::LDb_GP_cdnPt_V4 :
-  case Hexagon::LDub_GP_cdnPt_V4 :
-  case Hexagon::LDh_GP_cdnPt_V4 :
-  case Hexagon::LDuh_GP_cdnPt_V4 :
-  case Hexagon::LDw_GP_cdnPt_V4 :
-  case Hexagon::STrid_GP_cdnPt_V4 :
-  case Hexagon::STrib_GP_cdnPt_V4 :
-  case Hexagon::STrih_GP_cdnPt_V4 :
-  case Hexagon::STriw_GP_cdnPt_V4 :
-  case Hexagon::STd_GP_cdnPt_V4 :
-  case Hexagon::STb_GP_cdnPt_V4 :
-  case Hexagon::STh_GP_cdnPt_V4 :
-  case Hexagon::STw_GP_cdnPt_V4 :
-    return true;
-
-  case Hexagon::TFR_cNotPt:
-  case Hexagon::TFR_cdnNotPt:
-  case Hexagon::TFRI_cNotPt:
-  case Hexagon::TFRI_cdnNotPt:
-  case Hexagon::STrib_cNotPt :
-  case Hexagon::STrib_cdnNotPt_V4 :
-  case Hexagon::STrib_indexed_cNotPt :
-  case Hexagon::STrib_indexed_cdnNotPt_V4 :
-  case Hexagon::STrib_indexed_shl_cNotPt_V4 :
-  case Hexagon::STrib_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::POST_STbri_cNotPt :
-  case Hexagon::POST_STbri_cdnNotPt_V4 :
-  case Hexagon::STrih_cNotPt :
-  case Hexagon::STrih_cdnNotPt_V4 :
-  case Hexagon::STrih_indexed_cNotPt :
-  case Hexagon::STrih_indexed_cdnNotPt_V4 :
-  case Hexagon::STrih_indexed_shl_cNotPt_V4 :
-  case Hexagon::STrih_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::POST_SThri_cNotPt :
-  case Hexagon::POST_SThri_cdnNotPt_V4 :
-  case Hexagon::STriw_cNotPt :
-  case Hexagon::STriw_cdnNotPt_V4 :
-  case Hexagon::STriw_indexed_cNotPt :
-  case Hexagon::STriw_indexed_cdnNotPt_V4 :
-  case Hexagon::STriw_indexed_shl_cNotPt_V4 :
-  case Hexagon::STriw_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::POST_STwri_cNotPt :
-  case Hexagon::POST_STwri_cdnNotPt_V4 :
-  case Hexagon::STrib_imm_cNotPt_V4 :
-  case Hexagon::STrib_imm_cdnNotPt_V4 :
-  case Hexagon::STrid_cNotPt :
-  case Hexagon::STrid_cdnNotPt_V4 :
-  case Hexagon::STrid_indexed_cdnNotPt_V4 :
-  case Hexagon::STrid_indexed_cNotPt :
-  case Hexagon::STrid_indexed_shl_cNotPt_V4 :
-  case Hexagon::STrid_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::POST_STdri_cNotPt :
-  case Hexagon::POST_STdri_cdnNotPt_V4 :
-  case Hexagon::STrih_imm_cNotPt_V4 :
-  case Hexagon::STrih_imm_cdnNotPt_V4 :
-  case Hexagon::STriw_imm_cNotPt_V4 :
-  case Hexagon::STriw_imm_cdnNotPt_V4 :
-  case Hexagon::JMP_cdnNotPt :
-  case Hexagon::LDrid_cNotPt :
-  case Hexagon::LDrid_cdnNotPt :
-  case Hexagon::LDrid_indexed_cNotPt :
-  case Hexagon::LDrid_indexed_cdnNotPt :
-  case Hexagon::POST_LDrid_cNotPt :
-  case Hexagon::POST_LDrid_cdnNotPt_V4 :
-  case Hexagon::LDriw_cNotPt :
-  case Hexagon::LDriw_cdnNotPt :
-  case Hexagon::LDriw_indexed_cNotPt :
-  case Hexagon::LDriw_indexed_cdnNotPt :
-  case Hexagon::POST_LDriw_cNotPt :
-  case Hexagon::POST_LDriw_cdnNotPt_V4 :
-  case Hexagon::LDrih_cNotPt :
-  case Hexagon::LDrih_cdnNotPt :
-  case Hexagon::LDrih_indexed_cNotPt :
-  case Hexagon::LDrih_indexed_cdnNotPt :
-  case Hexagon::POST_LDrih_cNotPt :
-  case Hexagon::POST_LDrih_cdnNotPt_V4 :
-  case Hexagon::LDrib_cNotPt :
-  case Hexagon::LDrib_cdnNotPt :
-  case Hexagon::LDrib_indexed_cNotPt :
-  case Hexagon::LDrib_indexed_cdnNotPt :
-  case Hexagon::POST_LDrib_cNotPt :
-  case Hexagon::POST_LDrib_cdnNotPt_V4 :
-  case Hexagon::LDriuh_cNotPt :
-  case Hexagon::LDriuh_cdnNotPt :
-  case Hexagon::LDriuh_indexed_cNotPt :
-  case Hexagon::LDriuh_indexed_cdnNotPt :
-  case Hexagon::POST_LDriuh_cNotPt :
-  case Hexagon::POST_LDriuh_cdnNotPt_V4 :
-  case Hexagon::LDriub_cNotPt :
-  case Hexagon::LDriub_cdnNotPt :
-  case Hexagon::LDriub_indexed_cNotPt :
-  case Hexagon::LDriub_indexed_cdnNotPt :
-  case Hexagon::POST_LDriub_cNotPt :
-  case Hexagon::POST_LDriub_cdnNotPt_V4 :
-  case Hexagon::LDrid_indexed_cNotPt_V4 :
-  case Hexagon::LDrid_indexed_cdnNotPt_V4 :
-  case Hexagon::LDrid_indexed_shl_cNotPt_V4 :
-  case Hexagon::LDrid_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::LDrib_indexed_cNotPt_V4 :
-  case Hexagon::LDrib_indexed_cdnNotPt_V4 :
-  case Hexagon::LDrib_indexed_shl_cNotPt_V4 :
-  case Hexagon::LDrib_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::LDriub_indexed_cNotPt_V4 :
-  case Hexagon::LDriub_indexed_cdnNotPt_V4 :
-  case Hexagon::LDriub_indexed_shl_cNotPt_V4 :
-  case Hexagon::LDriub_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::LDrih_indexed_cNotPt_V4 :
-  case Hexagon::LDrih_indexed_cdnNotPt_V4 :
-  case Hexagon::LDrih_indexed_shl_cNotPt_V4 :
-  case Hexagon::LDrih_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::LDriuh_indexed_cNotPt_V4 :
-  case Hexagon::LDriuh_indexed_cdnNotPt_V4 :
-  case Hexagon::LDriuh_indexed_shl_cNotPt_V4 :
-  case Hexagon::LDriuh_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::LDriw_indexed_cNotPt_V4 :
-  case Hexagon::LDriw_indexed_cdnNotPt_V4 :
-  case Hexagon::LDriw_indexed_shl_cNotPt_V4 :
-  case Hexagon::LDriw_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::ADD_ri_cNotPt :
-  case Hexagon::ADD_ri_cdnNotPt :
-  case Hexagon::ADD_rr_cNotPt :
-  case Hexagon::ADD_rr_cdnNotPt :
-  case Hexagon::XOR_rr_cNotPt :
-  case Hexagon::XOR_rr_cdnNotPt :
-  case Hexagon::AND_rr_cNotPt :
-  case Hexagon::AND_rr_cdnNotPt :
-  case Hexagon::OR_rr_cNotPt :
-  case Hexagon::OR_rr_cdnNotPt :
-  case Hexagon::SUB_rr_cNotPt :
-  case Hexagon::SUB_rr_cdnNotPt :
-  case Hexagon::COMBINE_rr_cNotPt :
-  case Hexagon::COMBINE_rr_cdnNotPt :
-  case Hexagon::ASLH_cNotPt_V4 :
-  case Hexagon::ASLH_cdnNotPt_V4 :
-  case Hexagon::ASRH_cNotPt_V4 :
-  case Hexagon::ASRH_cdnNotPt_V4 :
-  case Hexagon::SXTB_cNotPt_V4 :
-  case Hexagon::SXTB_cdnNotPt_V4 :
-  case Hexagon::SXTH_cNotPt_V4 :
-  case Hexagon::SXTH_cdnNotPt_V4 :
-  case Hexagon::ZXTB_cNotPt_V4 :
-  case Hexagon::ZXTB_cdnNotPt_V4 :
-  case Hexagon::ZXTH_cNotPt_V4 :
-  case Hexagon::ZXTH_cdnNotPt_V4 :
-
-  case Hexagon::LDrid_GP_cNotPt_V4 :
-  case Hexagon::LDrib_GP_cNotPt_V4 :
-  case Hexagon::LDriub_GP_cNotPt_V4 :
-  case Hexagon::LDrih_GP_cNotPt_V4 :
-  case Hexagon::LDriuh_GP_cNotPt_V4 :
-  case Hexagon::LDriw_GP_cNotPt_V4 :
-  case Hexagon::LDd_GP_cNotPt_V4 :
-  case Hexagon::LDb_GP_cNotPt_V4 :
-  case Hexagon::LDub_GP_cNotPt_V4 :
-  case Hexagon::LDh_GP_cNotPt_V4 :
-  case Hexagon::LDuh_GP_cNotPt_V4 :
-  case Hexagon::LDw_GP_cNotPt_V4 :
-  case Hexagon::STrid_GP_cNotPt_V4 :
-  case Hexagon::STrib_GP_cNotPt_V4 :
-  case Hexagon::STrih_GP_cNotPt_V4 :
-  case Hexagon::STriw_GP_cNotPt_V4 :
-  case Hexagon::STd_GP_cNotPt_V4 :
-  case Hexagon::STb_GP_cNotPt_V4 :
-  case Hexagon::STh_GP_cNotPt_V4 :
-  case Hexagon::STw_GP_cNotPt_V4 :
-  case Hexagon::LDrid_GP_cdnNotPt_V4 :
-  case Hexagon::LDrib_GP_cdnNotPt_V4 :
-  case Hexagon::LDriub_GP_cdnNotPt_V4 :
-  case Hexagon::LDrih_GP_cdnNotPt_V4 :
-  case Hexagon::LDriuh_GP_cdnNotPt_V4 :
-  case Hexagon::LDriw_GP_cdnNotPt_V4 :
-  case Hexagon::LDd_GP_cdnNotPt_V4 :
-  case Hexagon::LDb_GP_cdnNotPt_V4 :
-  case Hexagon::LDub_GP_cdnNotPt_V4 :
-  case Hexagon::LDh_GP_cdnNotPt_V4 :
-  case Hexagon::LDuh_GP_cdnNotPt_V4 :
-  case Hexagon::LDw_GP_cdnNotPt_V4 :
-  case Hexagon::STrid_GP_cdnNotPt_V4 :
-  case Hexagon::STrib_GP_cdnNotPt_V4 :
-  case Hexagon::STrih_GP_cdnNotPt_V4 :
-  case Hexagon::STriw_GP_cdnNotPt_V4 :
-  case Hexagon::STd_GP_cdnNotPt_V4 :
-  case Hexagon::STb_GP_cdnNotPt_V4 :
-  case Hexagon::STh_GP_cdnNotPt_V4 :
-  case Hexagon::STw_GP_cdnNotPt_V4 :
-    return false;
-
-  default:
-    assert (false && "Unknown predicate sense of the instruction");
-  }
-  // return *some value* to avoid compiler warning
-  return false;
-}
-
-bool HexagonPacketizerList::isDotNewInst(MachineInstr* MI) {
-  if (isNewValueInst(MI))
-    return true;
-
-  switch (MI->getOpcode()) {
-  case Hexagon::TFR_cdnNotPt:
-  case Hexagon::TFR_cdnPt:
-  case Hexagon::TFRI_cdnNotPt:
-  case Hexagon::TFRI_cdnPt:
-  case Hexagon::LDrid_cdnPt :
-  case Hexagon::LDrid_cdnNotPt :
-  case Hexagon::LDrid_indexed_cdnPt :
-  case Hexagon::LDrid_indexed_cdnNotPt :
-  case Hexagon::POST_LDrid_cdnPt_V4 :
-  case Hexagon::POST_LDrid_cdnNotPt_V4 :
-  case Hexagon::LDriw_cdnPt :
-  case Hexagon::LDriw_cdnNotPt :
-  case Hexagon::LDriw_indexed_cdnPt :
-  case Hexagon::LDriw_indexed_cdnNotPt :
-  case Hexagon::POST_LDriw_cdnPt_V4 :
-  case Hexagon::POST_LDriw_cdnNotPt_V4 :
-  case Hexagon::LDrih_cdnPt :
-  case Hexagon::LDrih_cdnNotPt :
-  case Hexagon::LDrih_indexed_cdnPt :
-  case Hexagon::LDrih_indexed_cdnNotPt :
-  case Hexagon::POST_LDrih_cdnPt_V4 :
-  case Hexagon::POST_LDrih_cdnNotPt_V4 :
-  case Hexagon::LDrib_cdnPt :
-  case Hexagon::LDrib_cdnNotPt :
-  case Hexagon::LDrib_indexed_cdnPt :
-  case Hexagon::LDrib_indexed_cdnNotPt :
-  case Hexagon::POST_LDrib_cdnPt_V4 :
-  case Hexagon::POST_LDrib_cdnNotPt_V4 :
-  case Hexagon::LDriuh_cdnPt :
-  case Hexagon::LDriuh_cdnNotPt :
-  case Hexagon::LDriuh_indexed_cdnPt :
-  case Hexagon::LDriuh_indexed_cdnNotPt :
-  case Hexagon::POST_LDriuh_cdnPt_V4 :
-  case Hexagon::POST_LDriuh_cdnNotPt_V4 :
-  case Hexagon::LDriub_cdnPt :
-  case Hexagon::LDriub_cdnNotPt :
-  case Hexagon::LDriub_indexed_cdnPt :
-  case Hexagon::LDriub_indexed_cdnNotPt :
-  case Hexagon::POST_LDriub_cdnPt_V4 :
-  case Hexagon::POST_LDriub_cdnNotPt_V4 :
-
-  case Hexagon::LDrid_indexed_cdnPt_V4 :
-  case Hexagon::LDrid_indexed_cdnNotPt_V4 :
-  case Hexagon::LDrid_indexed_shl_cdnPt_V4 :
-  case Hexagon::LDrid_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::LDrib_indexed_cdnPt_V4 :
-  case Hexagon::LDrib_indexed_cdnNotPt_V4 :
-  case Hexagon::LDrib_indexed_shl_cdnPt_V4 :
-  case Hexagon::LDrib_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::LDriub_indexed_cdnPt_V4 :
-  case Hexagon::LDriub_indexed_cdnNotPt_V4 :
-  case Hexagon::LDriub_indexed_shl_cdnPt_V4 :
-  case Hexagon::LDriub_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::LDrih_indexed_cdnPt_V4 :
-  case Hexagon::LDrih_indexed_cdnNotPt_V4 :
-  case Hexagon::LDrih_indexed_shl_cdnPt_V4 :
-  case Hexagon::LDrih_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::LDriuh_indexed_cdnPt_V4 :
-  case Hexagon::LDriuh_indexed_cdnNotPt_V4 :
-  case Hexagon::LDriuh_indexed_shl_cdnPt_V4 :
-  case Hexagon::LDriuh_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::LDriw_indexed_cdnPt_V4 :
-  case Hexagon::LDriw_indexed_cdnNotPt_V4 :
-  case Hexagon::LDriw_indexed_shl_cdnPt_V4 :
-  case Hexagon::LDriw_indexed_shl_cdnNotPt_V4 :
-
-// Coditional add
-  case Hexagon::ADD_ri_cdnPt:
-  case Hexagon::ADD_ri_cdnNotPt:
-  case Hexagon::ADD_rr_cdnPt:
-  case Hexagon::ADD_rr_cdnNotPt:
-
-  // Conditional logical operations
-  case Hexagon::XOR_rr_cdnPt :
-  case Hexagon::XOR_rr_cdnNotPt :
-  case Hexagon::AND_rr_cdnPt :
-  case Hexagon::AND_rr_cdnNotPt :
-  case Hexagon::OR_rr_cdnPt :
-  case Hexagon::OR_rr_cdnNotPt :
-
-  // Conditonal subtract
-  case Hexagon::SUB_rr_cdnPt :
-  case Hexagon::SUB_rr_cdnNotPt :
-
-  // Conditional combine
-  case Hexagon::COMBINE_rr_cdnPt :
-  case Hexagon::COMBINE_rr_cdnNotPt :
-
-  // Conditional shift operations
-  case Hexagon::ASLH_cdnPt_V4:
-  case Hexagon::ASLH_cdnNotPt_V4:
-  case Hexagon::ASRH_cdnPt_V4:
-  case Hexagon::ASRH_cdnNotPt_V4:
-  case Hexagon::SXTB_cdnPt_V4:
-  case Hexagon::SXTB_cdnNotPt_V4:
-  case Hexagon::SXTH_cdnPt_V4:
-  case Hexagon::SXTH_cdnNotPt_V4:
-  case Hexagon::ZXTB_cdnPt_V4:
-  case Hexagon::ZXTB_cdnNotPt_V4:
-  case Hexagon::ZXTH_cdnPt_V4:
-  case Hexagon::ZXTH_cdnNotPt_V4:
-
-  // Conditional stores
-  case Hexagon::STrib_imm_cdnPt_V4 :
-  case Hexagon::STrib_imm_cdnNotPt_V4 :
-  case Hexagon::STrib_cdnPt_V4 :
-  case Hexagon::STrib_cdnNotPt_V4 :
-  case Hexagon::STrib_indexed_cdnPt_V4 :
-  case Hexagon::STrib_indexed_cdnNotPt_V4 :
-  case Hexagon::POST_STbri_cdnPt_V4 :
-  case Hexagon::POST_STbri_cdnNotPt_V4 :
-  case Hexagon::STrib_indexed_shl_cdnPt_V4 :
-  case Hexagon::STrib_indexed_shl_cdnNotPt_V4 :
-
-  // Store doubleword conditionally
-  case Hexagon::STrid_indexed_cdnPt_V4 :
-  case Hexagon::STrid_indexed_cdnNotPt_V4 :
-  case Hexagon::STrid_indexed_shl_cdnPt_V4 :
-  case Hexagon::STrid_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::POST_STdri_cdnPt_V4 :
-  case Hexagon::POST_STdri_cdnNotPt_V4 :
-
-  // Store halfword conditionally
-  case Hexagon::STrih_cdnPt_V4 :
-  case Hexagon::STrih_cdnNotPt_V4 :
-  case Hexagon::STrih_indexed_cdnPt_V4 :
-  case Hexagon::STrih_indexed_cdnNotPt_V4 :
-  case Hexagon::STrih_imm_cdnPt_V4 :
-  case Hexagon::STrih_imm_cdnNotPt_V4 :
-  case Hexagon::STrih_indexed_shl_cdnPt_V4 :
-  case Hexagon::STrih_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::POST_SThri_cdnPt_V4 :
-  case Hexagon::POST_SThri_cdnNotPt_V4 :
-
-  // Store word conditionally
-  case Hexagon::STriw_cdnPt_V4 :
-  case Hexagon::STriw_cdnNotPt_V4 :
-  case Hexagon::STriw_indexed_cdnPt_V4 :
-  case Hexagon::STriw_indexed_cdnNotPt_V4 :
-  case Hexagon::STriw_imm_cdnPt_V4 :
-  case Hexagon::STriw_imm_cdnNotPt_V4 :
-  case Hexagon::STriw_indexed_shl_cdnPt_V4 :
-  case Hexagon::STriw_indexed_shl_cdnNotPt_V4 :
-  case Hexagon::POST_STwri_cdnPt_V4 :
-  case Hexagon::POST_STwri_cdnNotPt_V4 :
-
-  case Hexagon::LDd_GP_cdnPt_V4:
-  case Hexagon::LDd_GP_cdnNotPt_V4:
-  case Hexagon::LDb_GP_cdnPt_V4:
-  case Hexagon::LDb_GP_cdnNotPt_V4:
-  case Hexagon::LDub_GP_cdnPt_V4:
-  case Hexagon::LDub_GP_cdnNotPt_V4:
-  case Hexagon::LDh_GP_cdnPt_V4:
-  case Hexagon::LDh_GP_cdnNotPt_V4:
-  case Hexagon::LDuh_GP_cdnPt_V4:
-  case Hexagon::LDuh_GP_cdnNotPt_V4:
-  case Hexagon::LDw_GP_cdnPt_V4:
-  case Hexagon::LDw_GP_cdnNotPt_V4:
-  case Hexagon::LDrid_GP_cdnPt_V4:
-  case Hexagon::LDrid_GP_cdnNotPt_V4:
-  case Hexagon::LDrib_GP_cdnPt_V4:
-  case Hexagon::LDrib_GP_cdnNotPt_V4:
-  case Hexagon::LDriub_GP_cdnPt_V4:
-  case Hexagon::LDriub_GP_cdnNotPt_V4:
-  case Hexagon::LDrih_GP_cdnPt_V4:
-  case Hexagon::LDrih_GP_cdnNotPt_V4:
-  case Hexagon::LDriuh_GP_cdnPt_V4:
-  case Hexagon::LDriuh_GP_cdnNotPt_V4:
-  case Hexagon::LDriw_GP_cdnPt_V4:
-  case Hexagon::LDriw_GP_cdnNotPt_V4:
-
-  case Hexagon::STrid_GP_cdnPt_V4:
-  case Hexagon::STrid_GP_cdnNotPt_V4:
-  case Hexagon::STrib_GP_cdnPt_V4:
-  case Hexagon::STrib_GP_cdnNotPt_V4:
-  case Hexagon::STrih_GP_cdnPt_V4:
-  case Hexagon::STrih_GP_cdnNotPt_V4:
-  case Hexagon::STriw_GP_cdnPt_V4:
-  case Hexagon::STriw_GP_cdnNotPt_V4:
-  case Hexagon::STd_GP_cdnPt_V4:
-  case Hexagon::STd_GP_cdnNotPt_V4:
-  case Hexagon::STb_GP_cdnPt_V4:
-  case Hexagon::STb_GP_cdnNotPt_V4:
-  case Hexagon::STh_GP_cdnPt_V4:
-  case Hexagon::STh_GP_cdnNotPt_V4:
-  case Hexagon::STw_GP_cdnPt_V4:
-  case Hexagon::STw_GP_cdnNotPt_V4:
-
-    return true;
-  }
-  return false;
-}
-
-static MachineOperand& GetPostIncrementOperand(MachineInstr *MI,
-                                               const HexagonInstrInfo *QII) {
-  assert(QII->isPostIncrement(MI) && "Not a post increment operation.");
-#ifndef NDEBUG
-  // Post Increment means duplicates. Use dense map to find duplicates in the
-  // list. Caution: Densemap initializes with the minimum of 64 buckets,
-  // whereas there are at most 5 operands in the post increment.
-  DenseMap DefRegsSet;
-  for(unsigned opNum = 0; opNum < MI->getNumOperands(); opNum++)
-    if (MI->getOperand(opNum).isReg() &&
-        MI->getOperand(opNum).isDef()) {
-      DefRegsSet[MI->getOperand(opNum).getReg()] = 1;
-    }
-
-  for(unsigned opNum = 0; opNum < MI->getNumOperands(); opNum++)
-    if (MI->getOperand(opNum).isReg() &&
-        MI->getOperand(opNum).isUse()) {
-      if (DefRegsSet[MI->getOperand(opNum).getReg()]) {
-        return MI->getOperand(opNum);
-      }
-    }
-#else
-  if (MI->getDesc().mayLoad()) {
-    // The 2nd operand is always the post increment operand in load.
-    assert(MI->getOperand(1).isReg() &&
-                "Post increment operand has be to a register.");
-    return (MI->getOperand(1));
-  }
-  if (MI->getDesc().mayStore()) {
-    // The 1st operand is always the post increment operand in store.
-    assert(MI->getOperand(0).isReg() &&
-                "Post increment operand has be to a register.");
-    return (MI->getOperand(0));
-  }
-#endif
-  // we should never come here.
-  llvm_unreachable("mayLoad or mayStore not set for Post Increment operation");
-}
-
-// get the value being stored
-static MachineOperand& GetStoreValueOperand(MachineInstr *MI) {
-  // value being stored is always the last operand.
-  return (MI->getOperand(MI->getNumOperands()-1));
-}
-
-// can be new value store?
-// Following restrictions are to be respected in convert a store into
-// a new value store.
-// 1. If an instruction uses auto-increment, its address register cannot
-//    be a new-value register. Arch Spec 5.4.2.1
-// 2. If an instruction uses absolute-set addressing mode,
-//    its address register cannot be a new-value register.
-//    Arch Spec 5.4.2.1.TODO: This is not enabled as
-//    as absolute-set address mode patters are not implemented.
-// 3. If an instruction produces a 64-bit result, its registers cannot be used
-//    as new-value registers. Arch Spec 5.4.2.2.
-// 4. If the instruction that sets a new-value register is conditional, then
-//    the instruction that uses the new-value register must also be conditional,
-//    and both must always have their predicates evaluate identically.
-//    Arch Spec 5.4.2.3.
-// 5. There is an implied restriction of a packet can not have another store,
-//    if there is a  new value store in the packet. Corollary, if there is
-//    already a store in a packet, there can not be a new value store.
-//    Arch Spec: 3.4.4.2
-bool HexagonPacketizerList::CanPromoteToNewValueStore( MachineInstr *MI,
-                MachineInstr *PacketMI, unsigned DepReg,
-                std::map  MIToSUnit)
-{
-  // Make sure we are looking at the store
-  if (!IsNewifyStore(MI))
-    return false;
-
-  // Make sure there is dependency and can be new value'ed
-  if (GetStoreValueOperand(MI).isReg() &&
-      GetStoreValueOperand(MI).getReg() != DepReg)
-    return false;
-
-  const HexagonRegisterInfo* QRI = (const HexagonRegisterInfo *) TM.getRegisterInfo();
-  const MCInstrDesc& MCID = PacketMI->getDesc();
-  // first operand is always the result
-
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-  const TargetRegisterClass* PacketRC = QII->getRegClass(MCID, 0, QRI);
-
-  // if there is already an store in the packet, no can do new value store
-  // Arch Spec 3.4.4.2.
-  for (std::vector::iterator VI = CurrentPacketMIs.begin(),
-         VE = CurrentPacketMIs.end();
-       (VI != VE); ++VI) {
-    SUnit* PacketSU = MIToSUnit[*VI];
-    if (PacketSU->getInstr()->getDesc().mayStore() ||
-        // if we have mayStore = 1 set on ALLOCFRAME and DEALLOCFRAME,
-        // then we don't need this
-        PacketSU->getInstr()->getOpcode() == Hexagon::ALLOCFRAME ||
-        PacketSU->getInstr()->getOpcode() == Hexagon::DEALLOCFRAME)
-      return false;
-  }
-
-  if (PacketRC == Hexagon::DoubleRegsRegisterClass) {
-    // new value store constraint: double regs can not feed into new value store
-    // arch spec section: 5.4.2.2
-    return false;
-  }
-
-  // Make sure it's NOT the post increment register that we are going to
-  // new value.
-  if (QII->isPostIncrement(MI) &&
-      MI->getDesc().mayStore() &&
-      GetPostIncrementOperand(MI, QII).getReg() == DepReg) {
-    return false;
-  }
-
-  if (QII->isPostIncrement(PacketMI) &&
-      PacketMI->getDesc().mayLoad() &&
-      GetPostIncrementOperand(PacketMI, QII).getReg() == DepReg) {
-    // if source is post_inc, or absolute-set addressing,
-    // it can not feed into new value store
-    //  r3 = memw(r2++#4)
-    //  memw(r30 + #-1404) = r2.new -> can not be new value store
-    // arch spec section: 5.4.2.1
-    return false;
-  }
-
-  // If the source that feeds the store is predicated, new value store must also be
-  // also predicated.
-  if (QII->isPredicated(PacketMI)) {
-    if (!QII->isPredicated(MI))
-      return false;
-
-    // Check to make sure that they both will have their predicates
-    // evaluate identically
-    unsigned predRegNumSrc;
-    unsigned predRegNumDst;
-    const TargetRegisterClass* predRegClass;
-
-    // Get predicate register used in the source instruction
-    for(unsigned opNum = 0; opNum < PacketMI->getNumOperands(); opNum++) {
-      if ( PacketMI->getOperand(opNum).isReg())
-      predRegNumSrc = PacketMI->getOperand(opNum).getReg();
-      predRegClass = QRI->getMinimalPhysRegClass(predRegNumSrc);
-      if (predRegClass == Hexagon::PredRegsRegisterClass) {
-        break;
-      }
-    }
-    assert ((predRegClass == Hexagon::PredRegsRegisterClass ) &&
-        ("predicate register not found in a predicated PacketMI instruction"));
-
-    // Get predicate register used in new-value store instruction
-    for(unsigned opNum = 0; opNum < MI->getNumOperands(); opNum++) {
-      if ( MI->getOperand(opNum).isReg())
-      predRegNumDst = MI->getOperand(opNum).getReg();
-      predRegClass = QRI->getMinimalPhysRegClass(predRegNumDst);
-      if (predRegClass == Hexagon::PredRegsRegisterClass) {
-        break;
-      }
-    }
-    assert ((predRegClass == Hexagon::PredRegsRegisterClass ) &&
-            ("predicate register not found in a predicated MI instruction"));
-
-    // New-value register producer and user (store) need to satisfy these
-    // constraints:
-    // 1) Both instructions should be predicated on the same register.
-    // 2) If producer of the new-value register is .new predicated then store
-    // should also be .new predicated and if producer is not .new predicated
-    // then store should not be .new predicated.
-    // 3) Both new-value register producer and user should have same predicate
-    // sense, i.e, either both should be negated or both should be none negated.
-
-    if (( predRegNumDst != predRegNumSrc) ||
-          isDotNewInst(PacketMI) != isDotNewInst(MI)  ||
-          GetPredicateSense(MI, QII) != GetPredicateSense(PacketMI, QII)) {
-      return false;
-    }
-  }
-
-  // Make sure that other than the new-value register no other store instruction
-  // register has been modified in the same packet. Predicate registers can be
-  // modified by they should not be modified between the producer and the store
-  // instruction as it will make them both conditional on different values.
-  // We already know this to be true for all the instructions before and
-  // including PacketMI. Howerver, we need to perform the check for the
-  // remaining instructions in the packet.
-
-  std::vector::iterator VI;
-  std::vector::iterator VE;
-  unsigned StartCheck = 0;
-
-  for (VI=CurrentPacketMIs.begin(), VE = CurrentPacketMIs.end();
-      (VI != VE); ++VI) {
-    SUnit* TempSU = MIToSUnit[*VI];
-    MachineInstr* TempMI = TempSU->getInstr();
-
-    // Following condition is true for all the instructions until PacketMI is
-    // reached (StartCheck is set to 0 before the for loop).
-    // StartCheck flag is 1 for all the instructions after PacketMI.
-    if (TempMI != PacketMI && !StartCheck) // start processing only after
-      continue;                            // encountering PacketMI
-
-    StartCheck = 1;
-    if (TempMI == PacketMI) // We don't want to check PacketMI for dependence
-      continue;
-
-    for(unsigned opNum = 0; opNum < MI->getNumOperands(); opNum++) {
-      if (MI->getOperand(opNum).isReg() &&
-          TempSU->getInstr()->modifiesRegister(MI->getOperand(opNum).getReg(), QRI))
-        return false;
-    }
-  }
-
-  // Make sure that for non POST_INC stores:
-  // 1. The only use of reg is DepReg and no other registers.
-  //    This handles V4 base+index registers.
-  //    The following store can not be dot new.
-  //    Eg.   r0 = add(r0, #3)a
-  //          memw(r1+r0<<#2) = r0
-  if (!QII->isPostIncrement(MI) &&
-      GetStoreValueOperand(MI).isReg() &&
-      GetStoreValueOperand(MI).getReg() == DepReg) {
-    for(unsigned opNum = 0; opNum < MI->getNumOperands()-1; opNum++) {
-      if (MI->getOperand(opNum).isReg() &&
-          MI->getOperand(opNum).getReg() == DepReg) {
-        return false;
-      }
-    }
-    // 2. If data definition is because of implicit definition of the register,
-    //    do not newify the store. Eg.
-    //    %R9 = ZXTH %R12, %D6, %R12
-    //    STrih_indexed %R8, 2, %R12; mem:ST2[%scevgep343]
-    for(unsigned opNum = 0; opNum < PacketMI->getNumOperands(); opNum++) {
-      if (PacketMI->getOperand(opNum).isReg() &&
-          PacketMI->getOperand(opNum).getReg() == DepReg &&
-          PacketMI->getOperand(opNum).isDef() &&
-          PacketMI->getOperand(opNum).isImplicit()) {
-        return false;
-      }
-    }
-  }
-
-  // Can be dot new store.
-  return true;
-}
-
-// can this MI to promoted to either
-// new value store or new value jump
-bool HexagonPacketizerList::CanPromoteToNewValue( MachineInstr *MI,
-                SUnit *PacketSU, unsigned DepReg,
-                std::map  MIToSUnit,
-                MachineBasicBlock::iterator &MII)
-{
-
-  const HexagonRegisterInfo* QRI = (const HexagonRegisterInfo *) TM.getRegisterInfo();
-  if (!QRI->Subtarget.hasV4TOps() ||
-      !IsNewifyStore(MI))
-    return false;
-
-  MachineInstr *PacketMI = PacketSU->getInstr();
-
-  // Check to see the store can be new value'ed.
-  if (CanPromoteToNewValueStore(MI, PacketMI, DepReg, MIToSUnit))
-    return true;
-
-  // Check to see the compare/jump can be new value'ed.
-  // This is done as a pass on its own. Don't need to check it here.
-  return false;
-}
-
-// Check to see if an instruction can be dot new
-// There are three kinds.
-// 1. dot new on predicate - V2/V3/V4
-// 2. dot new on stores NV/ST - V4
-// 3. dot new on jump NV/J - V4 -- This is generated in a pass.
-bool HexagonPacketizerList::CanPromoteToDotNew( MachineInstr *MI,
-                              SUnit *PacketSU, unsigned DepReg,
-                              std::map  MIToSUnit,
-                              MachineBasicBlock::iterator &MII,
-                              const TargetRegisterClass* RC )
-{
-  // already a dot new instruction
-  if (isDotNewInst(MI) && !IsNewifyStore(MI))
-    return false;
-
-  if (!isNewifiable(MI))
-    return false;
-
-  // predicate .new
-  if (RC == Hexagon::PredRegsRegisterClass && isCondInst(MI))
-      return true;
-  else if (RC != Hexagon::PredRegsRegisterClass &&
-      !IsNewifyStore(MI)) // MI is not a new-value store
-    return false;
-  else {
-    // Create a dot new machine instruction to see if resources can be
-    // allocated. If not, bail out now.
-    const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-    int NewOpcode = GetDotNewOp(MI->getOpcode());
-    const MCInstrDesc &desc = QII->get(NewOpcode);
-    DebugLoc dl;
-    MachineInstr *NewMI = MI->getParent()->getParent()->CreateMachineInstr(desc, dl);
-    bool ResourcesAvailable = ResourceTracker->canReserveResources(NewMI);
-    MI->getParent()->getParent()->DeleteMachineInstr(NewMI);
-
-    if (!ResourcesAvailable)
-      return false;
-
-    // new value store only
-    // new new value jump generated as a passes
-    if (!CanPromoteToNewValue(MI, PacketSU, DepReg, MIToSUnit, MII)) {
-      return false;
-    }
-  }
-  return true;
-}
-
-// Go through the packet instructions and search for anti dependency
-// between them and DepReg from MI
-// Consider this case:
-// Trying to add
-// a) %R1 = TFRI_cdNotPt %P3, 2
-// to this packet:
-// {
-//   b) %P0 = OR_pp %P3, %P0
-//   c) %P3 = TFR_PdRs %R23
-//   d) %R1 = TFRI_cdnPt %P3, 4
-//  }
-// The P3 from a) and d) will be complements after
-// a)'s P3 is converted to .new form
-// Anti Dep between c) and b) is irrelevant for this case
-bool HexagonPacketizerList::RestrictingDepExistInPacket (MachineInstr* MI,
-      unsigned DepReg,
-      std::map  MIToSUnit) {
-
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-  SUnit* PacketSUDep = MIToSUnit[MI];
-
-  for (std::vector::iterator VIN = CurrentPacketMIs.begin(),
-       VEN = CurrentPacketMIs.end(); (VIN != VEN); ++VIN) {
-
-    // We only care for dependencies to predicated instructions
-    if(!QII->isPredicated(*VIN)) continue;
-
-    // Scheduling Unit for current insn in the packet
-    SUnit* PacketSU = MIToSUnit[*VIN];
-
-    // Look at dependencies between current members of the packet
-    // and predicate defining instruction MI.
-    // Make sure that dependency is on the exact register
-    // we care about.
-    if (PacketSU->isSucc(PacketSUDep)) {
-      for (unsigned i = 0; i < PacketSU->Succs.size(); ++i) {
-        if ((PacketSU->Succs[i].getSUnit() == PacketSUDep) &&
-            (PacketSU->Succs[i].getKind() == SDep::Anti) &&
-            (PacketSU->Succs[i].getReg() == DepReg)) {
-          return true;
-        }
-      }
-    }
-  }
-
-  return false;
-}
-
-
-// Given two predicated instructions, this function detects whether
-// the predicates are complements
-bool HexagonPacketizerList::ArePredicatesComplements (MachineInstr* MI1,
-     MachineInstr* MI2, std::map  MIToSUnit) {
-
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-  // Currently can only reason about conditional transfers
-  if (!QII->isConditionalTransfer(MI1) || !QII->isConditionalTransfer(MI2)) {
-    return false;
-  }
-
-  // Scheduling unit for candidate
-  SUnit* SU = MIToSUnit[MI1];
-
-  // One corner case deals with the following scenario:
-  // Trying to add
-  // a) %R24 = TFR_cPt %P0, %R25
-  // to this packet:
-  //
-  // {
-  //   b) %R25 = TFR_cNotPt %P0, %R24
-  //   c) %P0 = CMPEQri %R26, 1
-  // }
-  //
-  // On general check a) and b) are complements, but
-  // presence of c) will convert a) to .new form, and
-  // then it is not a complement
-  // We attempt to detect it by analyzing  existing
-  // dependencies in the packet
-
-  // Analyze relationships between all existing members of the packet.
-  // Look for Anti dependecy on the same predicate reg
-  // as used in the candidate
-  for (std::vector::iterator VIN = CurrentPacketMIs.begin(),
-       VEN = CurrentPacketMIs.end(); (VIN != VEN); ++VIN) {
-
-    // Scheduling Unit for current insn in the packet
-    SUnit* PacketSU = MIToSUnit[*VIN];
-
-    // If this instruction in the packet is succeeded by the candidate...
-    if (PacketSU->isSucc(SU)) {
-      for (unsigned i = 0; i < PacketSU->Succs.size(); ++i) {
-        // The corner case exist when there is true data
-        // dependency between candidate and one of current
-        // packet members, this dep is on predicate reg, and
-        // there already exist anti dep on the same pred in
-        // the packet.
-        if (PacketSU->Succs[i].getSUnit() == SU &&
-            Hexagon::PredRegsRegisterClass->contains(
-              PacketSU->Succs[i].getReg()) &&
-            PacketSU->Succs[i].getKind() == SDep::Data &&
-            // Here I know that *VIN is predicate setting instruction
-            // with true data dep to candidate on the register
-            // we care about - c) in the above example.
-            // Now I need to see if there is an anti dependency
-            // from c) to any other instruction in the
-            // same packet on the pred reg of interest
-            RestrictingDepExistInPacket(*VIN,PacketSU->Succs[i].getReg(),
-                                        MIToSUnit)) {
-           return false;
-        }
-      }
-    }
-  }
-
-  // If the above case does not apply, check regular
-  // complement condition.
-  // Check that the predicate register is the same and
-  // that the predicate sense is different
-  // We also need to differentiate .old vs. .new:
-  // !p0 is not complimentary to p0.new
-  return ((MI1->getOperand(1).getReg() == MI2->getOperand(1).getReg()) &&
-          (GetPredicateSense(MI1, QII) != GetPredicateSense(MI2, QII)) &&
-          (isDotNewInst(MI1) == isDotNewInst(MI2)));
-}
-
-// initPacketizerState - Initialize packetizer flags
-void HexagonPacketizerList::initPacketizerState() {
-
-  Dependence = false;
-  PromotedToDotNew = false;
-  GlueToNewValueJump = false;
-  GlueAllocframeStore = false;
-  FoundSequentialDependence = false;
-
-  return;
-}
-
-// ignorePseudoInstruction - Ignore bundling of pseudo instructions.
-bool HexagonPacketizerList::ignorePseudoInstruction(MachineInstr *MI,
-                                                    MachineBasicBlock *MBB) {
-  if (MI->isDebugValue())
-    return true;
-
-  // We must print out inline assembly
-  if (MI->isInlineAsm())
-    return false;
-
-  // We check if MI has any functional units mapped to it.
-  // If it doesn't, we ignore the instruction.
-  const MCInstrDesc& TID = MI->getDesc();
-  unsigned SchedClass = TID.getSchedClass();
-  const InstrStage* IS = ResourceTracker->getInstrItins()->beginStage(SchedClass);
-  unsigned FuncUnits = IS->getUnits();
-  return !FuncUnits;
-}
-
-// isSoloInstruction: - Returns true for instructions that must be
-// scheduled in their own packet.
-bool HexagonPacketizerList::isSoloInstruction(MachineInstr *MI) {
-
-  if (MI->isInlineAsm())
-    return true;
-
-  if (MI->isEHLabel())
-    return true;
-
-  // From Hexagon V4 Programmer's Reference Manual 3.4.4 Grouping constraints:
-  // trap, pause, barrier, icinva, isync, and syncht are solo instructions.
-  // They must not be grouped with other instructions in a packet.
-  if (IsSchedBarrier(MI))
-    return true;
-
-  return false;
-}
-
-// isLegalToPacketizeTogether:
-// SUI is the current instruction that is out side of the current packet.
-// SUJ is the current instruction inside the current packet against which that
-// SUI will be packetized.
-bool HexagonPacketizerList::isLegalToPacketizeTogether(SUnit *SUI, SUnit *SUJ) {
-  MachineInstr *I = SUI->getInstr();
-  MachineInstr *J = SUJ->getInstr();
-  assert(I && J && "Unable to packetize null instruction!");
-
-  const MCInstrDesc &MCIDI = I->getDesc();
-  const MCInstrDesc &MCIDJ = J->getDesc();
-
-  MachineBasicBlock::iterator II = I;
-
-  const unsigned FrameSize = MF.getFrameInfo()->getStackSize();
-  const HexagonRegisterInfo* QRI = (const HexagonRegisterInfo *) TM.getRegisterInfo();
-  const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-
-  // Inline asm cannot go in the packet.
-  if (I->getOpcode() == Hexagon::INLINEASM)
-    llvm_unreachable("Should not meet inline asm here!");
-
-  if (isSoloInstruction(I))
-    llvm_unreachable("Should not meet solo instr here!");
-
-  // A save callee-save register function call can only be in a packet
-  // with instructions that don't write to the callee-save registers.
-  if ((QII->isSaveCalleeSavedRegsCall(I) &&
-       DoesModifyCalleeSavedReg(J, QRI)) ||
-      (QII->isSaveCalleeSavedRegsCall(J) &&
-       DoesModifyCalleeSavedReg(I, QRI))) {
-    Dependence = true;
-    return false;
-  }
-
-  // Two control flow instructions cannot go in the same packet.
-  if (IsControlFlow(I) && IsControlFlow(J)) {
-    Dependence = true;
-    return false;
-  }
-
-  // A LoopN instruction cannot appear in the same packet as a jump or call.
-  if (IsLoopN(I) && (   IsDirectJump(J)
-                     || MCIDJ.isCall()
-                     || QII->isDeallocRet(J))) {
-    Dependence = true;
-    return false;
-  }
-  if (IsLoopN(J) && (   IsDirectJump(I)
-                     || MCIDI.isCall()
-                     || QII->isDeallocRet(I))) {
-    Dependence = true;
-    return false;
-  }
-
-  // dealloc_return cannot appear in the same packet as a conditional or
-  // unconditional jump.
-  if (QII->isDeallocRet(I) && (   MCIDJ.isBranch()
-                               || MCIDJ.isCall()
-                               || MCIDJ.isBarrier())) {
-    Dependence = true;
-    return false;
-  }
-
-
-  // V4 allows dual store. But does not allow second store, if the
-  // first store is not in SLOT0. New value store, new value jump,
-  // dealloc_return and memop always take SLOT0.
-  // Arch spec 3.4.4.2
-  if (QRI->Subtarget.hasV4TOps()) {
-
-    if (MCIDI.mayStore() && MCIDJ.mayStore() && isNewValueInst(J)) {
-      Dependence = true;
-      return false;
-    }
-
-    if (   (QII->isMemOp(J) && MCIDI.mayStore())
-        || (MCIDJ.mayStore() && QII->isMemOp(I))
-        || (QII->isMemOp(J) && QII->isMemOp(I))) {
-      Dependence = true;
-      return false;
-    }
-
-    //if dealloc_return
-    if (MCIDJ.mayStore() && QII->isDeallocRet(I)){
-      Dependence = true;
-      return false;
-    }
-
-    // If an instruction feeds new value jump, glue it.
-    MachineBasicBlock::iterator NextMII = I;
-    ++NextMII;
-    MachineInstr *NextMI = NextMII;
-
-    if (QII->isNewValueJump(NextMI)) {
-
-      bool secondRegMatch = false;
-      bool maintainNewValueJump = false;
-
-      if (NextMI->getOperand(1).isReg() &&
-          I->getOperand(0).getReg() == NextMI->getOperand(1).getReg()) {
-        secondRegMatch = true;
-        maintainNewValueJump = true;
-      }
-
-      if (!secondRegMatch &&
-           I->getOperand(0).getReg() == NextMI->getOperand(0).getReg()) {
-        maintainNewValueJump = true;
-      }
-
-      for (std::vector::iterator
-            VI = CurrentPacketMIs.begin(),
-             VE = CurrentPacketMIs.end();
-           (VI != VE && maintainNewValueJump); ++VI) {
-        SUnit* PacketSU = MIToSUnit[*VI];
-
-        // NVJ can not be part of the dual jump - Arch Spec: section 7.8
-        if (PacketSU->getInstr()->getDesc().isCall()) {
-          Dependence = true;
-          break;
-        }
-        // Validate
-        // 1. Packet does not have a store in it.
-        // 2. If the first operand of the nvj is newified, and the second
-        //    operand is also a reg, it (second reg) is not defined in
-        //    the same packet.
-        // 3. If the second operand of the nvj is newified, (which means
-        //    first operand is also a reg), first reg is not defined in
-        //    the same packet.
-        if (PacketSU->getInstr()->getDesc().mayStore()               ||
-            PacketSU->getInstr()->getOpcode() == Hexagon::ALLOCFRAME ||
-            // Check #2.
-            (!secondRegMatch && NextMI->getOperand(1).isReg() &&
-             PacketSU->getInstr()->modifiesRegister(
-                               NextMI->getOperand(1).getReg(), QRI)) ||
-            // Check #3.
-            (secondRegMatch &&
-             PacketSU->getInstr()->modifiesRegister(
-                               NextMI->getOperand(0).getReg(), QRI))) {
-          Dependence = true;
-          break;
-        }
-      }
-      if (!Dependence)
-        GlueToNewValueJump = true;
-      else
-        return false;
-    }
-  }
-
-  if (SUJ->isSucc(SUI)) {
-    for (unsigned i = 0;
-         (i < SUJ->Succs.size()) && !FoundSequentialDependence;
-         ++i) {
-
-      if (SUJ->Succs[i].getSUnit() != SUI) {
-        continue;
-      }
-
-      SDep::Kind DepType = SUJ->Succs[i].getKind();
-
-      // For direct calls:
-      // Ignore register dependences for call instructions for
-      // packetization purposes except for those due to r31 and
-      // predicate registers.
-      //
-      // For indirect calls:
-      // Same as direct calls + check for true dependences to the register
-      // used in the indirect call.
-      //
-      // We completely ignore Order dependences for call instructions
-      //
-      // For returns:
-      // Ignore register dependences for return instructions like jumpr,
-      // dealloc return unless we have dependencies on the explicit uses
-      // of the registers used by jumpr (like r31) or dealloc return
-      // (like r29 or r30).
-      //
-      // TODO: Currently, jumpr is handling only return of r31. So, the
-      // following logic (specificaly IsCallDependent) is working fine.
-      // We need to enable jumpr for register other than r31 and then,
-      // we need to rework the last part, where it handles indirect call
-      // of that (IsCallDependent) function. Bug 6216 is opened for this.
-      //
-      unsigned DepReg;
-      const TargetRegisterClass* RC;
-      if (DepType == SDep::Data) {
-        DepReg = SUJ->Succs[i].getReg();
-        RC = QRI->getMinimalPhysRegClass(DepReg);
-      }
-      if ((MCIDI.isCall() || MCIDI.isReturn()) &&
-          (!IsRegDependence(DepType) ||
-            !IsCallDependent(I, DepType, SUJ->Succs[i].getReg()))) {
-        /* do nothing */
-      }
-
-      // For instructions that can be promoted to dot-new, try to promote.
-      else if ((DepType == SDep::Data) &&
-               CanPromoteToDotNew(I, SUJ, DepReg, MIToSUnit, II, RC) &&
-               PromoteToDotNew(I, DepType, II, RC)) {
-        PromotedToDotNew = true;
-        /* do nothing */
-      }
-
-      else if ((DepType == SDep::Data) &&
-               (QII->isNewValueJump(I))) {
-        /* do nothing */
-      }
-
-      // For predicated instructions, if the predicates are complements
-      // then there can be no dependence.
-      else if (QII->isPredicated(I) &&
-               QII->isPredicated(J) &&
-          ArePredicatesComplements(I, J, MIToSUnit)) {
-        /* do nothing */
-
-      }
-      else if (IsDirectJump(I) &&
-               !MCIDJ.isBranch() &&
-               !MCIDJ.isCall() &&
-               (DepType == SDep::Order)) {
-        // Ignore Order dependences between unconditional direct branches
-        // and non-control-flow instructions
-        /* do nothing */
-      }
-      else if (MCIDI.isConditionalBranch() && (DepType != SDep::Data) &&
-               (DepType != SDep::Output)) {
-        // Ignore all dependences for jumps except for true and output
-        // dependences
-        /* do nothing */
-      }
-
-      // Ignore output dependences due to superregs. We can
-      // write to two different subregisters of R1:0 for instance
-      // in the same cycle
-      //
-
-      //
-      // Let the
-      // If neither I nor J defines DepReg, then this is a
-      // superfluous output dependence. The dependence must be of the
-      // form:
-      //  R0 = ...
-      //  R1 = ...
-      // and there is an output dependence between the two instructions
-      // with
-      // DepReg = D0
-      // We want to ignore these dependences.
-      // Ideally, the dependence constructor should annotate such
-      // dependences. We can then avoid this relatively expensive check.
-      //
-      else if (DepType == SDep::Output) {
-        // DepReg is the register that's responsible for the dependence.
-        unsigned DepReg = SUJ->Succs[i].getReg();
-
-        // Check if I and J really defines DepReg.
-        if (I->definesRegister(DepReg) ||
-            J->definesRegister(DepReg)) {
-          FoundSequentialDependence = true;
-          break;
-        }
-      }
-
-      // We ignore Order dependences for
-      // 1. Two loads unless they are volatile.
-      // 2. Two stores in V4 unless they are volatile.
-      else if ((DepType == SDep::Order) &&
-               !I->hasVolatileMemoryRef() &&
-               !J->hasVolatileMemoryRef()) {
-        if (QRI->Subtarget.hasV4TOps() &&
-            // hexagonv4 allows dual store.
-            MCIDI.mayStore() && MCIDJ.mayStore()) {
-          /* do nothing */
-        }
-        // store followed by store-- not OK on V2
-        // store followed by load -- not OK on all (OK if addresses
-        // are not aliased)
-        // load followed by store -- OK on all
-        // load followed by load  -- OK on all
-        else if ( !MCIDJ.mayStore()) {
-          /* do nothing */
-        }
-        else {
-          FoundSequentialDependence = true;
-          break;
-        }
-      }
-
-      // For V4, special case ALLOCFRAME. Even though there is dependency
-      // between ALLOCAFRAME and subsequent store, allow it to be
-      // packetized in a same packet. This implies that the store is using
-      // caller's SP. Hense, offset needs to be updated accordingly.
-      else if (DepType == SDep::Data
-               && QRI->Subtarget.hasV4TOps()
-               && J->getOpcode() == Hexagon::ALLOCFRAME
-               && (I->getOpcode() == Hexagon::STrid
-                   || I->getOpcode() == Hexagon::STriw
-                   || I->getOpcode() == Hexagon::STrib)
-               && I->getOperand(0).getReg() == QRI->getStackRegister()
-               && QII->isValidOffset(I->getOpcode(),
-                                     I->getOperand(1).getImm() -
-                                     (FrameSize + HEXAGON_LRFP_SIZE)))
-      {
-        GlueAllocframeStore = true;
-        // Since this store is to be glued with allocframe in the same
-        // packet, it will use SP of the previous stack frame, i.e
-        // caller's SP. Therefore, we need to recalculate offset according
-        // to this change.
-        I->getOperand(1).setImm(I->getOperand(1).getImm() -
-                                        (FrameSize + HEXAGON_LRFP_SIZE));
-      }
-
-      //
-      // Skip over anti-dependences. Two instructions that are
-      // anti-dependent can share a packet
-      //
-      else if (DepType != SDep::Anti) {
-        FoundSequentialDependence = true;
-        break;
-      }
-    }
-
-    if (FoundSequentialDependence) {
-      Dependence = true;
-      return false;
-    }
-  }
-
-  return true;
-}
-
-// isLegalToPruneDependencies
-bool HexagonPacketizerList::isLegalToPruneDependencies(SUnit *SUI, SUnit *SUJ) {
-  MachineInstr *I = SUI->getInstr();
-  assert(I && SUJ->getInstr() && "Unable to packetize null instruction!");
-
-  const unsigned FrameSize = MF.getFrameInfo()->getStackSize();
-
-  if (Dependence) {
-
-    // Check if the instruction was promoted to a dot-new. If so, demote it
-    // back into a dot-old.
-    if (PromotedToDotNew) {
-      DemoteToDotOld(I);
-    }
-
-    // Check if the instruction (must be a store) was glued with an Allocframe
-    // instruction. If so, restore its offset to its original value, i.e. use
-    // curent SP instead of caller's SP.
-    if (GlueAllocframeStore) {
-      I->getOperand(1).setImm(I->getOperand(1).getImm() +
-                                             FrameSize + HEXAGON_LRFP_SIZE);
-    }
-
-    return false;
-  }
-  return true;
-}
-
-MachineBasicBlock::iterator HexagonPacketizerList::addToPacket(MachineInstr *MI) {
-
-    MachineBasicBlock::iterator MII = MI;
-    MachineBasicBlock *MBB = MI->getParent();
-
-    const HexagonInstrInfo *QII = (const HexagonInstrInfo *) TII;
-
-    if (GlueToNewValueJump) {
-
-      ++MII;
-      MachineInstr *nvjMI = MII;
-      assert(ResourceTracker->canReserveResources(MI));
-      ResourceTracker->reserveResources(MI);
-      if (QII->isExtended(MI) &&
-          !tryAllocateResourcesForConstExt(MI)) {
-        endPacket(MBB, MI);
-        ResourceTracker->reserveResources(MI);
-        assert(canReserveResourcesForConstExt(MI) &&
-               "Ensure that there is a slot");
-        reserveResourcesForConstExt(MI);
-        // Reserve resources for new value jump constant extender.
-        assert(canReserveResourcesForConstExt(MI) &&
-               "Ensure that there is a slot");
-        reserveResourcesForConstExt(nvjMI);
-        assert(ResourceTracker->canReserveResources(nvjMI) &&
-               "Ensure that there is a slot");
-
-      } else if (   // Extended instruction takes two slots in the packet.
-        // Try reserve and allocate 4-byte in the current packet first.
-        (QII->isExtended(nvjMI)
-            && (!tryAllocateResourcesForConstExt(nvjMI)
-                || !ResourceTracker->canReserveResources(nvjMI)))
-        || // For non-extended instruction, no need to allocate extra 4 bytes.
-        (!QII->isExtended(nvjMI) && !ResourceTracker->canReserveResources(nvjMI)))
-      {
-        endPacket(MBB, MI);
-        // A new and empty packet starts.
-        // We are sure that the resources requirements can be satisfied.
-        // Therefore, do not need to call "canReserveResources" anymore.
-        ResourceTracker->reserveResources(MI);
-        if (QII->isExtended(nvjMI))
-          reserveResourcesForConstExt(nvjMI);
-      }
-      // Here, we are sure that "reserveResources" would succeed.
-      ResourceTracker->reserveResources(nvjMI);
-      CurrentPacketMIs.push_back(MI);
-      CurrentPacketMIs.push_back(nvjMI);
-    } else {
-      if (   QII->isExtended(MI)
-          && (   !tryAllocateResourcesForConstExt(MI)
-              || !ResourceTracker->canReserveResources(MI)))
-      {
-        endPacket(MBB, MI);
-        // Check if the instruction was promoted to a dot-new. If so, demote it
-        // back into a dot-old
-        if (PromotedToDotNew) {
-          DemoteToDotOld(MI);
-        }
-        reserveResourcesForConstExt(MI);
-      }
-      // In case that "MI" is not an extended insn,
-      // the resource availability has already been checked.
-      ResourceTracker->reserveResources(MI);
-      CurrentPacketMIs.push_back(MI);
-    }
-    return MII;
-}
-
-//===----------------------------------------------------------------------===//
-//                         Public Constructor Functions
-//===----------------------------------------------------------------------===//
-
-FunctionPass *llvm::createHexagonPacketizer() {
-  return new HexagonPacketizer();
-}
-
diff --git a/lib/Target/Hexagon/InstPrinter/HexagonInstPrinter.cpp b/lib/Target/Hexagon/InstPrinter/HexagonInstPrinter.cpp
index 75d6bfb0813..47384cd533f 100644
--- a/lib/Target/Hexagon/InstPrinter/HexagonInstPrinter.cpp
+++ b/lib/Target/Hexagon/InstPrinter/HexagonInstPrinter.cpp
@@ -15,7 +15,6 @@
 #include "Hexagon.h"
 #include "HexagonAsmPrinter.h"
 #include "HexagonInstPrinter.h"
-#include "HexagonMCInst.h"
 #include "llvm/MC/MCInst.h"
 #include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCExpr.h"
@@ -38,50 +37,20 @@ StringRef HexagonInstPrinter::getRegName(unsigned RegNo) const {
 
 void HexagonInstPrinter::printInst(const MCInst *MI, raw_ostream &O,
                                    StringRef Annot) {
-  printInst((const HexagonMCInst*)(MI), O, Annot);
-}
-
-void HexagonInstPrinter::printInst(const HexagonMCInst *MI, raw_ostream &O,
-                                   StringRef Annot) {
   const char packetPadding[] = "      ";
   const char startPacket = '{',
              endPacket = '}';
   // TODO: add outer HW loop when it's supported too.
   if (MI->getOpcode() == Hexagon::ENDLOOP0) {
-    // Ending a harware loop is different from ending an regular packet.
-    assert(MI->isEndPacket() && "Loop end must also end the packet");
+    MCInst Nop;
 
-    if (MI->isStartPacket()) {
-      // There must be a packet to end a loop.
-      // FIXME: when shuffling is always run, this shouldn't be needed.
-      HexagonMCInst Nop;
-      StringRef NoAnnot;
-
-      Nop.setOpcode (Hexagon::NOP);
-      Nop.setStartPacket (MI->isStartPacket());
-      printInst (&Nop, O, NoAnnot);
-    }
-
-    // Close the packet.
-    if (MI->isEndPacket())
-      O << packetPadding << endPacket;
-
-    printInstruction(MI, O);
-  }
-  else {
-    // Prefix the insn opening the packet.
-    if (MI->isStartPacket())
-      O << packetPadding << startPacket << '\n';
-
-    printInstruction(MI, O);
-
-    // Suffix the insn closing the packet.
-    if (MI->isEndPacket())
-      // Suffix the packet in a new line always, since the GNU assembler has
-      // issues with a closing brace on the same line as CONST{32,64}.
-      O << '\n' << packetPadding << endPacket;
+    O << packetPadding << startPacket << '\n';
+    Nop.setOpcode(Hexagon::NOP);
+    printInstruction(&Nop, O);
+    O << packetPadding << endPacket;
   }
 
+  printInstruction(MI, O);
   printAnnotation(O, Annot);
 }
 
@@ -100,18 +69,18 @@ void HexagonInstPrinter::printOperand(const MCInst *MI, unsigned OpNo,
   }
 }
 
-void HexagonInstPrinter::printImmOperand(const MCInst *MI, unsigned OpNo,
-                                         raw_ostream &O) const {
+void HexagonInstPrinter::printImmOperand
+  (const MCInst *MI, unsigned OpNo, raw_ostream &O) const {
   O << MI->getOperand(OpNo).getImm();
 }
 
 void HexagonInstPrinter::printExtOperand(const MCInst *MI, unsigned OpNo,
-                                         raw_ostream &O) const {
+                                                raw_ostream &O) const {
   O << MI->getOperand(OpNo).getImm();
 }
 
-void HexagonInstPrinter::printUnsignedImmOperand(const MCInst *MI, unsigned OpNo,
-                                                 raw_ostream &O) const {
+void HexagonInstPrinter::printUnsignedImmOperand
+  (const MCInst *MI, unsigned OpNo, raw_ostream &O) const {
   O << MI->getOperand(OpNo).getImm();
 }
 
@@ -120,13 +89,13 @@ void HexagonInstPrinter::printNegImmOperand(const MCInst *MI, unsigned OpNo,
   O << -MI->getOperand(OpNo).getImm();
 }
 
-void HexagonInstPrinter::printNOneImmOperand(const MCInst *MI, unsigned OpNo,
-                                             raw_ostream &O) const {
+void HexagonInstPrinter::printNOneImmOperand
+  (const MCInst *MI, unsigned OpNo, raw_ostream &O) const {
   O << -1;
 }
 
-void HexagonInstPrinter::printMEMriOperand(const MCInst *MI, unsigned OpNo,
-                                           raw_ostream &O) const {
+void HexagonInstPrinter::printMEMriOperand
+  (const MCInst *MI, unsigned OpNo, raw_ostream &O) const {
   const MCOperand& MO0 = MI->getOperand(OpNo);
   const MCOperand& MO1 = MI->getOperand(OpNo + 1);
 
@@ -134,8 +103,8 @@ void HexagonInstPrinter::printMEMriOperand(const MCInst *MI, unsigned OpNo,
   O << " + #" << MO1.getImm();
 }
 
-void HexagonInstPrinter::printFrameIndexOperand(const MCInst *MI, unsigned OpNo,
-                                                raw_ostream &O) const {
+void HexagonInstPrinter::printFrameIndexOperand
+  (const MCInst *MI, unsigned OpNo, raw_ostream &O) const {
   const MCOperand& MO0 = MI->getOperand(OpNo);
   const MCOperand& MO1 = MI->getOperand(OpNo + 1);
 
diff --git a/lib/Target/Hexagon/InstPrinter/HexagonInstPrinter.h b/lib/Target/Hexagon/InstPrinter/HexagonInstPrinter.h
index 3ce7dfcbdbe..dad4334c3eb 100644
--- a/lib/Target/Hexagon/InstPrinter/HexagonInstPrinter.h
+++ b/lib/Target/Hexagon/InstPrinter/HexagonInstPrinter.h
@@ -14,7 +14,6 @@
 #ifndef HEXAGONINSTPRINTER_H
 #define HEXAGONINSTPRINTER_H
 
-#include "HexagonMCInst.h"
 #include "llvm/MC/MCInstPrinter.h"
 
 namespace llvm {
@@ -26,7 +25,6 @@ namespace llvm {
       : MCInstPrinter(MAI, MII, MRI) {}
 
     virtual void printInst(const MCInst *MI, raw_ostream &O, StringRef Annot);
-    void printInst(const HexagonMCInst *MI, raw_ostream &O, StringRef Annot);
     virtual StringRef getOpcodeName(unsigned Opcode) const;
     void printInstruction(const MCInst *MI, raw_ostream &O);
     StringRef getRegName(unsigned RegNo) const;
diff --git a/lib/Target/Hexagon/MCTargetDesc/CMakeLists.txt b/lib/Target/Hexagon/MCTargetDesc/CMakeLists.txt
index c9c5a6eadf8..8e3da99404e 100644
--- a/lib/Target/Hexagon/MCTargetDesc/CMakeLists.txt
+++ b/lib/Target/Hexagon/MCTargetDesc/CMakeLists.txt
@@ -1,6 +1,6 @@
 add_llvm_library(LLVMHexagonDesc
-  HexagonMCAsmInfo.cpp
   HexagonMCTargetDesc.cpp
+  HexagonMCAsmInfo.cpp
   )
 
 add_dependencies(LLVMHexagonDesc HexagonCommonTableGen)
diff --git a/lib/Target/Hexagon/MCTargetDesc/HexagonBaseInfo.h b/lib/Target/Hexagon/MCTargetDesc/HexagonBaseInfo.h
index 7221e906342..ed55c3c1c15 100644
--- a/lib/Target/Hexagon/MCTargetDesc/HexagonBaseInfo.h
+++ b/lib/Target/Hexagon/MCTargetDesc/HexagonBaseInfo.h
@@ -23,41 +23,14 @@ namespace llvm {
 /// instruction info tracks.
 ///
 namespace HexagonII {
+
   // *** The code below must match HexagonInstrFormat*.td *** //
 
-  // Insn types.
-  // *** Must match HexagonInstrFormat*.td ***
-  enum Type {
-    TypePSEUDO = 0,
-    TypeALU32  = 1,
-    TypeCR     = 2,
-    TypeJR     = 3,
-    TypeJ      = 4,
-    TypeLD     = 5,
-    TypeST     = 6,
-    TypeSYSTEM = 7,
-    TypeXTYPE  = 8,
-    TypeMEMOP  = 9,
-    TypeNV     = 10,
-    TypePREFIX = 30, // Such as extenders.
-    TypeMARKER = 31  // Such as end of a HW loop.
-  };
-
-
-
   // MCInstrDesc TSFlags
-  // *** Must match HexagonInstrFormat*.td ***
   enum {
-    // This 5-bit field describes the insn type.
-    TypePos  = 0,
-    TypeMask = 0x1f,
-
-    // Solo instructions.
-    SoloPos  = 5,
-    SoloMask = 0x1,
 
     // Predicated instructions.
-    PredicatedPos  = 6,
+    PredicatedPos  = 1,
     PredicatedMask = 0x1
   };
 
diff --git a/lib/Target/MBlaze/MBlazeCallingConv.td b/lib/Target/MBlaze/MBlazeCallingConv.td
index 4962573f96a..00a4219d047 100644
--- a/lib/Target/MBlaze/MBlazeCallingConv.td
+++ b/lib/Target/MBlaze/MBlazeCallingConv.td
@@ -9,10 +9,6 @@
 // This describes the calling conventions for MBlaze architecture.
 //===----------------------------------------------------------------------===//
 
-/// CCIfSubtarget - Match if the current subtarget has a feature F.
-class CCIfSubtarget:
-  CCIf().", F), A>;
-
 //===----------------------------------------------------------------------===//
 // MBlaze ABI Calling Convention
 //===----------------------------------------------------------------------===//
diff --git a/lib/Target/MBlaze/MBlazeTargetMachine.cpp b/lib/Target/MBlaze/MBlazeTargetMachine.cpp
index dd7de9bff36..62393d0920b 100644
--- a/lib/Target/MBlaze/MBlazeTargetMachine.cpp
+++ b/lib/Target/MBlaze/MBlazeTargetMachine.cpp
@@ -68,7 +68,7 @@ TargetPassConfig *MBlazeTargetMachine::createPassConfig(PassManagerBase &PM) {
 // Install an instruction selector pass using
 // the ISelDag to gen MBlaze code.
 bool MBlazePassConfig::addInstSelector() {
-  PM.add(createMBlazeISelDag(getMBlazeTargetMachine()));
+  PM->add(createMBlazeISelDag(getMBlazeTargetMachine()));
   return false;
 }
 
@@ -76,6 +76,6 @@ bool MBlazePassConfig::addInstSelector() {
 // machine code is emitted. return true if -print-machineinstrs should
 // print out the code after the passes.
 bool MBlazePassConfig::addPreEmitPass() {
-  PM.add(createMBlazeDelaySlotFillerPass(getMBlazeTargetMachine()));
+  PM->add(createMBlazeDelaySlotFillerPass(getMBlazeTargetMachine()));
   return true;
 }
diff --git a/lib/Target/MSP430/MSP430TargetMachine.cpp b/lib/Target/MSP430/MSP430TargetMachine.cpp
index 9f2eda13d7f..3acf96bb7d2 100644
--- a/lib/Target/MSP430/MSP430TargetMachine.cpp
+++ b/lib/Target/MSP430/MSP430TargetMachine.cpp
@@ -60,12 +60,12 @@ TargetPassConfig *MSP430TargetMachine::createPassConfig(PassManagerBase &PM) {
 
 bool MSP430PassConfig::addInstSelector() {
   // Install an instruction selector.
-  PM.add(createMSP430ISelDag(getMSP430TargetMachine(), getOptLevel()));
+  PM->add(createMSP430ISelDag(getMSP430TargetMachine(), getOptLevel()));
   return false;
 }
 
 bool MSP430PassConfig::addPreEmitPass() {
   // Must run branch selection immediately preceding the asm printer.
-  PM.add(createMSP430BranchSelectionPass());
+  PM->add(createMSP430BranchSelectionPass());
   return false;
 }
diff --git a/lib/Target/Mips/CMakeLists.txt b/lib/Target/Mips/CMakeLists.txt
index 13d17e4e529..0500c5dc383 100644
--- a/lib/Target/Mips/CMakeLists.txt
+++ b/lib/Target/Mips/CMakeLists.txt
@@ -2,12 +2,14 @@ set(LLVM_TARGET_DEFINITIONS Mips.td)
 
 tablegen(LLVM MipsGenRegisterInfo.inc -gen-register-info)
 tablegen(LLVM MipsGenInstrInfo.inc -gen-instr-info)
+tablegen(LLVM MipsGenDisassemblerTables.inc -gen-disassembler)
 tablegen(LLVM MipsGenCodeEmitter.inc -gen-emitter)
 tablegen(LLVM MipsGenMCCodeEmitter.inc -gen-emitter -mc-emitter)
 tablegen(LLVM MipsGenAsmWriter.inc -gen-asm-writer)
 tablegen(LLVM MipsGenDAGISel.inc -gen-dag-isel)
 tablegen(LLVM MipsGenCallingConv.inc -gen-callingconv)
 tablegen(LLVM MipsGenSubtargetInfo.inc -gen-subtarget)
+tablegen(LLVM MipsGenEDInfo.inc -gen-enhanced-disassembly-info)
 add_public_tablegen_target(MipsCommonTableGen)
 
 add_llvm_target(MipsCodeGen
@@ -32,6 +34,7 @@ add_llvm_target(MipsCodeGen
   )
 
 add_subdirectory(InstPrinter)
+add_subdirectory(Disassembler)
 add_subdirectory(TargetInfo)
 add_subdirectory(MCTargetDesc)
 add_subdirectory(AsmParser)
diff --git a/lib/Target/Mips/Disassembler/CMakeLists.txt b/lib/Target/Mips/Disassembler/CMakeLists.txt
new file mode 100644
index 00000000000..fe1dc75776f
--- /dev/null
+++ b/lib/Target/Mips/Disassembler/CMakeLists.txt
@@ -0,0 +1,15 @@
+include_directories( ${CMAKE_CURRENT_BINARY_DIR}/.. ${CMAKE_CURRENT_SOURCE_DIR}/.. )
+
+add_llvm_library(LLVMMipsDisassembler
+  MipsDisassembler.cpp
+  )
+
+# workaround for hanging compilation on MSVC9 and 10
+if( MSVC_VERSION EQUAL 1400 OR MSVC_VERSION EQUAL 1500 OR MSVC_VERSION EQUAL 1600 )
+set_property(
+  SOURCE MipsDisassembler.cpp
+  PROPERTY COMPILE_FLAGS "/Od"
+  )
+endif()
+
+add_dependencies(LLVMMipsDisassembler MipsCommonTableGen)
diff --git a/lib/Target/Mips/Disassembler/LLVMBuild.txt b/lib/Target/Mips/Disassembler/LLVMBuild.txt
new file mode 100644
index 00000000000..048ad0ddac5
--- /dev/null
+++ b/lib/Target/Mips/Disassembler/LLVMBuild.txt
@@ -0,0 +1,23 @@
+;===- ./lib/Target/Mips/Disassembler/LLVMBuild.txt --------------*- Conf -*--===;
+;
+;                     The LLVM Compiler Infrastructure
+;
+; This file is distributed under the University of Illinois Open Source
+; License. See LICENSE.TXT for details.
+;
+;===------------------------------------------------------------------------===;
+;
+; This is an LLVMBuild description file for the components in this subdirectory.
+;
+; For more information on the LLVMBuild system, please see:
+;
+;   http://llvm.org/docs/LLVMBuild.html
+;
+;===------------------------------------------------------------------------===;
+
+[component_0]
+type = Library
+name = MipsDisassembler
+parent = Mips
+required_libraries = MC Support MipsInfo
+add_to_library_groups = Mips
diff --git a/utils/json-bench/Makefile b/lib/Target/Mips/Disassembler/Makefile
similarity index 51%
rename from utils/json-bench/Makefile
rename to lib/Target/Mips/Disassembler/Makefile
index 6651626f683..a78feba1f8d 100644
--- a/utils/json-bench/Makefile
+++ b/lib/Target/Mips/Disassembler/Makefile
@@ -1,4 +1,4 @@
-##===- utils/FileCheck/Makefile ----------------------------*- Makefile -*-===##
+##===- lib/Target/Mips/Disassembler/Makefile ----------------*- Makefile -*-===##
 #
 #                     The LLVM Compiler Infrastructure
 #
@@ -7,15 +7,10 @@
 #
 ##===----------------------------------------------------------------------===##
 
-LEVEL = ../..
-TOOLNAME = json-bench
-USEDLIBS = LLVMSupport.a
+LEVEL = ../../../..
+LIBRARYNAME = LLVMMipsDisassembler
 
-# This tool has no plugins, optimize startup time.
-TOOL_NO_EXPORTS = 1
-
-# Don't install this utility
-NO_INSTALL = 1
+# Hack: we need to include 'main' Mips target directory to grab private headers
+CPP.Flags += -I$(PROJ_OBJ_DIR)/.. -I$(PROJ_SRC_DIR)/..
 
 include $(LEVEL)/Makefile.common
-
diff --git a/lib/Target/Mips/Disassembler/MipsDisassembler.cpp b/lib/Target/Mips/Disassembler/MipsDisassembler.cpp
new file mode 100644
index 00000000000..78dbc069490
--- /dev/null
+++ b/lib/Target/Mips/Disassembler/MipsDisassembler.cpp
@@ -0,0 +1,552 @@
+//===- MipsDisassembler.cpp - Disassembler for Mips -------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file is part of the Mips Disassembler.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Mips.h"
+#include "MipsSubtarget.h"
+#include "llvm/MC/EDInstInfo.h"
+#include "llvm/MC/MCDisassembler.h"
+#include "llvm/Support/MemoryObject.h"
+#include "llvm/Support/TargetRegistry.h"
+#include "llvm/MC/MCSubtargetInfo.h"
+#include "llvm/MC/MCInst.h"
+#include "llvm/MC/MCRegisterInfo.h"
+#include "llvm/Support/MathExtras.h"
+
+
+#include "MipsGenEDInfo.inc"
+
+using namespace llvm;
+
+typedef MCDisassembler::DecodeStatus DecodeStatus;
+
+/// MipsDisassembler - a disasembler class for Mips32.
+class MipsDisassembler : public MCDisassembler {
+public:
+  /// Constructor     - Initializes the disassembler.
+  ///
+  MipsDisassembler(const MCSubtargetInfo &STI, bool bigEndian) :
+    MCDisassembler(STI), isBigEndian(bigEndian) {
+  }
+
+  ~MipsDisassembler() {
+  }
+
+  /// getInstruction - See MCDisassembler.
+  DecodeStatus getInstruction(MCInst &instr,
+                              uint64_t &size,
+                              const MemoryObject ®ion,
+                              uint64_t address,
+                              raw_ostream &vStream,
+                              raw_ostream &cStream) const;
+
+  /// getEDInfo - See MCDisassembler.
+  const EDInstInfo *getEDInfo() const;
+
+private:
+  bool isBigEndian;
+};
+
+
+/// Mips64Disassembler - a disasembler class for Mips64.
+class Mips64Disassembler : public MCDisassembler {
+public:
+  /// Constructor     - Initializes the disassembler.
+  ///
+  Mips64Disassembler(const MCSubtargetInfo &STI, bool bigEndian) :
+    MCDisassembler(STI), isBigEndian(bigEndian) {
+  }
+
+  ~Mips64Disassembler() {
+  }
+
+  /// getInstruction - See MCDisassembler.
+  DecodeStatus getInstruction(MCInst &instr,
+                              uint64_t &size,
+                              const MemoryObject ®ion,
+                              uint64_t address,
+                              raw_ostream &vStream,
+                              raw_ostream &cStream) const;
+
+  /// getEDInfo - See MCDisassembler.
+  const EDInstInfo *getEDInfo() const;
+
+private:
+  bool isBigEndian;
+};
+
+const EDInstInfo *MipsDisassembler::getEDInfo() const {
+  return instInfoMips;
+}
+
+const EDInstInfo *Mips64Disassembler::getEDInfo() const {
+  return instInfoMips;
+}
+
+// Decoder tables for Mips register
+static const unsigned CPURegsTable[] = {
+  Mips::ZERO, Mips::AT, Mips::V0, Mips::V1,
+  Mips::A0, Mips::A1, Mips::A2, Mips::A3,
+  Mips::T0, Mips::T1, Mips::T2, Mips::T3,
+  Mips::T4, Mips::T5, Mips::T6, Mips::T7,
+  Mips::S0, Mips::S1, Mips::S2, Mips::S3,
+  Mips::S4, Mips::S5, Mips::S6, Mips::S7,
+  Mips::T8, Mips::T9, Mips::K0, Mips::K1,
+  Mips::GP, Mips::SP, Mips::FP, Mips::RA
+};
+
+static const unsigned FGR32RegsTable[] = {
+  Mips::F0, Mips::F1, Mips::F2, Mips::F3,
+  Mips::F4, Mips::F5, Mips::F6, Mips::F7,
+  Mips::F8, Mips::F9, Mips::F10, Mips::F11,
+  Mips::F12, Mips::F13, Mips::F14, Mips::F15,
+  Mips::F16, Mips::F17, Mips::F18, Mips::F18,
+  Mips::F20, Mips::F21, Mips::F22, Mips::F23,
+  Mips::F24, Mips::F25, Mips::F26, Mips::F27,
+  Mips::F28, Mips::F29, Mips::F30, Mips::F31
+};
+
+static const unsigned CPU64RegsTable[] = {
+  Mips::ZERO_64, Mips::AT_64, Mips::V0_64, Mips::V1_64,
+  Mips::A0_64, Mips::A1_64, Mips::A2_64, Mips::A3_64,
+  Mips::T0_64, Mips::T1_64, Mips::T2_64, Mips::T3_64,
+  Mips::T4_64, Mips::T5_64, Mips::T6_64, Mips::T7_64,
+  Mips::S0_64, Mips::S1_64, Mips::S2_64, Mips::S3_64,
+  Mips::S4_64, Mips::S5_64, Mips::S6_64, Mips::S7_64,
+  Mips::T8_64, Mips::T9_64, Mips::K0_64, Mips::K1_64,
+  Mips::GP_64, Mips::SP_64, Mips::FP_64, Mips::RA_64
+};
+
+static const unsigned FGR64RegsTable[] = {
+  Mips::D0_64,  Mips::D1_64,  Mips::D2_64,  Mips::D3_64,
+  Mips::D4_64,  Mips::D5_64,  Mips::D6_64,  Mips::D7_64,
+  Mips::D8_64,  Mips::D9_64,  Mips::D10_64, Mips::D11_64,
+  Mips::D12_64, Mips::D13_64, Mips::D14_64, Mips::D15_64,
+  Mips::D16_64, Mips::D17_64, Mips::D18_64, Mips::D19_64,
+  Mips::D20_64, Mips::D21_64, Mips::D22_64, Mips::D23_64,
+  Mips::D24_64, Mips::D25_64, Mips::D26_64, Mips::D27_64,
+  Mips::D28_64, Mips::D29_64, Mips::D30_64, Mips::D31_64
+};
+
+static const unsigned AFGR64RegsTable[] = {
+  Mips::D0,  Mips::D1,  Mips::D2,  Mips::D3,
+  Mips::D4,  Mips::D5,  Mips::D6,  Mips::D7,
+  Mips::D8,  Mips::D9,  Mips::D10, Mips::D11,
+  Mips::D12, Mips::D13, Mips::D14, Mips::D15
+};
+
+// Forward declare these because the autogenerated code will reference them.
+// Definitions are further down.
+static DecodeStatus DecodeCPU64RegsRegisterClass(MCInst &Inst,
+                                                 unsigned RegNo,
+                                                 uint64_t Address,
+                                                 const void *Decoder);
+
+static DecodeStatus DecodeCPURegsRegisterClass(MCInst &Inst,
+                                               unsigned RegNo,
+                                               uint64_t Address,
+                                               const void *Decoder);
+
+static DecodeStatus DecodeFGR64RegisterClass(MCInst &Inst,
+                                             unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder);
+
+static DecodeStatus DecodeFGR32RegisterClass(MCInst &Inst,
+                                             unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder);
+
+static DecodeStatus DecodeCCRRegisterClass(MCInst &Inst,
+                                           unsigned RegNo,
+                                           uint64_t Address,
+                                           const void *Decoder);
+
+static DecodeStatus DecodeHWRegsRegisterClass(MCInst &Inst,
+                                              unsigned Insn,
+                                              uint64_t Address,
+                                              const void *Decoder);
+
+static DecodeStatus DecodeAFGR64RegisterClass(MCInst &Inst,
+                                              unsigned RegNo,
+                                              uint64_t Address,
+                                              const void *Decoder);
+
+static DecodeStatus DecodeHWRegs64RegisterClass(MCInst &Inst,
+                                                unsigned Insn,
+                                                uint64_t Address,
+                                                const void *Decoder);
+
+static DecodeStatus DecodeBranchTarget(MCInst &Inst,
+                                       unsigned Offset,
+                                       uint64_t Address,
+                                       const void *Decoder);
+
+static DecodeStatus DecodeBC1(MCInst &Inst,
+                              unsigned Insn,
+                              uint64_t Address,
+                              const void *Decoder);
+
+
+static DecodeStatus DecodeJumpTarget(MCInst &Inst,
+                                     unsigned Insn,
+                                     uint64_t Address,
+                                     const void *Decoder);
+
+static DecodeStatus DecodeMem(MCInst &Inst,
+                              unsigned Insn,
+                              uint64_t Address,
+                              const void *Decoder);
+
+static DecodeStatus DecodeFMem(MCInst &Inst, unsigned Insn,
+                               uint64_t Address,
+                               const void *Decoder);
+
+static DecodeStatus DecodeSimm16(MCInst &Inst,
+                                 unsigned Insn,
+                                 uint64_t Address,
+                                 const void *Decoder);
+
+static DecodeStatus DecodeCondCode(MCInst &Inst,
+                                   unsigned Insn,
+                                   uint64_t Address,
+                                   const void *Decoder);
+
+static DecodeStatus DecodeInsSize(MCInst &Inst,
+                                  unsigned Insn,
+                                  uint64_t Address,
+                                  const void *Decoder);
+
+static DecodeStatus DecodeExtSize(MCInst &Inst,
+                                  unsigned Insn,
+                                  uint64_t Address,
+                                  const void *Decoder);
+
+namespace llvm {
+extern Target TheMipselTarget, TheMipsTarget, TheMips64Target,
+              TheMips64elTarget;
+}
+
+static MCDisassembler *createMipsDisassembler(
+                       const Target &T,
+                       const MCSubtargetInfo &STI) {
+  return new MipsDisassembler(STI,true);
+}
+
+static MCDisassembler *createMipselDisassembler(
+                       const Target &T,
+                       const MCSubtargetInfo &STI) {
+  return new MipsDisassembler(STI,false);
+}
+
+static MCDisassembler *createMips64Disassembler(
+                       const Target &T,
+                       const MCSubtargetInfo &STI) {
+  return new Mips64Disassembler(STI,true);
+}
+
+static MCDisassembler *createMips64elDisassembler(
+                       const Target &T,
+                       const MCSubtargetInfo &STI) {
+  return new Mips64Disassembler(STI, false);
+}
+
+extern "C" void LLVMInitializeMipsDisassembler() {
+  // Register the disassembler.
+  TargetRegistry::RegisterMCDisassembler(TheMipsTarget,
+                                         createMipsDisassembler);
+  TargetRegistry::RegisterMCDisassembler(TheMipselTarget,
+                                         createMipselDisassembler);
+  TargetRegistry::RegisterMCDisassembler(TheMips64Target,
+                                         createMips64Disassembler);
+  TargetRegistry::RegisterMCDisassembler(TheMips64elTarget,
+                                         createMips64elDisassembler);
+}
+
+
+#include "MipsGenDisassemblerTables.inc"
+
+  /// readInstruction - read four bytes from the MemoryObject
+  /// and return 32 bit word sorted according to the given endianess
+static DecodeStatus readInstruction32(const MemoryObject ®ion,
+                                      uint64_t address,
+                                      uint64_t &size,
+                                      uint32_t &insn,
+                                      bool isBigEndian) {
+  uint8_t Bytes[4];
+
+  // We want to read exactly 4 Bytes of data.
+  if (region.readBytes(address, 4, (uint8_t*)Bytes, NULL) == -1) {
+    size = 0;
+    return MCDisassembler::Fail;
+  }
+
+  if (isBigEndian) {
+    // Encoded as a big-endian 32-bit word in the stream.
+    insn = (Bytes[3] <<  0) |
+           (Bytes[2] <<  8) |
+           (Bytes[1] << 16) |
+           (Bytes[0] << 24);
+  }
+  else {
+    // Encoded as a small-endian 32-bit word in the stream.
+    insn = (Bytes[0] <<  0) |
+           (Bytes[1] <<  8) |
+           (Bytes[2] << 16) |
+           (Bytes[3] << 24);
+  }
+
+  return MCDisassembler::Success;
+}
+
+DecodeStatus
+MipsDisassembler::getInstruction(MCInst &instr,
+                                 uint64_t &Size,
+                                 const MemoryObject &Region,
+                                 uint64_t Address,
+                                 raw_ostream &vStream,
+                                 raw_ostream &cStream) const {
+  uint32_t Insn;
+
+  DecodeStatus Result = readInstruction32(Region, Address, Size,
+                                          Insn, isBigEndian);
+  if (Result == MCDisassembler::Fail)
+    return MCDisassembler::Fail;
+
+  // Calling the auto-generated decoder function.
+  Result = decodeMipsInstruction32(instr, Insn, Address, this, STI);
+  if (Result != MCDisassembler::Fail) {
+    Size = 4;
+    return Result;
+  }
+
+  return MCDisassembler::Fail;
+}
+
+DecodeStatus
+Mips64Disassembler::getInstruction(MCInst &instr,
+                                   uint64_t &Size,
+                                   const MemoryObject &Region,
+                                   uint64_t Address,
+                                   raw_ostream &vStream,
+                                   raw_ostream &cStream) const {
+  uint32_t Insn;
+
+  DecodeStatus Result = readInstruction32(Region, Address, Size,
+                                          Insn, isBigEndian);
+  if (Result == MCDisassembler::Fail)
+    return MCDisassembler::Fail;
+
+  // Calling the auto-generated decoder function.
+  Result = decodeMips64Instruction32(instr, Insn, Address, this, STI);
+  if (Result != MCDisassembler::Fail) {
+    Size = 4;
+    return Result;
+  }
+  // If we fail to decode in Mips64 decoder space we can try in Mips32
+  Result = decodeMipsInstruction32(instr, Insn, Address, this, STI);
+  if (Result != MCDisassembler::Fail) {
+    Size = 4;
+    return Result;
+  }
+
+  return MCDisassembler::Fail;
+}
+
+static DecodeStatus DecodeCPU64RegsRegisterClass(MCInst &Inst,
+                                                 unsigned RegNo,
+                                                 uint64_t Address,
+                                                 const void *Decoder) {
+
+  if (RegNo > 31)
+    return MCDisassembler::Fail;
+
+  Inst.addOperand(MCOperand::CreateReg(CPU64RegsTable[RegNo]));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeCPURegsRegisterClass(MCInst &Inst,
+                                               unsigned RegNo,
+                                               uint64_t Address,
+                                               const void *Decoder) {
+  if (RegNo > 31)
+    return MCDisassembler::Fail;
+
+  Inst.addOperand(MCOperand::CreateReg(CPURegsTable[RegNo]));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeFGR64RegisterClass(MCInst &Inst,
+                                             unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder) {
+  if (RegNo > 31)
+    return MCDisassembler::Fail;
+
+  Inst.addOperand(MCOperand::CreateReg(FGR64RegsTable[RegNo]));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeFGR32RegisterClass(MCInst &Inst,
+                                             unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder) {
+  if (RegNo > 31)
+    return MCDisassembler::Fail;
+
+  Inst.addOperand(MCOperand::CreateReg(FGR32RegsTable[RegNo]));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeCCRRegisterClass(MCInst &Inst,
+                                           unsigned RegNo,
+                                           uint64_t Address,
+                                           const void *Decoder) {
+  Inst.addOperand(MCOperand::CreateReg(RegNo));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeMem(MCInst &Inst,
+                              unsigned Insn,
+                              uint64_t Address,
+                              const void *Decoder) {
+  int Offset = SignExtend32<16>(Insn & 0xffff);
+  int Reg = (int)fieldFromInstruction32(Insn, 16, 5);
+  int Base = (int)fieldFromInstruction32(Insn, 21, 5);
+
+  if(Inst.getOpcode() == Mips::SC){
+    Inst.addOperand(MCOperand::CreateReg(CPURegsTable[Reg]));
+  }
+
+  Inst.addOperand(MCOperand::CreateReg(CPURegsTable[Reg]));
+  Inst.addOperand(MCOperand::CreateReg(CPURegsTable[Base]));
+  Inst.addOperand(MCOperand::CreateImm(Offset));
+
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeFMem(MCInst &Inst,
+                               unsigned Insn,
+                               uint64_t Address,
+                               const void *Decoder) {
+  int Offset = SignExtend32<16>(Insn & 0xffff);
+  int Reg = (int)fieldFromInstruction32(Insn, 16, 5);
+  int Base = (int)fieldFromInstruction32(Insn, 21, 5);
+
+  Inst.addOperand(MCOperand::CreateReg(FGR64RegsTable[Reg]));
+  Inst.addOperand(MCOperand::CreateReg(CPURegsTable[Base]));
+  Inst.addOperand(MCOperand::CreateImm(Offset));
+
+  return MCDisassembler::Success;
+}
+
+
+static DecodeStatus DecodeHWRegsRegisterClass(MCInst &Inst,
+                                              unsigned RegNo,
+                                              uint64_t Address,
+                                              const void *Decoder) {
+  // Currently only hardware register 29 is supported.
+  if (RegNo != 29)
+    return  MCDisassembler::Fail;
+  Inst.addOperand(MCOperand::CreateReg(Mips::HWR29));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeCondCode(MCInst &Inst,
+                                   unsigned Insn,
+                                   uint64_t Address,
+                                   const void *Decoder) {
+  int CondCode = Insn & 0xf;
+  Inst.addOperand(MCOperand::CreateImm(CondCode));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeAFGR64RegisterClass(MCInst &Inst,
+                                              unsigned RegNo,
+                                              uint64_t Address,
+                                              const void *Decoder) {
+  if (RegNo > 31)
+    return MCDisassembler::Fail;
+
+  Inst.addOperand(MCOperand::CreateReg(AFGR64RegsTable[RegNo]));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeHWRegs64RegisterClass(MCInst &Inst,
+                                                unsigned RegNo,
+                                                uint64_t Address,
+                                                const void *Decoder) {
+  //Currently only hardware register 29 is supported
+  if (RegNo != 29)
+    return  MCDisassembler::Fail;
+  Inst.addOperand(MCOperand::CreateReg(Mips::HWR29));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeBranchTarget(MCInst &Inst,
+                                       unsigned Offset,
+                                       uint64_t Address,
+                                       const void *Decoder) {
+  unsigned BranchOffset = Offset & 0xffff;
+  BranchOffset = SignExtend32<18>(BranchOffset << 2) + 4;
+  Inst.addOperand(MCOperand::CreateImm(BranchOffset));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeBC1(MCInst &Inst,
+                              unsigned Insn,
+                              uint64_t Address,
+                              const void *Decoder) {
+  unsigned BranchOffset = Insn & 0xffff;
+  BranchOffset = SignExtend32<18>(BranchOffset << 2) + 4;
+  Inst.addOperand(MCOperand::CreateImm(BranchOffset));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeJumpTarget(MCInst &Inst,
+                                     unsigned Insn,
+                                     uint64_t Address,
+                                     const void *Decoder) {
+
+  unsigned JumpOffset = fieldFromInstruction32(Insn, 0, 26) << 2;
+  Inst.addOperand(MCOperand::CreateImm(JumpOffset));
+  return MCDisassembler::Success;
+}
+
+
+static DecodeStatus DecodeSimm16(MCInst &Inst,
+                                 unsigned Insn,
+                                 uint64_t Address,
+                                 const void *Decoder) {
+  Inst.addOperand(MCOperand::CreateImm(SignExtend32<16>(Insn)));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeInsSize(MCInst &Inst,
+                                  unsigned Insn,
+                                  uint64_t Address,
+                                  const void *Decoder) {
+  // First we need to grab the pos(lsb) from MCInst.
+  int Pos = Inst.getOperand(2).getImm();
+  int Size = (int) Insn - Pos + 1;
+  Inst.addOperand(MCOperand::CreateImm(SignExtend32<16>(Size)));
+  return MCDisassembler::Success;
+}
+
+static DecodeStatus DecodeExtSize(MCInst &Inst,
+                                  unsigned Insn,
+                                  uint64_t Address,
+                                  const void *Decoder) {
+  int Size = (int) Insn  + 1;
+  Inst.addOperand(MCOperand::CreateImm(SignExtend32<16>(Size)));
+  return MCDisassembler::Success;
+}
diff --git a/lib/Target/Mips/LLVMBuild.txt b/lib/Target/Mips/LLVMBuild.txt
index abbed8c90fc..a95d6bc1352 100644
--- a/lib/Target/Mips/LLVMBuild.txt
+++ b/lib/Target/Mips/LLVMBuild.txt
@@ -16,7 +16,7 @@
 ;===------------------------------------------------------------------------===;
 
 [common]
-subdirectories = AsmParser InstPrinter MCTargetDesc TargetInfo
+subdirectories = AsmParser Disassembler InstPrinter MCTargetDesc TargetInfo
 
 [component_0]
 type = TargetGroup
@@ -24,6 +24,7 @@ name = Mips
 parent = Target
 has_asmparser = 1
 has_asmprinter = 1
+has_disassembler = 1
 has_jit = 1
 
 [component_1]
diff --git a/lib/Target/Mips/MCTargetDesc/MipsAsmBackend.cpp b/lib/Target/Mips/MCTargetDesc/MipsAsmBackend.cpp
index e79be336362..9b4caf65cbf 100644
--- a/lib/Target/Mips/MCTargetDesc/MipsAsmBackend.cpp
+++ b/lib/Target/Mips/MCTargetDesc/MipsAsmBackend.cpp
@@ -12,7 +12,6 @@
 //===----------------------------------------------------------------------===//
 //
 
-#include "MipsBaseInfo.h"
 #include "MipsFixupKinds.h"
 #include "MCTargetDesc/MipsMCTargetDesc.h"
 #include "llvm/MC/MCAsmBackend.h"
@@ -85,9 +84,8 @@ public:
                   uint64_t Value) const {
     MCFixupKind Kind = Fixup.getKind();
     Value = adjustFixupValue((unsigned)Kind, Value);
-    int64_t SymOffset = MipsGetSymAndOffset(Fixup).second;
 
-    if (!Value && !SymOffset)
+    if (!Value)
       return; // Doesn't change encoding.
 
     // Where do we start in the object
@@ -118,7 +116,7 @@ public:
     }
 
     uint64_t Mask = ((uint64_t)(-1) >> (64 - getFixupKindInfo(Kind).TargetSize));
-    CurVal |= (Value + SymOffset) & Mask;
+    CurVal |= Value & Mask;
 
     // Write out the fixed up bytes back to the code/data bits.
     for (unsigned i = 0; i != NumBytes; ++i) {
diff --git a/lib/Target/Mips/MCTargetDesc/MipsMCCodeEmitter.cpp b/lib/Target/Mips/MCTargetDesc/MipsMCCodeEmitter.cpp
index 27954b174ed..4ed2be0f430 100644
--- a/lib/Target/Mips/MCTargetDesc/MipsMCCodeEmitter.cpp
+++ b/lib/Target/Mips/MCTargetDesc/MipsMCCodeEmitter.cpp
@@ -194,7 +194,7 @@ getMachineOpValue(const MCInst &MI, const MCOperand &MO,
 
   assert (Kind == MCExpr::SymbolRef);
     
-  Mips::Fixups FixupKind;
+  Mips::Fixups FixupKind = Mips::Fixups(0);
 
   switch(cast(Expr)->getKind()) {
   case MCSymbolRefExpr::VK_Mips_GPREL:
diff --git a/lib/Target/Mips/MCTargetDesc/MipsMCTargetDesc.cpp b/lib/Target/Mips/MCTargetDesc/MipsMCTargetDesc.cpp
index 3c544f6aec9..f634f082be5 100644
--- a/lib/Target/Mips/MCTargetDesc/MipsMCTargetDesc.cpp
+++ b/lib/Target/Mips/MCTargetDesc/MipsMCTargetDesc.cpp
@@ -34,6 +34,38 @@
 
 using namespace llvm;
 
+static std::string ParseMipsTriple(StringRef TT, StringRef CPU) {
+  std::string MipsArchFeature;
+  size_t DashPosition = 0;
+  StringRef TheTriple;
+
+  // Let's see if there is a dash, like mips-unknown-linux.
+  DashPosition = TT.find('-');
+
+  if (DashPosition == StringRef::npos) {
+    // No dash, we check the string size.
+    TheTriple = TT.substr(0);
+  } else {
+    // We are only interested in substring before dash.
+    TheTriple = TT.substr(0,DashPosition);
+  }
+
+  if (TheTriple == "mips" || TheTriple == "mipsel") {
+    if (CPU.empty() || CPU == "mips32") {
+      MipsArchFeature = "+mips32";
+    } else if (CPU == "mips32r2") {
+      MipsArchFeature = "+mips32r2";
+    }
+  } else {
+      if (CPU.empty() || CPU == "mips64") {
+        MipsArchFeature = "+mips64";
+      } else if (CPU == "mips64r2") {
+        MipsArchFeature = "+mips64r2";
+      }
+  }
+  return MipsArchFeature;
+}
+
 static MCInstrInfo *createMipsMCInstrInfo() {
   MCInstrInfo *X = new MCInstrInfo();
   InitMipsMCInstrInfo(X);
@@ -48,8 +80,15 @@ static MCRegisterInfo *createMipsMCRegisterInfo(StringRef TT) {
 
 static MCSubtargetInfo *createMipsMCSubtargetInfo(StringRef TT, StringRef CPU,
                                                   StringRef FS) {
+  std::string ArchFS = ParseMipsTriple(TT,CPU);
+  if (!FS.empty()) {
+    if (!ArchFS.empty())
+      ArchFS = ArchFS + "," + FS.str();
+    else
+      ArchFS = FS;
+  }
   MCSubtargetInfo *X = new MCSubtargetInfo();
-  InitMipsMCSubtargetInfo(X, TT, CPU, FS);
+  InitMipsMCSubtargetInfo(X, TT, CPU, ArchFS);
   return X;
 }
 
diff --git a/lib/Target/Mips/Makefile b/lib/Target/Mips/Makefile
index 168635c96be..596f07145a2 100644
--- a/lib/Target/Mips/Makefile
+++ b/lib/Target/Mips/Makefile
@@ -15,9 +15,9 @@ TARGET = Mips
 BUILT_SOURCES = MipsGenRegisterInfo.inc MipsGenInstrInfo.inc \
                 MipsGenAsmWriter.inc MipsGenCodeEmitter.inc \
                 MipsGenDAGISel.inc MipsGenCallingConv.inc \
-                MipsGenSubtargetInfo.inc MipsGenMCCodeEmitter.inc
-
-DIRS = InstPrinter AsmParser TargetInfo MCTargetDesc
+                MipsGenSubtargetInfo.inc MipsGenMCCodeEmitter.inc \
+                MipsGenEDInfo.inc MipsGenDisassemblerTables.inc
+DIRS = InstPrinter Disassembler AsmParser TargetInfo MCTargetDesc
 
 include $(LEVEL)/Makefile.common
 
diff --git a/lib/Target/Mips/Mips64InstrInfo.td b/lib/Target/Mips/Mips64InstrInfo.td
index 427e8d97ad9..0382869255f 100644
--- a/lib/Target/Mips/Mips64InstrInfo.td
+++ b/lib/Target/Mips/Mips64InstrInfo.td
@@ -36,6 +36,7 @@ def immZExt6 : ImmLeaf;
 //===----------------------------------------------------------------------===//
 // Shifts
 // 64-bit shift instructions.
+let DecoderNamespace = "Mips64" in {
 class shift_rotate_imm64 func, bits<5> isRotate, string instr_asm,
                          SDNode OpNode>:
   shift_rotate_imm func, string instr_asm, InstrItinClass itin>:
 
 multiclass Atomic2Ops64 {
   def #NAME# : Atomic2Ops, Requires<[NotN64]>;
-  def _P8    : Atomic2Ops, Requires<[IsN64]>;
+  def _P8    : Atomic2Ops, Requires<[IsN64]> {
+    let isCodeGenOnly = 1;
+  }
 }
 
 multiclass AtomicCmpSwap64  {
   def #NAME# : AtomicCmpSwap, Requires<[NotN64]>;
   def _P8    : AtomicCmpSwap,
-               Requires<[IsN64]>;
+               Requires<[IsN64]> {
+    let isCodeGenOnly = 1;
+  }
 }
-
-let usesCustomInserter = 1, Predicates = [HasMips64] in {
+}
+let usesCustomInserter = 1, Predicates = [HasMips64],
+  DecoderNamespace = "Mips64" in {
   defm ATOMIC_LOAD_ADD_I64  : Atomic2Ops64;
   defm ATOMIC_LOAD_SUB_I64  : Atomic2Ops64;
   defm ATOMIC_LOAD_AND_I64  : Atomic2Ops64;
@@ -72,7 +78,7 @@ let usesCustomInserter = 1, Predicates = [HasMips64] in {
 //===----------------------------------------------------------------------===//
 // Instruction definition
 //===----------------------------------------------------------------------===//
-
+let DecoderNamespace = "Mips64" in {
 /// Arithmetic Instructions (ALU Immediate)
 def DADDiu   : ArithLogicI<0x19, "daddiu", add, simm16_64, immSExt16,
                            CPU64Regs>;
@@ -97,16 +103,17 @@ def NOR64    : LogicNOR<0x00, 0x27, "nor", CPU64Regs>;
 def DSLL     : shift_rotate_imm64<0x38, 0x00, "dsll", shl>;
 def DSRL     : shift_rotate_imm64<0x3a, 0x00, "dsrl", srl>;
 def DSRA     : shift_rotate_imm64<0x3b, 0x00, "dsra", sra>;
-def DSLLV    : shift_rotate_reg<0x24, 0x00, "dsllv", shl, CPU64Regs>;
-def DSRLV    : shift_rotate_reg<0x26, 0x00, "dsrlv", srl, CPU64Regs>;
-def DSRAV    : shift_rotate_reg<0x27, 0x00, "dsrav", sra, CPU64Regs>;
-
+def DSLLV    : shift_rotate_reg<0x14, 0x00, "dsllv", shl, CPU64Regs>;
+def DSRLV    : shift_rotate_reg<0x16, 0x00, "dsrlv", srl, CPU64Regs>;
+def DSRAV    : shift_rotate_reg<0x17, 0x00, "dsrav", sra, CPU64Regs>;
+}
 // Rotate Instructions
-let Predicates = [HasMips64r2] in {
+let Predicates = [HasMips64r2], DecoderNamespace = "Mips64" in {
   def DROTR    : shift_rotate_imm64<0x3a, 0x01, "drotr", rotr>;
   def DROTRV   : shift_rotate_reg<0x16, 0x01, "drotrv", rotr, CPU64Regs>;
 }
 
+let DecoderNamespace = "Mips64" in {
 /// Load and Store Instructions
 ///  aligned
 defm LB64    : LoadM64<0x20, "lb",  sextloadi8>;
@@ -132,9 +139,13 @@ defm USD       : StoreM64<0x3f, "usd", store_u, 1>;
 
 /// Load-linked, Store-conditional
 def LLD    : LLBase<0x34, "lld", CPU64Regs, mem>, Requires<[NotN64]>;
-def LLD_P8 : LLBase<0x34, "lld", CPU64Regs, mem64>, Requires<[IsN64]>;
+def LLD_P8 : LLBase<0x34, "lld", CPU64Regs, mem64>, Requires<[IsN64]> {
+  let isCodeGenOnly = 1;
+}
 def SCD    : SCBase<0x3c, "scd", CPU64Regs, mem>, Requires<[NotN64]>;
-def SCD_P8 : SCBase<0x3c, "scd", CPU64Regs, mem64>, Requires<[IsN64]>;
+def SCD_P8 : SCBase<0x3c, "scd", CPU64Regs, mem64>, Requires<[IsN64]> {
+  let isCodeGenOnly = 1;
+}
 
 /// Jump and Branch Instructions
 def JR64   : JumpFR<0x00, 0x08, "jr", CPU64Regs>;
@@ -142,11 +153,13 @@ def BEQ64  : CBranch<0x04, "beq", seteq, CPU64Regs>;
 def BNE64  : CBranch<0x05, "bne", setne, CPU64Regs>;
 def BGEZ64 : CBranchZero<0x01, 1, "bgez", setge, CPU64Regs>;
 def BGTZ64 : CBranchZero<0x07, 0, "bgtz", setgt, CPU64Regs>;
-def BLEZ64 : CBranchZero<0x07, 0, "blez", setle, CPU64Regs>;
+def BLEZ64 : CBranchZero<0x06, 0, "blez", setle, CPU64Regs>;
 def BLTZ64 : CBranchZero<0x01, 0, "bltz", setlt, CPU64Regs>;
-
+}
+let DecoderNamespace = "Mips64" in
 def JALR64 : JumpLinkReg<0x00, 0x09, "jalr", CPU64Regs>;
 
+let DecoderNamespace = "Mips64" in {
 /// Multiply and Divide Instructions.
 def DMULT    : Mult64<0x1c, "dmult", IIImul>;
 def DMULTu   : Mult64<0x1d, "dmultu", IIImul>;
@@ -171,11 +184,13 @@ def DSBH : SubwordSwap<0x24, 0x2, "dsbh", CPU64Regs>;
 def DSHD : SubwordSwap<0x24, 0x5, "dshd", CPU64Regs>;
 
 def LEA_ADDiu64 : EffectiveAddress<"daddiu\t$rt, $addr", CPU64Regs, mem_ea_64>;
-
-let Uses = [SP_64] in
+}
+let Uses = [SP_64], DecoderNamespace = "Mips64" in
 def DynAlloc64 : EffectiveAddress<"daddiu\t$rt, $addr", CPU64Regs, mem_ea_64>,
-                 Requires<[IsN64]>;
-
+                 Requires<[IsN64]> {
+  let isCodeGenOnly = 1;
+}
+let DecoderNamespace = "Mips64" in {
 def RDHWR64 : ReadHardware;
 
 def DEXT : ExtBase<3, "dext", CPU64Regs>;
@@ -183,12 +198,12 @@ def DINS : InsBase<7, "dins", CPU64Regs>;
 
 def DSLL64_32 : FR<0x3c, 0x00, (outs CPU64Regs:$rd), (ins CPURegs:$rt),
                    "dsll\t$rd, $rt, 32", [], IIAlu>;
-
 def SLL64_32 : FR<0x0, 0x00, (outs CPU64Regs:$rd), (ins CPURegs:$rt),
                   "sll\t$rd, $rt, 0", [], IIAlu>;
+let isCodeGenOnly = 1 in
 def SLL64_64 : FR<0x0, 0x00, (outs CPU64Regs:$rd), (ins CPU64Regs:$rt),
                   "sll\t$rd, $rt, 0", [], IIAlu>;
-
+}
 //===----------------------------------------------------------------------===//
 //  Arbitrary patterns that map to one or more instructions
 //===----------------------------------------------------------------------===//
diff --git a/lib/Target/Mips/MipsCondMov.td b/lib/Target/Mips/MipsCondMov.td
index 075a3e807b1..da336804e51 100644
--- a/lib/Target/Mips/MipsCondMov.td
+++ b/lib/Target/Mips/MipsCondMov.td
@@ -95,45 +95,65 @@ multiclass MovnPats;
-let Predicates = [HasMips64] in {
+let Predicates = [HasMips64],DecoderNamespace = "Mips64" in {
   def MOVZ_I_I64   : CondMovIntInt;
-  def MOVZ_I64_I   : CondMovIntInt;
-  def MOVZ_I64_I64 : CondMovIntInt;
+  def MOVZ_I64_I   : CondMovIntInt {
+    let isCodeGenOnly = 1;
+  }
+  def MOVZ_I64_I64 : CondMovIntInt {
+    let isCodeGenOnly = 1;
+  }
 }
 
 def MOVN_I_I     : CondMovIntInt;
-let Predicates = [HasMips64] in {
+let Predicates = [HasMips64],DecoderNamespace = "Mips64" in {
   def MOVN_I_I64   : CondMovIntInt;
-  def MOVN_I64_I   : CondMovIntInt;
-  def MOVN_I64_I64 : CondMovIntInt;
+  def MOVN_I64_I   : CondMovIntInt {
+    let isCodeGenOnly = 1;
+  }
+  def MOVN_I64_I64 : CondMovIntInt {
+    let isCodeGenOnly = 1;
+  }
 }
 
 def MOVZ_I_S   : CondMovIntFP;
 def MOVZ_I64_S : CondMovIntFP,
-                 Requires<[HasMips64]>;
+                 Requires<[HasMips64]> {
+  let DecoderNamespace = "Mips64";
+}
 
 def MOVN_I_S   : CondMovIntFP;
 def MOVN_I64_S : CondMovIntFP,
-                 Requires<[HasMips64]>;
+                 Requires<[HasMips64]> {
+  let DecoderNamespace = "Mips64";
+}
 
 let Predicates = [NotFP64bit] in {
   def MOVZ_I_D32   : CondMovIntFP;
   def MOVN_I_D32   : CondMovIntFP;
 }
-let Predicates = [IsFP64bit] in {
+let Predicates = [IsFP64bit],DecoderNamespace = "Mips64" in {
   def MOVZ_I_D64   : CondMovIntFP;
-  def MOVZ_I64_D64 : CondMovIntFP;
+  def MOVZ_I64_D64 : CondMovIntFP {
+    let isCodeGenOnly = 1;
+  }
   def MOVN_I_D64   : CondMovIntFP;
-  def MOVN_I64_D64 : CondMovIntFP;
+  def MOVN_I64_D64 : CondMovIntFP {
+    let isCodeGenOnly = 1;
+  }
 }
 
 def MOVT_I   : CondMovFPInt;
 def MOVT_I64 : CondMovFPInt,
-               Requires<[HasMips64]>;
+               Requires<[HasMips64]> {
+  let DecoderNamespace = "Mips64";
+}
 
 def MOVF_I   : CondMovFPInt;
 def MOVF_I64 : CondMovFPInt,
-               Requires<[HasMips64]>;
+               Requires<[HasMips64]> {
+  let DecoderNamespace = "Mips64";
+}
 
 def MOVT_S : CondMovFPFP;
 def MOVF_S : CondMovFPFP;
@@ -142,7 +162,7 @@ let Predicates = [NotFP64bit] in {
   def MOVT_D32 : CondMovFPFP;
   def MOVF_D32 : CondMovFPFP;
 }
-let Predicates = [IsFP64bit] in {
+let Predicates = [IsFP64bit], DecoderNamespace = "Mips64" in {
   def MOVT_D64 : CondMovFPFP;
   def MOVF_D64 : CondMovFPFP;
 }
diff --git a/lib/Target/Mips/MipsISelLowering.cpp b/lib/Target/Mips/MipsISelLowering.cpp
index 6a23bc3d1d7..ace47ab0791 100644
--- a/lib/Target/Mips/MipsISelLowering.cpp
+++ b/lib/Target/Mips/MipsISelLowering.cpp
@@ -34,6 +34,8 @@
 #include "llvm/CodeGen/ValueTypes.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/raw_ostream.h"
+
 using namespace llvm;
 
 // If I is a shifted mask, set the size (Size) and the first bit of the
diff --git a/lib/Target/Mips/MipsInstrFPU.td b/lib/Target/Mips/MipsInstrFPU.td
index b6559452fec..14d8f1e7ba9 100644
--- a/lib/Target/Mips/MipsInstrFPU.td
+++ b/lib/Target/Mips/MipsInstrFPU.td
@@ -47,17 +47,17 @@ def MipsExtractElementF64 : SDNode<"MipsISD::ExtractElementF64",
                                    SDT_MipsExtractElementF64>;
 
 // Operand for printing out a condition code.
-let PrintMethod = "printFCCOperand" in
+let PrintMethod = "printFCCOperand", DecoderMethod = "DecodeCondCode" in
   def condcode : Operand;
 
 //===----------------------------------------------------------------------===//
 // Feature predicates.
 //===----------------------------------------------------------------------===//
 
-def IsFP64bit        : Predicate<"Subtarget.isFP64bit()">;
-def NotFP64bit       : Predicate<"!Subtarget.isFP64bit()">;
-def IsSingleFloat    : Predicate<"Subtarget.isSingleFloat()">;
-def IsNotSingleFloat : Predicate<"!Subtarget.isSingleFloat()">;
+def IsFP64bit        : Predicate<"Subtarget.isFP64bit()">, AssemblerPredicate<"FeatureFP64Bit">;
+def NotFP64bit       : Predicate<"!Subtarget.isFP64bit()">, AssemblerPredicate<"!FeatureFP64Bit">;
+def IsSingleFloat    : Predicate<"Subtarget.isSingleFloat()">, AssemblerPredicate<"FeatureSingleFloat">;
+def IsNotSingleFloat : Predicate<"!Subtarget.isSingleFloat()">, AssemblerPredicate<"!FeatureSingleFloat">;
 
 // FP immediate patterns.
 def fpimm0 : PatLeaf<(fpimm), [{
@@ -83,6 +83,7 @@ def fpimm0neg : PatLeaf<(fpimm), [{
 //===----------------------------------------------------------------------===//
 
 // FP load.
+let DecoderMethod = "DecodeFMem" in {
 class FPLoad op, string opstr, RegisterClass RC, Operand MemOpnd>:
   FMem op, string opstr, RegisterClass RC, Operand MemOpnd>:
   FMem;
-
+}
 // FP indexed load.
 class FPIdxLoad funct, string opstr, RegisterClass DRC,
                 RegisterClass PRC, PatFrag FOp>:
@@ -118,11 +119,13 @@ multiclass FFR1_W_M funct, string opstr> {
   def _D32 : FFR1,
              Requires<[NotFP64bit]>;
   def _D64 : FFR1,
-             Requires<[IsFP64bit]>;
+             Requires<[IsFP64bit]> {
+    let DecoderNamespace = "Mips64";
+  }
 }
 
 // Instructions that convert an FP value to 64-bit fixed point.
-let Predicates = [IsFP64bit] in
+let Predicates = [IsFP64bit], DecoderNamespace = "Mips64" in
 multiclass FFR1_L_M funct, string opstr> {
   def _S   : FFR1;
   def _D64 : FFR1;
@@ -134,7 +137,9 @@ multiclass FFR1P_M funct, string opstr, SDNode OpNode> {
   def _D32 : FFR1P,
              Requires<[NotFP64bit]>;
   def _D64 : FFR1P,
-             Requires<[IsFP64bit]>;
+             Requires<[IsFP64bit]> {
+    let DecoderNamespace = "Mips64";
+  }
 }
 
 multiclass FFR2P_M funct, string opstr, SDNode OpNode, bit isComm = 0> {
@@ -143,9 +148,11 @@ multiclass FFR2P_M funct, string opstr, SDNode OpNode, bit isComm = 0> {
   def _D32 : FFR2P,
              Requires<[NotFP64bit]>;
   def _D64 : FFR2P,
-             Requires<[IsFP64bit]>;
+             Requires<[IsFP64bit]> {
+    let DecoderNamespace = "Mips64";
   }
 }
+}
 
 // FP madd/msub/nmadd/nmsub instruction classes.
 class FMADDSUB funct, bits<3> fmt, string opstr, string fmtstr,
@@ -172,9 +179,11 @@ defm CEIL_L  : FFR1_L_M<0xa, "ceil">;
 defm FLOOR_W : FFR1_W_M<0xf, "floor">;
 defm FLOOR_L : FFR1_L_M<0xb, "floor">;
 defm CVT_W   : FFR1_W_M<0x24, "cvt">;
-defm CVT_L   : FFR1_L_M<0x25, "cvt">;
+//defm CVT_L   : FFR1_L_M<0x25, "cvt">;
 
 def CVT_S_W : FFR1<0x20, 20, "cvt", "s.w", FGR32, FGR32>;
+def CVT_L_S : FFR1<0x25, 16, "cvt", "l.s", FGR64, FGR32>;
+def CVT_L_D64: FFR1<0x25, 17, "cvt", "l.d", FGR64, FGR64>;
 
 let Predicates = [NotFP64bit] in {
   def CVT_S_D32 : FFR1<0x20, 17, "cvt", "s.d", FGR32, AFGR64>;
@@ -182,7 +191,7 @@ let Predicates = [NotFP64bit] in {
   def CVT_D32_S : FFR1<0x21, 16, "cvt", "d.s", AFGR64, FGR32>;
 }
 
-let Predicates = [IsFP64bit] in {
+let Predicates = [IsFP64bit], DecoderNamespace = "Mips64" in {
  def CVT_S_D64 : FFR1<0x20, 17, "cvt", "s.d", FGR32, FGR64>;
  def CVT_S_L   : FFR1<0x20, 21, "cvt", "s.l", FGR32, FGR64>;
  def CVT_D64_W : FFR1<0x21, 20, "cvt", "d.w", FGR64, FGR32>;
@@ -235,14 +244,20 @@ def FMOV_S   : FFR1<0x6, 16, "mov", "s", FGR32, FGR32>;
 def FMOV_D32 : FFR1<0x6, 17, "mov", "d", AFGR64, AFGR64>,
                Requires<[NotFP64bit]>;
 def FMOV_D64 : FFR1<0x6, 17, "mov", "d", FGR64, FGR64>,
-               Requires<[IsFP64bit]>;
+               Requires<[IsFP64bit]> {
+  let DecoderNamespace = "Mips64";
+}
 
 /// Floating Point Memory Instructions
-let Predicates = [IsN64] in {
+let Predicates = [IsN64], DecoderNamespace = "Mips64" in {
   def LWC1_P8   : FPLoad<0x31, "lwc1", FGR32, mem64>;
   def SWC1_P8   : FPStore<0x39, "swc1", FGR32, mem64>;
-  def LDC164_P8 : FPLoad<0x35, "ldc1", FGR64, mem64>;
-  def SDC164_P8 : FPStore<0x3d, "sdc1", FGR64, mem64>;
+  def LDC164_P8 : FPLoad<0x35, "ldc1", FGR64, mem64> {
+    let isCodeGenOnly =1;
+  }
+  def SDC164_P8 : FPStore<0x3d, "sdc1", FGR64, mem64> {
+    let isCodeGenOnly =1;
+  }
 }
 
 let Predicates = [NotN64] in {
@@ -250,7 +265,7 @@ let Predicates = [NotN64] in {
   def SWC1   : FPStore<0x39, "swc1", FGR32, mem>;
 }
 
-let Predicates = [NotN64, HasMips64] in {
+let Predicates = [NotN64, HasMips64], DecoderNamespace = "Mips64" in {
   def LDC164 : FPLoad<0x35, "ldc1", FGR64, mem>;
   def SDC164 : FPStore<0x3d, "sdc1", FGR64, mem>;
 }
@@ -273,13 +288,13 @@ let Predicates = [HasMips32r2, NotMips64] in {
   def SDXC1 : FPIdxStore<0x9, "sdxc1", AFGR64, CPURegs, store_a>;
 }
 
-let Predicates = [HasMips64, NotN64] in {
+let Predicates = [HasMips64, NotN64], DecoderNamespace="Mips64" in {
   def LDXC164 : FPIdxLoad<0x1, "ldxc1", FGR64, CPURegs, load_a>;
   def SDXC164 : FPIdxStore<0x9, "sdxc1", FGR64, CPURegs, store_a>;
 }
 
 // n64
-let Predicates = [IsN64] in {
+let Predicates = [IsN64], isCodeGenOnly=1 in {
   def LWXC1_P8   : FPIdxLoad<0x0, "lwxc1", FGR32, CPU64Regs, load_a>;
   def LUXC1_P8   : FPIdxLoad<0x5, "luxc1", FGR32, CPU64Regs, load_u>;
   def LDXC164_P8 : FPIdxLoad<0x1, "ldxc1", FGR64, CPU64Regs, load_a>;
@@ -314,12 +329,12 @@ let Predicates = [HasMips32r2, NotFP64bit, NoNaNsFPMath] in {
   def NMSUB_D32 : FNMADDSUB<0x7, 1, "nmsub", "d", fsub, AFGR64>;
 }
 
-let Predicates = [HasMips32r2, IsFP64bit] in {
+let Predicates = [HasMips32r2, IsFP64bit], isCodeGenOnly=1 in {
   def MADD_D64 : FMADDSUB<0x4, 1, "madd", "d", fadd, FGR64>;
   def MSUB_D64 : FMADDSUB<0x5, 1, "msub", "d", fsub, FGR64>;
 }
 
-let Predicates = [HasMips32r2, IsFP64bit, NoNaNsFPMath] in {
+let Predicates = [HasMips32r2, IsFP64bit, NoNaNsFPMath], isCodeGenOnly=1 in {
   def NMADD_D64 : FNMADDSUB<0x6, 1, "nmadd", "d", fadd, FGR64>;
   def NMSUB_D64 : FNMADDSUB<0x7, 1, "nmsub", "d", fsub, FGR64>;
 }
@@ -342,9 +357,10 @@ let isBranch=1, isTerminator=1, hasDelaySlot=1, base=0x8, Uses=[FCR31] in
   let Inst{16} = tf;
 }
 
+let DecoderMethod = "DecodeBC1" in {
 def BC1F  : FBRANCH<0, 0, MIPS_BRANCH_F,  "bc1f">;
 def BC1T  : FBRANCH<0, 1, MIPS_BRANCH_T,  "bc1t">;
-
+}
 //===----------------------------------------------------------------------===//
 // Floating Point Flag Conditions
 //===----------------------------------------------------------------------===//
@@ -376,7 +392,9 @@ class FCMP fmt, RegisterClass RC, string typestr> :
 let Defs=[FCR31] in {
   def FCMP_S32 : FCMP<0x10, FGR32, "s">;
   def FCMP_D32 : FCMP<0x11, AFGR64, "d">, Requires<[NotFP64bit]>;
-  def FCMP_D64 : FCMP<0x11, FGR64, "d">, Requires<[IsFP64bit]>;
+  def FCMP_D64 : FCMP<0x11, FGR64, "d">, Requires<[IsFP64bit]> {
+    let DecoderNamespace = "Mips64";
+  }
 }
 
 //===----------------------------------------------------------------------===//
@@ -438,13 +456,13 @@ let Predicates = [IsFP64bit] in {
 
 // Patterns for unaligned floating point loads and stores.
 let Predicates = [HasMips32r2Or64, NotN64] in {
-  def : Pat<(f32 (load_u CPURegs:$addr)), (LUXC1 CPURegs:$addr, ZERO)>; 
+  def : Pat<(f32 (load_u CPURegs:$addr)), (LUXC1 CPURegs:$addr, ZERO)>;
   def : Pat<(store_u FGR32:$src, CPURegs:$addr),
             (SUXC1 FGR32:$src, CPURegs:$addr, ZERO)>;
 }
 
 let Predicates = [IsN64] in {
-  def : Pat<(f32 (load_u CPU64Regs:$addr)), (LUXC1_P8 CPU64Regs:$addr, ZERO_64)>; 
+  def : Pat<(f32 (load_u CPU64Regs:$addr)), (LUXC1_P8 CPU64Regs:$addr, ZERO_64)>;
   def : Pat<(store_u FGR32:$src, CPU64Regs:$addr),
             (SUXC1_P8 FGR32:$src, CPU64Regs:$addr, ZERO_64)>;
 }
diff --git a/lib/Target/Mips/MipsInstrFormats.td b/lib/Target/Mips/MipsInstrFormats.td
index 455530389eb..841eba0ec0a 100644
--- a/lib/Target/Mips/MipsInstrFormats.td
+++ b/lib/Target/Mips/MipsInstrFormats.td
@@ -45,6 +45,8 @@ class MipsInst pattern,
 
   let Namespace = "Mips";
 
+  let Size = 4;
+
   bits<6> Opcode = 0;
 
   // Top 6 bits are the 'opcode' field
@@ -64,6 +66,10 @@ class MipsInst pattern,
 
   // TSFlags layout should be kept in sync with MipsInstrInfo.h.
   let TSFlags{3-0}   = FormBits;
+
+  let DecoderNamespace = "Mips";
+
+  field bits<32> SoftFail = 0;
 }
 
 // Mips Pseudo Instructions Format
diff --git a/lib/Target/Mips/MipsInstrInfo.td b/lib/Target/Mips/MipsInstrInfo.td
index be74f8e5230..873d2bd99ae 100644
--- a/lib/Target/Mips/MipsInstrInfo.td
+++ b/lib/Target/Mips/MipsInstrInfo.td
@@ -121,21 +121,36 @@ def MipsIns :  SDNode<"MipsISD::Ins", SDT_Ins>;
 //===----------------------------------------------------------------------===//
 // Mips Instruction Predicate Definitions.
 //===----------------------------------------------------------------------===//
-def HasSEInReg  : Predicate<"Subtarget.hasSEInReg()">;
-def HasBitCount : Predicate<"Subtarget.hasBitCount()">;
-def HasSwap     : Predicate<"Subtarget.hasSwap()">;
-def HasCondMov  : Predicate<"Subtarget.hasCondMov()">;
-def HasMips32    : Predicate<"Subtarget.hasMips32()">;
-def HasMips32r2  : Predicate<"Subtarget.hasMips32r2()">;
-def HasMips64    : Predicate<"Subtarget.hasMips64()">;
-def HasMips32r2Or64 : Predicate<"Subtarget.hasMips32r2Or64()">;
-def NotMips64    : Predicate<"!Subtarget.hasMips64()">;
-def HasMips64r2  : Predicate<"Subtarget.hasMips64r2()">;
-def IsN64       : Predicate<"Subtarget.isABI_N64()">;
-def NotN64      : Predicate<"!Subtarget.isABI_N64()">;
-def RelocStatic : Predicate<"TM.getRelocationModel() == Reloc::Static">;
-def RelocPIC    : Predicate<"TM.getRelocationModel() == Reloc::PIC_">;
-def NoNaNsFPMath : Predicate<"TM.Options.NoNaNsFPMath">;
+def HasSEInReg  :     Predicate<"Subtarget.hasSEInReg()">,
+                      AssemblerPredicate<"FeatureSEInReg">;
+def HasBitCount :     Predicate<"Subtarget.hasBitCount()">,
+                      AssemblerPredicate<"FeatureBitCount">;
+def HasSwap     :     Predicate<"Subtarget.hasSwap()">,
+                      AssemblerPredicate<"FeatureSwap">;
+def HasCondMov  :     Predicate<"Subtarget.hasCondMov()">,
+                      AssemblerPredicate<"FeatureCondMov">;
+def HasMips32    :    Predicate<"Subtarget.hasMips32()">,
+                      AssemblerPredicate<"FeatureMips32">;
+def HasMips32r2  :    Predicate<"Subtarget.hasMips32r2()">,
+                      AssemblerPredicate<"FeatureMips32r2">;
+def HasMips64    :    Predicate<"Subtarget.hasMips64()">,
+                      AssemblerPredicate<"FeatureMips64">;
+def HasMips32r2Or64 : Predicate<"Subtarget.hasMips32r2Or64()">,
+                      AssemblerPredicate<"FeatureMips32r2,FeatureMips64">;
+def NotMips64    :    Predicate<"!Subtarget.hasMips64()">,
+                      AssemblerPredicate<"!FeatureMips64">;
+def HasMips64r2  :    Predicate<"Subtarget.hasMips64r2()">,
+                      AssemblerPredicate<"FeatureMips64r2">;
+def IsN64       :     Predicate<"Subtarget.isABI_N64()">,
+                      AssemblerPredicate<"FeatureN64">;
+def NotN64      :     Predicate<"!Subtarget.isABI_N64()">,
+                      AssemblerPredicate<"!FeatureN64">;
+def RelocStatic :     Predicate<"TM.getRelocationModel() == Reloc::Static">,
+                      AssemblerPredicate<"FeatureMips32">;
+def RelocPIC    :     Predicate<"TM.getRelocationModel() == Reloc::PIC_">,
+                      AssemblerPredicate<"FeatureMips32">;
+def NoNaNsFPMath :    Predicate<"TM.Options.NoNaNsFPMath">,
+                      AssemblerPredicate<"FeatureMips32">;
 
 //===----------------------------------------------------------------------===//
 // Mips Operand, Complex Patterns and Transformations Definitions.
@@ -148,12 +163,15 @@ def jmptarget   : Operand {
 def brtarget    : Operand {
   let EncoderMethod = "getBranchTargetOpValue";
   let OperandType = "OPERAND_PCREL";
+  let DecoderMethod = "DecodeBranchTarget";
 }
 def calltarget  : Operand {
   let EncoderMethod = "getJumpTargetOpValue";
 }
 def calltarget64: Operand;
-def simm16      : Operand;
+def simm16      : Operand {
+  let DecoderMethod= "DecodeSimm16";
+}
 def simm16_64   : Operand;
 def shamt       : Operand;
 
@@ -189,11 +207,13 @@ def mem_ea_64 : Operand {
 // size operand of ext instruction
 def size_ext : Operand {
   let EncoderMethod = "getSizeExtEncoding";
+  let DecoderMethod = "DecodeExtSize";
 }
 
 // size operand of ins instruction
 def size_ins : Operand {
   let EncoderMethod = "getSizeInsEncoding";
+  let DecoderMethod = "DecodeInsSize";
 }
 
 // Transformation Function - get the lower 16 bits.
@@ -295,6 +315,7 @@ class ArithLogicR op, bits<6> func, string instr_asm, SDNode OpNode,
      [(set RC:$rd, (OpNode RC:$rs, RC:$rt))], itin> {
   let shamt = 0;
   let isCommutable = isComm;
+  let isReMaterializable = 1;
 }
 
 class ArithOverflowR op, bits<6> func, string instr_asm,
@@ -310,7 +331,9 @@ class ArithLogicI op, string instr_asm, SDNode OpNode,
                   Operand Od, PatLeaf imm_type, RegisterClass RC> :
   FI;
+     [(set RC:$rt, (OpNode RC:$rs, imm_type:$imm16))], IIAlu> {
+  let isReMaterializable = 1;
+}
 
 class ArithOverflowI op, string instr_asm, SDNode OpNode,
                      Operand Od, PatLeaf imm_type, RegisterClass RC> :
@@ -366,6 +389,7 @@ class LoadUpper op, string instr_asm, RegisterClass RC, Operand Imm>:
      !strconcat(instr_asm, "\t$rt, $imm16"), [], IIAlu> {
   let rs = 0;
   let neverHasSideEffects = 1;
+  let isReMaterializable = 1;
 }
 
 class FMem op, dag outs, dag ins, string asmstr, list pattern,
@@ -373,6 +397,7 @@ class FMem op, dag outs, dag ins, string asmstr, list pattern,
   bits<21> addr;
   let Inst{25-21} = addr{20-16};
   let Inst{15-0}  = addr{15-0};
+  let DecoderMethod = "DecodeMem";
 }
 
 // Memory Load/Store
@@ -407,7 +432,10 @@ multiclass LoadM32 op, string instr_asm, PatFrag OpNode,
   def #NAME# : LoadM,
                Requires<[NotN64]>;
   def _P8    : LoadM,
-               Requires<[IsN64]>;
+               Requires<[IsN64]> {
+    let DecoderNamespace = "Mips64";
+    let isCodeGenOnly = 1;
+  }
 }
 
 // 64-bit load.
@@ -416,7 +444,10 @@ multiclass LoadM64 op, string instr_asm, PatFrag OpNode,
   def #NAME# : LoadM,
                Requires<[NotN64]>;
   def _P8    : LoadM,
-               Requires<[IsN64]>;
+               Requires<[IsN64]> {
+    let DecoderNamespace = "Mips64";
+    let isCodeGenOnly = 1;
+  }
 }
 
 // 32-bit load.
@@ -424,7 +455,10 @@ multiclass LoadUnAlign32 op> {
   def #NAME# : LoadUnAlign,
                Requires<[NotN64]>;
   def _P8    : LoadUnAlign,
-               Requires<[IsN64]>;
+               Requires<[IsN64]> {
+    let DecoderNamespace = "Mips64";
+    let isCodeGenOnly = 1;
+  }
 }
 // 32-bit store.
 multiclass StoreM32 op, string instr_asm, PatFrag OpNode,
@@ -432,7 +466,10 @@ multiclass StoreM32 op, string instr_asm, PatFrag OpNode,
   def #NAME# : StoreM,
                Requires<[NotN64]>;
   def _P8    : StoreM,
-               Requires<[IsN64]>;
+               Requires<[IsN64]> {
+    let DecoderNamespace = "Mips64";
+    let isCodeGenOnly = 1;
+  }
 }
 
 // 64-bit store.
@@ -441,7 +478,10 @@ multiclass StoreM64 op, string instr_asm, PatFrag OpNode,
   def #NAME# : StoreM,
                Requires<[NotN64]>;
   def _P8    : StoreM,
-               Requires<[IsN64]>;
+               Requires<[IsN64]> {
+    let DecoderNamespace = "Mips64";
+    let isCodeGenOnly = 1;
+  }
 }
 
 // 32-bit store.
@@ -449,7 +489,10 @@ multiclass StoreUnAlign32 op> {
   def #NAME# : StoreUnAlign,
                Requires<[NotN64]>;
   def _P8    : StoreUnAlign,
-               Requires<[IsN64]>;
+               Requires<[IsN64]> {
+    let DecoderNamespace = "Mips64";
+    let isCodeGenOnly = 1;
+  }
 }
 
 // Conditional Branch
@@ -499,6 +542,7 @@ class JumpFJ op, string instr_asm>:
   let isBarrier=1;
   let hasDelaySlot = 1;
   let Predicates = [RelocStatic];
+  let DecoderMethod = "DecodeJumpTarget";
 }
 
 // Unconditional branch
@@ -529,7 +573,9 @@ let isCall=1, hasDelaySlot=1 in {
   class JumpLink op, string instr_asm>:
     FJ;
+       IIBranch> {
+       let DecoderMethod = "DecodeJumpTarget";
+       }
 
   class JumpLinkReg op, bits<6> func, string instr_asm,
                     RegisterClass RC>:
@@ -685,7 +731,9 @@ class Atomic2Ops {
   def #NAME# : Atomic2Ops, Requires<[NotN64]>;
-  def _P8    : Atomic2Ops, Requires<[IsN64]>;
+  def _P8    : Atomic2Ops, Requires<[IsN64]> {
+    let DecoderNamespace = "Mips64";
+  }
 }
 
 // Atomic Compare & Swap.
@@ -697,7 +745,9 @@ class AtomicCmpSwap  {
   def #NAME# : AtomicCmpSwap, Requires<[NotN64]>;
-  def _P8    : AtomicCmpSwap, Requires<[IsN64]>;
+  def _P8    : AtomicCmpSwap, Requires<[IsN64]> {
+    let DecoderNamespace = "Mips64";
+  }
 }
 
 class LLBase Opc, string opstring, RegisterClass RC, Operand Mem> :
@@ -868,9 +918,14 @@ def SYNC : MipsInst<(outs), (ins i32imm:$stype), "sync $stype",
 
 /// Load-linked, Store-conditional
 def LL    : LLBase<0x30, "ll", CPURegs, mem>, Requires<[NotN64]>;
-def LL_P8 : LLBase<0x30, "ll", CPURegs, mem64>, Requires<[IsN64]>;
+def LL_P8 : LLBase<0x30, "ll", CPURegs, mem64>, Requires<[IsN64]> {
+  let DecoderNamespace = "Mips64";
+}
+
 def SC    : SCBase<0x38, "sc", CPURegs, mem>, Requires<[NotN64]>;
-def SC_P8 : SCBase<0x38, "sc", CPURegs, mem64>, Requires<[IsN64]>;
+def SC_P8 : SCBase<0x38, "sc", CPURegs, mem64>, Requires<[IsN64]> {
+  let DecoderNamespace = "Mips64";
+}
 
 /// Jump and Branch Instructions
 def J       : JumpFJ<0x02, "j">;
@@ -888,7 +943,7 @@ def JALR : JumpLinkReg<0x00, 0x09, "jalr", CPURegs>;
 def BGEZAL  : BranchLink<"bgezal", 0x11, CPURegs>;
 def BLTZAL  : BranchLink<"bltzal", 0x10, CPURegs>;
 
-let isReturn=1, isTerminator=1, hasDelaySlot=1,
+let isReturn=1, isTerminator=1, hasDelaySlot=1, isCodeGenOnly=1,
     isBarrier=1, hasCtrlDep=1, rd=0, rt=0, shamt=0 in
   def RET : FR <0x00, 0x08, (outs), (ins CPURegs:$target),
                 "jr\t$target", [(MipsRet CPURegs:$target)], IIBranch>;
@@ -923,13 +978,17 @@ let addr=0 in
 // instructions. The same not happens for stack address copies, so an
 // add op with mem ComplexPattern is used and the stack address copy
 // can be matched. It's similar to Sparc LEA_ADDRi
-def LEA_ADDiu : EffectiveAddress<"addiu\t$rt, $addr", CPURegs, mem_ea>;
+def LEA_ADDiu : EffectiveAddress<"addiu\t$rt, $addr", CPURegs, mem_ea> {
+  let isCodeGenOnly = 1;
+}
 
 // DynAlloc node points to dynamically allocated stack space.
 // $sp is added to the list of implicitly used registers to prevent dead code
 // elimination from removing instructions that modify $sp.
 let Uses = [SP] in
-def DynAlloc : EffectiveAddress<"addiu\t$rt, $addr", CPURegs, mem_ea>;
+def DynAlloc : EffectiveAddress<"addiu\t$rt, $addr", CPURegs, mem_ea> {
+  let isCodeGenOnly = 1;
+}
 
 // MADD*/MSUB*
 def MADD  : MArithR<0, "madd", MipsMAdd, 1>;
diff --git a/lib/Target/Mips/MipsTargetMachine.cpp b/lib/Target/Mips/MipsTargetMachine.cpp
index ad022311ed7..858723bad9d 100644
--- a/lib/Target/Mips/MipsTargetMachine.cpp
+++ b/lib/Target/Mips/MipsTargetMachine.cpp
@@ -117,18 +117,16 @@ TargetPassConfig *MipsTargetMachine::createPassConfig(PassManagerBase &PM) {
 
 // Install an instruction selector pass using
 // the ISelDag to gen Mips code.
-bool MipsPassConfig::addInstSelector()
-{
-  PM.add(createMipsISelDag(getMipsTargetMachine()));
+bool MipsPassConfig::addInstSelector() {
+  PM->add(createMipsISelDag(getMipsTargetMachine()));
   return false;
 }
 
 // Implemented by targets that want to run passes immediately before
 // machine code is emitted. return true if -print-machineinstrs should
 // print out the code after the passes.
-bool MipsPassConfig::addPreEmitPass()
-{
-  PM.add(createMipsDelaySlotFillerPass(getMipsTargetMachine()));
+bool MipsPassConfig::addPreEmitPass() {
+  PM->add(createMipsDelaySlotFillerPass(getMipsTargetMachine()));
   return true;
 }
 
@@ -136,12 +134,12 @@ bool MipsPassConfig::addPreRegAlloc() {
   // Do not restore $gp if target is Mips64.
   // In N32/64, $gp is a callee-saved register.
   if (!getMipsSubtarget().hasMips64())
-    PM.add(createMipsEmitGPRestorePass(getMipsTargetMachine()));
+    PM->add(createMipsEmitGPRestorePass(getMipsTargetMachine()));
   return true;
 }
 
 bool MipsPassConfig::addPreSched2() {
-  PM.add(createMipsExpandPseudoPass(getMipsTargetMachine()));
+  PM->add(createMipsExpandPseudoPass(getMipsTargetMachine()));
   return true;
 }
 
diff --git a/lib/Target/PTX/PTXTargetMachine.cpp b/lib/Target/PTX/PTXTargetMachine.cpp
index c55a658dc37..97b8de1a0b4 100644
--- a/lib/Target/PTX/PTXTargetMachine.cpp
+++ b/lib/Target/PTX/PTXTargetMachine.cpp
@@ -130,7 +130,7 @@ TargetPassConfig *PTXTargetMachine::createPassConfig(PassManagerBase &PM) {
 }
 
 bool PTXPassConfig::addInstSelector() {
-  PM.add(createPTXISelDag(getPTXTargetMachine(), getOptLevel()));
+  PM->add(createPTXISelDag(getPTXTargetMachine(), getOptLevel()));
   return false;
 }
 
@@ -145,7 +145,7 @@ void PTXPassConfig::addOptimizedRegAlloc(FunctionPass *RegAllocPass) {
 
 bool PTXPassConfig::addPostRegAlloc() {
   // PTXMFInfoExtract must after register allocation!
-  //PM.add(createPTXMFInfoExtract(getPTXTargetMachine()));
+  //PM->add(createPTXMFInfoExtract(getPTXTargetMachine()));
   return false;
 }
 
@@ -159,7 +159,7 @@ void PTXPassConfig::addMachineLateOptimization() {
 }
 
 bool PTXPassConfig::addPreEmitPass() {
-  PM.add(createPTXMFInfoExtract(getPTXTargetMachine(), getOptLevel()));
-  PM.add(createPTXFPRoundingModePass(getPTXTargetMachine(), getOptLevel()));
+  PM->add(createPTXMFInfoExtract(getPTXTargetMachine(), getOptLevel()));
+  PM->add(createPTXFPRoundingModePass(getPTXTargetMachine(), getOptLevel()));
   return true;
 }
diff --git a/lib/Target/PowerPC/PPCCallingConv.td b/lib/Target/PowerPC/PPCCallingConv.td
index 9883c2e4299..b2b53648561 100644
--- a/lib/Target/PowerPC/PPCCallingConv.td
+++ b/lib/Target/PowerPC/PPCCallingConv.td
@@ -12,10 +12,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-/// CCIfSubtarget - Match if the current subtarget has a feature F.
-class CCIfSubtarget
- : CCIf().", F), A>;
-
 //===----------------------------------------------------------------------===//
 // Return Value Calling Convention
 //===----------------------------------------------------------------------===//
diff --git a/lib/Target/PowerPC/PPCTargetMachine.cpp b/lib/Target/PowerPC/PPCTargetMachine.cpp
index d1139766991..50f3db8b27f 100644
--- a/lib/Target/PowerPC/PPCTargetMachine.cpp
+++ b/lib/Target/PowerPC/PPCTargetMachine.cpp
@@ -98,13 +98,13 @@ TargetPassConfig *PPCTargetMachine::createPassConfig(PassManagerBase &PM) {
 
 bool PPCPassConfig::addInstSelector() {
   // Install an instruction selector.
-  PM.add(createPPCISelDag(getPPCTargetMachine()));
+  PM->add(createPPCISelDag(getPPCTargetMachine()));
   return false;
 }
 
 bool PPCPassConfig::addPreEmitPass() {
   // Must run branch selection immediately preceding the asm printer.
-  PM.add(createPPCBranchSelectionPass());
+  PM->add(createPPCBranchSelectionPass());
   return false;
 }
 
diff --git a/lib/Target/Sparc/SparcTargetMachine.cpp b/lib/Target/Sparc/SparcTargetMachine.cpp
index 6f313562c10..cc253077a98 100644
--- a/lib/Target/Sparc/SparcTargetMachine.cpp
+++ b/lib/Target/Sparc/SparcTargetMachine.cpp
@@ -59,7 +59,7 @@ TargetPassConfig *SparcTargetMachine::createPassConfig(PassManagerBase &PM) {
 }
 
 bool SparcPassConfig::addInstSelector() {
-  PM.add(createSparcISelDag(getSparcTargetMachine()));
+  PM->add(createSparcISelDag(getSparcTargetMachine()));
   return false;
 }
 
@@ -67,8 +67,8 @@ bool SparcPassConfig::addInstSelector() {
 /// passes immediately before machine code is emitted.  This should return
 /// true if -print-machineinstrs should print out the code after the passes.
 bool SparcPassConfig::addPreEmitPass(){
-  PM.add(createSparcFPMoverPass(getSparcTargetMachine()));
-  PM.add(createSparcDelaySlotFillerPass(getSparcTargetMachine()));
+  PM->add(createSparcFPMoverPass(getSparcTargetMachine()));
+  PM->add(createSparcDelaySlotFillerPass(getSparcTargetMachine()));
   return true;
 }
 
diff --git a/lib/Target/TargetLibraryInfo.cpp b/lib/Target/TargetLibraryInfo.cpp
index 269958fd7f1..ec95ad4deea 100644
--- a/lib/Target/TargetLibraryInfo.cpp
+++ b/lib/Target/TargetLibraryInfo.cpp
@@ -56,7 +56,7 @@ const char* TargetLibraryInfo::StandardNames[LibFunc::NumLibFuncs] =
     "exp2f",
     "expm1",
     "expm1l",
-    "expl1f",
+    "expm1f",
     "fabs",
     "fabsl",
     "fabsf",
@@ -95,6 +95,9 @@ const char* TargetLibraryInfo::StandardNames[LibFunc::NumLibFuncs] =
     "rint",
     "rintf",
     "rintl",
+    "round",
+    "roundf",
+    "roundl",
     "sin",
     "sinl",
     "sinf",
@@ -155,6 +158,81 @@ static void initialize(TargetLibraryInfo &TLI, const Triple &T) {
     TLI.setUnavailable(LibFunc::siprintf);
     TLI.setUnavailable(LibFunc::fiprintf);
   }
+
+  if (T.getOS() == Triple::Win32) {
+    // Win32 does not support long double
+    TLI.setUnavailable(LibFunc::acosl);
+    TLI.setUnavailable(LibFunc::asinl);
+    TLI.setUnavailable(LibFunc::atanl);
+    TLI.setUnavailable(LibFunc::atan2l);
+    TLI.setUnavailable(LibFunc::ceill);
+    TLI.setUnavailable(LibFunc::copysignl);
+    TLI.setUnavailable(LibFunc::cosl);
+    TLI.setUnavailable(LibFunc::coshl);
+    TLI.setUnavailable(LibFunc::expl);
+    TLI.setUnavailable(LibFunc::fabsf); // Win32 and Win64 both lack fabsf
+    TLI.setUnavailable(LibFunc::fabsl);
+    TLI.setUnavailable(LibFunc::floorl);
+    TLI.setUnavailable(LibFunc::fmodl);
+    TLI.setUnavailable(LibFunc::logl);
+    TLI.setUnavailable(LibFunc::powl);
+    TLI.setUnavailable(LibFunc::sinl);
+    TLI.setUnavailable(LibFunc::sinhl);
+    TLI.setUnavailable(LibFunc::sqrtl);
+    TLI.setUnavailable(LibFunc::tanl);
+    TLI.setUnavailable(LibFunc::tanhl);
+
+    // Win32 only has C89 math
+    TLI.setUnavailable(LibFunc::exp2);
+    TLI.setUnavailable(LibFunc::exp2f);
+    TLI.setUnavailable(LibFunc::exp2l);
+    TLI.setUnavailable(LibFunc::expm1);
+    TLI.setUnavailable(LibFunc::expm1f);
+    TLI.setUnavailable(LibFunc::expm1l);
+    TLI.setUnavailable(LibFunc::log2);
+    TLI.setUnavailable(LibFunc::log2f);
+    TLI.setUnavailable(LibFunc::log2l);
+    TLI.setUnavailable(LibFunc::log1p);
+    TLI.setUnavailable(LibFunc::log1pf);
+    TLI.setUnavailable(LibFunc::log1pl);
+    TLI.setUnavailable(LibFunc::nearbyint);
+    TLI.setUnavailable(LibFunc::nearbyintf);
+    TLI.setUnavailable(LibFunc::nearbyintl);
+    TLI.setUnavailable(LibFunc::rint);
+    TLI.setUnavailable(LibFunc::rintf);
+    TLI.setUnavailable(LibFunc::rintl);
+    TLI.setUnavailable(LibFunc::round);
+    TLI.setUnavailable(LibFunc::roundf);
+    TLI.setUnavailable(LibFunc::roundl);
+    TLI.setUnavailable(LibFunc::trunc);
+    TLI.setUnavailable(LibFunc::truncf);
+    TLI.setUnavailable(LibFunc::truncl);
+
+    // Win32 provides some C99 math with mangled names
+    TLI.setAvailableWithName(LibFunc::copysign, "_copysign");
+
+    if (T.getArch() == Triple::x86) {
+      // Win32 on x86 implements single-precision math functions as macros
+      TLI.setUnavailable(LibFunc::acosf);
+      TLI.setUnavailable(LibFunc::asinf);
+      TLI.setUnavailable(LibFunc::atanf);
+      TLI.setUnavailable(LibFunc::atan2f);
+      TLI.setUnavailable(LibFunc::ceilf);
+      TLI.setUnavailable(LibFunc::copysignf);
+      TLI.setUnavailable(LibFunc::cosf);
+      TLI.setUnavailable(LibFunc::coshf);
+      TLI.setUnavailable(LibFunc::expf);
+      TLI.setUnavailable(LibFunc::floorf);
+      TLI.setUnavailable(LibFunc::fmodf);
+      TLI.setUnavailable(LibFunc::logf);
+      TLI.setUnavailable(LibFunc::powf);
+      TLI.setUnavailable(LibFunc::sinf);
+      TLI.setUnavailable(LibFunc::sinhf);
+      TLI.setUnavailable(LibFunc::sqrtf);
+      TLI.setUnavailable(LibFunc::tanf);
+      TLI.setUnavailable(LibFunc::tanhf);
+    }
+  }
 }
 
 
diff --git a/lib/Target/X86/Utils/X86ShuffleDecode.cpp b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
index 32c722acc43..a802333002d 100644
--- a/lib/Target/X86/Utils/X86ShuffleDecode.cpp
+++ b/lib/Target/X86/Utils/X86ShuffleDecode.cpp
@@ -169,6 +169,9 @@ void DecodeUNPCKLMask(EVT VT, SmallVectorImpl &ShuffleMask) {
 
 void DecodeVPERM2X128Mask(EVT VT, unsigned Imm,
                           SmallVectorImpl &ShuffleMask) {
+  if (Imm & 0x88)
+    return; // Not a shuffle
+
   unsigned HalfSize = VT.getVectorNumElements()/2;
   unsigned FstHalfBegin = (Imm & 0x3) * HalfSize;
   unsigned SndHalfBegin = ((Imm >> 4) & 0x3) * HalfSize;
diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 9b83aade338..04299f30080 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -2935,6 +2935,7 @@ static SDValue getTargetShuffleNode(unsigned Opc, DebugLoc dl, EVT VT,
   case X86ISD::PSHUFHW:
   case X86ISD::PSHUFLW:
   case X86ISD::VPERMILP:
+  case X86ISD::VPERMI:
     return DAG.getNode(Opc, dl, VT, V1, DAG.getConstant(TargetMask, MVT::i8));
   }
 }
@@ -3976,6 +3977,27 @@ unsigned X86::getInsertVINSERTF128Immediate(SDNode *N) {
   return Index / NumElemsPerChunk;
 }
 
+/// getShuffleCLImmediate - Return the appropriate immediate to shuffle
+/// the specified VECTOR_SHUFFLE mask with VPERMQ and VPERMPD instructions.
+/// Handles 256-bit.
+static unsigned getShuffleCLImmediate(ShuffleVectorSDNode *N) {
+  EVT VT = N->getValueType(0);
+
+  unsigned NumElts = VT.getVectorNumElements();
+
+  assert((VT.is256BitVector() && NumElts == 4) &&
+         "Unsupported vector type for VPERMQ/VPERMPD");
+
+  unsigned Mask = 0;
+  for (unsigned i = 0; i != NumElts; ++i) {
+    int Elt = N->getMaskElt(i);
+    if (Elt < 0)
+      continue;
+    Mask |= Elt << (i*2);
+  }
+
+  return Mask;
+}
 /// isZeroNode - Returns true if Elt is a constant zero or a floating point
 /// constant +0.0.
 bool X86::isZeroNode(SDValue Elt) {
@@ -4408,6 +4430,7 @@ static bool getTargetShuffleMask(SDNode *N, EVT VT,
   case X86ISD::VPERM2X128:
     ImmN = N->getOperand(N->getNumOperands()-1);
     DecodeVPERM2X128Mask(VT, cast(ImmN)->getZExtValue(), Mask);
+    if (Mask.empty()) return false;
     break;
   case X86ISD::MOVDDUP:
   case X86ISD::MOVLHPD:
@@ -6628,6 +6651,23 @@ X86TargetLowering::LowerVECTOR_SHUFFLE(SDValue Op, SelectionDAG &DAG) const {
   if (BlendOp.getNode())
     return BlendOp;
 
+  if (V2IsUndef && HasAVX2 && (VT == MVT::v8i32 || VT == MVT::v8f32)) {
+    SmallVector permclMask;
+    for (unsigned i = 0; i != 8; ++i) {
+      permclMask.push_back(DAG.getConstant((M[i]>=0) ? M[i] : 0, MVT::i32));
+    }
+    SDValue Mask = DAG.getNode(ISD::BUILD_VECTOR, dl, MVT::v8i32,
+                               &permclMask[0], 8);
+    // Bitcast is for VPERMPS since mask is v8i32 but node takes v8f32
+    return DAG.getNode(X86ISD::VPERMV, dl, VT,
+                       DAG.getNode(ISD::BITCAST, dl, VT, Mask), V1);
+  }
+
+  if (V2IsUndef && HasAVX2 && (VT == MVT::v4i64 || VT == MVT::v4f64))
+    return getTargetShuffleNode(X86ISD::VPERMI, dl, VT, V1,
+                                getShuffleCLImmediate(SVOp), DAG);
+
+
   //===--------------------------------------------------------------------===//
   // Since no target specific shuffle was selected for this generic one,
   // lower it into other known shuffles. FIXME: this isn't true yet, but
@@ -9552,12 +9592,12 @@ X86TargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op, SelectionDAG &DAG) const
   case Intrinsic::x86_avx2_vperm2i128:
     return DAG.getNode(X86ISD::VPERM2X128, dl, Op.getValueType(),
                        Op.getOperand(1), Op.getOperand(2), Op.getOperand(3));
-  case Intrinsic::x86_avx_vpermil_ps:
-  case Intrinsic::x86_avx_vpermil_pd:
-  case Intrinsic::x86_avx_vpermil_ps_256:
-  case Intrinsic::x86_avx_vpermil_pd_256:
-    return DAG.getNode(X86ISD::VPERMILP, dl, Op.getValueType(),
-                       Op.getOperand(1), Op.getOperand(2));
+  case Intrinsic::x86_avx2_permd:
+  case Intrinsic::x86_avx2_permps:
+    // Operands intentionally swapped. Mask is last operand to intrinsic,
+    // but second operand for node/intruction.
+    return DAG.getNode(X86ISD::VPERMV, dl, Op.getValueType(),
+                       Op.getOperand(2), Op.getOperand(1));
 
   // ptest and testp intrinsics. The intrinsic these come from are designed to
   // return an integer value, not just an instruction so lower it to the ptest
@@ -11141,6 +11181,8 @@ const char *X86TargetLowering::getTargetNodeName(unsigned Opcode) const {
   case X86ISD::VBROADCAST:         return "X86ISD::VBROADCAST";
   case X86ISD::VPERMILP:           return "X86ISD::VPERMILP";
   case X86ISD::VPERM2X128:         return "X86ISD::VPERM2X128";
+  case X86ISD::VPERMV:             return "X86ISD::VPERMV";
+  case X86ISD::VPERMI:             return "X86ISD::VPERMI";
   case X86ISD::PMULUDQ:            return "X86ISD::PMULUDQ";
   case X86ISD::VASTART_SAVE_XMM_REGS: return "X86ISD::VASTART_SAVE_XMM_REGS";
   case X86ISD::VAARG_64:           return "X86ISD::VAARG_64";
@@ -11298,14 +11340,15 @@ X86TargetLowering::EmitAtomicBitwiseWithCustomInserter(MachineInstr *bInstr,
                                                        unsigned notOpc,
                                                        unsigned EAXreg,
                                                  const TargetRegisterClass *RC,
-                                                       bool invSrc) const {
+                                                       bool Invert) const {
   // For the atomic bitwise operator, we generate
   //   thisMBB:
   //   newMBB:
   //     ld  t1 = [bitinstr.addr]
   //     op  t2 = t1, [bitinstr.val]
+  //     not t3 = t2  (if Invert)
   //     mov EAX = t1
-  //     lcs dest = [bitinstr.addr], t2  [EAX is implicit]
+  //     lcs dest = [bitinstr.addr], t3  [EAX is implicit]
   //     bz  newMBB
   //     fallthrough -->nextMBB
   const TargetInstrInfo *TII = getTargetMachine().getInstrInfo();
@@ -11353,13 +11396,6 @@ X86TargetLowering::EmitAtomicBitwiseWithCustomInserter(MachineInstr *bInstr,
   for (int i=0; i <= lastAddrIndx; ++i)
     (*MIB).addOperand(*argOpers[i]);
 
-  unsigned tt = F->getRegInfo().createVirtualRegister(RC);
-  if (invSrc) {
-    MIB = BuildMI(newMBB, dl, TII->get(notOpc), tt).addReg(t1);
-  }
-  else
-    tt = t1;
-
   unsigned t2 = F->getRegInfo().createVirtualRegister(RC);
   assert((argOpers[valArgIndx]->isReg() ||
           argOpers[valArgIndx]->isImm()) &&
@@ -11368,16 +11404,23 @@ X86TargetLowering::EmitAtomicBitwiseWithCustomInserter(MachineInstr *bInstr,
     MIB = BuildMI(newMBB, dl, TII->get(regOpc), t2);
   else
     MIB = BuildMI(newMBB, dl, TII->get(immOpc), t2);
-  MIB.addReg(tt);
+  MIB.addReg(t1);
   (*MIB).addOperand(*argOpers[valArgIndx]);
 
+  unsigned t3 = F->getRegInfo().createVirtualRegister(RC);
+  if (Invert) {
+    MIB = BuildMI(newMBB, dl, TII->get(notOpc), t3).addReg(t2);
+  }
+  else
+    t3 = t2;
+
   MIB = BuildMI(newMBB, dl, TII->get(TargetOpcode::COPY), EAXreg);
   MIB.addReg(t1);
 
   MIB = BuildMI(newMBB, dl, TII->get(CXchgOpc));
   for (int i=0; i <= lastAddrIndx; ++i)
     (*MIB).addOperand(*argOpers[i]);
-  MIB.addReg(t2);
+  MIB.addReg(t3);
   assert(bInstr->hasOneMemOperand() && "Unexpected number of memoperand");
   (*MIB).setMemRefs(bInstr->memoperands_begin(),
                     bInstr->memoperands_end());
@@ -11400,7 +11443,7 @@ X86TargetLowering::EmitAtomicBit6432WithCustomInserter(MachineInstr *bInstr,
                                                        unsigned regOpcH,
                                                        unsigned immOpcL,
                                                        unsigned immOpcH,
-                                                       bool invSrc) const {
+                                                       bool Invert) const {
   // For the atomic bitwise operator, we generate
   //   thisMBB (instructions are in pairs, except cmpxchg8b)
   //     ld t1,t2 = [bitinstr.addr]
@@ -11408,6 +11451,7 @@ X86TargetLowering::EmitAtomicBit6432WithCustomInserter(MachineInstr *bInstr,
   //     out1, out2 = phi (thisMBB, t1/t2) (newMBB, t3/t4)
   //     op  t5, t6 <- out1, out2, [bitinstr.val]
   //      (for SWAP, substitute:  mov t5, t6 <- [bitinstr.val])
+  //     neg t7, t8 < t5, t6  (if Invert)
   //     mov ECX, EBX <- t5, t6
   //     mov EAX, EDX <- t1, t2
   //     cmpxchg8b [bitinstr.addr]  [EAX, EDX, EBX, ECX implicit]
@@ -11491,16 +11535,9 @@ X86TargetLowering::EmitAtomicBit6432WithCustomInserter(MachineInstr *bInstr,
     .addReg(t2).addMBB(thisMBB).addReg(t4).addMBB(newMBB);
 
   // The subsequent operations should be using the destination registers of
-  //the PHI instructions.
-  if (invSrc) {
-    t1 = F->getRegInfo().createVirtualRegister(RC);
-    t2 = F->getRegInfo().createVirtualRegister(RC);
-    MIB = BuildMI(newMBB, dl, TII->get(NotOpc), t1).addReg(dest1Oper.getReg());
-    MIB = BuildMI(newMBB, dl, TII->get(NotOpc), t2).addReg(dest2Oper.getReg());
-  } else {
-    t1 = dest1Oper.getReg();
-    t2 = dest2Oper.getReg();
-  }
+  // the PHI instructions.
+  t1 = dest1Oper.getReg();
+  t2 = dest2Oper.getReg();
 
   int valArgIndx = lastAddrIndx + 1;
   assert((argOpers[valArgIndx]->isReg() ||
@@ -11527,15 +11564,26 @@ X86TargetLowering::EmitAtomicBit6432WithCustomInserter(MachineInstr *bInstr,
     MIB.addReg(t2);
   (*MIB).addOperand(*argOpers[valArgIndx + 1]);
 
+  unsigned t7, t8;
+  if (Invert) {
+    t7 = F->getRegInfo().createVirtualRegister(RC);
+    t8 = F->getRegInfo().createVirtualRegister(RC);
+    MIB = BuildMI(newMBB, dl, TII->get(NotOpc), t7).addReg(t5);
+    MIB = BuildMI(newMBB, dl, TII->get(NotOpc), t8).addReg(t6);
+  } else {
+    t7 = t5;
+    t8 = t6;
+  }
+
   MIB = BuildMI(newMBB, dl, TII->get(TargetOpcode::COPY), X86::EAX);
   MIB.addReg(t1);
   MIB = BuildMI(newMBB, dl, TII->get(TargetOpcode::COPY), X86::EDX);
   MIB.addReg(t2);
 
   MIB = BuildMI(newMBB, dl, TII->get(TargetOpcode::COPY), X86::EBX);
-  MIB.addReg(t5);
+  MIB.addReg(t7);
   MIB = BuildMI(newMBB, dl, TII->get(TargetOpcode::COPY), X86::ECX);
-  MIB.addReg(t6);
+  MIB.addReg(t8);
 
   MIB = BuildMI(newMBB, dl, TII->get(X86::LCMPXCHG8B));
   for (int i=0; i <= lastAddrIndx; ++i)
diff --git a/lib/Target/X86/X86ISelLowering.h b/lib/Target/X86/X86ISelLowering.h
index 4e0073365a7..09116e88af5 100644
--- a/lib/Target/X86/X86ISelLowering.h
+++ b/lib/Target/X86/X86ISelLowering.h
@@ -285,6 +285,8 @@ namespace llvm {
       UNPCKL,
       UNPCKH,
       VPERMILP,
+      VPERMV,
+      VPERMI,
       VPERM2X128,
       VBROADCAST,
 
@@ -855,7 +857,7 @@ namespace llvm {
                                                     unsigned notOpc,
                                                     unsigned EAXreg,
                                               const TargetRegisterClass *RC,
-                                                    bool invSrc = false) const;
+                                                    bool Invert = false) const;
 
     MachineBasicBlock *EmitAtomicBit6432WithCustomInserter(
                                                     MachineInstr *BInstr,
@@ -864,7 +866,7 @@ namespace llvm {
                                                     unsigned regOpcH,
                                                     unsigned immOpcL,
                                                     unsigned immOpcH,
-                                                    bool invSrc = false) const;
+                                                    bool Invert = false) const;
 
     /// Utility function to emit atomic min and max.  It takes the min/max
     /// instruction to expand, the associated basic block, and the associated
diff --git a/lib/Target/X86/X86InstrFragmentsSIMD.td b/lib/Target/X86/X86InstrFragmentsSIMD.td
index 041a64f336f..35801e43229 100644
--- a/lib/Target/X86/X86InstrFragmentsSIMD.td
+++ b/lib/Target/X86/X86InstrFragmentsSIMD.td
@@ -155,6 +155,8 @@ def X86Unpckl : SDNode<"X86ISD::UNPCKL", SDTShuff2Op>;
 def X86Unpckh : SDNode<"X86ISD::UNPCKH", SDTShuff2Op>;
 
 def X86VPermilp  : SDNode<"X86ISD::VPERMILP", SDTShuff2OpI>;
+def X86VPermv    : SDNode<"X86ISD::VPERMV",   SDTShuff2Op>;
+def X86VPermi    : SDNode<"X86ISD::VPERMI",   SDTShuff2OpI>;
 
 def X86VPerm2x128 : SDNode<"X86ISD::VPERM2X128", SDTShuff3OpI>;
 
diff --git a/lib/Target/X86/X86InstrInfo.cpp b/lib/Target/X86/X86InstrInfo.cpp
index 307c96b8c43..b12c1db0fe1 100644
--- a/lib/Target/X86/X86InstrInfo.cpp
+++ b/lib/Target/X86/X86InstrInfo.cpp
@@ -1049,9 +1049,9 @@ X86InstrInfo::X86InstrInfo(X86TargetMachine &tm)
     { X86::VPCMPGTWYrr,       X86::VPCMPGTWYrm,        TB_ALIGN_32 },
     { X86::VPERM2I128rr,      X86::VPERM2I128rm,       TB_ALIGN_32 },
     { X86::VPERMDYrr,         X86::VPERMDYrm,          TB_ALIGN_32 },
-    { X86::VPERMPDYrr,        X86::VPERMPDYrm,         TB_ALIGN_32 },
+    { X86::VPERMPDYri,        X86::VPERMPDYmi,         TB_ALIGN_32 },
     { X86::VPERMPSYrr,        X86::VPERMPSYrm,         TB_ALIGN_32 },
-    { X86::VPERMQYrr,         X86::VPERMQYrm,          TB_ALIGN_32 },
+    { X86::VPERMQYri,         X86::VPERMQYmi,          TB_ALIGN_32 },
     { X86::VPHADDDYrr,        X86::VPHADDDYrm,         TB_ALIGN_32 },
     { X86::VPHADDSWrr256,     X86::VPHADDSWrm256,      TB_ALIGN_32 },
     { X86::VPHADDWYrr,        X86::VPHADDWYrm,         TB_ALIGN_32 },
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 408ab16778d..65e3c1e19fa 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -6742,6 +6742,16 @@ let Predicates = [HasAVX] in {
   def : Pat<(v4f64 (X86Blendpd (v4f64 VR256:$src1), (v4f64 VR256:$src2),
                                (imm:$mask))),
             (VBLENDPDYrri VR256:$src2, VR256:$src1, imm:$mask)>;
+
+  def : Pat<(v8i16 (X86Blendpw (v8i16 VR128:$src1), (v8i16 VR128:$src2),
+                               (imm:$mask))),
+            (VPBLENDWrri VR128:$src2, VR128:$src1, imm:$mask)>;
+  def : Pat<(v4f32 (X86Blendps (v4f32 VR128:$src1), (v4f32 VR128:$src2),
+                               (imm:$mask))),
+            (VBLENDPSrri VR128:$src2, VR128:$src1, imm:$mask)>;
+  def : Pat<(v2f64 (X86Blendpd (v2f64 VR128:$src1), (v2f64 VR128:$src2),
+                               (imm:$mask))),
+            (VBLENDPDrri VR128:$src2, VR128:$src1, imm:$mask)>;
 }
 
 let Predicates = [HasAVX2] in {
@@ -6802,13 +6812,13 @@ let Predicates = [HasSSE41] in {
 
   def : Pat<(v8i16 (X86Blendpw (v8i16 VR128:$src1), (v8i16 VR128:$src2),
                                (imm:$mask))),
-            (VPBLENDWrri VR128:$src2, VR128:$src1, imm:$mask)>;
+            (PBLENDWrri VR128:$src2, VR128:$src1, imm:$mask)>;
   def : Pat<(v4f32 (X86Blendps (v4f32 VR128:$src1), (v4f32 VR128:$src2),
                                (imm:$mask))),
-            (VBLENDPSrri VR128:$src2, VR128:$src1, imm:$mask)>;
+            (BLENDPSrri VR128:$src2, VR128:$src1, imm:$mask)>;
   def : Pat<(v2f64 (X86Blendpd (v2f64 VR128:$src1), (v2f64 VR128:$src2),
                                (imm:$mask))),
-            (VBLENDPDrri VR128:$src2, VR128:$src1, imm:$mask)>;
+            (BLENDPDrri VR128:$src2, VR128:$src1, imm:$mask)>;
 
 }
 
@@ -7725,45 +7735,47 @@ def : Pat<(v4i32 (X86VBroadcast (loadi32 addr:$src))),
 //
 
 multiclass avx2_perm opc, string OpcodeStr, PatFrag mem_frag,
-                     Intrinsic Int> {
+                     ValueType OpVT> {
   def Yrr : AVX28I, VEX_4V;
+                   [(set VR256:$dst,
+                     (OpVT (X86VPermv VR256:$src1, VR256:$src2)))]>, VEX_4V;
   def Yrm : AVX28I,
+                   [(set VR256:$dst,
+                     (OpVT (X86VPermv VR256:$src1,
+                            (bitconvert (mem_frag addr:$src2)))))]>,
                    VEX_4V;
 }
 
-defm VPERMD : avx2_perm<0x36, "vpermd", memopv4i64, int_x86_avx2_permd>;
+defm VPERMD : avx2_perm<0x36, "vpermd", memopv4i64, v8i32>;
 let ExeDomain = SSEPackedSingle in
-defm VPERMPS : avx2_perm<0x16, "vpermps", memopv8f32, int_x86_avx2_permps>;
+defm VPERMPS : avx2_perm<0x16, "vpermps", memopv8f32, v8f32>;
 
 multiclass avx2_perm_imm opc, string OpcodeStr, PatFrag mem_frag,
-                         Intrinsic Int> {
-  def Yrr : AVX2AIi8 {
+  def Yri : AVX2AIi8, VEX;
-  def Yrm : AVX2AIi8, VEX;
+  def Ymi : AVX2AIi8,
-                     VEX;
+                     [(set VR256:$dst,
+                       (OpVT (X86VPermi (mem_frag addr:$src1),
+                              (i8 imm:$src2))))]>, VEX;
 }
 
-defm VPERMQ : avx2_perm_imm<0x00, "vpermq", memopv4i64, int_x86_avx2_permq>,
-                            VEX_W;
+defm VPERMQ : avx2_perm_imm<0x00, "vpermq", memopv4i64, v4i64>, VEX_W;
 let ExeDomain = SSEPackedDouble in
-defm VPERMPD : avx2_perm_imm<0x01, "vpermpd", memopv4f64, int_x86_avx2_permpd>,
-                             VEX_W;
+defm VPERMPD : avx2_perm_imm<0x01, "vpermpd", memopv4f64, v4f64>, VEX_W;
 
 //===----------------------------------------------------------------------===//
 // VPERM2I128 - Permute Floating-Point Values in 128-bit chunks
diff --git a/lib/Target/X86/X86Subtarget.cpp b/lib/Target/X86/X86Subtarget.cpp
index 452dd7eba32..ed1a40965aa 100644
--- a/lib/Target/X86/X86Subtarget.cpp
+++ b/lib/Target/X86/X86Subtarget.cpp
@@ -424,7 +424,9 @@ bool X86Subtarget::enablePostRAScheduler(
            CodeGenOpt::Level OptLevel,
            TargetSubtargetInfo::AntiDepBreakMode& Mode,
            RegClassVector& CriticalPathRCs) const {
-  Mode = TargetSubtargetInfo::ANTIDEP_CRITICAL;
+  //TODO: change back to ANTIDEP_CRITICAL when the
+  // X86 subtarget properly sets up post RA liveness.
+  Mode = TargetSubtargetInfo::ANTIDEP_NONE;
   CriticalPathRCs.clear();
   return PostRAScheduler && OptLevel >= CodeGenOpt::Default;
 }
diff --git a/lib/Target/X86/X86TargetMachine.cpp b/lib/Target/X86/X86TargetMachine.cpp
index f4b7a6277ad..89c388415bd 100644
--- a/lib/Target/X86/X86TargetMachine.cpp
+++ b/lib/Target/X86/X86TargetMachine.cpp
@@ -145,34 +145,34 @@ TargetPassConfig *X86TargetMachine::createPassConfig(PassManagerBase &PM) {
 
 bool X86PassConfig::addInstSelector() {
   // Install an instruction selector.
-  PM.add(createX86ISelDag(getX86TargetMachine(), getOptLevel()));
+  PM->add(createX86ISelDag(getX86TargetMachine(), getOptLevel()));
 
   // For 32-bit, prepend instructions to set the "global base reg" for PIC.
   if (!getX86Subtarget().is64Bit())
-    PM.add(createGlobalBaseRegPass());
+    PM->add(createGlobalBaseRegPass());
 
   return false;
 }
 
 bool X86PassConfig::addPreRegAlloc() {
-  PM.add(createX86MaxStackAlignmentHeuristicPass());
+  PM->add(createX86MaxStackAlignmentHeuristicPass());
   return false;  // -print-machineinstr shouldn't print after this.
 }
 
 bool X86PassConfig::addPostRegAlloc() {
-  PM.add(createX86FloatingPointStackifierPass());
+  PM->add(createX86FloatingPointStackifierPass());
   return true;  // -print-machineinstr should print after this.
 }
 
 bool X86PassConfig::addPreEmitPass() {
   bool ShouldPrint = false;
   if (getOptLevel() != CodeGenOpt::None && getX86Subtarget().hasSSE2()) {
-    PM.add(createExecutionDependencyFixPass(&X86::VR128RegClass));
+    PM->add(createExecutionDependencyFixPass(&X86::VR128RegClass));
     ShouldPrint = true;
   }
 
   if (getX86Subtarget().hasAVX() && UseVZeroUpper) {
-    PM.add(createX86IssueVZeroUpperPass());
+    PM->add(createX86IssueVZeroUpperPass());
     ShouldPrint = true;
   }
 
diff --git a/lib/Target/XCore/XCoreTargetMachine.cpp b/lib/Target/XCore/XCoreTargetMachine.cpp
index f65297e54a7..5afd5a1affc 100644
--- a/lib/Target/XCore/XCoreTargetMachine.cpp
+++ b/lib/Target/XCore/XCoreTargetMachine.cpp
@@ -55,7 +55,7 @@ TargetPassConfig *XCoreTargetMachine::createPassConfig(PassManagerBase &PM) {
 }
 
 bool XCorePassConfig::addInstSelector() {
-  PM.add(createXCoreISelDag(getXCoreTargetMachine(), getOptLevel()));
+  PM->add(createXCoreISelDag(getXCoreTargetMachine(), getOptLevel()));
   return false;
 }
 
diff --git a/lib/Transforms/IPO/Internalize.cpp b/lib/Transforms/IPO/Internalize.cpp
index cd29e7a7a7d..fb5869ede2b 100644
--- a/lib/Transforms/IPO/Internalize.cpp
+++ b/lib/Transforms/IPO/Internalize.cpp
@@ -123,6 +123,8 @@ bool InternalizePass::runOnModule(Module &M) {
   bool Changed = false;
 
   // Never internalize functions which code-gen might insert.
+  // FIXME: We should probably add this (and the __stack_chk_guard) via some
+  // type of call-back in CodeGen.
   ExternalNames.insert("__stack_chk_fail");
 
   // Mark all functions not in the api as internal.
diff --git a/lib/Transforms/IPO/PassManagerBuilder.cpp b/lib/Transforms/IPO/PassManagerBuilder.cpp
index a1b0a4580bf..43b4ab5efa4 100644
--- a/lib/Transforms/IPO/PassManagerBuilder.cpp
+++ b/lib/Transforms/IPO/PassManagerBuilder.cpp
@@ -35,6 +35,11 @@ using namespace llvm;
 static cl::opt
 RunVectorization("vectorize", cl::desc("Run vectorization passes"));
 
+static cl::opt
+UseGVNAfterVectorization("use-gvn-after-vectorization",
+  cl::init(false), cl::Hidden,
+  cl::desc("Run GVN instead of Early CSE after vectorization passes"));
+
 PassManagerBuilder::PassManagerBuilder() {
     OptLevel = 2;
     SizeLevel = 0;
@@ -182,8 +187,10 @@ void PassManagerBuilder::populateModulePassManager(PassManagerBase &MPM) {
   if (Vectorize) {
     MPM.add(createBBVectorizePass());
     MPM.add(createInstructionCombiningPass());
-    if (OptLevel > 1)
-      MPM.add(createGVNPass());                 // Remove redundancies
+    if (OptLevel > 1 && UseGVNAfterVectorization)
+      MPM.add(createGVNPass());                   // Remove redundancies
+    else
+      MPM.add(createEarlyCSEPass());              // Catch trivial redundancies
   }
 
   MPM.add(createAggressiveDCEPass());         // Delete dead instructions
diff --git a/lib/Transforms/Scalar/LoopStrengthReduce.cpp b/lib/Transforms/Scalar/LoopStrengthReduce.cpp
index d57ec22f44a..b085b006e4d 100644
--- a/lib/Transforms/Scalar/LoopStrengthReduce.cpp
+++ b/lib/Transforms/Scalar/LoopStrengthReduce.cpp
@@ -77,6 +77,12 @@
 #include 
 using namespace llvm;
 
+/// MaxIVUsers is an arbitrary threshold that provides an early opportunitiy for
+/// bail out. This threshold is far beyond the number of users that LSR can
+/// conceivably solve, so it should not affect generated code, but catches the
+/// worst cases before LSR burns too much compile time and stack space.
+static const unsigned MaxIVUsers = 200;
+
 // Temporary flag to cleanup congruent phis after LSR phi expansion.
 // It's currently disabled until we can determine whether it's truly useful or
 // not. The flag should be removed after the v3.0 release.
@@ -4102,7 +4108,7 @@ LSRInstance::HoistInsertPosition(BasicBlock::iterator IP,
       // Attempt to find an insert position in the middle of the block,
       // instead of at the end, so that it can be used for other expansions.
       if (IDom == Inst->getParent() &&
-          (!BetterPos || DT.dominates(BetterPos, Inst)))
+          (!BetterPos || !DT.dominates(Inst, BetterPos)))
         BetterPos = llvm::next(BasicBlock::iterator(Inst));
     }
     if (!AllDominate)
@@ -4519,6 +4525,17 @@ LSRInstance::LSRInstance(const TargetLowering *tli, Loop *l, Pass *P)
   // If there's no interesting work to be done, bail early.
   if (IU.empty()) return;
 
+  // If there's too much analysis to be done, bail early. We won't be able to
+  // model the problem anyway.
+  unsigned NumUsers = 0;
+  for (IVUsers::const_iterator UI = IU.begin(), E = IU.end(); UI != E; ++UI) {
+    if (++NumUsers > MaxIVUsers) {
+      DEBUG(dbgs() << "LSR skipping loop, too many IV Users in " << *L
+            << "\n");
+      return;
+    }
+  }
+
 #ifndef NDEBUG
   // All dominating loops must have preheaders, or SCEVExpander may not be able
   // to materialize an AddRecExpr whose Start is an outer AddRecExpr.
diff --git a/lib/Transforms/Scalar/LoopUnswitch.cpp b/lib/Transforms/Scalar/LoopUnswitch.cpp
index ee232687ffd..930980f528a 100644
--- a/lib/Transforms/Scalar/LoopUnswitch.cpp
+++ b/lib/Transforms/Scalar/LoopUnswitch.cpp
@@ -624,11 +624,10 @@ bool LoopUnswitch::IsTrivialUnswitchCondition(Value *Cond, Constant **Val,
 /// LoopCond == Val to simplify the loop.  If we decide that this is profitable,
 /// unswitch the loop, reprocess the pieces, then return true.
 bool LoopUnswitch::UnswitchIfProfitable(Value *LoopCond, Constant *Val) {
-
   Function *F = loopHeader->getParent();
-
   Constant *CondVal = 0;
   BasicBlock *ExitBlock = 0;
+
   if (IsTrivialUnswitchCondition(LoopCond, &CondVal, &ExitBlock)) {
     // If the condition is trivial, always unswitch. There is no code growth
     // for this case.
@@ -688,8 +687,8 @@ void LoopUnswitch::EmitPreheaderBranchOnCondition(Value *LIC, Constant *Val,
 
   // If either edge is critical, split it. This helps preserve LoopSimplify
   // form for enclosing loops.
-  SplitCriticalEdge(BI, 0, this);
-  SplitCriticalEdge(BI, 1, this);
+  SplitCriticalEdge(BI, 0, this, false, false, true);
+  SplitCriticalEdge(BI, 1, this, false, false, true);
 }
 
 /// UnswitchTrivialCondition - Given a loop that has a trivial unswitchable
diff --git a/lib/Transforms/Scalar/ObjCARC.cpp b/lib/Transforms/Scalar/ObjCARC.cpp
index 40b0b206168..7e3e69b78d3 100644
--- a/lib/Transforms/Scalar/ObjCARC.cpp
+++ b/lib/Transforms/Scalar/ObjCARC.cpp
@@ -162,6 +162,7 @@ namespace {
     IC_MoveWeak,            ///< objc_moveWeak (derived)
     IC_CopyWeak,            ///< objc_copyWeak (derived)
     IC_DestroyWeak,         ///< objc_destroyWeak (derived)
+    IC_StoreStrong,         ///< objc_storeStrong (derived)
     IC_CallOrUser,          ///< could call objc_release and/or "use" pointers
     IC_Call,                ///< could call objc_release
     IC_User,                ///< could "use" a pointer
@@ -262,6 +263,7 @@ static InstructionClass GetFunctionClass(const Function *F) {
               return StringSwitch(F->getName())
                      .Case("objc_storeWeak",             IC_StoreWeak)
                      .Case("objc_initWeak",              IC_InitWeak)
+                     .Case("objc_storeStrong",           IC_StoreStrong)
                      .Default(IC_CallOrUser);
             // Second argument is i8**.
             if (PointerType *Pte1 = dyn_cast(ETy1))
@@ -618,22 +620,35 @@ static bool DoesObjCBlockEscape(const Value *BlockPtr) {
       const User *UUser = *UI;
       // Special - Use by a call (callee or argument) is not considered
       // to be an escape.
-      if (isa(UUser) || isa(UUser))
-        continue;
-      // Use by an instruction which copies the value is an escape if the
-      // result is an escape.
-      if (isa(UUser) || isa(UUser) ||
-          isa(UUser) || isa(UUser)) {
-        Worklist.push_back(UUser);
+      switch (GetBasicInstructionClass(UUser)) {
+      case IC_StoreWeak:
+      case IC_InitWeak:
+      case IC_StoreStrong:
+      case IC_Autorelease:
+      case IC_AutoreleaseRV:
+        // These special functions make copies of their pointer arguments.
+        return true;
+      case IC_User:
+      case IC_None:
+        // Use by an instruction which copies the value is an escape if the
+        // result is an escape.
+        if (isa(UUser) || isa(UUser) ||
+            isa(UUser) || isa(UUser)) {
+          Worklist.push_back(UUser);
+          continue;
+        }
+        // Use by a load is not an escape.
+        if (isa(UUser))
+          continue;
+        // Use by a store is not an escape if the use is the address.
+        if (const StoreInst *SI = dyn_cast(UUser))
+          if (V != SI->getValueOperand())
+            continue;
+        break;
+      default:
+        // Regular calls and other stuff are not considered escapes.
         continue;
       }
-      // Use by a load is not an escape.
-      if (isa(UUser))
-        continue;
-      // Use by a store is not an escape if the use is the address.
-      if (const StoreInst *SI = dyn_cast(UUser))
-        if (V != SI->getValueOperand())
-          continue;
       // Otherwise, conservatively assume an escape.
       return true;
     }
@@ -883,7 +898,7 @@ bool ObjCARCExpand::runOnFunction(Function &F) {
       // These calls return their argument verbatim, as a low-level
       // optimization. However, this makes high-level optimizations
       // harder. Undo any uses of this optimization that the front-end
-      // emitted here. We'll redo them in a later pass.
+      // emitted here. We'll redo them in the contract pass.
       Changed = true;
       Inst->replaceAllUsesWith(cast(Inst)->getArgOperand(0));
       break;
@@ -997,7 +1012,11 @@ bool ObjCARCAPElim::runOnModule(Module &M) {
     return false;
 
   // Find the llvm.global_ctors variable, as the first step in
-  // identifying the global constructors.
+  // identifying the global constructors. In theory, unnecessary autorelease
+  // pools could occur anywhere, but in practice it's pretty rare. Global
+  // ctors are a place where autorelease pools get inserted automatically,
+  // so it's pretty common for them to be unnecessary, and it's pretty
+  // profitable to eliminate them.
   GlobalVariable *GV = M.getGlobalVariable("llvm.global_ctors");
   if (!GV)
     return false;
@@ -2263,6 +2282,7 @@ void ObjCARCOpt::OptimizeIndividualCalls(Function &F) {
     case IC_DestroyWeak: {
       CallInst *CI = cast(Inst);
       if (isNullOrUndef(CI->getArgOperand(0))) {
+        Changed = true;
         Type *Ty = CI->getArgOperand(0)->getType();
         new StoreInst(UndefValue::get(cast(Ty)->getElementType()),
                       Constant::getNullValue(Ty),
@@ -2278,6 +2298,7 @@ void ObjCARCOpt::OptimizeIndividualCalls(Function &F) {
       CallInst *CI = cast(Inst);
       if (isNullOrUndef(CI->getArgOperand(0)) ||
           isNullOrUndef(CI->getArgOperand(1))) {
+        Changed = true;
         Type *Ty = CI->getArgOperand(0)->getType();
         new StoreInst(UndefValue::get(cast(Ty)->getElementType()),
                       Constant::getNullValue(Ty),
@@ -3165,6 +3186,8 @@ void ObjCARCOpt::MoveCalls(Value *Arg,
   }
 }
 
+/// PerformCodePlacement - Identify pairings between the retains and releases,
+/// and delete and/or move them.
 bool
 ObjCARCOpt::PerformCodePlacement(DenseMap
                                    &BBStates,
@@ -3178,6 +3201,7 @@ ObjCARCOpt::PerformCodePlacement(DenseMap
   SmallVector NewReleases;
   SmallVector DeadInsts;
 
+  // Visit each retain.
   for (MapVector::const_iterator I = Retains.begin(),
        E = Retains.end(); I != E; ++I) {
     Value *V = I->first;
@@ -3651,6 +3675,7 @@ bool ObjCARCOpt::doInitialization(Module &M) {
   if (!EnableARCOpts)
     return false;
 
+  // If nothing in the Module uses ARC, don't do anything.
   Run = ModuleHasARC(M);
   if (!Run)
     return false;
@@ -3985,6 +4010,7 @@ void ObjCARCContract::ContractRelease(Instruction *Release,
 }
 
 bool ObjCARCContract::doInitialization(Module &M) {
+  // If nothing in the Module uses ARC, don't do anything.
   Run = ModuleHasARC(M);
   if (!Run)
     return false;
@@ -4060,6 +4086,7 @@ bool ObjCARCContract::runOnFunction(Function &F) {
       --BBI;
       while (isNoopInstruction(BBI)) --BBI;
       if (&*BBI == GetObjCArg(Inst)) {
+        Changed = true;
         InlineAsm *IA =
           InlineAsm::get(FunctionType::get(Type::getVoidTy(Inst->getContext()),
                                            /*isVarArg=*/false),
@@ -4109,6 +4136,13 @@ bool ObjCARCContract::runOnFunction(Function &F) {
         Use &U = UI.getUse();
         unsigned OperandNo = UI.getOperandNo();
         ++UI; // Increment UI now, because we may unlink its element.
+
+        // If the call's return value dominates a use of the call's argument
+        // value, rewrite the use to use the return value. We check for
+        // reachability here because an unreachable call is considered to
+        // trivially dominate itself, which would lead us to rewriting its
+        // argument in terms of its return value, which would lead to
+        // infinite loops in GetObjCArg.
         if (DT->isReachableFromEntry(U) &&
             DT->dominates(Inst, U)) {
           Changed = true;
@@ -4123,6 +4157,9 @@ bool ObjCARCContract::runOnFunction(Function &F) {
             if (Replacement->getType() != UseTy)
               Replacement = new BitCastInst(Replacement, UseTy, "",
                                             &BB->back());
+            // While we're here, rewrite all edges for this PHI, rather
+            // than just one use at a time, to minimize the number of
+            // bitcasts we emit.
             for (unsigned i = 0, e = PHI->getNumIncomingValues();
                  i != e; ++i)
               if (PHI->getIncomingBlock(i) == BB) {
diff --git a/lib/Transforms/Scalar/Reassociate.cpp b/lib/Transforms/Scalar/Reassociate.cpp
index cb408a137ea..5de00d1ab5e 100644
--- a/lib/Transforms/Scalar/Reassociate.cpp
+++ b/lib/Transforms/Scalar/Reassociate.cpp
@@ -559,7 +559,8 @@ static unsigned FindInOperandList(SmallVectorImpl &Ops, unsigned i,
 
 /// EmitAddTreeOfValues - Emit a tree of add instructions, summing Ops together
 /// and returning the result.  Insert the tree before I.
-static Value *EmitAddTreeOfValues(Instruction *I, SmallVectorImpl &Ops){
+static Value *EmitAddTreeOfValues(Instruction *I,
+                                  SmallVectorImpl &Ops){
   if (Ops.size() == 1) return Ops.back();
   
   Value *V1 = Ops.back();
@@ -833,7 +834,7 @@ Value *Reassociate::OptimizeAdd(Instruction *I,
     // from an expression will drop a use of maxocc, and this can cause 
     // RemoveFactorFromExpression on successive values to behave differently.
     Instruction *DummyInst = BinaryOperator::CreateAdd(MaxOccVal, MaxOccVal);
-    SmallVector NewMulOps;
+    SmallVector NewMulOps;
     for (unsigned i = 0; i != Ops.size(); ++i) {
       // Only try to remove factors from expressions we're allowed to.
       BinaryOperator *BOp = dyn_cast(Ops[i].Op);
diff --git a/lib/Transforms/Scalar/SimplifyLibCalls.cpp b/lib/Transforms/Scalar/SimplifyLibCalls.cpp
index 9c49ec1c84d..f7b69411b1d 100644
--- a/lib/Transforms/Scalar/SimplifyLibCalls.cpp
+++ b/lib/Transforms/Scalar/SimplifyLibCalls.cpp
@@ -1583,21 +1583,16 @@ void SimplifyLibCalls::InitOptimizations() {
   Optimizations["llvm.exp2.f64"] = &Exp2;
   Optimizations["llvm.exp2.f32"] = &Exp2;
 
-#ifdef HAVE_FLOORF
-  Optimizations["floor"] = &UnaryDoubleFP;
-#endif
-#ifdef HAVE_CEILF
-  Optimizations["ceil"] = &UnaryDoubleFP;
-#endif
-#ifdef HAVE_ROUNDF
-  Optimizations["round"] = &UnaryDoubleFP;
-#endif
-#ifdef HAVE_RINTF
-  Optimizations["rint"] = &UnaryDoubleFP;
-#endif
-#ifdef HAVE_NEARBYINTF
-  Optimizations["nearbyint"] = &UnaryDoubleFP;
-#endif
+  if (TLI->has(LibFunc::floor) && TLI->has(LibFunc::floorf))
+    Optimizations["floor"] = &UnaryDoubleFP;
+  if (TLI->has(LibFunc::ceil) && TLI->has(LibFunc::ceilf))
+    Optimizations["ceil"] = &UnaryDoubleFP;
+  if (TLI->has(LibFunc::round) && TLI->has(LibFunc::roundf))
+    Optimizations["round"] = &UnaryDoubleFP;
+  if (TLI->has(LibFunc::rint) && TLI->has(LibFunc::rintf))
+    Optimizations["rint"] = &UnaryDoubleFP;
+  if (TLI->has(LibFunc::nearbyint) && TLI->has(LibFunc::nearbyintf))
+    Optimizations["nearbyint"] = &UnaryDoubleFP;
 
   // Integer Optimizations
   Optimizations["ffs"] = &FFS;
diff --git a/lib/Transforms/Utils/BreakCriticalEdges.cpp b/lib/Transforms/Utils/BreakCriticalEdges.cpp
index f752d7981c5..2a8e9b834e3 100644
--- a/lib/Transforms/Utils/BreakCriticalEdges.cpp
+++ b/lib/Transforms/Utils/BreakCriticalEdges.cpp
@@ -117,33 +117,38 @@ bool llvm::isCriticalEdge(const TerminatorInst *TI, unsigned SuccNum,
   return false;
 }
 
-/// CreatePHIsForSplitLoopExit - When a loop exit edge is split, LCSSA form
+/// createPHIsForSplitLoopExit - When a loop exit edge is split, LCSSA form
 /// may require new PHIs in the new exit block. This function inserts the
-/// new PHIs, as needed.  Preds is a list of preds inside the loop, SplitBB
+/// new PHIs, as needed. Preds is a list of preds inside the loop, SplitBB
 /// is the new loop exit block, and DestBB is the old loop exit, now the
 /// successor of SplitBB.
-static void CreatePHIsForSplitLoopExit(SmallVectorImpl &Preds,
+static void createPHIsForSplitLoopExit(SmallVectorImpl &Preds,
                                        BasicBlock *SplitBB,
                                        BasicBlock *DestBB) {
   // SplitBB shouldn't have anything non-trivial in it yet.
-  assert(SplitBB->getFirstNonPHI() == SplitBB->getTerminator() &&
-         "SplitBB has non-PHI nodes!");
+  assert((SplitBB->getFirstNonPHI() == SplitBB->getTerminator() ||
+          SplitBB->isLandingPad()) && "SplitBB has non-PHI nodes!");
 
-  // For each PHI in the destination block...
+  // For each PHI in the destination block.
   for (BasicBlock::iterator I = DestBB->begin();
        PHINode *PN = dyn_cast(I); ++I) {
     unsigned Idx = PN->getBasicBlockIndex(SplitBB);
     Value *V = PN->getIncomingValue(Idx);
+
     // If the input is a PHI which already satisfies LCSSA, don't create
     // a new one.
     if (const PHINode *VP = dyn_cast(V))
       if (VP->getParent() == SplitBB)
         continue;
+
     // Otherwise a new PHI is needed. Create one and populate it.
-    PHINode *NewPN = PHINode::Create(PN->getType(), Preds.size(), "split",
-                                     SplitBB->getTerminator());
+    PHINode *NewPN =
+      PHINode::Create(PN->getType(), Preds.size(), "split",
+                      SplitBB->isLandingPad() ?
+                      SplitBB->begin() : SplitBB->getTerminator());
     for (unsigned i = 0, e = Preds.size(); i != e; ++i)
       NewPN->addIncoming(V, Preds[i]);
+
     // Update the original PHI.
     PN->setIncomingValue(Idx, NewPN);
   }
@@ -168,7 +173,8 @@ static void CreatePHIsForSplitLoopExit(SmallVectorImpl &Preds,
 ///
 BasicBlock *llvm::SplitCriticalEdge(TerminatorInst *TI, unsigned SuccNum,
                                     Pass *P, bool MergeIdenticalEdges,
-                                    bool DontDeleteUselessPhis) {
+                                    bool DontDeleteUselessPhis,
+                                    bool SplitLandingPads) {
   if (!isCriticalEdge(TI, SuccNum, MergeIdenticalEdges)) return 0;
 
   assert(!isa(TI) &&
@@ -338,7 +344,7 @@ BasicBlock *llvm::SplitCriticalEdge(TerminatorInst *TI, unsigned SuccNum,
         if (P->mustPreserveAnalysisID(LCSSAID)) {
           SmallVector OrigPred;
           OrigPred.push_back(TIBB);
-          CreatePHIsForSplitLoopExit(OrigPred, NewBB, DestBB);
+          createPHIsForSplitLoopExit(OrigPred, NewBB, DestBB);
         }
 
         // For each unique exit block...
@@ -371,10 +377,19 @@ BasicBlock *llvm::SplitCriticalEdge(TerminatorInst *TI, unsigned SuccNum,
           // getUniqueExitBlocks above because that depends on LoopSimplify
           // form, which we're in the process of restoring!
           if (!Preds.empty() && HasPredOutsideOfLoop) {
-            BasicBlock *NewExitBB =
-              SplitBlockPredecessors(Exit, Preds, "split", P);
-            if (P->mustPreserveAnalysisID(LCSSAID))
-              CreatePHIsForSplitLoopExit(Preds, NewExitBB, Exit);
+            if (!Exit->isLandingPad()) {
+              BasicBlock *NewExitBB =
+                SplitBlockPredecessors(Exit, Preds, "split", P);
+              if (P->mustPreserveAnalysisID(LCSSAID))
+                createPHIsForSplitLoopExit(Preds, NewExitBB, Exit);
+            } else if (SplitLandingPads) {
+              SmallVector NewBBs;
+              SplitLandingPadPredecessors(Exit, Preds,
+                                          ".split1", ".split2",
+                                          P, NewBBs);
+              if (P->mustPreserveAnalysisID(LCSSAID))
+                createPHIsForSplitLoopExit(Preds, NewBBs[0], Exit);
+            }
           }
         }
       }
diff --git a/lib/Transforms/Vectorize/BBVectorize.cpp b/lib/Transforms/Vectorize/BBVectorize.cpp
index 286b54f2f06..9d62306dce5 100644
--- a/lib/Transforms/Vectorize/BBVectorize.cpp
+++ b/lib/Transforms/Vectorize/BBVectorize.cpp
@@ -83,6 +83,10 @@ static cl::opt
 NoFloats("bb-vectorize-no-floats", cl::init(false), cl::Hidden,
   cl::desc("Don't try to vectorize floating-point values"));
 
+static cl::opt
+NoPointers("bb-vectorize-no-pointers", cl::init(false), cl::Hidden,
+  cl::desc("Don't try to vectorize pointer values"));
+
 static cl::opt
 NoCasts("bb-vectorize-no-casts", cl::init(false), cl::Hidden,
   cl::desc("Don't try to vectorize casting (conversion) operations"));
@@ -95,6 +99,14 @@ static cl::opt
 NoFMA("bb-vectorize-no-fma", cl::init(false), cl::Hidden,
   cl::desc("Don't try to vectorize the fused-multiply-add intrinsic"));
 
+static cl::opt
+NoSelect("bb-vectorize-no-select", cl::init(false), cl::Hidden,
+  cl::desc("Don't try to vectorize select instructions"));
+
+static cl::opt
+NoGEP("bb-vectorize-no-gep", cl::init(false), cl::Hidden,
+  cl::desc("Don't try to vectorize getelementptr instructions"));
+
 static cl::opt
 NoMemOps("bb-vectorize-no-mem-ops", cl::init(false), cl::Hidden,
   cl::desc("Don't try to vectorize loads and stores"));
@@ -546,11 +558,21 @@ namespace {
         return false;
 
       Type *SrcTy = C->getSrcTy();
-      if (!SrcTy->isSingleValueType() || SrcTy->isPointerTy())
+      if (!SrcTy->isSingleValueType())
         return false;
 
       Type *DestTy = C->getDestTy();
-      if (!DestTy->isSingleValueType() || DestTy->isPointerTy())
+      if (!DestTy->isSingleValueType())
+        return false;
+    } else if (isa(I)) {
+      if (!Config.VectorizeSelect)
+        return false;
+    } else if (GetElementPtrInst *G = dyn_cast(I)) {
+      if (!Config.VectorizeGEP)
+        return false;
+
+      // Currently, vector GEPs exist only with one index.
+      if (G->getNumIndices() != 1)
         return false;
     } else if (!(I->isBinaryOp() || isa(I) ||
         isa(I) || isa(I))) {
@@ -590,6 +612,11 @@ namespace {
         && (T1->isFPOrFPVectorTy() || T2->isFPOrFPVectorTy()))
       return false;
 
+    if ((!Config.VectorizePointers || TD == 0) &&
+        (T1->getScalarType()->isPointerTy() ||
+         T2->getScalarType()->isPointerTy()))
+      return false;
+
     if (T1->getPrimitiveSizeInBits() > Config.VectorBits/2 ||
         T2->getPrimitiveSizeInBits() > Config.VectorBits/2)
       return false;
@@ -828,16 +855,33 @@ namespace {
                       std::vector &PairableInsts,
                       std::multimap &ConnectedPairs,
                       ValuePair P) {
+    StoreInst *SI, *SJ;
+
     // For each possible pairing for this variable, look at the uses of
     // the first value...
     for (Value::use_iterator I = P.first->use_begin(),
          E = P.first->use_end(); I != E; ++I) {
+      if (isa(*I)) {
+        // A pair cannot be connected to a load because the load only takes one
+        // operand (the address) and it is a scalar even after vectorization.
+        continue;
+      } else if ((SI = dyn_cast(*I)) &&
+                 P.first == SI->getPointerOperand()) {
+        // Similarly, a pair cannot be connected to a store through its
+        // pointer operand.
+        continue;
+      }
+
       VPIteratorPair IPairRange = CandidatePairs.equal_range(*I);
 
       // For each use of the first variable, look for uses of the second
       // variable...
       for (Value::use_iterator J = P.second->use_begin(),
            E2 = P.second->use_end(); J != E2; ++J) {
+        if ((SJ = dyn_cast(*J)) &&
+            P.second == SJ->getPointerOperand())
+          continue;
+
         VPIteratorPair JPairRange = CandidatePairs.equal_range(*J);
 
         // Look for :
@@ -853,6 +897,10 @@ namespace {
       // Look for cases where just the first value in the pair is used by
       // both members of another pair (splatting).
       for (Value::use_iterator J = P.first->use_begin(); J != E; ++J) {
+        if ((SJ = dyn_cast(*J)) &&
+            P.first == SJ->getPointerOperand())
+          continue;
+
         if (isSecondInIteratorPair(*J, IPairRange))
           ConnectedPairs.insert(VPPair(P, ValuePair(*I, *J)));
       }
@@ -863,9 +911,19 @@ namespace {
     // both members of another pair (splatting).
     for (Value::use_iterator I = P.second->use_begin(),
          E = P.second->use_end(); I != E; ++I) {
+      if (isa(*I))
+        continue;
+      else if ((SI = dyn_cast(*I)) &&
+               P.second == SI->getPointerOperand())
+        continue;
+
       VPIteratorPair IPairRange = CandidatePairs.equal_range(*I);
 
       for (Value::use_iterator J = P.second->use_begin(); J != E; ++J) {
+        if ((SJ = dyn_cast(*J)) &&
+            P.second == SJ->getPointerOperand())
+          continue;
+
         if (isSecondInIteratorPair(*J, IPairRange))
           ConnectedPairs.insert(VPPair(P, ValuePair(*I, *J)));
       }
@@ -1891,9 +1949,12 @@ VectorizeConfig::VectorizeConfig() {
   VectorBits = ::VectorBits;
   VectorizeInts = !::NoInts;
   VectorizeFloats = !::NoFloats;
+  VectorizePointers = !::NoPointers;
   VectorizeCasts = !::NoCasts;
   VectorizeMath = !::NoMath;
   VectorizeFMA = !::NoFMA;
+  VectorizeSelect = !::NoSelect;
+  VectorizeGEP = !::NoGEP;
   VectorizeMemOps = !::NoMemOps;
   AlignedOnly = ::AlignedOnly;
   ReqChainDepth= ::ReqChainDepth;
diff --git a/lib/VMCore/AutoUpgrade.cpp b/lib/VMCore/AutoUpgrade.cpp
index ea3d4ba2243..2e16372fb1c 100644
--- a/lib/VMCore/AutoUpgrade.cpp
+++ b/lib/VMCore/AutoUpgrade.cpp
@@ -18,9 +18,6 @@
 #include "llvm/LLVMContext.h"
 #include "llvm/Module.h"
 #include "llvm/IntrinsicInst.h"
-#include "llvm/ADT/DenseMap.h"
-#include "llvm/ADT/SmallPtrSet.h"
-#include "llvm/ADT/SmallVector.h"
 #include "llvm/Support/CallSite.h"
 #include "llvm/Support/CFG.h"
 #include "llvm/Support/ErrorHandling.h"
@@ -59,7 +56,8 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
     if (Name.startswith("x86.sse2.pcmpeq.") ||
         Name.startswith("x86.sse2.pcmpgt.") ||
         Name.startswith("x86.avx2.pcmpeq.") ||
-        Name.startswith("x86.avx2.pcmpgt.")) {
+        Name.startswith("x86.avx2.pcmpgt.") ||
+        Name.startswith("x86.avx.vpermil.")) {
       NewFn = 0;
       return true;
     }
@@ -121,7 +119,42 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
       // need to sign extend since icmp returns vector of i1
       Rep = Builder.CreateSExt(Rep, CI->getType(), "");
     } else {
-      llvm_unreachable("Unknown function for CallInst upgrade.");
+      bool PD128 = false, PD256 = false, PS128 = false, PS256 = false;
+      if (Name.startswith("llvm.x86.avx.vpermil.pd.256"))
+        PD256 = true;
+      else if (Name.startswith("llvm.x86.avx.vpermil.pd"))
+        PD128 = true;
+      else if (Name.startswith("llvm.x86.avx.vpermil.ps.256"))
+        PS256 = true;
+      else if (Name.startswith("llvm.x86.avx.vpermil.ps"))
+        PS128 = true;
+
+      if (PD256 || PD128 || PS256 || PS128) {
+        Value *Op0 = CI->getArgOperand(0);
+        unsigned Imm = cast(CI->getArgOperand(1))->getZExtValue();
+        SmallVector Idxs;
+
+        if (PD128)
+          for (unsigned i = 0; i != 2; ++i)
+            Idxs.push_back(Builder.getInt32((Imm >> i) & 0x1));
+        else if (PD256)
+          for (unsigned l = 0; l != 4; l+=2)
+            for (unsigned i = 0; i != 2; ++i)
+              Idxs.push_back(Builder.getInt32(((Imm >> (l+i)) & 0x1) + l));
+        else if (PS128)
+          for (unsigned i = 0; i != 4; ++i)
+            Idxs.push_back(Builder.getInt32((Imm >> (2 * i)) & 0x3));
+        else if (PS256)
+          for (unsigned l = 0; l != 8; l+=4)
+            for (unsigned i = 0; i != 4; ++i)
+              Idxs.push_back(Builder.getInt32(((Imm >> (2 * i)) & 0x3) + l));
+        else
+          llvm_unreachable("Unexpected function");
+
+        Rep = Builder.CreateShuffleVector(Op0, Op0, ConstantVector::get(Idxs));
+      } else {
+        llvm_unreachable("Unknown function for CallInst upgrade.");
+      }
     }
 
     CI->replaceAllUsesWith(Rep);
diff --git a/lib/VMCore/Instructions.cpp b/lib/VMCore/Instructions.cpp
index 8db6ac9a33f..6c5db328764 100644
--- a/lib/VMCore/Instructions.cpp
+++ b/lib/VMCore/Instructions.cpp
@@ -2002,6 +2002,23 @@ bool BinaryOperator::isExact() const {
   return cast(this)->isExact();
 }
 
+//===----------------------------------------------------------------------===//
+//                             FPMathOperator Class
+//===----------------------------------------------------------------------===//
+
+/// getFPAccuracy - Get the maximum error permitted by this operation in ULPs.
+/// An accuracy of 0.0 means that the operation should be performed with the
+/// default precision.
+float FPMathOperator::getFPAccuracy() const {
+  const MDNode *MD =
+    cast(this)->getMetadata(LLVMContext::MD_fpmath);
+  if (!MD)
+    return 0.0;
+  ConstantFP *Accuracy = cast(MD->getOperand(0));
+  return Accuracy->getValueAPF().convertToFloat();
+}
+
+
 //===----------------------------------------------------------------------===//
 //                                CastInst Class
 //===----------------------------------------------------------------------===//
diff --git a/lib/VMCore/LLVMContext.cpp b/lib/VMCore/LLVMContext.cpp
index 68c56212bc6..f07f0b39392 100644
--- a/lib/VMCore/LLVMContext.cpp
+++ b/lib/VMCore/LLVMContext.cpp
@@ -44,9 +44,9 @@ LLVMContext::LLVMContext() : pImpl(new LLVMContextImpl(*this)) {
   unsigned ProfID = getMDKindID("prof");
   assert(ProfID == MD_prof && "prof kind id drifted"); (void)ProfID;
 
-  // Create the 'fpaccuracy' metadata kind.
-  unsigned FPAccuracyID = getMDKindID("fpaccuracy");
-  assert(FPAccuracyID == MD_fpaccuracy && "fpaccuracy kind id drifted");
+  // Create the 'fpmath' metadata kind.
+  unsigned FPAccuracyID = getMDKindID("fpmath");
+  assert(FPAccuracyID == MD_fpmath && "fpmath kind id drifted");
   (void)FPAccuracyID;
 
   // Create the 'range' metadata kind.
diff --git a/lib/VMCore/Module.cpp b/lib/VMCore/Module.cpp
index e8bc6dbe970..3c67191e09c 100644
--- a/lib/VMCore/Module.cpp
+++ b/lib/VMCore/Module.cpp
@@ -434,7 +434,7 @@ bool Module::MaterializeAllPermanently(std::string *ErrInfo) {
 //
 
 
-// dropAllReferences() - This function causes all the subelementss to "let go"
+// dropAllReferences() - This function causes all the subelements to "let go"
 // of all references that they are maintaining.  This allows one to 'delete' a
 // whole module at a time, even though there may be circular references... first
 // all references are dropped, and all use counts go to zero.  Then everything
diff --git a/lib/VMCore/Verifier.cpp b/lib/VMCore/Verifier.cpp
index 96492e44d56..47baef3e29d 100644
--- a/lib/VMCore/Verifier.cpp
+++ b/lib/VMCore/Verifier.cpp
@@ -1653,16 +1653,18 @@ void Verifier::visitInstruction(Instruction &I) {
     }
   }
 
-  if (MDNode *MD = I.getMetadata(LLVMContext::MD_fpaccuracy)) {
+  if (MDNode *MD = I.getMetadata(LLVMContext::MD_fpmath)) {
     Assert1(I.getType()->isFPOrFPVectorTy(),
-            "fpaccuracy requires a floating point result!", &I);
-    Assert1(MD->getNumOperands() == 1, "fpaccuracy takes one operand!", &I);
-    ConstantFP *Op = dyn_cast_or_null(MD->getOperand(0));
-    Assert1(Op, "fpaccuracy ULPs not a floating point number!", &I);
-    APFloat ULPs = Op->getValueAPF();
-    Assert1(ULPs.isNormal() || ULPs.isZero(),
-            "fpaccuracy ULPs not a normal number!", &I);
-    Assert1(!ULPs.isNegative(), "fpaccuracy ULPs is negative!", &I);
+            "fpmath requires a floating point result!", &I);
+    Assert1(MD->getNumOperands() == 1, "fpmath takes one operand!", &I);
+    Value *Op0 = MD->getOperand(0);
+    if (ConstantFP *CFP0 = dyn_cast_or_null(Op0)) {
+      APFloat Accuracy = CFP0->getValueAPF();
+      Assert1(Accuracy.isNormal() && !Accuracy.isNegative(),
+              "fpmath accuracy not a positive number!", &I);
+    } else {
+      Assert1(false, "invalid fpmath accuracy!", &I);
+    }
   }
 
   MDNode *MD = I.getMetadata(LLVMContext::MD_range);
diff --git a/test/Analysis/ScalarEvolution/nsw-offset.ll b/test/Analysis/ScalarEvolution/nsw-offset.ll
index a919319bdcd..8969a5ad4ce 100644
--- a/test/Analysis/ScalarEvolution/nsw-offset.ll
+++ b/test/Analysis/ScalarEvolution/nsw-offset.ll
@@ -23,7 +23,7 @@ bb:                                               ; preds = %bb.nph, %bb1
   %1 = sext i32 %i.01 to i64                      ;  [#uses=1]
 
 ; CHECK: %2 = getelementptr inbounds double* %d, i64 %1
-; CHECK: -->  {%d,+,16}<%bb>
+; CHECK: -->  {%d,+,16}<%bb>
   %2 = getelementptr inbounds double* %d, i64 %1  ;  [#uses=1]
 
   %3 = load double* %2, align 8                   ;  [#uses=1]
@@ -37,7 +37,7 @@ bb:                                               ; preds = %bb.nph, %bb1
   %8 = sext i32 %7 to i64                         ;  [#uses=1]
 
 ; CHECK: %9 = getelementptr inbounds double* %q, i64 %8
-; CHECK: {(8 + %q),+,16}<%bb>
+; CHECK: {(8 + %q),+,16}<%bb>
   %9 = getelementptr inbounds double* %q, i64 %8  ;  [#uses=1]
 
 ; Artificially repeat the above three instructions, this time using
@@ -49,7 +49,7 @@ bb:                                               ; preds = %bb.nph, %bb1
   %t8 = sext i32 %t7 to i64                         ;  [#uses=1]
 
 ; CHECK: %t9 = getelementptr inbounds double* %q, i64 %t8
-; CHECK: {(8 + %q),+,16}<%bb>
+; CHECK: {(8 + %q),+,16}<%bb>
   %t9 = getelementptr inbounds double* %q, i64 %t8  ;  [#uses=1]
 
   %10 = load double* %9, align 8                  ;  [#uses=1]
diff --git a/test/Analysis/ScalarEvolution/nsw.ll b/test/Analysis/ScalarEvolution/nsw.ll
index 288b6facd94..659cf4f8da9 100644
--- a/test/Analysis/ScalarEvolution/nsw.ll
+++ b/test/Analysis/ScalarEvolution/nsw.ll
@@ -92,10 +92,10 @@ for.body.i.i:                                     ; preds = %entry, %for.body.i.
 ; CHECK: {1,+,1}<%for.body.i.i>
   %ptrincdec.i.i = getelementptr inbounds i32* %begin, i64 %tmp
 ; CHECK: %ptrincdec.i.i =
-; CHECK: {(4 + %begin),+,4}<%for.body.i.i>
+; CHECK: {(4 + %begin),+,4}<%for.body.i.i>
   %__first.addr.08.i.i = getelementptr inbounds i32* %begin, i64 %indvar.i.i
 ; CHECK: %__first.addr.08.i.i
-; CHECK: {%begin,+,4}<%for.body.i.i>
+; CHECK: {%begin,+,4}<%for.body.i.i>
   store i32 0, i32* %__first.addr.08.i.i, align 4
   %cmp.i.i = icmp eq i32* %ptrincdec.i.i, %end
   br i1 %cmp.i.i, label %_ZSt4fillIPiiEvT_S1_RKT0_.exit, label %for.body.i.i
diff --git a/test/CodeGen/ARM/2011-03-23-PeepholeBug.ll b/test/CodeGen/ARM/2011-03-23-PeepholeBug.ll
index 7c9af6f5e59..0fe88bd0ed7 100644
--- a/test/CodeGen/ARM/2011-03-23-PeepholeBug.ll
+++ b/test/CodeGen/ARM/2011-03-23-PeepholeBug.ll
@@ -26,7 +26,7 @@ bb2:                                              ; preds = %bb1, %entry
 ; CHECK: bb2
 ; CHECK: subs [[REG:r[0-9]+]], #1
 ; CHECK: cmp [[REG]], #0
-; CHECK: bgt
+; CHECK: ble
   %indvar = phi i32 [ %indvar.next, %bb1 ], [ 0, %entry ]
   %tries.0 = sub i32 2147483647, %indvar
   %tmp1 = icmp sgt i32 %tries.0, 0
diff --git a/test/CodeGen/ARM/fusedMAC.ll b/test/CodeGen/ARM/fusedMAC.ll
index a8b3999d2bf..802d1b8b393 100644
--- a/test/CodeGen/ARM/fusedMAC.ll
+++ b/test/CodeGen/ARM/fusedMAC.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=arm -mattr=+neon,+vfp4 | FileCheck %s
+; RUN: llc < %s -mtriple=armv7-eabi -mattr=+neon,+vfp4 | FileCheck %s
 ; Check generated fused MAC and MLS.
 
 define double @fusedMACTest1(double %d1, double %d2, double %d3) {
diff --git a/test/CodeGen/ARM/ldr_post.ll b/test/CodeGen/ARM/ldr_post.ll
index 97a48e1377e..8ddf025dbf1 100644
--- a/test/CodeGen/ARM/ldr_post.ll
+++ b/test/CodeGen/ARM/ldr_post.ll
@@ -1,7 +1,9 @@
-; RUN: llc < %s -march=arm | \
-; RUN:   grep {ldr.*\\\[.*\],} | count 1
+; RUN: llc < %s -march=arm | FileCheck %s
 
-define i32 @test(i32 %a, i32 %b, i32 %c) {
+; CHECK: test1:
+; CHECK: ldr {{.*, \[.*]}}, -r2
+; CHECK-NOT: ldr
+define i32 @test1(i32 %a, i32 %b, i32 %c) {
         %tmp1 = mul i32 %a, %b          ;  [#uses=2]
         %tmp2 = inttoptr i32 %tmp1 to i32*              ;  [#uses=1]
         %tmp3 = load i32* %tmp2         ;  [#uses=1]
@@ -10,3 +12,14 @@ define i32 @test(i32 %a, i32 %b, i32 %c) {
         ret i32 %tmp5
 }
 
+; CHECK: test2:
+; CHECK: ldr {{.*, \[.*\]}}, #-16
+; CHECK-NOT: ldr
+define i32 @test2(i32 %a, i32 %b) {
+        %tmp1 = mul i32 %a, %b          ;  [#uses=2]
+        %tmp2 = inttoptr i32 %tmp1 to i32*              ;  [#uses=1]
+        %tmp3 = load i32* %tmp2         ;  [#uses=1]
+        %tmp4 = sub i32 %tmp1, 16               ;  [#uses=1]
+        %tmp5 = mul i32 %tmp4, %tmp3            ;  [#uses=1]
+        ret i32 %tmp5
+}
diff --git a/test/CodeGen/ARM/ldr_pre.ll b/test/CodeGen/ARM/ldr_pre.ll
index 7c442845682..e904e5fd2cd 100644
--- a/test/CodeGen/ARM/ldr_pre.ll
+++ b/test/CodeGen/ARM/ldr_pre.ll
@@ -1,6 +1,8 @@
-; RUN: llc < %s -march=arm | \
-; RUN:   grep {ldr.*\\!} | count 2
+; RUN: llc < %s -march=arm | FileCheck %s
 
+; CHECK: test1:
+; CHECK: ldr {{.*!}}
+; CHECK-NOT: ldr
 define i32* @test1(i32* %X, i32* %dest) {
         %Y = getelementptr i32* %X, i32 4               ;  [#uses=2]
         %A = load i32* %Y               ;  [#uses=1]
@@ -8,6 +10,9 @@ define i32* @test1(i32* %X, i32* %dest) {
         ret i32* %Y
 }
 
+; CHECK: test2:
+; CHECK: ldr {{.*!}}
+; CHECK-NOT: ldr
 define i32 @test2(i32 %a, i32 %b, i32 %c) {
         %tmp1 = sub i32 %a, %b          ;  [#uses=2]
         %tmp2 = inttoptr i32 %tmp1 to i32*              ;  [#uses=1]
@@ -16,4 +21,3 @@ define i32 @test2(i32 %a, i32 %b, i32 %c) {
         %tmp5 = add i32 %tmp4, %tmp3            ;  [#uses=1]
         ret i32 %tmp5
 }
-
diff --git a/test/CodeGen/ARM/tail-opts.ll b/test/CodeGen/ARM/tail-opts.ll
index 3dc77e2a808..220b0f17373 100644
--- a/test/CodeGen/ARM/tail-opts.ll
+++ b/test/CodeGen/ARM/tail-opts.ll
@@ -16,11 +16,11 @@ declare i8* @choose(i8*, i8*)
 
 ; CHECK: tail_duplicate_me:
 ; CHECK:      qux
-; CHECK:      qux
 ; CHECK:      movw r{{[0-9]+}}, :lower16:_GHJK
 ; CHECK:      movt r{{[0-9]+}}, :upper16:_GHJK
 ; CHECK:      str r
 ; CHECK-NEXT: bx r
+; CHECK:      qux
 ; CHECK:      movw r{{[0-9]+}}, :lower16:_GHJK
 ; CHECK:      movt r{{[0-9]+}}, :upper16:_GHJK
 ; CHECK:      str r
diff --git a/test/CodeGen/ARM/vector-extend-narrow.ll b/test/CodeGen/ARM/vector-extend-narrow.ll
index 5e9239f2563..1ec36da38f7 100644
--- a/test/CodeGen/ARM/vector-extend-narrow.ll
+++ b/test/CodeGen/ARM/vector-extend-narrow.ll
@@ -44,3 +44,17 @@ define <4 x i8> @h(<4 x float> %v) {
   %1 = fptoui <4 x float> %v to <4 x i8>
   ret <4 x i8> %1
 }
+
+; CHECK: i:
+define <4 x i8> @i(<4 x i8>* %x) {
+  ; CHECK: vldr
+  ; CHECK: vmovl.s8
+  ; CHECK: vmovl.s16
+  ; CHECK: vrecpe
+  ; CHECK: vrecps
+  ; CHECK: vmul
+  ; CHECK: vmovn
+  %1 = load <4 x i8>* %x, align 4
+  %2 = sdiv <4 x i8> zeroinitializer, %1
+  ret <4 x i8> %2
+}
diff --git a/test/CodeGen/ARM/widen-vmovs.ll b/test/CodeGen/ARM/widen-vmovs.ll
index 2cffda31790..679e3f43473 100644
--- a/test/CodeGen/ARM/widen-vmovs.ll
+++ b/test/CodeGen/ARM/widen-vmovs.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -widen-vmovs -mcpu=cortex-a8 -verify-machineinstrs | FileCheck %s
+; RUN: llc < %s -widen-vmovs -mcpu=cortex-a8 -verify-machineinstrs -disable-code-place | FileCheck %s
 target triple = "thumbv7-apple-ios"
 
 ; The 1.0e+10 constant is loaded from the constant pool and kept in a register.
@@ -10,6 +10,7 @@ target triple = "thumbv7-apple-ios"
 ; CHECK: , [[DN]]
 ; CHECK: %for.body.i
 ; CHECK: vadd.f32 [[DL]], [[DL]], [[DN]]
+; CHECK: %rInnerproduct.exit
 ;
 ; This test is verifying:
 ; - The VMOVS widening is happening.
diff --git a/test/CodeGen/CellSPU/2009-01-01-BrCond.ll b/test/CodeGen/CellSPU/2009-01-01-BrCond.ll
index 58e3190454f..35422311c57 100644
--- a/test/CodeGen/CellSPU/2009-01-01-BrCond.ll
+++ b/test/CodeGen/CellSPU/2009-01-01-BrCond.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=cellspu -o - | grep brnz
+; RUN: llc < %s -march=cellspu -o - | grep brz
 ; PR3274
 
 target datalayout = "E-p:32:32:128-i1:8:128-i8:8:128-i16:16:128-i32:32:128-i64:32:128-f32:32:128-f64:64:128-v64:64:64-v128:128:128-a0:0:128-s0:128:128"
diff --git a/test/CodeGen/Mips/analyzebranch.ll b/test/CodeGen/Mips/analyzebranch.ll
index 8f0bdf286c5..bc5bcc391ba 100644
--- a/test/CodeGen/Mips/analyzebranch.ll
+++ b/test/CodeGen/Mips/analyzebranch.ll
@@ -26,9 +26,9 @@ return:                                           ; preds = %if.else, %if.end6
 
 define void @f1(float %f) nounwind {
 entry:
-; CHECK: bc1t $BB1_2
+; CHECK: bc1f $BB1_1
 ; CHECK: nop
-; CHECK: # BB#1:      
+; CHECK: # BB#2:
   %cmp = fcmp une float %f, 0.000000e+00
   br i1 %cmp, label %if.then, label %if.end
 
diff --git a/test/CodeGen/Mips/eh.ll b/test/CodeGen/Mips/eh.ll
index c3facdbc555..2e2f9a451ed 100644
--- a/test/CodeGen/Mips/eh.ll
+++ b/test/CodeGen/Mips/eh.ll
@@ -26,7 +26,7 @@ entry:
 lpad:                                             ; preds = %entry
 ; CHECK-EL:  # %lpad
 ; CHECK-EL:  lw  $gp
-; CHECK-EL:  beq $5
+; CHECK-EL:  bne $5
 
   %exn.val = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
            catch i8* bitcast (i8** @_ZTId to i8*)
diff --git a/test/CodeGen/Mips/fpbr.ll b/test/CodeGen/Mips/fpbr.ll
index 0a6478b0f8f..a136557cc4a 100644
--- a/test/CodeGen/Mips/fpbr.ll
+++ b/test/CodeGen/Mips/fpbr.ll
@@ -45,7 +45,7 @@ if.end:                                           ; preds = %if.else, %if.then
 define void @func2(float %f2, float %f3) nounwind {
 entry:
 ; CHECK: c.ole.s
-; CHECK: bc1f
+; CHECK: bc1t
   %cmp = fcmp ugt float %f2, %f3
   br i1 %cmp, label %if.else, label %if.then
 
@@ -102,7 +102,7 @@ if.end:                                           ; preds = %if.else, %if.then
 define void @func5(double %f2, double %f3) nounwind {
 entry:
 ; CHECK: c.ole.d
-; CHECK: bc1f
+; CHECK: bc1t
   %cmp = fcmp ugt double %f2, %f3
   br i1 %cmp, label %if.else, label %if.then
 
diff --git a/test/CodeGen/PowerPC/ppc-vaarg-agg.ll b/test/CodeGen/PowerPC/ppc-vaarg-agg.ll
new file mode 100644
index 00000000000..d5ea044599e
--- /dev/null
+++ b/test/CodeGen/PowerPC/ppc-vaarg-agg.ll
@@ -0,0 +1,46 @@
+target datalayout = "E-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v128:128:128-n32"
+target triple = "powerpc-montavista-linux-gnuspe"
+; RUN: llc < %s -march=ppc32 | FileCheck %s
+
+%struct.__va_list_tag.0.9.18.23.32.41.48.55.62.67.72.77.82.87.90.93.96.101.105 = type { i8, i8, i16, i8*, i8* }
+
+define fastcc void @test1(%struct.__va_list_tag.0.9.18.23.32.41.48.55.62.67.72.77.82.87.90.93.96.101.105* %args) {
+entry:
+  br i1 undef, label %repeat, label %maxlen_reached
+
+repeat:                                           ; preds = %entry
+  switch i32 undef, label %sw.bb323 [
+    i32 77, label %sw.bb72
+    i32 111, label %sw.bb309
+    i32 80, label %sw.bb313
+    i32 117, label %sw.bb326
+    i32 88, label %sw.bb321
+  ]
+
+sw.bb72:                                          ; preds = %repeat
+  unreachable
+
+sw.bb309:                                         ; preds = %repeat
+  unreachable
+
+sw.bb313:                                         ; preds = %repeat
+  unreachable
+
+sw.bb321:                                         ; preds = %repeat
+  unreachable
+
+sw.bb323:                                         ; preds = %repeat
+  %0 = va_arg %struct.__va_list_tag.0.9.18.23.32.41.48.55.62.67.72.77.82.87.90.93.96.101.105* %args, i32
+  unreachable
+
+sw.bb326:                                         ; preds = %repeat
+  unreachable
+
+maxlen_reached:                                   ; preds = %entry
+  ret void
+}
+
+; If the SD nodes are not cleaup up correctly, then this can fail to compile
+; with an error like:  Cannot select: ch = setlt [ID=6]
+; CHECK: @test1
+
diff --git a/test/CodeGen/Thumb2/thumb2-branch.ll b/test/CodeGen/Thumb2/thumb2-branch.ll
index 27d8e8fb408..f1c097c1892 100644
--- a/test/CodeGen/Thumb2/thumb2-branch.ll
+++ b/test/CodeGen/Thumb2/thumb2-branch.ll
@@ -58,8 +58,8 @@ define i32 @f4(i32 %a, i32 %b, i32* %v) {
 entry:
 ; CHECK: f4:
 ; CHECK: blo LBB
-        %tmp = icmp ult i32 %a, %b              ;  [#uses=1]
-        br i1 %tmp, label %return, label %cond_true
+        %tmp = icmp uge i32 %a, %b              ;  [#uses=1]
+        br i1 %tmp, label %cond_true, label %return
 
 cond_true:              ; preds = %entry
         fence seq_cst
diff --git a/test/CodeGen/Thumb2/thumb2-ifcvt2.ll b/test/CodeGen/Thumb2/thumb2-ifcvt2.ll
index f577f79d691..5aa9a735f25 100644
--- a/test/CodeGen/Thumb2/thumb2-ifcvt2.ll
+++ b/test/CodeGen/Thumb2/thumb2-ifcvt2.ll
@@ -29,13 +29,13 @@ declare i32 @bar(...)
 define fastcc i32 @CountTree(%struct.quad_struct* %tree) {
 entry:
 ; CHECK: CountTree:
-; CHECK: it eq
-; CHECK: cmpeq
-; CHECK: bne
-; CHECK: cmp
 ; CHECK: itt eq
 ; CHECK: moveq
 ; CHECK: popeq
+; CHECK: bne
+; CHECK: cmp
+; CHECK: it eq
+; CHECK: cmpeq
 	br label %tailrecurse
 
 tailrecurse:		; preds = %bb, %entry
@@ -83,7 +83,7 @@ define fastcc void @t2() nounwind {
 entry:
 ; CHECK: t2:
 ; CHECK: cmp r0, #0
-; CHECK: beq
+; CHECK: %growMapping.exit
 	br i1 undef, label %bb.i.i3, label %growMapping.exit
 
 bb.i.i3:		; preds = %entry
diff --git a/test/CodeGen/Thumb2/thumb2-jtb.ll b/test/CodeGen/Thumb2/thumb2-jtb.ll
index f5a56e5ace0..7e1655f6c25 100644
--- a/test/CodeGen/Thumb2/thumb2-jtb.ll
+++ b/test/CodeGen/Thumb2/thumb2-jtb.ll
@@ -3,11 +3,19 @@
 ; Do not use tbb / tbh if any destination is before the jumptable.
 ; rdar://7102917
 
-define i16 @main__getopt_internal_2E_exit_2E_ce(i32) nounwind {
+define i16 @main__getopt_internal_2E_exit_2E_ce(i32, i1 %b) nounwind {
+entry:
+  br i1 %b, label %codeRepl127.exitStub, label %newFuncRoot
+
 newFuncRoot:
 	br label %_getopt_internal.exit.ce
 
 codeRepl127.exitStub:		; preds = %_getopt_internal.exit.ce
+  ; Add an explicit edge back to before the jump table to ensure this block
+  ; is placed first.
+  br i1 %b, label %newFuncRoot, label %codeRepl127.exitStub.exit
+
+codeRepl127.exitStub.exit:
 	ret i16 0
 
 parse_options.exit.loopexit.exitStub:		; preds = %_getopt_internal.exit.ce
diff --git a/test/CodeGen/X86/2006-10-19-SwitchUnnecessaryBranching.ll b/test/CodeGen/X86/2006-10-19-SwitchUnnecessaryBranching.ll
index 88e8b4a4fd9..d583e5964dc 100644
--- a/test/CodeGen/X86/2006-10-19-SwitchUnnecessaryBranching.ll
+++ b/test/CodeGen/X86/2006-10-19-SwitchUnnecessaryBranching.ll
@@ -6,8 +6,8 @@
 define i32 @test(i32 %argc, i8** %argv) nounwind {
 entry:
 ; CHECK: cmpl	$2
-; CHECK-NEXT: je
-; CHECK-NEXT: %entry
+; CHECK-NEXT: jne
+; CHECK-NEXT: %bb2
 
 	switch i32 %argc, label %UnifiedReturnBlock [
 		 i32 1, label %bb
diff --git a/test/CodeGen/X86/2008-05-01-InvalidOrdCompare.ll b/test/CodeGen/X86/2008-05-01-InvalidOrdCompare.ll
index a708224dd0d..4160b203e36 100644
--- a/test/CodeGen/X86/2008-05-01-InvalidOrdCompare.ll
+++ b/test/CodeGen/X86/2008-05-01-InvalidOrdCompare.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -enable-unsafe-fp-math -march=x86 | grep jnp
+; RUN: llc < %s -enable-unsafe-fp-math -march=x86 | grep jp
 ; rdar://5902801
 
 declare void @test2()
diff --git a/test/CodeGen/X86/2010-08-04-MaskedSignedCompare.ll b/test/CodeGen/X86/2010-08-04-MaskedSignedCompare.ll
index 1919d2ef34a..12a8274fb56 100644
--- a/test/CodeGen/X86/2010-08-04-MaskedSignedCompare.ll
+++ b/test/CodeGen/X86/2010-08-04-MaskedSignedCompare.ll
@@ -17,7 +17,7 @@ entry:
 
 ; CHECK: andl	$150
 ; CHECK-NEXT: testb
-; CHECK-NEXT: jg
+; CHECK-NEXT: jle
 
 entry.if.end_crit_edge:                           ; preds = %entry
   %tmp4.pre = load i32* @g_38                     ;  [#uses=1]
diff --git a/test/CodeGen/X86/2010-11-18-SelectOfExtload.ll b/test/CodeGen/X86/2010-11-18-SelectOfExtload.ll
index a1074b6b8f3..6d54c7e2982 100644
--- a/test/CodeGen/X86/2010-11-18-SelectOfExtload.ll
+++ b/test/CodeGen/X86/2010-11-18-SelectOfExtload.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=x86 | FileCheck %s
+; RUN: llc < %s -march=x86 -mattr=+cmov | FileCheck %s
 ; Both values were being zero extended.
 @u = external global i8
 @s = external global i8
diff --git a/test/CodeGen/X86/2011-09-14-valcoalesce.ll b/test/CodeGen/X86/2011-09-14-valcoalesce.ll
index 1068d1b704b..a5ec614a943 100644
--- a/test/CodeGen/X86/2011-09-14-valcoalesce.ll
+++ b/test/CodeGen/X86/2011-09-14-valcoalesce.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=x86 | FileCheck %s
+; RUN: llc < %s -march=x86 -disable-code-place | FileCheck %s
 ;
 ; Test RegistersDefinedFromSameValue. We have multiple copies of the same vreg:
 ; while.body85.i:
diff --git a/test/CodeGen/X86/2012-04-26-sdglue.ll b/test/CodeGen/X86/2012-04-26-sdglue.ll
new file mode 100644
index 00000000000..9543587747a
--- /dev/null
+++ b/test/CodeGen/X86/2012-04-26-sdglue.ll
@@ -0,0 +1,46 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=core-avx2 -mattr=+avx | FileCheck %s
+; rdar://11314175: SD Scheduler, BuildSchedUnits assert:
+;                  N->getNodeId() == -1 && "Node already inserted!
+
+; It's hard to test for the ISEL condition because CodeGen optimizes
+; away the bugpointed code. Just ensure the basics are still there.
+;CHECK: func:
+;CHECK: vmovups
+;CHECK: vpshufd
+;CHECK: vpshufd
+;CHECK: vmulps
+;CHECK: vmulps
+;CHECK: ret
+
+define void @func() nounwind ssp {
+  %tmp = load <4 x float>* null, align 1
+  %tmp14 = getelementptr <4 x float>* null, i32 2
+  %tmp15 = load <4 x float>* %tmp14, align 1
+  %tmp16 = shufflevector <4 x float> %tmp, <4 x float> , <8 x i32> 
+  %tmp17 = call <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float> %tmp16, <4 x float> undef, i8 1)
+  %tmp18 = bitcast <4 x float> %tmp to <16 x i8>
+  %tmp19 = shufflevector <16 x i8> %tmp18, <16 x i8> undef, <16 x i32> 
+  %tmp20 = bitcast <16 x i8> %tmp19 to <4 x float>
+  %tmp21 = bitcast <4 x float> %tmp15 to <16 x i8>
+  %tmp22 = shufflevector <16 x i8> undef, <16 x i8> %tmp21, <16 x i32> 
+  %tmp23 = bitcast <16 x i8> %tmp22 to <4 x float>
+  %tmp24 = shufflevector <4 x float> %tmp20, <4 x float> , <8 x i32> 
+  %tmp25 = call <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float> %tmp24, <4 x float> %tmp23, i8 1)
+  %tmp26 = fmul <8 x float> %tmp17, undef
+  %tmp27 = fmul <8 x float> %tmp25, undef
+  %tmp28 = fadd <8 x float> %tmp26, %tmp27
+  %tmp29 = fadd <8 x float> %tmp28, undef
+  %tmp30 = shufflevector <8 x float> %tmp29, <8 x float> undef, <4 x i32> 
+  %tmp31 = fmul <4 x float> undef, %tmp30
+  %tmp32 = call <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float> zeroinitializer, <4 x float> %tmp31, i8 1)
+  %tmp33 = fadd <8 x float> undef, %tmp32
+  %tmp34 = call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %tmp33, <8 x float> undef) nounwind
+  %tmp35 = fsub <8 x float> %tmp34, undef
+  %tmp36 = call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> zeroinitializer, <8 x float> %tmp35) nounwind
+  store <8 x float> %tmp36, <8 x float>* undef, align 32
+  ret void
+}
+
+declare <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float>, <4 x float>, i8) nounwind readnone
+
+declare <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float>, <8 x float>) nounwind readnone
diff --git a/test/CodeGen/X86/GC/cg-O0.ll b/test/CodeGen/X86/GC/cg-O0.ll
new file mode 100644
index 00000000000..b4929425e94
--- /dev/null
+++ b/test/CodeGen/X86/GC/cg-O0.ll
@@ -0,0 +1,17 @@
+; RUN: llc < %s -O0
+
+define i32 @main() {
+entry:
+  call void @f()
+  ret i32 0
+}
+
+define void @f() gc "ocaml" {
+entry:
+  %ptr.stackref = alloca i8*
+  %gcroot = bitcast i8** %ptr.stackref to i8**
+  call void @llvm.gcroot(i8** %gcroot, i8* null)
+  ret void
+}
+
+declare void @llvm.gcroot(i8**, i8*) nounwind
diff --git a/test/CodeGen/X86/atom-sched.ll b/test/CodeGen/X86/atom-sched.ll
index 2301dfc020a..4dd9a9e3481 100644
--- a/test/CodeGen/X86/atom-sched.ll
+++ b/test/CodeGen/X86/atom-sched.ll
@@ -1,5 +1,9 @@
+; XFAIL: *
 ; RUN: llc <%s -O2 -mcpu=atom -march=x86 -relocation-model=static | FileCheck -check-prefix=atom %s
 ; RUN: llc <%s -O2 -mcpu=core2 -march=x86 -relocation-model=static | FileCheck %s
+;
+; FIXME: Atom's scheduler is temporarily disabled.
+; XFAIL: *
 
 @a = common global i32 0, align 4
 @b = common global i32 0, align 4
diff --git a/test/CodeGen/X86/atomic_op.ll b/test/CodeGen/X86/atomic_op.ll
index 972dab292ae..7c5abe2095c 100644
--- a/test/CodeGen/X86/atomic_op.ll
+++ b/test/CodeGen/X86/atomic_op.ll
@@ -13,6 +13,7 @@ entry:
 	%xort = alloca i32		;  [#uses=2]
 	%old = alloca i32		;  [#uses=18]
 	%temp = alloca i32		;  [#uses=2]
+	%temp64 = alloca i64
 	store i32 %argc, i32* %argc.addr
 	store i8** %argv, i8*** %argv.addr
 	store i32 0, i32* %val1
@@ -106,6 +107,25 @@ entry:
         ; CHECK: cmpxchgl
   %17 = cmpxchg i32* %val2, i32 1976, i32 1 monotonic
 	store i32 %17, i32* %old
+        ; CHECK: movl	$1401, %[[R17mask:[a-z]*]]
+        ; CHECK: movl	[[R17atomic:.*]], %eax
+        ; CHECK: movl	%eax, %[[R17newval:[a-z]*]]
+        ; CHECK: andl	%[[R17mask]], %[[R17newval]]
+        ; CHECK: notl	%[[R17newval]]
+        ; CHECK: lock
+        ; CHECK: cmpxchgl	%[[R17newval]], [[R17atomic]]
+        ; CHECK: jne
+        ; CHECK: movl	%eax,
+  %18 = atomicrmw nand i32* %val2, i32 1401 monotonic
+  store i32 %18, i32* %old
+        ; CHECK: andl
+        ; CHECK: andl
+        ; CHECK: notl
+        ; CHECK: notl
+        ; CHECK: lock
+        ; CHECK: cmpxchg8b
+  %19 = atomicrmw nand i64* %temp64, i64 17361641481138401520 monotonic
+  store i64 %19, i64* %temp64
 	ret void
 }
 
diff --git a/test/CodeGen/X86/avx2-intrinsics-x86.ll b/test/CodeGen/X86/avx2-intrinsics-x86.ll
index 1fb41c02b9e..3f27a0291b4 100644
--- a/test/CodeGen/X86/avx2-intrinsics-x86.ll
+++ b/test/CodeGen/X86/avx2-intrinsics-x86.ll
@@ -800,22 +800,6 @@ define <8 x float> @test_x86_avx2_permps(<8 x float> %a0, <8 x float> %a1) {
 declare <8 x float> @llvm.x86.avx2.permps(<8 x float>, <8 x float>) nounwind readonly
 
 
-define <4 x i64> @test_x86_avx2_permq(<4 x i64> %a0) {
-  ; CHECK: vpermq
-  %res = call <4 x i64> @llvm.x86.avx2.permq(<4 x i64> %a0, i8 7) ; <<4 x i64>> [#uses=1]
-  ret <4 x i64> %res
-}
-declare <4 x i64> @llvm.x86.avx2.permq(<4 x i64>, i8) nounwind readonly
-
-
-define <4 x double> @test_x86_avx2_permpd(<4 x double> %a0) {
-  ; CHECK: vpermpd
-  %res = call <4 x double> @llvm.x86.avx2.permpd(<4 x double> %a0, i8 7) ; <<4 x double>> [#uses=1]
-  ret <4 x double> %res
-}
-declare <4 x double> @llvm.x86.avx2.permpd(<4 x double>, i8) nounwind readonly
-
-
 define <4 x i64> @test_x86_avx2_vperm2i128(<4 x i64> %a0, <4 x i64> %a1) {
   ; CHECK: vperm2i128
   %res = call <4 x i64> @llvm.x86.avx2.vperm2i128(<4 x i64> %a0, <4 x i64> %a1, i8 1) ; <<4 x i64>> [#uses=1]
diff --git a/test/CodeGen/X86/avx2-vperm.ll b/test/CodeGen/X86/avx2-vperm.ll
new file mode 100755
index 00000000000..d576d0e3741
--- /dev/null
+++ b/test/CodeGen/X86/avx2-vperm.ll
@@ -0,0 +1,34 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=core-avx2 -mattr=+avx2 | FileCheck %s
+
+define <8 x i32> @perm_cl_int_8x32(<8 x i32> %A) nounwind readnone {
+entry:
+; CHECK: perm_cl_int_8x32
+; CHECK: vpermd
+  %B = shufflevector <8 x i32> %A, <8 x i32> undef, <8 x i32> 
+  ret <8 x i32> %B
+}
+
+
+define <8 x float> @perm_cl_fp_8x32(<8 x float> %A) nounwind readnone {
+entry:
+; CHECK: perm_cl_fp_8x32
+; CHECK: vpermps
+  %B = shufflevector <8 x float> %A, <8 x float> undef, <8 x i32> 
+  ret <8 x float> %B
+}
+
+define <4 x i64> @perm_cl_int_4x64(<4 x i64> %A) nounwind readnone {
+entry:
+; CHECK: perm_cl_int_4x64
+; CHECK: vpermq
+  %B = shufflevector <4 x i64> %A, <4 x i64> undef, <4 x i32> 
+  ret <4 x i64> %B
+}
+
+define <4 x double> @perm_cl_fp_4x64(<4 x double> %A) nounwind readnone {
+entry:
+; CHECK: perm_cl_fp_4x64
+; CHECK: vpermpd
+  %B = shufflevector <4 x double> %A, <4 x double> undef, <4 x i32> 
+  ret <4 x double> %B
+}
diff --git a/test/CodeGen/X86/block-placement.ll b/test/CodeGen/X86/block-placement.ll
index 167d522d47d..fc7b6383b8b 100644
--- a/test/CodeGen/X86/block-placement.ll
+++ b/test/CodeGen/X86/block-placement.ll
@@ -1,4 +1,4 @@
-; RUN: llc -mtriple=i686-linux -enable-block-placement < %s | FileCheck %s
+; RUN: llc -mtriple=i686-linux < %s | FileCheck %s
 
 declare void @error(i32 %i, i32 %a, i32 %b)
 
@@ -76,11 +76,11 @@ define i32 @test_loop_cold_blocks(i32 %i, i32* %a) {
 ; Check that we sink cold loop blocks after the hot loop body.
 ; CHECK: test_loop_cold_blocks:
 ; CHECK: %entry
+; CHECK: %unlikely1
+; CHECK: %unlikely2
 ; CHECK: %body1
 ; CHECK: %body2
 ; CHECK: %body3
-; CHECK: %unlikely1
-; CHECK: %unlikely2
 ; CHECK: %exit
 
 entry:
@@ -122,14 +122,14 @@ define i32 @test_loop_early_exits(i32 %i, i32* %a) {
 ; Check that we sink early exit blocks out of loop bodies.
 ; CHECK: test_loop_early_exits:
 ; CHECK: %entry
+; CHECK: %body1
 ; CHECK: %body2
 ; CHECK: %body3
 ; CHECK: %body4
-; CHECK: %body1
+; CHECK: %exit
 ; CHECK: %bail1
 ; CHECK: %bail2
 ; CHECK: %bail3
-; CHECK: %exit
 
 entry:
   br label %body1
@@ -199,6 +199,36 @@ exit:
   ret i32 %base
 }
 
+define i32 @test_no_loop_rotate(i32 %i, i32* %a) {
+; Check that we don't try to rotate a loop which is already laid out with
+; fallthrough opportunities into the top and out of the bottom.
+; CHECK: test_no_loop_rotate:
+; CHECK: %entry
+; CHECK: %body0
+; CHECK: %body1
+; CHECK: %exit
+
+entry:
+  br label %body0
+
+body0:
+  %iv = phi i32 [ 0, %entry ], [ %next, %body1 ]
+  %base = phi i32 [ 0, %entry ], [ %sum, %body1 ]
+  %arrayidx = getelementptr inbounds i32* %a, i32 %iv
+  %0 = load i32* %arrayidx
+  %sum = add nsw i32 %0, %base
+  %bailcond1 = icmp eq i32 %sum, 42
+  br i1 %bailcond1, label %exit, label %body1
+
+body1:
+  %next = add i32 %iv, 1
+  %exitcond = icmp eq i32 %next, %i
+  br i1 %exitcond, label %exit, label %body0
+
+exit:
+  ret i32 %base
+}
+
 define void @test_loop_rotate_reversed_blocks() {
 ; This test case (greatly reduced from an Olden bencmark) ensures that the loop
 ; rotate implementation doesn't assume that loops are laid out in a particular
@@ -348,7 +378,6 @@ define void @unnatural_cfg2() {
 ; CHECK: %entry
 ; CHECK: %loop.body1
 ; CHECK: %loop.body2
-; CHECK: %loop.header
 ; CHECK: %loop.body3
 ; CHECK: %loop.inner1.begin
 ; The end block is folded with %loop.body3...
@@ -356,6 +385,7 @@ define void @unnatural_cfg2() {
 ; CHECK: %loop.body4
 ; CHECK: %loop.inner2.begin
 ; The loop.inner2.end block is folded
+; CHECK: %loop.header
 ; CHECK: %bail
 
 entry:
@@ -928,3 +958,126 @@ entry:
 exit:
   ret void
 }
+
+define void @benchmark_heapsort(i32 %n, double* nocapture %ra) {
+; This test case comes from the heapsort benchmark, and exemplifies several
+; important aspects to block placement in the presence of loops:
+; 1) Loop rotation needs to *ensure* that the desired exiting edge can be
+;    a fallthrough.
+; 2) The exiting edge from the loop which is rotated to be laid out at the
+;    bottom of the loop needs to be exiting into the nearest enclosing loop (to
+;    which there is an exit). Otherwise, we force that enclosing loop into
+;    strange layouts that are siginificantly less efficient, often times maing
+;    it discontiguous.
+;
+; CHECK: @benchmark_heapsort
+; CHECK: %entry
+; First rotated loop top.
+; CHECK: .align
+; CHECK: %while.end
+; CHECK: %for.cond
+; CHECK: %if.then
+; CHECK: %if.else
+; CHECK: %if.end10
+; Second rotated loop top
+; CHECK: .align
+; CHECK: %if.then24
+; CHECK: %while.cond.outer
+; Third rotated loop top
+; CHECK: .align
+; CHECK: %while.cond
+; CHECK: %while.body
+; CHECK: %land.lhs.true
+; CHECK: %if.then19
+; CHECK: %if.then19
+; CHECK: %if.then8
+; CHECK: ret
+
+entry:
+  %shr = ashr i32 %n, 1
+  %add = add nsw i32 %shr, 1
+  %arrayidx3 = getelementptr inbounds double* %ra, i64 1
+  br label %for.cond
+
+for.cond:
+  %ir.0 = phi i32 [ %n, %entry ], [ %ir.1, %while.end ]
+  %l.0 = phi i32 [ %add, %entry ], [ %l.1, %while.end ]
+  %cmp = icmp sgt i32 %l.0, 1
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  %dec = add nsw i32 %l.0, -1
+  %idxprom = sext i32 %dec to i64
+  %arrayidx = getelementptr inbounds double* %ra, i64 %idxprom
+  %0 = load double* %arrayidx, align 8
+  br label %if.end10
+
+if.else:
+  %idxprom1 = sext i32 %ir.0 to i64
+  %arrayidx2 = getelementptr inbounds double* %ra, i64 %idxprom1
+  %1 = load double* %arrayidx2, align 8
+  %2 = load double* %arrayidx3, align 8
+  store double %2, double* %arrayidx2, align 8
+  %dec6 = add nsw i32 %ir.0, -1
+  %cmp7 = icmp eq i32 %dec6, 1
+  br i1 %cmp7, label %if.then8, label %if.end10
+
+if.then8:
+  store double %1, double* %arrayidx3, align 8
+  ret void
+
+if.end10:
+  %ir.1 = phi i32 [ %ir.0, %if.then ], [ %dec6, %if.else ]
+  %l.1 = phi i32 [ %dec, %if.then ], [ %l.0, %if.else ]
+  %rra.0 = phi double [ %0, %if.then ], [ %1, %if.else ]
+  %add31 = add nsw i32 %ir.1, 1
+  br label %while.cond.outer
+
+while.cond.outer:
+  %j.0.ph.in = phi i32 [ %l.1, %if.end10 ], [ %j.1, %if.then24 ]
+  %j.0.ph = shl i32 %j.0.ph.in, 1
+  br label %while.cond
+
+while.cond:
+  %j.0 = phi i32 [ %add31, %if.end20 ], [ %j.0.ph, %while.cond.outer ]
+  %cmp11 = icmp sgt i32 %j.0, %ir.1
+  br i1 %cmp11, label %while.end, label %while.body
+
+while.body:
+  %cmp12 = icmp slt i32 %j.0, %ir.1
+  br i1 %cmp12, label %land.lhs.true, label %if.end20
+
+land.lhs.true:
+  %idxprom13 = sext i32 %j.0 to i64
+  %arrayidx14 = getelementptr inbounds double* %ra, i64 %idxprom13
+  %3 = load double* %arrayidx14, align 8
+  %add15 = add nsw i32 %j.0, 1
+  %idxprom16 = sext i32 %add15 to i64
+  %arrayidx17 = getelementptr inbounds double* %ra, i64 %idxprom16
+  %4 = load double* %arrayidx17, align 8
+  %cmp18 = fcmp olt double %3, %4
+  br i1 %cmp18, label %if.then19, label %if.end20
+
+if.then19:
+  br label %if.end20
+
+if.end20:
+  %j.1 = phi i32 [ %add15, %if.then19 ], [ %j.0, %land.lhs.true ], [ %j.0, %while.body ]
+  %idxprom21 = sext i32 %j.1 to i64
+  %arrayidx22 = getelementptr inbounds double* %ra, i64 %idxprom21
+  %5 = load double* %arrayidx22, align 8
+  %cmp23 = fcmp olt double %rra.0, %5
+  br i1 %cmp23, label %if.then24, label %while.cond
+
+if.then24:
+  %idxprom27 = sext i32 %j.0.ph.in to i64
+  %arrayidx28 = getelementptr inbounds double* %ra, i64 %idxprom27
+  store double %5, double* %arrayidx28, align 8
+  br label %while.cond.outer
+
+while.end:
+  %idxprom33 = sext i32 %j.0.ph.in to i64
+  %arrayidx34 = getelementptr inbounds double* %ra, i64 %idxprom33
+  store double %rra.0, double* %arrayidx34, align 8
+  br label %for.cond
+}
diff --git a/test/CodeGen/X86/br-fold.ll b/test/CodeGen/X86/br-fold.ll
index 8af3bd1bc22..2c371949380 100644
--- a/test/CodeGen/X86/br-fold.ll
+++ b/test/CodeGen/X86/br-fold.ll
@@ -1,7 +1,7 @@
 ; RUN: llc -march=x86-64 < %s | FileCheck %s
 
 ; CHECK: orq
-; CHECK-NEXT: jne
+; CHECK-NEXT: LBB0_1
 
 @_ZN11xercesc_2_513SchemaSymbols21fgURI_SCHEMAFORSCHEMAE = external constant [33 x i16], align 32 ; <[33 x i16]*> [#uses=1]
 @_ZN11xercesc_2_56XMLUni16fgNotationStringE = external constant [9 x i16], align 16 ; <[9 x i16]*> [#uses=1]
diff --git a/test/CodeGen/X86/call-push.ll b/test/CodeGen/X86/call-push.ll
index 8cca10c8307..e69f8c1ebf7 100644
--- a/test/CodeGen/X86/call-push.ll
+++ b/test/CodeGen/X86/call-push.ll
@@ -7,8 +7,8 @@ define i32 @decode_byte(%struct.decode_t* %decode) nounwind {
 ; CHECK: decode_byte:
 ; CHECK: pushl
 ; CHECK: popl
-; CHECK: popl
 ; CHECK: jmp
+; CHECK: popl
 entry:
         %tmp2 = getelementptr %struct.decode_t* %decode, i32 0, i32 4           ;  [#uses=1]
         %tmp23 = bitcast i16* %tmp2 to i32*             ;  [#uses=1]
diff --git a/test/CodeGen/Generic/dbg-declare.ll b/test/CodeGen/X86/dbg-declare.ll
similarity index 93%
rename from test/CodeGen/Generic/dbg-declare.ll
rename to test/CodeGen/X86/dbg-declare.ll
index 01f7d6d4791..5d4cedc5c4e 100644
--- a/test/CodeGen/Generic/dbg-declare.ll
+++ b/test/CodeGen/X86/dbg-declare.ll
@@ -1,7 +1,5 @@
-; RUN: llc < %s -O0
+; RUN: llc < %s -O0 -mtriple x86_64-apple-darwin
 ; 
-target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
-target triple = "x86_64-apple-macosx10.8.0"
 
 define i32 @foo(i32* %x) nounwind uwtable ssp {
 entry:
diff --git a/test/CodeGen/X86/licm-dominance.ll b/test/CodeGen/X86/licm-dominance.ll
index 8a0958db0ef..019f8a32b6c 100644
--- a/test/CodeGen/X86/licm-dominance.ll
+++ b/test/CodeGen/X86/licm-dominance.ll
@@ -1,7 +1,7 @@
-; RUN: llc -asm-verbose=false < %s | FileCheck %s
+; RUN: llc -asm-verbose=true < %s | FileCheck %s
 
 ; MachineLICM should check dominance before hoisting instructions.
-; CHECK:	jne	LBB0_3
+; CHECK: ## in Loop:
 ; CHECK-NEXT:	xorb	%al, %al
 ; CHECK-NEXT:	testb	%al, %al
 
diff --git a/test/CodeGen/X86/loop-blocks.ll b/test/CodeGen/X86/loop-blocks.ll
index faba6300712..d14102fe245 100644
--- a/test/CodeGen/X86/loop-blocks.ll
+++ b/test/CodeGen/X86/loop-blocks.ll
@@ -41,6 +41,7 @@ done:
 ; CHECK-NEXT:   align
 ; CHECK-NEXT: .LBB1_4:
 ; CHECK-NEXT:   callq bar99
+; CHECK-NEXT:   align
 ; CHECK-NEXT: .LBB1_1:
 ; CHECK-NEXT:   callq body
 
@@ -75,19 +76,21 @@ exit:
 ; CHECK: yet_more_involved:
 ;      CHECK:   jmp .LBB2_1
 ; CHECK-NEXT:   align
-; CHECK-NEXT: .LBB2_4:
-; CHECK-NEXT:   callq bar99
+; CHECK-NEXT: .LBB2_5:
+; CHECK-NEXT:   callq block_a_true_func
+; CHECK-NEXT:   callq block_a_merge_func
+; CHECK-NEXT:   align
+; CHECK-NEXT: .LBB2_1:
+; CHECK-NEXT:   callq body
+;
+; LBB2_4
+;      CHECK:   callq bar99
 ; CHECK-NEXT:   callq get
 ; CHECK-NEXT:   cmpl $2999, %eax
 ; CHECK-NEXT:   jle .LBB2_5
 ; CHECK-NEXT:   callq block_a_false_func
 ; CHECK-NEXT:   callq block_a_merge_func
 ; CHECK-NEXT:   jmp .LBB2_1
-; CHECK-NEXT: .LBB2_5:
-; CHECK-NEXT:   callq block_a_true_func
-; CHECK-NEXT:   callq block_a_merge_func
-; CHECK-NEXT: .LBB2_1:
-; CHECK-NEXT:   callq body
 
 define void @yet_more_involved() nounwind {
 entry:
@@ -136,17 +139,22 @@ exit:
 ; CHECK-NEXT:   align
 ; CHECK-NEXT: .LBB3_7:
 ; CHECK-NEXT:   callq   bar100
-; CHECK-NEXT:   jmp     .LBB3_1
-; CHECK-NEXT: .LBB3_8:
-; CHECK-NEXT:   callq   bar101
-; CHECK-NEXT:   jmp     .LBB3_1
-; CHECK-NEXT: .LBB3_9:
-; CHECK-NEXT:   callq   bar102
-; CHECK-NEXT:   jmp     .LBB3_1
-; CHECK-NEXT: .LBB3_5:
-; CHECK-NEXT:   callq   loop_latch
+; CHECK-NEXT:   align
 ; CHECK-NEXT: .LBB3_1:
 ; CHECK-NEXT:   callq   loop_header
+;      CHECK:   jl .LBB3_7
+;      CHECK:   jge .LBB3_3
+; CHECK-NEXT:   callq   bar101
+; CHECK-NEXT:   jmp     .LBB3_1
+; CHECK-NEXT: .LBB3_3:
+;      CHECK:   jge .LBB3_4
+; CHECK-NEXT:   callq   bar102
+; CHECK-NEXT:   jmp     .LBB3_1
+; CHECK-NEXT: .LBB3_4:
+;      CHECK:   jl .LBB3_6
+; CHECK-NEXT:   callq   loop_latch
+; CHECK-NEXT:   jmp     .LBB3_1
+; CHECK-NEXT: .LBB3_6:
 
 define void @cfg_islands() nounwind {
 entry:
diff --git a/test/CodeGen/X86/machine-cp.ll b/test/CodeGen/X86/machine-cp.ll
index 54fa01c38fd..8e97b991d07 100644
--- a/test/CodeGen/X86/machine-cp.ll
+++ b/test/CodeGen/X86/machine-cp.ll
@@ -5,11 +5,11 @@
 define i32 @t1(i32 %a, i32 %b) nounwind  {
 entry:
 ; CHECK: t1:
-; CHECK: jne
+; CHECK: je [[LABEL:.*BB.*]]
   %cmp1 = icmp eq i32 %b, 0
   br i1 %cmp1, label %while.end, label %while.body
 
-; CHECK: BB
+; CHECK: [[LABEL]]:
 ; CHECK-NOT: mov
 ; CHECK: ret
 
diff --git a/test/CodeGen/X86/postra-licm.ll b/test/CodeGen/X86/postra-licm.ll
index 48c48aebe5b..01d6cbef1ee 100644
--- a/test/CodeGen/X86/postra-licm.ll
+++ b/test/CodeGen/X86/postra-licm.ll
@@ -70,8 +70,8 @@ bb26.preheader:                                   ; preds = %imix_test.exit
 bb23:                                             ; preds = %imix_test.exit
   unreachable
 ; Verify that there are no loads inside the loop.
-; X86-32: %bb26.preheader
 ; X86-32: .align 4
+; X86-32: %bb28
 ; X86-32-NOT: (%esp),
 ; X86-32-NOT: (%ebp),
 ; X86-32: jmp
diff --git a/test/CodeGen/X86/pr2659.ll b/test/CodeGen/X86/pr2659.ll
index 5dab5c9f5b8..8003588a2e8 100644
--- a/test/CodeGen/X86/pr2659.ll
+++ b/test/CodeGen/X86/pr2659.ll
@@ -18,11 +18,12 @@ forcond.preheader:              ; preds = %entry
 ; CHECK-NOT: xorl
 ; CHECK-NOT: movl
 ; CHECK-NOT: LBB
-; CHECK: jne
+; CHECK: je
 
 ; There should be no moves required in the for loop body.
 ; CHECK: %forbody
 ; CHECK-NOT: mov
+; CHECK: jbe
 
 ifthen:         ; preds = %entry
   ret i32 0
diff --git a/test/CodeGen/X86/select.ll b/test/CodeGen/X86/select.ll
index ce04e07854a..f465a4ffc58 100644
--- a/test/CodeGen/X86/select.ll
+++ b/test/CodeGen/X86/select.ll
@@ -75,9 +75,9 @@ define void @test6(i32 %C, <4 x float>* %A, <4 x float>* %B) nounwind {
 ; Verify that the fmul gets sunk into the one part of the diamond where it is
 ; needed.
 ; CHECK: test6:
-; CHECK: jne
-; CHECK: mulps
+; CHECK: je
 ; CHECK: ret
+; CHECK: mulps
 ; CHECK: ret
 }
 
diff --git a/test/CodeGen/X86/sibcall.ll b/test/CodeGen/X86/sibcall.ll
index a9a5420cbcd..2af355905dc 100644
--- a/test/CodeGen/X86/sibcall.ll
+++ b/test/CodeGen/X86/sibcall.ll
@@ -147,7 +147,7 @@ define i32 @t11(i32 %x, i32 %y, i32 %z.0, i32 %z.1, i32 %z.2) nounwind ssp {
 
 ; 32: t11:
 ; 32-NOT: subl ${{[0-9]+}}, %esp
-; 32: jne
+; 32: je
 ; 32-NOT: movl
 ; 32-NOT: addl ${{[0-9]+}}, %esp
 ; 32: jmp {{_?}}foo5
diff --git a/test/CodeGen/X86/sink-hoist.ll b/test/CodeGen/X86/sink-hoist.ll
index e13a81719ea..7957eb84967 100644
--- a/test/CodeGen/X86/sink-hoist.ll
+++ b/test/CodeGen/X86/sink-hoist.ll
@@ -7,8 +7,9 @@
 
 ; CHECK: foo:
 ; CHECK-NEXT: testb $1, %dil
-; CHECK-NEXT: je
+; CHECK-NEXT: jne
 ; CHECK-NEXT: divsd
+; CHECK-NEXT: movaps
 ; CHECK-NEXT: ret
 ; CHECK:      divsd
 
@@ -25,10 +26,10 @@ define double @foo(double %x, double %y, i1 %c) nounwind {
 
 ; CHECK: split:
 ; CHECK-NEXT: testb $1, %dil
-; CHECK-NEXT: je
-; CHECK-NEXT: divsd
+; CHECK-NEXT: jne
+; CHECK-NEXT: movaps
 ; CHECK-NEXT: ret
-; CHECK:      movaps
+; CHECK:      divsd
 ; CHECK-NEXT: ret
 define double @split(double %x, double %y, i1 %c) nounwind {
   %a = fdiv double %x, 3.2
diff --git a/test/CodeGen/X86/smul-with-overflow.ll b/test/CodeGen/X86/smul-with-overflow.ll
index 7c2e247c874..7ac3840482a 100644
--- a/test/CodeGen/X86/smul-with-overflow.ll
+++ b/test/CodeGen/X86/smul-with-overflow.ll
@@ -19,7 +19,7 @@ overflow:
   ret i1 false
 ; CHECK: test1:
 ; CHECK: imull
-; CHECK-NEXT: jo
+; CHECK-NEXT: jno
 }
 
 define i1 @test2(i32 %v1, i32 %v2) nounwind {
diff --git a/test/CodeGen/X86/sse41-blend.ll b/test/CodeGen/X86/sse41-blend.ll
index 78604a0e963..1a1017d2c17 100644
--- a/test/CodeGen/X86/sse41-blend.ll
+++ b/test/CodeGen/X86/sse41-blend.ll
@@ -80,3 +80,11 @@ define <2 x double> @B(<2 x double> %x, <2 x double> %y) {
   ret <2 x double> %min
 }
 
+; CHECK: float_crash
+define void @float_crash() nounwind {
+entry:
+  %merge205vector_func.i = select <4 x i1> undef, <4 x double> undef, <4 x double> undef
+  %extract214vector_func.i = extractelement <4 x double> %merge205vector_func.i, i32 0
+  store double %extract214vector_func.i, double addrspace(1)* undef, align 8
+  ret void
+}
diff --git a/test/CodeGen/X86/sub-with-overflow.ll b/test/CodeGen/X86/sub-with-overflow.ll
index 749b5db480f..db8313cecdc 100644
--- a/test/CodeGen/X86/sub-with-overflow.ll
+++ b/test/CodeGen/X86/sub-with-overflow.ll
@@ -20,7 +20,7 @@ overflow:
 
 ; CHECK: func1:
 ; CHECK: subl 20(%esp)
-; CHECK-NEXT: jo
+; CHECK-NEXT: jno
 }
 
 define i1 @func2(i32 %v1, i32 %v2) nounwind {
@@ -40,7 +40,7 @@ carry:
 
 ; CHECK: func2:
 ; CHECK: subl 20(%esp)
-; CHECK-NEXT: jb
+; CHECK-NEXT: jae
 }
 
 declare i32 @printf(i8*, ...) nounwind
diff --git a/test/CodeGen/X86/switch-bt.ll b/test/CodeGen/X86/switch-bt.ll
index 8e393422143..58a5c033854 100644
--- a/test/CodeGen/X86/switch-bt.ll
+++ b/test/CodeGen/X86/switch-bt.ll
@@ -5,11 +5,11 @@
 
 ;      CHECK: movabsq $2305843009482129440, %r
 ; CHECK-NEXT: btq %rax, %r
-; CHECK-NEXT: jb
-; CHECK-NEXT: movl  $671088640, %e
+; CHECK-NEXT: jae
+;     CHECK: movl  $671088640, %e
 ; CHECK-NEXT: btq %rax, %r
-; CHECK-NEXT: jb
-; CHECK-NEXT: testq %rax, %r
+; CHECK-NEXT: jae
+;      CHECK: testq %rax, %r
 ; CHECK-NEXT: j
 
 define void @test(i8* %l) nounwind {
@@ -60,7 +60,7 @@ define void @test2(i32 %x) nounwind ssp {
 ; CHECK-NEXT: movl $91
 ; CHECK-NOT: movl
 ; CHECK-NEXT: btl
-; CHECK-NEXT: jb
+; CHECK-NEXT: jae
 entry:
   switch i32 %x, label %if.end [
     i32 6, label %if.then
@@ -85,7 +85,7 @@ define void @test3(i32 %x) nounwind {
 ; CHECK: cmpl $5
 ; CHECK: ja
 ; CHECK: cmpl $4
-; CHECK: jne
+; CHECK: je
   switch i32 %x, label %if.end [
     i32 0, label %if.then
     i32 1, label %if.then
diff --git a/test/CodeGen/X86/tail-opts.ll b/test/CodeGen/X86/tail-opts.ll
index f1b9f20082f..6e20af5866e 100644
--- a/test/CodeGen/X86/tail-opts.ll
+++ b/test/CodeGen/X86/tail-opts.ll
@@ -113,15 +113,16 @@ altret:
 ; CHECK-NEXT:   jbe .LBB2_3
 ; CHECK-NEXT:   ucomiss %xmm{{[0-2]}}, %xmm{{[0-2]}}
 ; CHECK-NEXT:   ja .LBB2_4
-; CHECK-NEXT: .LBB2_2:
-; CHECK-NEXT:   movb $1, %al
-; CHECK-NEXT:   ret
+; CHECK-NEXT:   jmp .LBB2_2
 ; CHECK-NEXT: .LBB2_3:
 ; CHECK-NEXT:   ucomiss %xmm{{[0-2]}}, %xmm{{[0-2]}}
 ; CHECK-NEXT:   jbe .LBB2_2
 ; CHECK-NEXT: .LBB2_4:
 ; CHECK-NEXT:   xorb %al, %al
 ; CHECK-NEXT:   ret
+; CHECK-NEXT: .LBB2_2:
+; CHECK-NEXT:   movb $1, %al
+; CHECK-NEXT:   ret
 
 define i1 @dont_merge_oddly(float* %result) nounwind {
 entry:
@@ -336,10 +337,10 @@ return:
 
 ; CHECK: two:
 ; CHECK-NOT: XYZ
+; CHECK: ret
 ; CHECK: movl $0, XYZ(%rip)
 ; CHECK: movl $1, XYZ(%rip)
 ; CHECK-NOT: XYZ
-; CHECK: ret
 
 define void @two() nounwind optsize {
 entry:
diff --git a/test/CodeGen/X86/uint64-to-float.ll b/test/CodeGen/X86/uint64-to-float.ll
index e853e7717f1..ca764e7568f 100644
--- a/test/CodeGen/X86/uint64-to-float.ll
+++ b/test/CodeGen/X86/uint64-to-float.ll
@@ -7,13 +7,14 @@ target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f3
 target triple = "x86_64-apple-darwin10.0.0"
 
 ; CHECK: testq %rdi, %rdi
-; CHECK-NEXT: jns LBB0_2
+; CHECK-NEXT: js LBB0_1
+; CHECK: cvtsi2ss
+; CHECK-NEXT: ret
+; CHECK: LBB0_1
 ; CHECK: shrq
 ; CHECK-NEXT: andq
 ; CHECK-NEXT: orq
 ; CHECK-NEXT: cvtsi2ss
-; CHECK: LBB0_2
-; CHECK-NEXT: cvtsi2ss
 define float @test(i64 %a) {
 entry:
   %b = uitofp i64 %a to float
diff --git a/test/CodeGen/X86/vec_shuffle-20.ll b/test/CodeGen/X86/vec_shuffle-20.ll
index b6b8ba6f846..976cd1835b4 100644
--- a/test/CodeGen/X86/vec_shuffle-20.ll
+++ b/test/CodeGen/X86/vec_shuffle-20.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -o /dev/null -march=x86 -mcpu=corei7 -mtriple=i686-apple-darwin9 -stats -info-output-file - | grep asm-printer | grep 2
+; RUN: llc < %s -o /dev/null -march=x86 -mcpu=corei7 -mtriple=i686-apple-darwin9 -stats -info-output-file - | grep asm-printer | grep 3
 
 define <4 x float> @func(<4 x float> %fp0, <4 x float> %fp1) nounwind  {
 entry:
diff --git a/test/CodeGen/X86/xor-icmp.ll b/test/CodeGen/X86/xor-icmp.ll
index 34875ed8995..fd1b0064046 100644
--- a/test/CodeGen/X86/xor-icmp.ll
+++ b/test/CodeGen/X86/xor-icmp.ll
@@ -9,13 +9,13 @@ entry:
 ; X32-NOT: andb
 ; X32-NOT: shrb
 ; X32:     testb $64
-; X32:     jne
+; X32:     je
 
 ; X64:     t:
 ; X64-NOT: setne
 ; X64:     xorl
 ; X64:     testb $64
-; X64:     jne
+; X64:     je
   %0 = and i32 %a, 16384
   %1 = icmp ne i32 %0, 0
   %2 = and i32 %b, 16384
@@ -43,7 +43,7 @@ define i32 @t2(i32 %x, i32 %y) nounwind ssp {
 ; X32: cmpl
 ; X32: sete
 ; X32-NOT: xor
-; X32: jne
+; X32: je
 
 ; X64: t2:
 ; X64: testl
@@ -51,7 +51,7 @@ define i32 @t2(i32 %x, i32 %y) nounwind ssp {
 ; X64: testl
 ; X64: sete
 ; X64-NOT: xor
-; X64: jne
+; X64: je
 entry:
   %0 = icmp eq i32 %x, 0                          ;  [#uses=1]
   %1 = icmp eq i32 %y, 0                          ;  [#uses=1]
diff --git a/test/CodeGen/XCore/ashr.ll b/test/CodeGen/XCore/ashr.ll
index 4514fdb8bf3..03b6b1f1695 100644
--- a/test/CodeGen/XCore/ashr.ll
+++ b/test/CodeGen/XCore/ashr.ll
@@ -30,7 +30,7 @@ not_less:
 }
 ; CHECK: f1:
 ; CHECK-NEXT: ashr r0, r0, 32
-; CHECK-NEXT: bf r0
+; CHECK-NEXT: bt r0
 
 define i32 @f2(i32 %a) {
         %1 = icmp sge i32 %a, 0
@@ -51,9 +51,9 @@ define i32 @f3(i32 %a) {
 }
 ; CHECK: f3:
 ; CHECK-NEXT: ashr r0, r0, 32
-; CHECK-NEXT: bf r0
-; CHECK-NEXT: ldc r0, 10
-; CHECK: ldc r0, 17
+; CHECK-NEXT: bt r0
+; CHECK-NEXT: ldc r0, 17
+; CHECK: ldc r0, 10
 
 define i32 @f4(i32 %a) {
         %1 = icmp sge i32 %a, 0
@@ -62,9 +62,9 @@ define i32 @f4(i32 %a) {
 }
 ; CHECK: f4:
 ; CHECK-NEXT: ashr r0, r0, 32
-; CHECK-NEXT: bf r0
-; CHECK-NEXT: ldc r0, 17
-; CHECK: ldc r0, 10
+; CHECK-NEXT: bt r0
+; CHECK-NEXT: ldc r0, 10
+; CHECK: ldc r0, 17
 
 define i32 @f5(i32 %a) {
         %1 = icmp sge i32 %a, 0
diff --git a/test/MC/ARM/neon-add-encoding.s b/test/MC/ARM/neon-add-encoding.s
index 1fdfa4c13df..50c8f85449f 100644
--- a/test/MC/ARM/neon-add-encoding.s
+++ b/test/MC/ARM/neon-add-encoding.s
@@ -64,31 +64,86 @@
 	vhadd.u16	q8, q8, q9
 @ CHECK: vhadd.u32	q8, q8, q9      @ encoding: [0xe2,0x00,0x60,0xf3]
 	vhadd.u32	q8, q8, q9
-	
-@ CHECK: vrhadd.s8	d16, d16, d17   @ encoding: [0xa1,0x01,0x40,0xf2]
+
+
+	vhadd.s8	d11, d24
+	vhadd.s16	d12, d23
+	vhadd.s32	d13, d22
+	vhadd.u8	d14, d21
+	vhadd.u16	d15, d20
+	vhadd.u32	d16, d19
+	vhadd.s8	q1, q12
+	vhadd.s16	q2, q11
+	vhadd.s32	q3, q10
+	vhadd.u8	q4, q9
+	vhadd.u16	q5, q8
+	vhadd.u32	q6, q7
+
+@ CHECK: vhadd.s8	d11, d11, d24   @ encoding: [0x28,0xb0,0x0b,0xf2]
+@ CHECK: vhadd.s16	d12, d12, d23   @ encoding: [0x27,0xc0,0x1c,0xf2]
+@ CHECK: vhadd.s32	d13, d13, d22   @ encoding: [0x26,0xd0,0x2d,0xf2]
+@ CHECK: vhadd.u8	d14, d14, d21   @ encoding: [0x25,0xe0,0x0e,0xf3]
+@ CHECK: vhadd.u16	d15, d15, d20   @ encoding: [0x24,0xf0,0x1f,0xf3]
+@ CHECK: vhadd.u32	d16, d16, d19   @ encoding: [0xa3,0x00,0x60,0xf3]
+@ CHECK: vhadd.s8	q1, q1, q12     @ encoding: [0x68,0x20,0x02,0xf2]
+@ CHECK: vhadd.s16	q2, q2, q11     @ encoding: [0x66,0x40,0x14,0xf2]
+@ CHECK: vhadd.s32	q3, q3, q10     @ encoding: [0x64,0x60,0x26,0xf2]
+@ CHECK: vhadd.u8	q4, q4, q9      @ encoding: [0x62,0x80,0x08,0xf3]
+@ CHECK: vhadd.u16	q5, q5, q8      @ encoding: [0x60,0xa0,0x1a,0xf3]
+@ CHECK: vhadd.u32	q6, q6, q7      @ encoding: [0x4e,0xc0,0x2c,0xf3]
+
 	vrhadd.s8	d16, d16, d17
-@ CHECK: vrhadd.s16	d16, d16, d17   @ encoding: [0xa1,0x01,0x50,0xf2]
 	vrhadd.s16	d16, d16, d17
-@ CHECK: vrhadd.s32	d16, d16, d17   @ encoding: [0xa1,0x01,0x60,0xf2]
 	vrhadd.s32	d16, d16, d17
-@ CHECK: vrhadd.u8	d16, d16, d17   @ encoding: [0xa1,0x01,0x40,0xf3]
 	vrhadd.u8	d16, d16, d17
-@ CHECK: vrhadd.u16	d16, d16, d17   @ encoding: [0xa1,0x01,0x50,0xf3]
 	vrhadd.u16	d16, d16, d17
-@ CHECK: vrhadd.u32	d16, d16, d17   @ encoding: [0xa1,0x01,0x60,0xf3]
 	vrhadd.u32	d16, d16, d17
-@ CHECK: vrhadd.s8	q8, q8, q9      @ encoding: [0xe2,0x01,0x40,0xf2]
 	vrhadd.s8	q8, q8, q9
-@ CHECK: vrhadd.s16	q8, q8, q9      @ encoding: [0xe2,0x01,0x50,0xf2]
 	vrhadd.s16	q8, q8, q9
-@ CHECK: vrhadd.s32	q8, q8, q9      @ encoding: [0xe2,0x01,0x60,0xf2]
 	vrhadd.s32	q8, q8, q9
-@ CHECK: vrhadd.u8	q8, q8, q9      @ encoding: [0xe2,0x01,0x40,0xf3]
 	vrhadd.u8	q8, q8, q9
-@ CHECK: vrhadd.u16	q8, q8, q9      @ encoding: [0xe2,0x01,0x50,0xf3]
 	vrhadd.u16	q8, q8, q9
-@ CHECK: vrhadd.u32	q8, q8, q9      @ encoding: [0xe2,0x01,0x60,0xf3]
 	vrhadd.u32	q8, q8, q9
+        @ Two-operand forms.
+	vrhadd.s8	d16, d17
+	vrhadd.s16	d16, d17
+	vrhadd.s32	d16, d17
+	vrhadd.u8	d16, d17
+	vrhadd.u16	d16, d17
+	vrhadd.u32	d16, d17
+	vrhadd.s8	q8, q9
+	vrhadd.s16	q8, q9
+	vrhadd.s32	q8, q9
+	vrhadd.u8	q8, q9
+	vrhadd.u16	q8, q9
+	vrhadd.u32	q8, q9
+
+@ CHECK: vrhadd.s8	d16, d16, d17   @ encoding: [0xa1,0x01,0x40,0xf2]
+@ CHECK: vrhadd.s16	d16, d16, d17   @ encoding: [0xa1,0x01,0x50,0xf2]
+@ CHECK: vrhadd.s32	d16, d16, d17   @ encoding: [0xa1,0x01,0x60,0xf2]
+@ CHECK: vrhadd.u8	d16, d16, d17   @ encoding: [0xa1,0x01,0x40,0xf3]
+@ CHECK: vrhadd.u16	d16, d16, d17   @ encoding: [0xa1,0x01,0x50,0xf3]
+@ CHECK: vrhadd.u32	d16, d16, d17   @ encoding: [0xa1,0x01,0x60,0xf3]
+@ CHECK: vrhadd.s8	q8, q8, q9      @ encoding: [0xe2,0x01,0x40,0xf2]
+@ CHECK: vrhadd.s16	q8, q8, q9      @ encoding: [0xe2,0x01,0x50,0xf2]
+@ CHECK: vrhadd.s32	q8, q8, q9      @ encoding: [0xe2,0x01,0x60,0xf2]
+@ CHECK: vrhadd.u8	q8, q8, q9      @ encoding: [0xe2,0x01,0x40,0xf3]
+@ CHECK: vrhadd.u16	q8, q8, q9      @ encoding: [0xe2,0x01,0x50,0xf3]
+@ CHECK: vrhadd.u32	q8, q8, q9      @ encoding: [0xe2,0x01,0x60,0xf3]
+
+@ CHECK: vrhadd.s8	d16, d16, d17   @ encoding: [0xa1,0x01,0x40,0xf2]
+@ CHECK: vrhadd.s16	d16, d16, d17   @ encoding: [0xa1,0x01,0x50,0xf2]
+@ CHECK: vrhadd.s32	d16, d16, d17   @ encoding: [0xa1,0x01,0x60,0xf2]
+@ CHECK: vrhadd.u8	d16, d16, d17   @ encoding: [0xa1,0x01,0x40,0xf3]
+@ CHECK: vrhadd.u16	d16, d16, d17   @ encoding: [0xa1,0x01,0x50,0xf3]
+@ CHECK: vrhadd.u32	d16, d16, d17   @ encoding: [0xa1,0x01,0x60,0xf3]
+@ CHECK: vrhadd.s8	q8, q8, q9      @ encoding: [0xe2,0x01,0x40,0xf2]
+@ CHECK: vrhadd.s16	q8, q8, q9      @ encoding: [0xe2,0x01,0x50,0xf2]
+@ CHECK: vrhadd.s32	q8, q8, q9      @ encoding: [0xe2,0x01,0x60,0xf2]
+@ CHECK: vrhadd.u8	q8, q8, q9      @ encoding: [0xe2,0x01,0x40,0xf3]
+@ CHECK: vrhadd.u16	q8, q8, q9      @ encoding: [0xe2,0x01,0x50,0xf3]
+@ CHECK: vrhadd.u32	q8, q8, q9      @ encoding: [0xe2,0x01,0x60,0xf3]
+
 
 	vqadd.s8	d16, d16, d17
 	vqadd.s16	d16, d16, d17
diff --git a/test/MC/ARM/neon-shift-encoding.s b/test/MC/ARM/neon-shift-encoding.s
index 54ed173c92e..3c97f8b61fb 100644
--- a/test/MC/ARM/neon-shift-encoding.s
+++ b/test/MC/ARM/neon-shift-encoding.s
@@ -250,113 +250,124 @@ _foo:
 @ CHECK: vsli.64 q7, q7, #63            @ encoding: [0xde,0xe5,0xbf,0xf3]
 
 
-@ CHECK: vshll.s8	q8, d16, #7  @ encoding: [0x30,0x0a,0xcf,0xf2]
 	vshll.s8	q8, d16, #7
-@ CHECK: vshll.s16	q8, d16, #15  @ encoding: [0x30,0x0a,0xdf,0xf2]
 	vshll.s16	q8, d16, #15
-@ CHECK: vshll.s32	q8, d16, #31  @ encoding: [0x30,0x0a,0xff,0xf2]
 	vshll.s32	q8, d16, #31
-@ CHECK: vshll.u8	q8, d16, #7  @ encoding: [0x30,0x0a,0xcf,0xf3]
 	vshll.u8	q8, d16, #7
-@ CHECK: vshll.u16	q8, d16, #15  @ encoding: [0x30,0x0a,0xdf,0xf3]
 	vshll.u16	q8, d16, #15
-@ CHECK: vshll.u32	q8, d16, #31  @ encoding: [0x30,0x0a,0xff,0xf3]
 	vshll.u32	q8, d16, #31
-@ CHECK: vshll.i8	q8, d16, #8  @ encoding: [0x20,0x03,0xf2,0xf3]
 	vshll.i8	q8, d16, #8
-@ CHECK: vshll.i16	q8, d16, #16  @ encoding: [0x20,0x03,0xf6,0xf3]
 	vshll.i16	q8, d16, #16
-@ CHECK: vshll.i32	q8, d16, #32  @ encoding: [0x20,0x03,0xfa,0xf3]
 	vshll.i32	q8, d16, #32
-@ CHECK: vshrn.i16	d16, q8, #8  @ encoding: [0x30,0x08,0xc8,0xf2]
+
+@ CHECK: vshll.s8	q8, d16, #7  @ encoding: [0x30,0x0a,0xcf,0xf2]
+@ CHECK: vshll.s16	q8, d16, #15  @ encoding: [0x30,0x0a,0xdf,0xf2]
+@ CHECK: vshll.s32	q8, d16, #31  @ encoding: [0x30,0x0a,0xff,0xf2]
+@ CHECK: vshll.u8	q8, d16, #7  @ encoding: [0x30,0x0a,0xcf,0xf3]
+@ CHECK: vshll.u16	q8, d16, #15  @ encoding: [0x30,0x0a,0xdf,0xf3]
+@ CHECK: vshll.u32	q8, d16, #31  @ encoding: [0x30,0x0a,0xff,0xf3]
+@ CHECK: vshll.i8	q8, d16, #8  @ encoding: [0x20,0x03,0xf2,0xf3]
+@ CHECK: vshll.i16	q8, d16, #16  @ encoding: [0x20,0x03,0xf6,0xf3]
+@ CHECK: vshll.i32	q8, d16, #32  @ encoding: [0x20,0x03,0xfa,0xf3]
+
 	vshrn.i16	d16, q8, #8
-@ CHECK: vshrn.i32	d16, q8, #16  @ encoding: [0x30,0x08,0xd0,0xf2]
 	vshrn.i32	d16, q8, #16
-@ CHECK: vshrn.i64	d16, q8, #32  @ encoding: [0x30,0x08,0xe0,0xf2]
 	vshrn.i64	d16, q8, #32
-@ CHECK: vrshl.s8	d16, d17, d16  @ encoding: [0xa1,0x05,0x40,0xf2]
+
+@ CHECK: vshrn.i16	d16, q8, #8  @ encoding: [0x30,0x08,0xc8,0xf2]
+@ CHECK: vshrn.i32	d16, q8, #16  @ encoding: [0x30,0x08,0xd0,0xf2]
+@ CHECK: vshrn.i64	d16, q8, #32  @ encoding: [0x30,0x08,0xe0,0xf2]
+
 	vrshl.s8	d16, d17, d16
-@ CHECK: vrshl.s16	d16, d17, d16  @ encoding: [0xa1,0x05,0x50,0xf2]
 	vrshl.s16	d16, d17, d16
-@ CHECK: vrshl.s32	d16, d17, d16  @ encoding: [0xa1,0x05,0x60,0xf2]
 	vrshl.s32	d16, d17, d16
-@ CHECK: vrshl.s64	d16, d17, d16  @ encoding: [0xa1,0x05,0x70,0xf2]
 	vrshl.s64	d16, d17, d16
-@ CHECK: vrshl.u8	d16, d17, d16  @ encoding: [0xa1,0x05,0x40,0xf3]
 	vrshl.u8	d16, d17, d16
-@ CHECK: vrshl.u16	d16, d17, d16  @ encoding: [0xa1,0x05,0x50,0xf3]
 	vrshl.u16	d16, d17, d16
-@ CHECK: vrshl.u32	d16, d17, d16  @ encoding: [0xa1,0x05,0x60,0xf3]
 	vrshl.u32	d16, d17, d16
-@ CHECK: vrshl.u64	d16, d17, d16  @ encoding: [0xa1,0x05,0x70,0xf3]
 	vrshl.u64	d16, d17, d16
-@ CHECK: vrshl.s8	q8, q9, q8  @ encoding: [0xe2,0x05,0x40,0xf2]
 	vrshl.s8	q8, q9, q8
-@ CHECK: vrshl.s16	q8, q9, q8  @ encoding: [0xe2,0x05,0x50,0xf2]
 	vrshl.s16	q8, q9, q8
-@ CHECK: vrshl.s32	q8, q9, q8  @ encoding: [0xe2,0x05,0x60,0xf2]
 	vrshl.s32	q8, q9, q8
-@ CHECK: vrshl.s64	q8, q9, q8  @ encoding: [0xe2,0x05,0x70,0xf2]
 	vrshl.s64	q8, q9, q8
-@ CHECK: vrshl.u8	q8, q9, q8  @ encoding: [0xe2,0x05,0x40,0xf3]
 	vrshl.u8	q8, q9, q8
-@ CHECK: vrshl.u16	q8, q9, q8  @ encoding: [0xe2,0x05,0x50,0xf3]
 	vrshl.u16	q8, q9, q8
-@ CHECK: vrshl.u32	q8, q9, q8  @ encoding: [0xe2,0x05,0x60,0xf3]
 	vrshl.u32	q8, q9, q8
-@ CHECK: vrshl.u64	q8, q9, q8  @ encoding: [0xe2,0x05,0x70,0xf3]
 	vrshl.u64	q8, q9, q8
-@ CHECK: vrshr.s8	d16, d16, #8  @ encoding: [0x30,0x02,0xc8,0xf2]
+
+@ CHECK: vrshl.s8	d16, d17, d16  @ encoding: [0xa1,0x05,0x40,0xf2]
+@ CHECK: vrshl.s16	d16, d17, d16  @ encoding: [0xa1,0x05,0x50,0xf2]
+@ CHECK: vrshl.s32	d16, d17, d16  @ encoding: [0xa1,0x05,0x60,0xf2]
+@ CHECK: vrshl.s64	d16, d17, d16  @ encoding: [0xa1,0x05,0x70,0xf2]
+@ CHECK: vrshl.u8	d16, d17, d16  @ encoding: [0xa1,0x05,0x40,0xf3]
+@ CHECK: vrshl.u16	d16, d17, d16  @ encoding: [0xa1,0x05,0x50,0xf3]
+@ CHECK: vrshl.u32	d16, d17, d16  @ encoding: [0xa1,0x05,0x60,0xf3]
+@ CHECK: vrshl.u64	d16, d17, d16  @ encoding: [0xa1,0x05,0x70,0xf3]
+@ CHECK: vrshl.s8	q8, q9, q8  @ encoding: [0xe2,0x05,0x40,0xf2]
+@ CHECK: vrshl.s16	q8, q9, q8  @ encoding: [0xe2,0x05,0x50,0xf2]
+@ CHECK: vrshl.s32	q8, q9, q8  @ encoding: [0xe2,0x05,0x60,0xf2]
+@ CHECK: vrshl.s64	q8, q9, q8  @ encoding: [0xe2,0x05,0x70,0xf2]
+@ CHECK: vrshl.u8	q8, q9, q8  @ encoding: [0xe2,0x05,0x40,0xf3]
+@ CHECK: vrshl.u16	q8, q9, q8  @ encoding: [0xe2,0x05,0x50,0xf3]
+@ CHECK: vrshl.u32	q8, q9, q8  @ encoding: [0xe2,0x05,0x60,0xf3]
+@ CHECK: vrshl.u64	q8, q9, q8  @ encoding: [0xe2,0x05,0x70,0xf3]
+
 	vrshr.s8	d16, d16, #8
-@ CHECK: vrshr.s16	d16, d16, #16  @ encoding: [0x30,0x02,0xd0,0xf2]
 	vrshr.s16	d16, d16, #16
-@ CHECK: vrshr.s32	d16, d16, #32  @ encoding: [0x30,0x02,0xe0,0xf2]
 	vrshr.s32	d16, d16, #32
-@ CHECK: vrshr.s64	d16, d16, #64  @ encoding: [0xb0,0x02,0xc0,0xf2]
 	vrshr.s64	d16, d16, #64
-@ CHECK: vrshr.u8	d16, d16, #8  @ encoding: [0x30,0x02,0xc8,0xf3]
 	vrshr.u8	d16, d16, #8
-@ CHECK: vrshr.u16	d16, d16, #16  @ encoding: [0x30,0x02,0xd0,0xf3]
 	vrshr.u16	d16, d16, #16
-@ CHECK: vrshr.u32	d16, d16, #32  @ encoding: [0x30,0x02,0xe0,0xf3]
 	vrshr.u32	d16, d16, #32
-@ CHECK: vrshr.u64	d16, d16, #64  @ encoding: [0xb0,0x02,0xc0,0xf3]
 	vrshr.u64	d16, d16, #64
-@ CHECK: vrshr.s8	q8, q8, #8  @ encoding: [0x70,0x02,0xc8,0xf2]
 	vrshr.s8	q8, q8, #8
-@ CHECK: vrshr.s16	q8, q8, #16  @ encoding: [0x70,0x02,0xd0,0xf2]
 	vrshr.s16	q8, q8, #16
-@ CHECK: vrshr.s32	q8, q8, #32  @ encoding: [0x70,0x02,0xe0,0xf2]
 	vrshr.s32	q8, q8, #32
-@ CHECK: vrshr.s64	q8, q8, #64  @ encoding: [0xf0,0x02,0xc0,0xf2]
 	vrshr.s64	q8, q8, #64
-@ CHECK: vrshr.u8	q8, q8, #8  @ encoding: [0x70,0x02,0xc8,0xf3]
 	vrshr.u8	q8, q8, #8
-@ CHECK: vrshr.u16	q8, q8, #16  @ encoding: [0x70,0x02,0xd0,0xf3]
 	vrshr.u16	q8, q8, #16
-@ CHECK: vrshr.u32	q8, q8, #32  @ encoding: [0x70,0x02,0xe0,0xf3]
 	vrshr.u32	q8, q8, #32
-@ CHECK: vrshr.u64	q8, q8, #64  @ encoding: [0xf0,0x02,0xc0,0xf3]
 	vrshr.u64	q8, q8, #64
-@ CHECK: vrshrn.i16	d16, q8, #8  @ encoding: [0x70,0x08,0xc8,0xf2]
+
+@ CHECK: vrshr.s8	d16, d16, #8  @ encoding: [0x30,0x02,0xc8,0xf2]
+@ CHECK: vrshr.s16	d16, d16, #16  @ encoding: [0x30,0x02,0xd0,0xf2]
+@ CHECK: vrshr.s32	d16, d16, #32  @ encoding: [0x30,0x02,0xe0,0xf2]
+@ CHECK: vrshr.s64	d16, d16, #64  @ encoding: [0xb0,0x02,0xc0,0xf2]
+@ CHECK: vrshr.u8	d16, d16, #8  @ encoding: [0x30,0x02,0xc8,0xf3]
+@ CHECK: vrshr.u16	d16, d16, #16  @ encoding: [0x30,0x02,0xd0,0xf3]
+@ CHECK: vrshr.u32	d16, d16, #32  @ encoding: [0x30,0x02,0xe0,0xf3]
+@ CHECK: vrshr.u64	d16, d16, #64  @ encoding: [0xb0,0x02,0xc0,0xf3]
+@ CHECK: vrshr.s8	q8, q8, #8  @ encoding: [0x70,0x02,0xc8,0xf2]
+@ CHECK: vrshr.s16	q8, q8, #16  @ encoding: [0x70,0x02,0xd0,0xf2]
+@ CHECK: vrshr.s32	q8, q8, #32  @ encoding: [0x70,0x02,0xe0,0xf2]
+@ CHECK: vrshr.s64	q8, q8, #64  @ encoding: [0xf0,0x02,0xc0,0xf2]
+@ CHECK: vrshr.u8	q8, q8, #8  @ encoding: [0x70,0x02,0xc8,0xf3]
+@ CHECK: vrshr.u16	q8, q8, #16  @ encoding: [0x70,0x02,0xd0,0xf3]
+@ CHECK: vrshr.u32	q8, q8, #32  @ encoding: [0x70,0x02,0xe0,0xf3]
+@ CHECK: vrshr.u64	q8, q8, #64  @ encoding: [0xf0,0x02,0xc0,0xf3]
+
+
 	vrshrn.i16	d16, q8, #8
-@ CHECK: vrshrn.i32	d16, q8, #16  @ encoding: [0x70,0x08,0xd0,0xf2]
 	vrshrn.i32	d16, q8, #16
-@ CHECK: vrshrn.i64	d16, q8, #32  @ encoding: [0x70,0x08,0xe0,0xf2]
 	vrshrn.i64	d16, q8, #32
-@ CHECK: vqrshrn.s16	d16, q8, #4  @ encoding: [0x70,0x09,0xcc,0xf2]
 	vqrshrn.s16	d16, q8, #4
-@ CHECK: vqrshrn.s32	d16, q8, #13  @ encoding: [0x70,0x09,0xd3,0xf2]
 	vqrshrn.s32	d16, q8, #13
-@ CHECK: vqrshrn.s64	d16, q8, #13  @ encoding: [0x70,0x09,0xf3,0xf2]
 	vqrshrn.s64	d16, q8, #13
-@ CHECK: vqrshrn.u16	d16, q8, #4  @ encoding: [0x70,0x09,0xcc,0xf3]
 	vqrshrn.u16	d16, q8, #4
-@ CHECK: vqrshrn.u32	d16, q8, #13  @ encoding: [0x70,0x09,0xd3,0xf3]
 	vqrshrn.u32	d16, q8, #13
-@ CHECK: vqrshrn.u64	d16, q8, #13  @ encoding: [0x70,0x09,0xf3,0xf3]
 	vqrshrn.u64	d16, q8, #13
 
+@ CHECK: vrshrn.i16	d16, q8, #8  @ encoding: [0x70,0x08,0xc8,0xf2]
+@ CHECK: vrshrn.i32	d16, q8, #16  @ encoding: [0x70,0x08,0xd0,0xf2]
+@ CHECK: vrshrn.i64	d16, q8, #32  @ encoding: [0x70,0x08,0xe0,0xf2]
+@ CHECK: vqrshrn.s16	d16, q8, #4  @ encoding: [0x70,0x09,0xcc,0xf2]
+@ CHECK: vqrshrn.s32	d16, q8, #13  @ encoding: [0x70,0x09,0xd3,0xf2]
+@ CHECK: vqrshrn.s64	d16, q8, #13  @ encoding: [0x70,0x09,0xf3,0xf2]
+@ CHECK: vqrshrn.u16	d16, q8, #4  @ encoding: [0x70,0x09,0xcc,0xf3]
+@ CHECK: vqrshrn.u32	d16, q8, #13  @ encoding: [0x70,0x09,0xd3,0xf3]
+@ CHECK: vqrshrn.u64	d16, q8, #13  @ encoding: [0x70,0x09,0xf3,0xf3]
+
+
 @ Optional destination operand variants.
         vshl.s8 q4, q5
         vshl.s16 q4, q5
@@ -418,6 +429,41 @@ _foo:
 @ CHECK: vshl.i32	d4, d4, #17     @ encoding: [0x14,0x45,0xb1,0xf2]
 @ CHECK: vshl.i64	d4, d4, #43     @ encoding: [0x94,0x45,0xab,0xf2]
 
+        @ Two-operand VRSHL forms.
+	vrshl.s8	d11, d4
+	vrshl.s16	d12, d5
+	vrshl.s32	d13, d6
+	vrshl.s64	d14, d7
+	vrshl.u8	d15, d8
+	vrshl.u16	d16, d9
+	vrshl.u32	d17, d10
+	vrshl.u64	d18, d11
+	vrshl.s8	q1, q8
+	vrshl.s16	q2, q15
+	vrshl.s32	q3, q14
+	vrshl.s64	q4, q13
+	vrshl.u8	q5, q12
+	vrshl.u16	q6, q11
+	vrshl.u32	q7, q10
+	vrshl.u64	q8, q9
+
+@ CHECK: vrshl.s8	d11, d11, d4    @ encoding: [0x0b,0xb5,0x04,0xf2]
+@ CHECK: vrshl.s16	d12, d12, d5    @ encoding: [0x0c,0xc5,0x15,0xf2]
+@ CHECK: vrshl.s32	d13, d13, d6    @ encoding: [0x0d,0xd5,0x26,0xf2]
+@ CHECK: vrshl.s64	d14, d14, d7    @ encoding: [0x0e,0xe5,0x37,0xf2]
+@ CHECK: vrshl.u8	d15, d15, d8    @ encoding: [0x0f,0xf5,0x08,0xf3]
+@ CHECK: vrshl.u16	d16, d16, d9    @ encoding: [0x20,0x05,0x59,0xf3]
+@ CHECK: vrshl.u32	d17, d17, d10   @ encoding: [0x21,0x15,0x6a,0xf3]
+@ CHECK: vrshl.u64	d18, d18, d11   @ encoding: [0x22,0x25,0x7b,0xf3]
+@ CHECK: vrshl.s8	q1, q1, q8      @ encoding: [0xc2,0x25,0x00,0xf2]
+@ CHECK: vrshl.s16	q2, q2, q15     @ encoding: [0xc4,0x45,0x1e,0xf2]
+@ CHECK: vrshl.s32	q3, q3, q14     @ encoding: [0xc6,0x65,0x2c,0xf2]
+@ CHECK: vrshl.s64	q4, q4, q13     @ encoding: [0xc8,0x85,0x3a,0xf2]
+@ CHECK: vrshl.u8	q5, q5, q12     @ encoding: [0xca,0xa5,0x08,0xf3]
+@ CHECK: vrshl.u16	q6, q6, q11     @ encoding: [0xcc,0xc5,0x16,0xf3]
+@ CHECK: vrshl.u32	q7, q7, q10     @ encoding: [0xce,0xe5,0x24,0xf3]
+@ CHECK: vrshl.u64	q8, q8, q9      @ encoding: [0xe0,0x05,0x72,0xf3]
+
 
 @ Two-operand forms.
 	vshr.s8	d15, #8
diff --git a/test/MC/ARM/neon-sub-encoding.s b/test/MC/ARM/neon-sub-encoding.s
index 0622e192bc6..8eb38a5f711 100644
--- a/test/MC/ARM/neon-sub-encoding.s
+++ b/test/MC/ARM/neon-sub-encoding.s
@@ -132,3 +132,29 @@
 	vrsubhn.i32	d16, q8, q9
 @ CHECK: vrsubhn.i64	d16, q8, q9     @ encoding: [0xa2,0x06,0xe0,0xf3]
 	vrsubhn.i64	d16, q8, q9
+
+	vhsub.s8	d11, d24
+	vhsub.s16	d12, d23
+	vhsub.s32	d13, d22
+	vhsub.u8	d14, d21
+	vhsub.u16	d15, d20
+	vhsub.u32	d16, d19
+	vhsub.s8	q1, q12
+	vhsub.s16	q2, q11
+	vhsub.s32	q3, q10
+	vhsub.u8	q4, q9
+	vhsub.u16	q5, q8
+	vhsub.u32	q6, q7
+
+@ CHECK: vhsub.s8	d11, d11, d24   @ encoding: [0x28,0xb2,0x0b,0xf2]
+@ CHECK: vhsub.s16	d12, d12, d23   @ encoding: [0x27,0xc2,0x1c,0xf2]
+@ CHECK: vhsub.s32	d13, d13, d22   @ encoding: [0x26,0xd2,0x2d,0xf2]
+@ CHECK: vhsub.u8	d14, d14, d21   @ encoding: [0x25,0xe2,0x0e,0xf3]
+@ CHECK: vhsub.u16	d15, d15, d20   @ encoding: [0x24,0xf2,0x1f,0xf3]
+@ CHECK: vhsub.u32	d16, d16, d19   @ encoding: [0xa3,0x02,0x60,0xf3]
+@ CHECK: vhsub.s8	q1, q1, q12     @ encoding: [0x68,0x22,0x02,0xf2]
+@ CHECK: vhsub.s16	q2, q2, q11     @ encoding: [0x66,0x42,0x14,0xf2]
+@ CHECK: vhsub.s32	q3, q3, q10     @ encoding: [0x64,0x62,0x26,0xf2]
+@ CHECK: vhsub.u8	q4, q4, q9      @ encoding: [0x62,0x82,0x08,0xf3]
+@ CHECK: vhsub.u16	q5, q5, q8      @ encoding: [0x60,0xa2,0x1a,0xf3]
+@ CHECK: vhsub.u32	q6, q6, q7      @ encoding: [0x4e,0xc2,0x2c,0xf3]
diff --git a/test/MC/AsmParser/macro-args.s b/test/MC/AsmParser/macro-args.s
index 4b878999e42..13b197a55a8 100644
--- a/test/MC/AsmParser/macro-args.s
+++ b/test/MC/AsmParser/macro-args.s
@@ -18,3 +18,27 @@ bar
 
 // CHECK: .long 3
 // CHECK: .long 0
+
+
+.macro top
+    middle _$0, $1
+.endm
+.macro middle
+     $0:
+    .if $n > 1
+        bottom $1
+    .endif
+.endm
+.macro bottom
+    .set fred, $0
+.endm
+
+.text
+
+top foo
+top bar, 42
+
+// CHECK: _foo:
+// CHECK-NOT: fred
+// CHECK: _bar
+// CHECK-NEXT: fred = 42
diff --git a/test/MC/Disassembler/ARM/arm-tests.txt b/test/MC/Disassembler/ARM/arm-tests.txt
index ce1446b02b1..471076a19bd 100644
--- a/test/MC/Disassembler/ARM/arm-tests.txt
+++ b/test/MC/Disassembler/ARM/arm-tests.txt
@@ -321,3 +321,6 @@
 # CHECK: ldmgt	sp!, {r9}
 0x00 0x02 0xbd 0xc8
 
+# CHECK: cdp2	p10, #0, c6, c12, c0, #7
+0xe0 0x6a 0x0c 0xfe
+
diff --git a/test/MC/Disassembler/ARM/invalid-MRRC2-arm.txt b/test/MC/Disassembler/ARM/invalid-MRRC2-arm.txt
new file mode 100644
index 00000000000..aaae6ce2e40
--- /dev/null
+++ b/test/MC/Disassembler/ARM/invalid-MRRC2-arm.txt
@@ -0,0 +1,4 @@
+# RUN: llvm-mc --disassemble %s -triple=armv7-linux-gnueabi |& FileCheck %s
+
+# CHECK: invalid instruction encoding
+0x00 0x1a 0x50 0xfc
diff --git a/test/MC/Disassembler/ARM/neon.txt b/test/MC/Disassembler/ARM/neon.txt
index 58fe20eaa27..c5dbee3aa89 100644
--- a/test/MC/Disassembler/ARM/neon.txt
+++ b/test/MC/Disassembler/ARM/neon.txt
@@ -2243,3 +2243,40 @@
 # CHECK: vld4.16	{d8, d10, d12, d14}, [r4] 
 0x8f 0x81 0x24 0xf4
 # CHECK: vld4.32	{d8, d10, d12, d14}, [r4] 
+
+# rdar://11256967
+0x0f 0x0d 0xa2 0xf4
+# CHECK: vld2.8	{d0[], d1[]}, [r2]      
+0x4f 0x0d 0xa2 0xf4
+# CHECK: vld2.16	{d0[], d1[]}, [r2]      
+0x8f 0x0d 0xa2 0xf4
+# CHECK: vld2.32	{d0[], d1[]}, [r2]      
+0x0d 0x0d 0xa2 0xf4
+# CHECK: vld2.8	{d0[], d1[]}, [r2]!     
+0x4d 0x0d 0xa2 0xf4
+# CHECK: vld2.16	{d0[], d1[]}, [r2]!     
+0x8d 0x0d 0xa2 0xf4
+# CHECK: vld2.32	{d0[], d1[]}, [r2]!     
+0x03 0x0d 0xa2 0xf4
+# CHECK: vld2.8	{d0[], d1[]}, [r2], r3  
+0x43 0x0d 0xa2 0xf4
+# CHECK: vld2.16	{d0[], d1[]}, [r2], r3  
+0x83 0x0d 0xa2 0xf4
+# CHECK: vld2.32	{d0[], d1[]}, [r2], r3  
+0x2f 0x0d 0xa3 0xf4
+# CHECK: vld2.8	{d0[], d2[]}, [r3]      
+0x6f 0x0d 0xa3 0xf4
+# CHECK: vld2.16	{d0[], d2[]}, [r3]      
+0xaf 0x0d 0xa3 0xf4
+# CHECK: vld2.32	{d0[], d2[]}, [r3]      
+0x2d 0x0d 0xa3 0xf4
+# CHECK: vld2.8	{d0[], d2[]}, [r3]!     
+0x6d 0x0d 0xa3 0xf4
+# CHECK: vld2.16	{d0[], d2[]}, [r3]!     
+0xad 0x0d 0xa3 0xf4
+# CHECK: vld2.32	{d0[], d2[]}, [r3]!     
+0x24 0x0d 0xa3 0xf4
+# CHECK: vld2.8	{d0[], d2[]}, [r3], r4  
+0x64 0x0d 0xa3 0xf4
+0xa4 0x0d 0xa3 0xf4
+# CHECK: vld2.32	{d0[], d2[]}, [r3], r4  
diff --git a/test/MC/Disassembler/ARM/neont2.txt b/test/MC/Disassembler/ARM/neont2.txt
index efe7e60ddae..65cd2304149 100644
--- a/test/MC/Disassembler/ARM/neont2.txt
+++ b/test/MC/Disassembler/ARM/neont2.txt
@@ -1960,3 +1960,41 @@
 # CHECK: vld4.16	{d8, d10, d12, d14}, [r4] 
 0x24 0xf9 0x8f 0x81
 # CHECK: vld4.32	{d8, d10, d12, d14}, [r4] 
+
+# rdar://11256967
+0xa2 0xf9 0x0f 0x0d
+# CHECK: vld2.8	{d0[], d1[]}, [r2]      
+0xa2 0xf9 0x4f 0x0d
+# CHECK: vld2.16	{d0[], d1[]}, [r2]      
+0xa2 0xf9 0x8f 0x0d
+# CHECK: vld2.32	{d0[], d1[]}, [r2]      
+0xa2 0xf9 0x0d 0x0d
+# CHECK: vld2.8	{d0[], d1[]}, [r2]!     
+0xa2 0xf9 0x4d 0x0d
+# CHECK: vld2.16	{d0[], d1[]}, [r2]!     
+0xa2 0xf9 0x8d 0x0d
+# CHECK: vld2.32	{d0[], d1[]}, [r2]!     
+0xa2 0xf9 0x03 0x0d
+# CHECK: vld2.8	{d0[], d1[]}, [r2], r3  
+0xa2 0xf9 0x43 0x0d
+# CHECK: vld2.16	{d0[], d1[]}, [r2], r3  
+0xa2 0xf9 0x83 0x0d
+# CHECK: vld2.32	{d0[], d1[]}, [r2], r3  
+0xa3 0xf9 0x2f 0x0d
+# CHECK: vld2.8	{d0[], d2[]}, [r3]      
+0xa3 0xf9 0x6f 0x0d
+# CHECK: vld2.16	{d0[], d2[]}, [r3]      
+0xa3 0xf9 0xaf 0x0d
+# CHECK: vld2.32	{d0[], d2[]}, [r3]      
+0xa3 0xf9 0x2d 0x0d
+# CHECK: vld2.8	{d0[], d2[]}, [r3]!     
+0xa3 0xf9 0x6d 0x0d
+# CHECK: vld2.16	{d0[], d2[]}, [r3]!     
+0xa3 0xf9 0xad 0x0d
+# CHECK: vld2.32	{d0[], d2[]}, [r3]!     
+0xa3 0xf9 0x24 0x0d
+# CHECK: vld2.8	{d0[], d2[]}, [r3], r4  
+0xa3 0xf9 0x64 0x0d
+# CHECK: vld2.16	{d0[], d2[]}, [r3], r4  
+0xa3 0xf9 0xa4 0x0d
+# CHECK: vld2.32	{d0[], d2[]}, [r3], r4  
diff --git a/test/MC/Disassembler/ARM/unpredictable-AI1cmp-arm.txt b/test/MC/Disassembler/ARM/unpredictable-AI1cmp-arm.txt
new file mode 100644
index 00000000000..dac4390cde8
--- /dev/null
+++ b/test/MC/Disassembler/ARM/unpredictable-AI1cmp-arm.txt
@@ -0,0 +1,30 @@
+# RUN: llvm-mc --disassemble %s -triple=armv7-linux-gnueabi |& FileCheck %s
+
+# CHECK: potentially undefined
+# CHECK: 0x01 0x10 0x50 0x03
+0x01 0x10 0x50 0x03
+
+# CHECK: potentially undefined
+# CHECK: 0x82 0x10  0x50 0x01
+0x82 0x10 0x50 0x01
+
+# CHECK: potentially undefined
+# CHECK: 0x02 0x10 0x50 0x01
+0x02 0x10 0x50 0x01
+
+# CHECK: potentially undefined
+# CHECK: 0x1f 0x01 0x52 0x01
+0x1f 0x01 0x52 0x01
+
+# CHECK: potentially undefined
+# CHECK: 0x10 0x11 0x52 0x01
+0x10 0x11 0x52 0x01
+
+# CHECK: potentially undefined
+# CHECK: 0x10 0x0f 0x51 0x01
+0x10 0x0f 0x51 0x01
+
+# CHECK: potentially undefined
+# CHECK: 0x10 0x01 0x5f 0x01
+0x10 0x01 0x5f 0x01
+
diff --git a/test/MC/Disassembler/ARM/unpredictable-MRRC2-arm.txt b/test/MC/Disassembler/ARM/unpredictable-MRRC2-arm.txt
new file mode 100644
index 00000000000..26b286dbf4e
--- /dev/null
+++ b/test/MC/Disassembler/ARM/unpredictable-MRRC2-arm.txt
@@ -0,0 +1,13 @@
+# RUN: llvm-mc --disassemble %s -triple=armv7-linux-gnueabi |& FileCheck %s
+
+# CHECK: potentially undefined
+# CHECK: 0x00 0x10 0x51 0xfc
+0x00 0x10 0x51 0xfc
+
+# CHECK: potentially undefined
+# CHECK: 0x00 0xf0 0x41 0x0c
+0x00 0xf0 0x41 0x0c
+
+# CHECK: potentially undefined
+# CHECK: 0x00 0x00 0x4f 0x0c
+0x00 0x00 0x4f 0x0c
diff --git a/test/MC/Disassembler/ARM/unpredictable-MRS-arm.txt b/test/MC/Disassembler/ARM/unpredictable-MRS-arm.txt
new file mode 100644
index 00000000000..3e472cdbfb1
--- /dev/null
+++ b/test/MC/Disassembler/ARM/unpredictable-MRS-arm.txt
@@ -0,0 +1,18 @@
+# RUN: llvm-mc --disassemble %s -triple=arm-apple-darwin9 |& FileCheck %s
+
+# CHECK: warning: potentially undefined
+# CHECK: 0x00 0xf0 0x0f 0x01
+0x00 0xf0 0x0f 0x01
+
+# CHECK: warning: potentially undefined
+# CHECK: 0x00 0xf0 0x4f 0x01
+0x00 0xf0 0x4f 0x01
+
+# CHECK: warning: potentially undefined
+# CHECK: 0x0f 0x0d 0x01 0x01
+0x0f 0x0d 0x01 0x01
+
+# CHECK: warning: potentially undefined
+# CHECK: 0x0f 0x0d 0x40 0x01
+0x0f 0x0d 0x40 0x01
+
diff --git a/test/MC/Disassembler/ARM/unpredictable-swp-arm.txt b/test/MC/Disassembler/ARM/unpredictable-swp-arm.txt
new file mode 100644
index 00000000000..64bb171bf81
--- /dev/null
+++ b/test/MC/Disassembler/ARM/unpredictable-swp-arm.txt
@@ -0,0 +1,26 @@
+# RUN: llvm-mc --disassemble %s -triple=armv7-linux-gnueabi |& FileCheck %s
+
+# CHECK: potentially undefined
+# CHECK: 0x9f 0x10 0x03 0x01
+0x9f 0x10 0x03 0x01
+
+# CHECK: potentially undefined
+# CHECK: 0x90 0xf0 0x03 0x01
+0x90 0xf0 0x03 0x01
+
+# CHECK: potentially undefined
+# CHECK: 0x90 0x1f 0x03 0x01
+0x90 0x1f 0x03 0x01
+
+# CHECK: potentially undefined
+# CHECK: 0x90 0x10 0x0f 0x01
+0x90 0x10 0x0f 0x01
+
+# CHECK: potentially undefined
+# CHECK: 0x90 0x10 0x01 0x01
+0x90 0x10 0x01 0x01
+
+# CHECK: potentially undefined
+# CHECK: 0x90 0x10 0x00 0x01
+0x90 0x10 0x00 0x01
+
diff --git a/test/MC/Disassembler/Mips/mips32.txt b/test/MC/Disassembler/Mips/mips32.txt
new file mode 100644
index 00000000000..591d8c444af
--- /dev/null
+++ b/test/MC/Disassembler/Mips/mips32.txt
@@ -0,0 +1,421 @@
+# RUN: llvm-mc --disassemble %s -triple=mips-unknown-linux
+
+# CHECK: abs.d $f12,$f14
+0x46 0x20 0x39 0x85
+
+# CHECK: abs.s $f6,$f7
+0x46 0x00 0x39 0x85
+
+# CHECK: add t1,a2,a3
+0x00 0xc7 0x48 0x20
+
+# CHECK: add.d $f18,$f12,$f14
+0x46 0x27 0x32 0x40
+
+# CHECK: add.s $f9,$f6,$f7
+0x46 0x07 0x32 0x40
+
+# CHECK: addi t1,a2,17767
+0x20 0xc9 0x45 0x67
+
+# CHECK: addiu t1,a2,-15001
+0x24 0xc9 0xc5 0x67
+
+# CHECK: addu t1,a2,a3
+0x00 0xc7 0x48 0x21
+
+# CHECK: and t1,a2,a3
+0x00 0xc7 0x48 0x24
+
+# CHECK: andi t1,a2,0x4567
+0x30 0xc9 0x45 0x67
+
+# CHECK: b 00000534
+0x10 0x00 0x01 0x4c
+
+# CHECK: bal 00000534
+0x04 0x11 0x01 0x4c
+
+# CHECK: bc1f 00000534
+0x45 0x00 0x01 0x4c
+
+# CHECK: bc1t 00000534
+0x45 0x01 0x01 0x4c
+
+# CHECK: beq t1,a2,00000534
+0x11 0x26 0x01 0x4c
+
+# CHECK: bgez a2,00000534
+0x04 0xc1 0x01 0x4c
+
+# CHECK: bgezal a2,00000534
+0x04 0xd1 0x01 0x4c
+
+# CHECK: bgtz a2,00000534
+0x1c 0xc0 0x01 0x4c
+
+# CHECK: blez a2,00000534
+0x18 0xc0 0x01 0x4c
+
+# CHECK: bne t1,a2,00000534
+0x15 0x26 0x01 0x4c
+
+# CHECK: c.eq.d $f12,$f14
+0x46 0x27 0x30 0x32
+
+# CHECK: c.eq.s $f6,$f7
+0x46 0x07 0x30 0x32
+
+# CHECK: c.f.d $f12,$f14
+0x46 0x27 0x30 0x30
+
+# CHECK: c.f.s $f6,$f7
+0x46 0x07 0x30 0x30
+
+# CHECK: c.le.d $f12,$f14
+0x46 0x27 0x30 0x3e
+
+# CHECK: c.le.s $f6,$f7
+0x46 0x07 0x30 0x3e
+
+# CHECK: c.lt.d $f12,$f14
+0x46 0x27 0x30 0x3c
+
+# CHECK: c.lt.s $f6,$f7
+0x46 0x07 0x30 0x3c
+
+# CHECK: c.nge.d $f12,$f14
+0x46 0x27 0x30 0x3d
+
+# CHECK: c.nge.s $f6,$f7
+0x46 0x07 0x30 0x3d
+
+# CHECK: c.ngl.d $f12,$f14
+0x46 0x27 0x30 0x3b
+
+# CHECK: c.ngl.s $f6,$f7
+0x46 0x07 0x30 0x3b
+
+# CHECK: c.ngle.d $f12,$f14
+0x46 0x27 0x30 0x39
+
+# CHECK: c.ngle.s $f6,$f7
+0x46 0x07 0x30 0x39
+
+# CHECK: c.ngt.d $f12,$f14
+0x46 0x27 0x30 0x3f
+
+# CHECK: c.ngt.s $f6,$f7
+0x46 0x07 0x30 0x3f
+
+# CHECK: c.ole.d $f12,$f14
+0x46 0x27 0x30 0x36
+
+# CHECK: c.ole.s $f6,$f7
+0x46 0x07 0x30 0x36
+
+# CHECK: c.olt.d $f12,$f14
+0x46 0x27 0x30 0x34
+
+# CHECK: c.olt.s $f6,$f7
+0x46 0x07 0x30 0x34
+
+# CHECK: c.seq.d $f12,$f14
+0x46 0x27 0x30 0x3a
+
+# CHECK: c.seq.s $f6,$f7
+0x46 0x07 0x30 0x3a
+
+# CHECK: c.sf.d $f12,$f14
+0x46 0x27 0x30 0x38
+
+# CHECK: c.sf.s $f6,$f7
+0x46 0x07 0x30 0x38
+
+# CHECK: c.ueq.d $f12,$f14
+0x46 0x27 0x30 0x33
+
+# CHECK: c.ueq.s $f28,$f18
+0x46 0x12 0xe0 0x33
+
+# CHECK: c.ule.d $f12,$f14
+0x46 0x27 0x30 0x37
+
+# CHECK: c.ule.s $f6,$f7
+0x46 0x07 0x30 0x37
+
+# CHECK: c.ult.d $f12,$f14
+0x46 0x27 0x30 0x35
+
+# CHECK: c.ult.s $f6,$f7
+0x46 0x07 0x30 0x35
+
+# CHECK: c.un.d $f12,$f14
+0x46 0x27 0x30 0x31
+
+# CHECK: c.un.s $f6,$f7
+0x46 0x07 0x30 0x31
+
+# CHECK: ceil.w.d $f12,$f14
+0x46 0x20 0x39 0x8e
+
+# CHECK: ceil.w.s $f6,$f7
+0x46 0x00 0x39 0x8e
+
+# CHECK: cfc1 a2,$7
+0x44 0x46 0x38 0x00
+
+# CHECK: clo a2,a3
+0x70 0xe6 0x30 0x21
+
+# CHECK: clz a2,a3
+0x70 0xe6 0x30 0x20
+
+# CHECK: ctc1 a2,$7
+0x44 0xc6 0x38 0x00
+
+# CHECK: cvt.d.s $f6,$f7
+0x46 0x00 0x38 0xa1
+
+# CHECK: cvt.d.w $f12,$f14
+0x46 0x80 0x38 0xa1
+
+# CHECK: cvt.l.d $f12,$f14
+0x46 0x20 0x39 0xa5
+
+# CHECK: cvt.l.s $f6,$f7
+0x46 0x00 0x39 0xa5
+
+# CHECK: cvt.s.d $f12,$f14
+0x46 0x20 0x39 0xa0
+
+# CHECK: cvt.s.w $f6,$f7
+0x46 0x80 0x39 0xa0
+
+# CHECK: cvt.w.d $f12,$f14
+0x46 0x20 0x39 0xa4
+
+# CHECK: cvt.w.s $f6,$f7
+0x46 0x00 0x39 0xa4
+
+# CHECK: floor.w.d $f12,$f14
+0x46 0x20 0x39 0x8f
+
+# CHECK: floor.w.s $f6,$f7
+0x46 0x00 0x39 0x8f
+
+# CHECK: j 00000530
+0x08 0x00 0x01 0x4c
+
+# CHECK: jal 00000530
+0x0c 0x00 0x01 0x4c
+
+# CHECK: jalr a2,a3
+0x00 0xe0 0xf8 0x09
+
+# CHECK: jr a3
+0x00 0xe0 0x00 0x08
+
+# CHECK: lb  a0,9158(a1)
+0x80 0xa4 0x23 0xc6
+
+# CHECK: lbu a0,6(a1)
+0x90 0xa4 0x00 0x06
+
+# CHECK: ldc1  $f9,9158(a3)
+0xd4 0xe9 0x23 0xc6
+
+# CHECK: lh  a0,12(a1)
+0x84 0xa4 0x00 0x0c
+
+# CHECK: lh  a0,12(a1)
+0x84 0xa4 0x00 0x0c
+
+# CHECK: li  v1,17767
+0x24 0x03 0x45 0x67
+
+# CHECK: ll  t1,9158(a3)
+0xc0 0xe9 0x23 0xc6
+
+# CHECK: lui a2,0x4567
+0x3c 0x06 0x45 0x67
+
+# CHECK: lw  a0,24(a1)
+0x8c 0xa4 0x00 0x18
+
+# CHECK: lwc1  $f9,9158(a3)
+0xc4 0xe9 0x23 0xc6
+
+# CHECK: madd  a2,a3
+0x70 0xc7 0x00 0x00
+
+# CHECK: maddu a2,a3
+0x70 0xc7 0x00 0x01
+
+# CHECK: mfc1  a2,$f7
+0x44 0x06 0x38 0x00
+
+# CHECK: mfhi  a1
+0x00 0x00 0x28 0x10
+
+# CHECK: mflo  a1
+0x00 0x00 0x28 0x12
+
+# CHECK: mov.d $f6,$f7
+0x46 0x20 0x39 0x86
+
+# CHECK: mov.s $f6,$f7
+0x46 0x00 0x39 0x86
+
+# CHECK: move  a2,a1
+0x00 0xa0 0x30 0x21
+
+# CHECK: msub  a2,a3
+0x70 0xc7 0x00 0x04
+
+# CHECK: msubu a2,a3
+0x70 0xc7 0x00 0x05
+
+# CHECK: mtc1  a2,$f7
+0x44 0x86 0x38 0x00
+
+# CHECK: mthi  a3
+0x00 0xe0 0x00 0x11
+
+# CHECK: mtlo  a3
+0x00 0xe0 0x00 0x13
+
+# CHECK: mul.d $f9,$f12,$f14
+0x46 0x27 0x32 0x42
+
+# CHECK: mul.s $f9,$f6,$f7
+0x46 0x07 0x32 0x42
+
+# CHECK: mul t1,a2,a3
+0x70 0xc7 0x48 0x02
+
+# CHECK: mult  v1,a1
+0x00 0x65 0x00 0x18
+
+# CHECK: multu v1,a1
+0x00 0x65 0x00 0x19
+
+# CHECK: neg.d $f12,$f14
+0x46 0x20 0x39 0x87
+
+# CHECK: neg.s $f6,$f7
+0x46 0x00 0x39 0x87
+
+# CHECK: neg v1,a1
+0x00 0x05 0x18 0x22
+
+# CHECK: nop
+0x00 0x00 0x00 0x00
+
+# CHECK: nor t1,a2,a3
+0x00 0xc7 0x48 0x27
+
+# CHECK: not v1,a1
+0x00 0xa0 0x18 0x27
+
+# CHECK: or  v1,v1,a1
+0x00 0x65 0x18 0x25
+
+# CHECK: ori t1,a2,0x4567
+0x34 0xc9 0x45 0x67
+
+# CHECK: rdhwr a2,$29
+0x7c 0x06 0xe8 0x3b
+
+# CHECK: round.w.d $f12,$f14
+0x46 0x20 0x39 0x8c
+
+# CHECK: round.w.s $f6,$f7
+0x46 0x00 0x39 0x8c
+
+# CHECK: sb  a0,9158(a1)
+0xa0 0xa4 0x23 0xc6
+
+# CHECK: sb  a0,6(a1)
+0xa0 0xa4 0x00 0x06
+
+# CHECK: sc  t1,9158(a3)
+0xe0 0xe9 0x23 0xc6
+
+# CHECK: sdc1  $f9,9158(a3)
+0xf4 0xe9 0x23 0xc6
+
+# CHECK: sh  a0,9158(a1)
+0xa4 0xa4 0x23 0xc6
+
+# CHECK: sll a0,v1,0x7
+0x00 0x03 0x21 0xc0
+
+# CHECK: sllv  v0,v1,a1
+0x00 0xa3 0x10 0x04
+
+# CHECK: slt v1,v1,a1
+0x00 0x65 0x18 0x2a
+
+# CHECK: slti  v1,v1,103
+0x28 0x63 0x00 0x67
+
+# CHECK: sltiu v1,v1,103
+0x2c 0x63 0x00 0x67
+
+# CHECK: sltu  v1,v1,a1
+0x00 0x65 0x18 0x2b
+
+# CHECK: sqrt.d  $f12,$f14
+0x46 0x20 0x39 0x84
+
+# CHECK: sqrt.s  $f6,$f7
+0x46 0x00 0x39 0x84
+
+# CHECK: sra a0,v1,0x7
+0x00 0x03 0x21 0xc3
+
+# CHECK: sra a0,v1,0x7
+0x00 0x03 0x21 0xc3
+
+# CHECK: srav  v0,v1,a1
+0x00 0xa3 0x10 0x07
+
+# CHECK: srl a0,v1,0x7
+0x00 0x03 0x21 0xc2
+
+# CHECK: srlv  v0,v1,a1
+0x00 0xa3 0x10 0x06
+
+# CHECK: sub.d $f9,$f12,$f14
+0x46 0x27 0x32 0x41
+
+# CHECK: sub.s $f9,$f6,$f7
+0x46 0x07 0x32 0x41
+
+# CHECK: sub t1,a2,a3
+0x00 0xc7 0x48 0x22
+
+# CHECK: subu  a0,v1,a1
+0x00 0x65 0x20 0x23
+
+# CHECK: sw  a0,24(a1)
+0xac 0xa4 0x00 0x18
+
+# CHECK: swc1  $f9,9158(a3)
+0xe4 0xe9 0x23 0xc6
+
+# CHECK: sync  0x7
+0x00 0x00 0x01 0xcf
+
+# CHECK: trunc.w.d $f12,$f14
+0x46 0x20 0x39 0x8d
+
+# CHECK: trunc.w.s $f6,$f7
+0x46 0x00 0x39 0x8d
+
+# CHECK: xor v1,v1,a1
+0x00 0x65 0x18 0x26
+
+# CHECK: xori  t1,a2,0x4567
+0x38 0xc9 0x45 0x67
diff --git a/test/MC/Disassembler/Mips/mips32_le.txt b/test/MC/Disassembler/Mips/mips32_le.txt
new file mode 100644
index 00000000000..a5a3cfd095d
--- /dev/null
+++ b/test/MC/Disassembler/Mips/mips32_le.txt
@@ -0,0 +1,424 @@
+# RUN: llvm-mc --disassemble %s -triple=mipsel-unknown-linux
+
+# CHECK: abs.d $f12,$f14
+0x85 0x39 0x20 0x46
+
+# CHECK: abs.s $f6,$f7
+0x85 0x39 0x00 0x46
+
+# CHECK: add t1,a2,a3
+0x20 0x48 0xc7 0x00
+
+# CHECK: add.d $f18,$f12,$f14
+0x40 0x32 0x27 0x46
+
+# CHECK: add.s $f9,$f6,$f7
+0x40 0x32 0x07 0x46
+
+# CHECK: addi t1,a2,17767
+0x67 0x45 0xc9 0x20
+
+# CHECK: addiu t1,a2,-15001
+0x67 0xc5 0xc9 0x24
+
+# CHECK: addu t1,a2,a3
+0x21 0x48 0xc7 0x00
+
+# CHECK: and t1,a2,a3
+0x24 0x48 0xc7 0x00
+
+# CHECK: andi t1,a2,0x4567
+0x67 0x45 0xc9 0x30
+
+# CHECK: b 00000534
+0x4c 0x01 0x00 0x10
+
+# CHECK: bal 00000534
+0x4c 0x01 0x11 0x04
+
+# CHECK: bc1f 00000534
+0x4c 0x01 0x00 0x45
+
+# CHECK: bc1t 00000534
+0x4c 0x01 0x01 0x45
+
+# CHECK: beq t1,a2,00000534
+0x4c 0x01 0x26 0x11
+
+# CHECK: bgez a2,00000534
+0x4c 0x01 0xc1 0x04
+
+# CHECK: bgezal a2,00000534
+0x4c 0x01 0xd1 0x04
+
+# CHECK: bgtz a2,00000534
+0x4c 0x01 0xc0 0x1c
+
+# CHECK: blez a2,00000534
+0x4c 0x01 0xc0 0x18
+
+# CHECK: bne t1,a2,00000534
+0x4c 0x01 0x26 0x15
+
+# CHECK: c.eq.d $f12,$f14
+0x32 0x30 0x27 0x46
+
+# CHECK: c.eq.s $f6,$f7
+0x32 0x30 0x07 0x46
+
+# CHECK: c.f.d $f12,$f14
+0x30 0x30 0x27 0x46
+
+# CHECK: c.f.s $f6,$f7
+0x30 0x30 0x07 0x46
+
+# CHECK: c.le.d $f12,$f14
+0x3e 0x30 0x27 0x46
+
+# CHECK: c.le.s $f6,$f7
+0x3e 0x30 0x07 0x46
+
+# CHECK: c.lt.d $f12,$f14
+0x3c 0x30 0x27 0x46
+
+# CHECK: c.lt.s $f6,$f7
+0x3c 0x30 0x07 0x46
+
+# CHECK: c.nge.d $f12,$f14
+0x3d 0x30 0x27 0x46
+
+# CHECK: c.nge.s $f6,$f7
+0x3d 0x30 0x07 0x46
+
+# CHECK: c.ngl.d $f12,$f14
+0x3b 0x30 0x27 0x46
+
+# CHECK: c.ngl.s $f6,$f7
+0x3b 0x30 0x07 0x46
+
+# CHECK: c.ngle.d $f12,$f14
+0x39 0x30 0x27 0x46
+
+# CHECK: c.ngle.s $f6,$f7
+0x39 0x30 0x07 0x46
+
+# CHECK: c.ngt.d $f12,$f14
+0x3f 0x30 0x27 0x46
+
+# CHECK: c.ngt.s $f6,$f7
+0x3f 0x30 0x07 0x46
+
+# CHECK: c.ole.d $f12,$f14
+0x36 0x30 0x27 0x46
+
+# CHECK: c.ole.s $f6,$f7
+0x36 0x30 0x07 0x46
+
+# CHECK: c.olt.d $f12,$f14
+0x34 0x30 0x27 0x46
+
+# CHECK: c.olt.s $f6,$f7
+0x34 0x30 0x07 0x46
+
+# CHECK: c.seq.d $f12,$f14
+0x3a 0x30 0x27 0x46
+
+# CHECK: c.seq.s $f6,$f7
+0x3a 0x30 0x07 0x46
+
+# CHECK: c.sf.d $f12,$f14
+0x38 0x30 0x27 0x46
+
+# CHECK: c.sf.s $f6,$f7
+0x38 0x30 0x07 0x46
+
+# CHECK: c.ueq.d $f12,$f14
+0x33 0x30 0x27 0x46
+
+# CHECK: c.ueq.s $f28,$f18
+0x33 0xe0 0x12 0x46
+
+# CHECK: c.ule.d $f12,$f14
+0x37 0x30 0x27 0x46
+
+# CHECK: c.ule.s $f6,$f7
+0x37 0x30 0x07 0x46
+
+# CHECK: c.ult.d $f12,$f14
+0x35 0x30 0x27 0x46
+
+# CHECK: c.ult.s $f6,$f7
+0x35 0x30 0x07 0x46
+
+# CHECK: c.un.d $f12,$f14
+0x31 0x30 0x27 0x46
+
+# CHECK: c.un.s $f6,$f7
+0x31 0x30 0x07 0x46
+
+# CHECK: ceil.w.d $f12,$f14
+0x8e 0x38 0x20 0x46
+
+# CHECK: ceil.w.s $f6,$f7
+0x8e 0x38 0x00 0x46
+
+# CHECK: cfc1 a2,$7
+0x00 0x38 0x46 0x44
+
+# CHECK: clo a2,a3
+0x21 0x30 0xe6 0x70
+
+# CHECK: clz a2,a3
+0x20 0x30 0xe6 0x70
+
+# CHECK: ctc1 a2,$7
+0x00 0x38 0xc6 0x44
+
+# CHECK: cvt.d.s $f6,$f7
+0xa1 0x39 0x00 0x46
+
+# CHECK: cvt.d.w $f12,$f14
+0xa1 0x39 0x80 0x46
+
+# CHECK: cvt.l.d $f12,$f14
+0xa5 0x39 0x20 0x46
+
+# CHECK: cvt.l.s $f6,$f7
+0xa5 0x39 0x00 0x46
+
+# CHECK: cvt.s.d $f12,$f14
+0xa0 0x39 0x20 0x46
+
+# CHECK: cvt.s.w $f6,$f7
+0xa0 0x39 0x80 0x46
+
+# CHECK: cvt.w.d $f12,$f14
+0xa4 0x39 0x20 0x46
+
+# CHECK: cvt.w.s $f6,$f7
+0xa4 0x39 0x00 0x46
+
+# CHECK: floor.w.d $f12,$f14
+0x8f 0x39 0x20 0x46
+
+# CHECK: floor.w.s $f6,$f7
+0x8f 0x39 0x00 0x46
+
+# CHECK: j 00000530
+0x4c 0x01 0x00 0x08
+
+# CHECK: jal 00000530
+0x4c 0x01 0x00 0x0c
+
+# CHECK: jalr a2,a3
+0x09 0xf8 0xe0 0x00
+
+# CHECK: jr a3
+0x08 0x00 0xe0 0x00
+
+# CHECK: lb  a0,9158(a1)
+0xc6 0x23 0xa4 0x80
+
+# CHECK: lbu a0,6(a1)
+0x06 0x00 0xa4 0x90
+
+# CHECK: ldc1  $f9,9158(a3)
+0xc6 0x23 0xe9 0xd4
+
+# CHECK: lh  a0,12(a1)
+0x0c 0x00 0xa4 0x84
+
+# CHECK: lh  a0,12(a1)
+0x0c 0x00 0xa4 0x84
+
+# CHECK: li  v1,17767
+0x67 0x45 0x03 0x24
+
+# CHECK: ll  t1,9158(a3)
+0xc6 0x23 0xe9 0xc0
+
+# CHECK: lui a2,0x4567
+0x67 0x45 0x06 0x3c
+
+# CHECK: lw  a0,24(a1)
+0x18 0x00 0xa4 0x8c
+
+# CHECK lw at,-18316(v0)
+0x74 0xb8 0x41 0x8c
+
+# CHECK: lwc1  $f9,9158(a3)
+0xc6 0x23 0xe9 0xc4
+
+# CHECK: madd  a2,a3
+0x00 0x00 0xc7 0x70
+
+# CHECK: maddu a2,a3
+0x01 0x00 0xc7 0x70
+
+# CHECK: mfc1  a2,$f7
+0x00 0x38 0x06 0x44
+
+# CHECK: mfhi  a1
+0x10 0x28 0x00 0x00
+
+# CHECK: mflo  a1
+0x12 0x28 0x00 0x00
+
+# CHECK: mov.d $f12,$f14
+0x86 0x39 0x20 0x46
+
+# CHECK: mov.s $f6,$f7
+0x86 0x39 0x00 0x46
+
+# CHECK: move  a2,a1
+0x21 0x30 0xa0 0x00
+
+# CHECK: msub  a2,a3
+0x04 0x00 0xc7 0x70
+
+# CHECK: msubu a2,a3
+0x05 0x00 0xc7 0x70
+
+# CHECK: mtc1  a2,$f7
+0x00 0x38 0x86 0x44
+
+# CHECK: mthi  a3
+0x11 0x00 0xe0 0x00
+
+# CHECK: mtlo  a3
+0x13 0x00 0xe0 0x00
+
+# CHECK: mul.d $f9,$f12,$f14
+0x42 0x32 0x27 0x46
+
+# CHECK: mul.s $f9,$f6,$f7
+0x42 0x32 0x07 0x46
+
+# CHECK: mul t1,a2,a3
+0x02 0x48 0xc7 0x70
+
+# CHECK: mult  v1,a1
+0x18 0x00 0x65 0x00
+
+# CHECK: multu v1,a1
+0x19 0x00 0x65 0x00
+
+# CHECK: neg.d $f12,$f14
+0x87 0x39 0x20 0x46
+
+# CHECK: neg.s $f6,$f7
+0x87 0x39 0x00 0x46
+
+# CHECK: neg v1,a1
+0x22 0x18 0x05 0x00
+
+# CHECK: nop
+0x00 0x00 0x00 0x00
+
+# CHECK: nor t1,a2,a3
+0x27 0x48 0xc7 0x00
+
+# CHECK: not v1,a1
+0x27 0x18 0xa0 0x00
+
+# CHECK: or  v1,v1,a1
+0x25 0x18 0x65 0x00
+
+# CHECK: ori t1,a2,0x4567
+0x67 0x45 0xc9 0x34
+
+# CHECK: rdhwr a2,$29
+0x3b 0xe8 0x06 0x7c
+
+# CHECK: round.w.d $f12,$f14
+0x8c 0x39 0x20 0x46
+
+# CHECK: round.w.s $f6,$f7
+0x8c 0x39 0x00 0x46
+
+# CHECK: sb  a0,9158(a1)
+0xc6 0x23 0xa4 0xa0
+
+# CHECK: sb  a0,6(a1)
+0x06 0x00 0xa4 0xa0
+
+# CHECK: sc  t1,9158(a3)
+0xc6 0x23 0xe9 0xe0
+
+# CHECK: sdc1  $f9,9158(a3)
+0xc6 0x23 0xe9 0xf4
+
+# CHECK: sh  a0,9158(a1)
+0xc6 0x23 0xa4 0xa4
+
+# CHECK: sll a0,v1,0x7
+0xc0 0x21 0x03 0x00
+
+# CHECK: sllv  v0,v1,a1
+0x04 0x10 0xa3 0x00
+
+# CHECK: slt v1,v1,a1
+0x2a 0x18 0x65 0x00
+
+# CHECK: slti  v1,v1,103
+0x67 0x00 0x63 0x28
+
+# CHECK: sltiu v1,v1,103
+0x67 0x00 0x63 0x2c
+
+# CHECK: sltu  v1,v1,a1
+0x2b 0x18 0x65 0x00
+
+# CHECK: sqrt.d  $f12,$f14
+0x84 0x39 0x20 0x46
+
+# CHECK: sqrt.s  $f6,$f7
+0x84 0x39 0x00 0x46
+
+# CHECK: sra a0,v1,0x7
+0xc3 0x21 0x03 0x00
+
+# CHECK: sra a0,v1,0x7
+0xc3 0x21 0x03 0x00
+
+# CHECK: srav  v0,v1,a1
+0x07 0x10 0xa3 0x00
+
+# CHECK: srl a0,v1,0x7
+0xc2 0x21 0x03 0x00
+
+# CHECK: srlv  v0,v1,a1
+0x06 0x10 0xa3 0x00
+
+# CHECK: sub.d $f9,$f12,$f14
+0x41 0x32 0x27 0x46
+
+# CHECK: sub.s $f9,$f6,$f7
+0x41 0x32 0x07 0x46
+
+# CHECK: sub t1,a2,a3
+0x22 0x48 0xc7 0x00
+
+# CHECK: subu  a0,v1,a1
+0x23 0x20 0x65 0x00
+
+# CHECK: sw  a0,24(a1)
+0x18 0x00 0xa4 0xac
+
+# CHECK: swc1  $f9,9158(a3)
+0xc6 0x23 0xe9 0xe4
+
+# CHECK: sync  0x7
+0xcf 0x01 0x00 0x00
+
+# CHECK: trunc.w.d $f12,$f14
+0x8d 0x39 0x20 0x46
+
+# CHECK: trunc.w.s $f6,$f7
+0x8d 0x39 0x00 0x46
+
+# CHECK: xor v1,v1,a1
+0x26 0x18 0x65 0x00
+
+# CHECK: xori  t1,a2,0x4567
+0x67 0x45 0xc9 0x38
diff --git a/test/MC/Disassembler/Mips/mips32r2.txt b/test/MC/Disassembler/Mips/mips32r2.txt
new file mode 100644
index 00000000000..295ffd03895
--- /dev/null
+++ b/test/MC/Disassembler/Mips/mips32r2.txt
@@ -0,0 +1,439 @@
+# RUN: llvm-mc --disassemble %s -triple=mips-unknown-linux -mcpu=mips32r2
+
+# CHECK: abs.d $f12,$f14
+0x46 0x20 0x39 0x85
+
+# CHECK: abs.s $f6,$f7
+0x46 0x00 0x39 0x85
+
+# CHECK: add t1,a2,a3
+0x00 0xc7 0x48 0x20
+
+# CHECK: add.d $f18,$f12,$f14
+0x46 0x27 0x32 0x40
+
+# CHECK: add.s $f9,$f6,$f7
+0x46 0x07 0x32 0x40
+
+# CHECK: addi t1,a2,17767
+0x20 0xc9 0x45 0x67
+
+# CHECK: addiu t1,a2,-15001
+0x24 0xc9 0xc5 0x67
+
+# CHECK: addu t1,a2,a3
+0x00 0xc7 0x48 0x21
+
+# CHECK: and t1,a2,a3
+0x00 0xc7 0x48 0x24
+
+# CHECK: andi t1,a2,0x4567
+0x30 0xc9 0x45 0x67
+
+# CHECK: b 00000534
+0x10 0x00 0x01 0x4c
+
+# CHECK: bal 00000534
+0x04 0x11 0x01 0x4c
+
+# CHECK: bc1f 00000534
+0x45 0x00 0x01 0x4c
+
+# CHECK: bc1t 00000534
+0x45 0x01 0x01 0x4c
+
+# CHECK: beq t1,a2,00000534
+0x11 0x26 0x01 0x4c
+
+# CHECK: bgez a2,00000534
+0x04 0xc1 0x01 0x4c
+
+# CHECK: bgezal a2,00000534
+0x04 0xd1 0x01 0x4c
+
+# CHECK: bgtz a2,00000534
+0x1c 0xc0 0x01 0x4c
+
+# CHECK: blez a2,00000534
+0x18 0xc0 0x01 0x4c
+
+# CHECK: bne t1,a2,00000534
+0x15 0x26 0x01 0x4c
+
+# CHECK: c.eq.d $f12,$f14
+0x46 0x27 0x30 0x32
+
+# CHECK: c.eq.s $f6,$f7
+0x46 0x07 0x30 0x32
+
+# CHECK: c.f.d $f12,$f14
+0x46 0x27 0x30 0x30
+
+# CHECK: c.f.s $f6,$f7
+0x46 0x07 0x30 0x30
+
+# CHECK: c.le.d $f12,$f14
+0x46 0x27 0x30 0x3e
+
+# CHECK: c.le.s $f6,$f7
+0x46 0x07 0x30 0x3e
+
+# CHECK: c.lt.d $f12,$f14
+0x46 0x27 0x30 0x3c
+
+# CHECK: c.lt.s $f6,$f7
+0x46 0x07 0x30 0x3c
+
+# CHECK: c.nge.d $f12,$f14
+0x46 0x27 0x30 0x3d
+
+# CHECK: c.nge.s $f6,$f7
+0x46 0x07 0x30 0x3d
+
+# CHECK: c.ngl.d $f12,$f14
+0x46 0x27 0x30 0x3b
+
+# CHECK: c.ngl.s $f6,$f7
+0x46 0x07 0x30 0x3b
+
+# CHECK: c.ngle.d $f12,$f14
+0x46 0x27 0x30 0x39
+
+# CHECK: c.ngle.s $f6,$f7
+0x46 0x07 0x30 0x39
+
+# CHECK: c.ngt.d $f12,$f14
+0x46 0x27 0x30 0x3f
+
+# CHECK: c.ngt.s $f6,$f7
+0x46 0x07 0x30 0x3f
+
+# CHECK: c.ole.d $f12,$f14
+0x46 0x27 0x30 0x36
+
+# CHECK: c.ole.s $f6,$f7
+0x46 0x07 0x30 0x36
+
+# CHECK: c.olt.d $f12,$f14
+0x46 0x27 0x30 0x34
+
+# CHECK: c.olt.s $f6,$f7
+0x46 0x07 0x30 0x34
+
+# CHECK: c.seq.d $f12,$f14
+0x46 0x27 0x30 0x3a
+
+# CHECK: c.seq.s $f6,$f7
+0x46 0x07 0x30 0x3a
+
+# CHECK: c.sf.d $f12,$f14
+0x46 0x27 0x30 0x38
+
+# CHECK: c.sf.s $f6,$f7
+0x46 0x07 0x30 0x38
+
+# CHECK: c.ueq.d $f12,$f14
+0x46 0x27 0x30 0x33
+
+# CHECK: c.ueq.s $f28,$f18
+0x46 0x12 0xe0 0x33
+
+# CHECK: c.ule.d $f12,$f14
+0x46 0x27 0x30 0x37
+
+# CHECK: c.ule.s $f6,$f7
+0x46 0x07 0x30 0x37
+
+# CHECK: c.ult.d $f12,$f14
+0x46 0x27 0x30 0x35
+
+# CHECK: c.ult.s $f6,$f7
+0x46 0x07 0x30 0x35
+
+# CHECK: c.un.d $f12,$f14
+0x46 0x27 0x30 0x31
+
+# CHECK: c.un.s $f6,$f7
+0x46 0x07 0x30 0x31
+
+# CHECK: ceil.w.d $f12,$f14
+0x46 0x20 0x39 0x8e
+
+# CHECK: ceil.w.s $f6,$f7
+0x46 0x00 0x39 0x8e
+
+# CHECK: cfc1 a2,$7
+0x44 0x46 0x38 0x00
+
+# CHECK: clo a2,a3
+0x70 0xe6 0x30 0x21
+
+# CHECK: clz a2,a3
+0x70 0xe6 0x30 0x20
+
+# CHECK: ctc1 a2,$7
+0x44 0xc6 0x38 0x00
+
+# CHECK: cvt.d.s $f6,$f7
+0x46 0x00 0x38 0xa1
+
+# CHECK: cvt.d.w $f12,$f14
+0x46 0x80 0x38 0xa1
+
+# CHECK: cvt.l.d $f12,$f14
+0x46 0x20 0x39 0xa5
+
+# CHECK: cvt.l.s $f6,$f7
+0x46 0x00 0x39 0xa5
+
+# CHECK: cvt.s.d $f12,$f14
+0x46 0x20 0x39 0xa0
+
+# CHECK: cvt.s.w $f6,$f7
+0x46 0x80 0x39 0xa0
+
+# CHECK: cvt.w.d $f12,$f14
+0x46 0x20 0x39 0xa4
+
+# CHECK: cvt.w.s $f6,$f7
+0x46 0x00 0x39 0xa4
+
+# CHECK: floor.w.d $f12,$f14
+0x46 0x20 0x39 0x8f
+
+# CHECK: floor.w.s $f6,$f7
+0x46 0x00 0x39 0x8f
+
+# CHECK: ins s3,t1,0x6,0x7
+0x7d 0x33 0x61 0x84
+
+# CHECK: j 00000530
+0x08 0x00 0x01 0x4c
+
+# CHECK: jal 00000530
+0x0c 0x00 0x01 0x4c
+
+# CHECK: jalr a2,a3
+0x00 0xe0 0xf8 0x09
+
+# CHECK: jr a3
+0x00 0xe0 0x00 0x08
+
+# CHECK: lb  a0,9158(a1)
+0x80 0xa4 0x23 0xc6
+
+# CHECK: lbu a0,6(a1)
+0x90 0xa4 0x00 0x06
+
+# CHECK: ldc1  $f9,9158(a3)
+0xd4 0xe9 0x23 0xc6
+
+# CHECK: lh  a0,12(a1)
+0x84 0xa4 0x00 0x0c
+
+# CHECK: lh  a0,12(a1)
+0x84 0xa4 0x00 0x0c
+
+# CHECK: li  v1,17767
+0x24 0x03 0x45 0x67
+
+# CHECK: ll  t1,9158(a3)
+0xc0 0xe9 0x23 0xc6
+
+# CHECK: lui a2,0x4567
+0x3c 0x06 0x45 0x67
+
+# CHECK: lw  a0,24(a1)
+0x8c 0xa4 0x00 0x18
+
+# CHECK: lwc1  $f9,9158(a3)
+0xc4 0xe9 0x23 0xc6
+
+# CHECK: madd  a2,a3
+0x70 0xc7 0x00 0x00
+
+# CHECK: maddu a2,a3
+0x70 0xc7 0x00 0x01
+
+# CHECK: mfc1  a2,$f7
+0x44 0x06 0x38 0x00
+
+# CHECK: mfhi  a1
+0x00 0x00 0x28 0x10
+
+# CHECK: mflo  a1
+0x00 0x00 0x28 0x12
+
+# CHECK: mov.d $f6,$f7
+0x46 0x20 0x39 0x86
+
+# CHECK: mov.s $f6,$f7
+0x46 0x00 0x39 0x86
+
+# CHECK: move  a2,a1
+0x00 0xa0 0x30 0x21
+
+# CHECK: msub  a2,a3
+0x70 0xc7 0x00 0x04
+
+# CHECK: msubu a2,a3
+0x70 0xc7 0x00 0x05
+
+# CHECK: mtc1  a2,$f7
+0x44 0x86 0x38 0x00
+
+# CHECK: mthi  a3
+0x00 0xe0 0x00 0x11
+
+# CHECK: mtlo  a3
+0x00 0xe0 0x00 0x13
+
+# CHECK: mul.d $f9,$f12,$f14
+0x46 0x27 0x32 0x42
+
+# CHECK: mul.s $f9,$f6,$f7
+0x46 0x07 0x32 0x42
+
+# CHECK: mul t1,a2,a3
+0x70 0xc7 0x48 0x02
+
+# CHECK: mult  v1,a1
+0x00 0x65 0x00 0x18
+
+# CHECK: multu v1,a1
+0x00 0x65 0x00 0x19
+
+# CHECK: neg.d $f12,$f14
+0x46 0x20 0x39 0x87
+
+# CHECK: neg.s $f6,$f7
+0x46 0x00 0x39 0x87
+
+# CHECK: neg v1,a1
+0x00 0x05 0x18 0x22
+
+# CHECK: nop
+0x00 0x00 0x00 0x00
+
+# CHECK: nor t1,a2,a3
+0x00 0xc7 0x48 0x27
+
+# CHECK: not v1,a1
+0x00 0xa0 0x18 0x27
+
+# CHECK: or  v1,v1,a1
+0x00 0x65 0x18 0x25
+
+# CHECK: ori t1,a2,0x4567
+0x34 0xc9 0x45 0x67
+
+# CHECK: rdhwr a2,$29
+0x7c 0x06 0xe8 0x3b
+
+# CHECK: ror t1,a2,0x7
+0x00 0x26 0x49 0xc2
+
+# CHECK: rorv  t1,a2,a3
+0x00 0xe6 0x48 0x46
+
+# CHECK: round.w.d $f12,$f14
+0x46 0x20 0x39 0x8c
+
+# CHECK: round.w.s $f6,$f7
+0x46 0x00 0x39 0x8c
+
+# CHECK: sb  a0,9158(a1)
+0xa0 0xa4 0x23 0xc6
+
+# CHECK: sb  a0,6(a1)
+0xa0 0xa4 0x00 0x06
+
+# CHECK: sc  t1,9158(a3)
+0xe0 0xe9 0x23 0xc6
+
+# CHECK: sdc1  $f9,9158(a3)
+0xf4 0xe9 0x23 0xc6
+
+# CHECK: seb a2,a3
+0x7c 0x07 0x34 0x20
+
+# CHECK: seh a2,a3
+0x7c 0x07 0x36 0x20
+
+# CHECK: sh  a0,9158(a1)
+0xa4 0xa4 0x23 0xc6
+
+# CHECK: sll a0,v1,0x7
+0x00 0x03 0x21 0xc0
+
+# CHECK: sllv  v0,v1,a1
+0x00 0xa3 0x10 0x04
+
+# CHECK: slt v1,v1,a1
+0x00 0x65 0x18 0x2a
+
+# CHECK: slti  v1,v1,103
+0x28 0x63 0x00 0x67
+
+# CHECK: sltiu v1,v1,103
+0x2c 0x63 0x00 0x67
+
+# CHECK: sltu  v1,v1,a1
+0x00 0x65 0x18 0x2b
+
+# CHECK: sqrt.d  $f12,$f14
+0x46 0x20 0x39 0x84
+
+# CHECK: sqrt.s  $f6,$f7
+0x46 0x00 0x39 0x84
+
+# CHECK: sra a0,v1,0x7
+0x00 0x03 0x21 0xc3
+
+# CHECK: sra a0,v1,0x7
+0x00 0x03 0x21 0xc3
+
+# CHECK: srav  v0,v1,a1
+0x00 0xa3 0x10 0x07
+
+# CHECK: srl a0,v1,0x7
+0x00 0x03 0x21 0xc2
+
+# CHECK: srlv  v0,v1,a1
+0x00 0xa3 0x10 0x06
+
+# CHECK: sub.d $f9,$f12,$f14
+0x46 0x27 0x32 0x41
+
+# CHECK: sub.s $f9,$f6,$f7
+0x46 0x07 0x32 0x41
+
+# CHECK: sub t1,a2,a3
+0x00 0xc7 0x48 0x22
+
+# CHECK: subu  a0,v1,a1
+0x00 0x65 0x20 0x23
+
+# CHECK: sw  a0,24(a1)
+0xac 0xa4 0x00 0x18
+
+# CHECK: swc1  $f9,9158(a3)
+0xe4 0xe9 0x23 0xc6
+
+# CHECK: sync  0x7
+0x00 0x00 0x01 0xcf
+
+# CHECK: trunc.w.d $f12,$f14
+0x46 0x20 0x39 0x8d
+
+# CHECK: trunc.w.s $f6,$f7
+0x46 0x00 0x39 0x8d
+
+# CHECK: wsbh  a2,a3
+0x7c 0x07 0x30 0xa0
+
+# CHECK: xor v1,v1,a1
+0x00 0x65 0x18 0x26
+
+# CHECK: xori  t1,a2,0x4567
+0x38 0xc9 0x45 0x67
diff --git a/test/MC/Disassembler/Mips/mips32r2_le.txt b/test/MC/Disassembler/Mips/mips32r2_le.txt
new file mode 100644
index 00000000000..6d8be790f14
--- /dev/null
+++ b/test/MC/Disassembler/Mips/mips32r2_le.txt
@@ -0,0 +1,442 @@
+# RUN: llvm-mc --disassemble %s -triple=mipsel-unknown-linux -mcpu=mips32r2
+
+# CHECK: abs.d $f12,$f14
+0x85 0x39 0x20 0x46
+
+# CHECK: abs.s $f6,$f7
+0x85 0x39 0x00 0x46
+
+# CHECK: add t1,a2,a3
+0x20 0x48 0xc7 0x00
+
+# CHECK: add.d $f18,$f12,$f14
+0x40 0x32 0x27 0x46
+
+# CHECK: add.s $f9,$f6,$f7
+0x40 0x32 0x07 0x46
+
+# CHECK: addi t1,a2,17767
+0x67 0x45 0xc9 0x20
+
+# CHECK: addiu t1,a2,-15001
+0x67 0xc5 0xc9 0x24
+
+# CHECK: addu t1,a2,a3
+0x21 0x48 0xc7 0x00
+
+# CHECK: and t1,a2,a3
+0x24 0x48 0xc7 0x00
+
+# CHECK: andi t1,a2,0x4567
+0x67 0x45 0xc9 0x30
+
+# CHECK: b 00000534
+0x4c 0x01 0x00 0x10
+
+# CHECK: bal 00000534
+0x4c 0x01 0x11 0x04
+
+# CHECK: bc1f 00000534
+0x4c 0x01 0x00 0x45
+
+# CHECK: bc1t 00000534
+0x4c 0x01 0x01 0x45
+
+# CHECK: beq t1,a2,00000534
+0x4c 0x01 0x26 0x11
+
+# CHECK: bgez a2,00000534
+0x4c 0x01 0xc1 0x04
+
+# CHECK: bgezal a2,00000534
+0x4c 0x01 0xd1 0x04
+
+# CHECK: bgtz a2,00000534
+0x4c 0x01 0xc0 0x1c
+
+# CHECK: blez a2,00000534
+0x4c 0x01 0xc0 0x18
+
+# CHECK: bne t1,a2,00000534
+0x4c 0x01 0x26 0x15
+
+# CHECK: c.eq.d $f12,$f14
+0x32 0x30 0x27 0x46
+
+# CHECK: c.eq.s $f6,$f7
+0x32 0x30 0x07 0x46
+
+# CHECK: c.f.d $f12,$f14
+0x30 0x30 0x27 0x46
+
+# CHECK: c.f.s $f6,$f7
+0x30 0x30 0x07 0x46
+
+# CHECK: c.le.d $f12,$f14
+0x3e 0x30 0x27 0x46
+
+# CHECK: c.le.s $f6,$f7
+0x3e 0x30 0x07 0x46
+
+# CHECK: c.lt.d $f12,$f14
+0x3c 0x30 0x27 0x46
+
+# CHECK: c.lt.s $f6,$f7
+0x3c 0x30 0x07 0x46
+
+# CHECK: c.nge.d $f12,$f14
+0x3d 0x30 0x27 0x46
+
+# CHECK: c.nge.s $f6,$f7
+0x3d 0x30 0x07 0x46
+
+# CHECK: c.ngl.d $f12,$f14
+0x3b 0x30 0x27 0x46
+
+# CHECK: c.ngl.s $f6,$f7
+0x3b 0x30 0x07 0x46
+
+# CHECK: c.ngle.d $f12,$f14
+0x39 0x30 0x27 0x46
+
+# CHECK: c.ngle.s $f6,$f7
+0x39 0x30 0x07 0x46
+
+# CHECK: c.ngt.d $f12,$f14
+0x3f 0x30 0x27 0x46
+
+# CHECK: c.ngt.s $f6,$f7
+0x3f 0x30 0x07 0x46
+
+# CHECK: c.ole.d $f12,$f14
+0x36 0x30 0x27 0x46
+
+# CHECK: c.ole.s $f6,$f7
+0x36 0x30 0x07 0x46
+
+# CHECK: c.olt.d $f12,$f14
+0x34 0x30 0x27 0x46
+
+# CHECK: c.olt.s $f6,$f7
+0x34 0x30 0x07 0x46
+
+# CHECK: c.seq.d $f12,$f14
+0x3a 0x30 0x27 0x46
+
+# CHECK: c.seq.s $f6,$f7
+0x3a 0x30 0x07 0x46
+
+# CHECK: c.sf.d $f12,$f14
+0x38 0x30 0x27 0x46
+
+# CHECK: c.sf.s $f6,$f7
+0x38 0x30 0x07 0x46
+
+# CHECK: c.ueq.d $f12,$f14
+0x33 0x30 0x27 0x46
+
+# CHECK: c.ueq.s $f28,$f18
+0x33 0xe0 0x12 0x46
+
+# CHECK: c.ule.d $f12,$f14
+0x37 0x30 0x27 0x46
+
+# CHECK: c.ule.s $f6,$f7
+0x37 0x30 0x07 0x46
+
+# CHECK: c.ult.d $f12,$f14
+0x35 0x30 0x27 0x46
+
+# CHECK: c.ult.s $f6,$f7
+0x35 0x30 0x07 0x46
+
+# CHECK: c.un.d $f12,$f14
+0x31 0x30 0x27 0x46
+
+# CHECK: c.un.s $f6,$f7
+0x31 0x30 0x07 0x46
+
+# CHECK: ceil.w.d $f12,$f14
+0x8e 0x38 0x20 0x46
+
+# CHECK: ceil.w.s $f6,$f7
+0x8e 0x38 0x00 0x46
+
+# CHECK: cfc1 a2,$7
+0x00 0x38 0x46 0x44
+
+# CHECK: clo a2,a3
+0x21 0x30 0xe6 0x70
+
+# CHECK: clz a2,a3
+0x20 0x30 0xe6 0x70
+
+# CHECK: ctc1 a2,$7
+0x00 0x38 0xc6 0x44
+
+# CHECK: cvt.d.s $f6,$f7
+0xa1 0x39 0x00 0x46
+
+# CHECK: cvt.d.w $f12,$f14
+0xa1 0x39 0x80 0x46
+
+# CHECK: cvt.l.d $f12,$f14
+0xa5 0x39 0x20 0x46
+
+# CHECK: cvt.l.s $f6,$f7
+0xa5 0x39 0x00 0x46
+
+# CHECK: cvt.s.d $f12,$f14
+0xa0 0x39 0x20 0x46
+
+# CHECK: cvt.s.w $f6,$f7
+0xa0 0x39 0x80 0x46
+
+# CHECK: cvt.w.d $f12,$f14
+0xa4 0x39 0x20 0x46
+
+# CHECK: cvt.w.s $f6,$f7
+0xa4 0x39 0x00 0x46
+
+# CHECK: floor.w.d $f12,$f14
+0x8f 0x39 0x20 0x46
+
+# CHECK: floor.w.s $f6,$f7
+0x8f 0x39 0x00 0x46
+
+# CHECK: ins s3,t1,0x6,0x7
+0x84 0x61 0x33 0x7d
+
+# CHECK: j 00000530
+0x4c 0x01 0x00 0x08
+
+# CHECK: jal 00000530
+0x4c 0x01 0x00 0x0c
+
+# CHECK: jalr a2,a3
+0x09 0xf8 0xe0 0x00
+
+# CHECK: jr a3
+0x08 0x00 0xe0 0x00
+
+# CHECK: lb  a0,9158(a1)
+0xc6 0x23 0xa4 0x80
+
+# CHECK: lbu a0,6(a1)
+0x06 0x00 0xa4 0x90
+
+# CHECK: ldc1  $f9,9158(a3)
+0xc6 0x23 0xe9 0xd4
+
+# CHECK: lh  a0,12(a1)
+0x0c 0x00 0xa4 0x84
+
+# CHECK: lh  a0,12(a1)
+0x0c 0x00 0xa4 0x84
+
+# CHECK: li  v1,17767
+0x67 0x45 0x03 0x24
+
+# CHECK: ll  t1,9158(a3)
+0xc6 0x23 0xe9 0xc0
+
+# CHECK: lui a2,0x4567
+0x67 0x45 0x06 0x3c
+
+# CHECK: lw  a0,24(a1)
+0x18 0x00 0xa4 0x8c
+
+# CHECK lw at,-18316(v0)
+0x74 0xb8 0x41 0x8c
+
+# CHECK: lwc1  $f9,9158(a3)
+0xc6 0x23 0xe9 0xc4
+
+# CHECK: madd  a2,a3
+0x00 0x00 0xc7 0x70
+
+# CHECK: maddu a2,a3
+0x01 0x00 0xc7 0x70
+
+# CHECK: mfc1  a2,$f7
+0x00 0x38 0x06 0x44
+
+# CHECK: mfhi  a1
+0x10 0x28 0x00 0x00
+
+# CHECK: mflo  a1
+0x12 0x28 0x00 0x00
+
+# CHECK: mov.d $f12,$f14
+0x86 0x39 0x20 0x46
+
+# CHECK: mov.s $f6,$f7
+0x86 0x39 0x00 0x46
+
+# CHECK: move  a2,a1
+0x21 0x30 0xa0 0x00
+
+# CHECK: msub  a2,a3
+0x04 0x00 0xc7 0x70
+
+# CHECK: msubu a2,a3
+0x05 0x00 0xc7 0x70
+
+# CHECK: mtc1  a2,$f7
+0x00 0x38 0x86 0x44
+
+# CHECK: mthi  a3
+0x11 0x00 0xe0 0x00
+
+# CHECK: mtlo  a3
+0x13 0x00 0xe0 0x00
+
+# CHECK: mul.d $f9,$f12,$f14
+0x42 0x32 0x27 0x46
+
+# CHECK: mul.s $f9,$f6,$f7
+0x42 0x32 0x07 0x46
+
+# CHECK: mul t1,a2,a3
+0x02 0x48 0xc7 0x70
+
+# CHECK: mult  v1,a1
+0x18 0x00 0x65 0x00
+
+# CHECK: multu v1,a1
+0x19 0x00 0x65 0x00
+
+# CHECK: neg.d $f12,$f14
+0x87 0x39 0x20 0x46
+
+# CHECK: neg.s $f6,$f7
+0x87 0x39 0x00 0x46
+
+# CHECK: neg v1,a1
+0x22 0x18 0x05 0x00
+
+# CHECK: nop
+0x00 0x00 0x00 0x00
+
+# CHECK: nor t1,a2,a3
+0x27 0x48 0xc7 0x00
+
+# CHECK: not v1,a1
+0x27 0x18 0xa0 0x00
+
+# CHECK: or  v1,v1,a1
+0x25 0x18 0x65 0x00
+
+# CHECK: ori t1,a2,0x4567
+0x67 0x45 0xc9 0x34
+
+# CHECK: rdhwr a2,$29
+0x3b 0xe8 0x06 0x7c
+
+# CHECK: ror t1,a2,0x7
+0xc2 0x49 0x26 0x00
+
+# CHECK: rorv  t1,a2,a3
+0x46 0x48 0xe6 0x00
+
+# CHECK: round.w.d $f12,$f14
+0x8c 0x39 0x20 0x46
+
+# CHECK: round.w.s $f6,$f7
+0x8c 0x39 0x00 0x46
+
+# CHECK: sb  a0,9158(a1)
+0xc6 0x23 0xa4 0xa0
+
+# CHECK: sb  a0,6(a1)
+0x06 0x00 0xa4 0xa0
+
+# CHECK: sc  t1,9158(a3)
+0xc6 0x23 0xe9 0xe0
+
+# CHECK: sdc1  $f9,9158(a3)
+0xc6 0x23 0xe9 0xf4
+
+# CHECK: seb a2,a3
+0x20 0x34 0x07 0x7c
+
+# CHECK: seh a2,a3
+0x20 0x36 0x07 0x7c
+
+# CHECK: sh  a0,9158(a1)
+0xc6 0x23 0xa4 0xa4
+
+# CHECK: sll a0,v1,0x7
+0xc0 0x21 0x03 0x00
+
+# CHECK: sllv  v0,v1,a1
+0x04 0x10 0xa3 0x00
+
+# CHECK: slt v1,v1,a1
+0x2a 0x18 0x65 0x00
+
+# CHECK: slti  v1,v1,103
+0x67 0x00 0x63 0x28
+
+# CHECK: sltiu v1,v1,103
+0x67 0x00 0x63 0x2c
+
+# CHECK: sltu  v1,v1,a1
+0x2b 0x18 0x65 0x00
+
+# CHECK: sqrt.d  $f12,$f14
+0x84 0x39 0x20 0x46
+
+# CHECK: sqrt.s  $f6,$f7
+0x84 0x39 0x00 0x46
+
+# CHECK: sra a0,v1,0x7
+0xc3 0x21 0x03 0x00
+
+# CHECK: sra a0,v1,0x7
+0xc3 0x21 0x03 0x00
+
+# CHECK: srav  v0,v1,a1
+0x07 0x10 0xa3 0x00
+
+# CHECK: srl a0,v1,0x7
+0xc2 0x21 0x03 0x00
+
+# CHECK: srlv  v0,v1,a1
+0x06 0x10 0xa3 0x00
+
+# CHECK: sub.d $f9,$f12,$f14
+0x41 0x32 0x27 0x46
+
+# CHECK: sub.s $f9,$f6,$f7
+0x41 0x32 0x07 0x46
+
+# CHECK: sub t1,a2,a3
+0x22 0x48 0xc7 0x00
+
+# CHECK: subu  a0,v1,a1
+0x23 0x20 0x65 0x00
+
+# CHECK: sw  a0,24(a1)
+0x18 0x00 0xa4 0xac
+
+# CHECK: swc1  $f9,9158(a3)
+0xc6 0x23 0xe9 0xe4
+
+# CHECK: sync  0x7
+0xcf 0x01 0x00 0x00
+
+# CHECK: trunc.w.d $f12,$f14
+0x8d 0x39 0x20 0x46
+
+# CHECK: trunc.w.s $f6,$f7
+0x8d 0x39 0x00 0x46
+
+# CHECK: wsbh  a2,a3
+0xa0 0x30 0x07 0x7c
+
+# CHECK: xor v1,v1,a1
+0x26 0x18 0x65 0x00
+
+# CHECK: xori  t1,a2,0x4567
+0x67 0x45 0xc9 0x38
diff --git a/test/MC/Disassembler/Mips/mips64.txt b/test/MC/Disassembler/Mips/mips64.txt
new file mode 100644
index 00000000000..1c7447a815f
--- /dev/null
+++ b/test/MC/Disassembler/Mips/mips64.txt
@@ -0,0 +1,67 @@
+# RUN: llvm-mc --disassemble %s -triple=mips64-unknown-linux
+
+# CHECK: daddiu t3,k0,31949
+0x67 0x4b 0x7c 0xcd
+
+# CHECK: daddu k0,at,t3
+0x00 0x2b 0xd0 0x2d
+
+# CHECK: ddiv zero,k0,s6
+0x03 0x56 0x00 0x1e
+
+# CHECK: ddivu zero,t1,t8
+0x01 0x38 0x00 0x1f
+
+# CHECK: dmfc1 v0,$f14
+0x44 0x22 0x70 0x00
+
+# CHECK: dmtc1 s7,$f5
+0x44 0xb7 0x28 0x00
+
+# CHECK: dmult t3,k0
+0x01 0x7a 0x00 0x1c
+
+# CHECK: dmultu s7,t5
+0x02 0xed 0x00 0x1d
+
+# CHECK: dsll v1,t8,0x11
+0x00 0x18 0x1c 0x78
+
+# CHECK: dsllv gp,k1,t8
+0x03 0x1b 0xe0 0x14
+
+# CHECK: dsra at,at,0x1e
+0x00 0x01 0x0f 0xbb
+
+# CHECK: dsrav at,at,s8
+0x03 0xc1 0x08 0x17
+
+# CHECK: dsrl t2,gp,0x18
+0x00 0x1c 0x56 0x3a
+
+# CHECK: dsrlv gp,t2,s7
+0x02 0xea 0xe0 0x16
+
+# CHECK: dsubu gp,k1,t8
+0x03 0x78 0xe0 0x2f
+
+# CHECK: lw k1,-15155(at)
+0x8c 0x3b 0xc4 0xcd
+
+# CHECK: lui at,0x1
+0x3c 0x01 0x00 0x01
+
+# CHECK: lwu v1,-1746(v1)
+0x9c 0x63 0xf9 0x2e
+
+# CHECK: lui ra,0x1
+0x3c 0x1f 0x00 0x01
+
+# CHECK: sw k0,-15159(at)
+0xac 0x3a 0xc4 0xc9
+
+# CHECK: ld k0,3958(zero)
+0xdc 0x1a 0x0f 0x76
+
+# CHECK: sd a2,17767(zero)
+0xfc 0x06 0x45 0x67
diff --git a/test/MC/Disassembler/Mips/mips64_le.txt b/test/MC/Disassembler/Mips/mips64_le.txt
new file mode 100644
index 00000000000..dd87522b083
--- /dev/null
+++ b/test/MC/Disassembler/Mips/mips64_le.txt
@@ -0,0 +1,67 @@
+# RUN: llvm-mc --disassemble %s -triple=mips64el-unknown-linux
+
+# CHECK: daddiu t3,k0,31949
+0xcd 0x7c 0x4b 0x67
+
+# CHECK: daddu k0,at,t3
+0x2d 0xd0 0x2b 0x00
+
+# CHECK: ddiv zero,k0,s6
+0x1e 0x00 0x56 0x03
+
+# CHECK: ddivu zero,t1,t8
+0x1f 0x00 0x38 0x01
+
+# CHECK: dmfc1 v0,$f14
+0x00 0x70 0x22 0x44
+
+# CHECK: dmtc1 s7,$f5
+0x00 0x28 0xb7 0x44
+
+# CHECK: dmult t3,k0
+0x1c 0x00 0x7a 0x01
+
+# CHECK: dmultu s7,t5
+0x1d 0x00 0xed 0x02
+
+# CHECK: dsll v1,t8,0x11
+0x78 0x1c 0x18 0x00
+
+# CHECK: dsllv gp,k1,t8
+0x14 0xe0 0x1b 0x03
+
+# CHECK: dsra at,at,0x1e
+0xbb 0x0f 0x01 0x00
+
+# CHECK: dsrav at,at,s8
+0x17 0x08 0xc1 0x03
+
+# CHECK: dsrl t2,gp,0x18
+0x3a 0x56 0x1c 0x00
+
+# CHECK: dsrlv gp,t2,s7
+0x16 0xe0 0xea 0x02
+
+# CHECK: dsubu gp,k1,t8
+0x2f 0xe0 0x78 0x03
+
+# CHECK: lw k1,-15155(at)
+0xcd 0xc4 0x3b 0x8c
+
+# CHECK: lui at,0x1
+0x01 0x00 0x01 0x3c
+
+# CHECK: lwu v1,-1746(v1)
+0x2e 0xf9 0x63 0x9c
+
+# CHECK: lui ra,0x1
+0x01 0x00 0x1f 0x3c
+
+# CHECK: sw k0,-15159(at)
+0xc9 0xc4 0x3a 0xac
+
+# CHECK: ld k0,3958(zero)
+0x76 0x0f 0x1a 0xdc
+
+# CHECK: sd a2,17767(zero)
+0x67 0x45 0x06 0xfc
diff --git a/test/MC/Disassembler/Mips/mips64r2.txt b/test/MC/Disassembler/Mips/mips64r2.txt
new file mode 100644
index 00000000000..26bc94d65b5
--- /dev/null
+++ b/test/MC/Disassembler/Mips/mips64r2.txt
@@ -0,0 +1,91 @@
+# RUN: llvm-mc --disassemble %s -triple=mips64-unknown-linux -mattr +mips64r2
+
+# CHECK: daddiu t3,k0,31949
+0x67 0x4b 0x7c 0xcd
+
+# CHECK: daddu k0,at,t3
+0x00 0x2b 0xd0 0x2d
+
+# CHECK: ddiv zero,k0,s6
+0x03 0x56 0x00 0x1e
+
+# CHECK: ddivu zero,t1,t8
+0x01 0x38 0x00 0x1f
+
+# CHECK: dmfc1 v0,$f14
+0x44 0x22 0x70 0x00
+
+# CHECK: dmtc1 s7,$f5
+0x44 0xb7 0x28 0x00
+
+# CHECK: dmult t3,k0
+0x01 0x7a 0x00 0x1c
+
+# CHECK: dmultu s7,t5
+0x02 0xed 0x00 0x1d
+
+# CHECK: dsll v1,t8,0x11
+0x00 0x18 0x1c 0x78
+
+# CHECK: dsllv gp,k1,t8
+0x03 0x1b 0xe0 0x14
+
+# CHECK: dsra at,at,0x1e
+0x00 0x01 0x0f 0xbb
+
+# CHECK: dsrav at,at,s8
+0x03 0xc1 0x08 0x17
+
+# CHECK: dsrl t2,gp,0x18
+0x00 0x1c 0x56 0x3a
+
+# CHECK: dsrlv gp,t2,s7
+0x02 0xea 0xe0 0x16
+
+# CHECK: dsubu gp,k1,t8
+0x03 0x78 0xe0 0x2f
+
+# CHECK: lw k1,-15155(at)
+0x8c 0x3b 0xc4 0xcd
+
+# CHECK: lui at,0x1
+0x3c 0x01 0x00 0x01
+
+# CHECK: lwu v1,-1746(v1)
+0x9c 0x63 0xf9 0x2e
+
+# CHECK: lui ra,0x1
+0x3c 0x1f 0x00 0x01
+
+# CHECK: sw k0,-15159(at)
+0xac 0x3a 0xc4 0xc9
+
+# CHECK: ld k0,3958(zero)
+0xdc 0x1a 0x0f 0x76
+
+# CHECK: sd a2,17767(zero)
+0xfc 0x06 0x45 0x67
+
+# CHECK: dclo t1,t8
+0x73 0x09 0x48 0x25
+
+# CHECK: dclz k0,t1
+0x71 0x3a 0xd0 0x24
+
+# CHECK: dext a3,gp,0x1d,0x1f
+0x7f 0x87 0xf7 0x43
+
+# CHECK: dins s4,gp,0xf,0x1
+0x7f 0x94 0x7b 0xc7
+
+# CHECK: dsbh a3,gp
+0x7c 0x1c 0x38 0xa4
+
+# CHECK: dshd v1,t6
+0x7c 0x0e 0x19 0x64
+
+# CHECK: drotr s4,k1,0x6
+0x00 0x3b 0xa1 0xba
+
+# CHECK: drotrv t8,s7,a1
+0x00 0xb7 0xc0 0x56
diff --git a/test/MC/Disassembler/Mips/mips64r2_le.txt b/test/MC/Disassembler/Mips/mips64r2_le.txt
new file mode 100644
index 00000000000..81a7c66d1a2
--- /dev/null
+++ b/test/MC/Disassembler/Mips/mips64r2_le.txt
@@ -0,0 +1,91 @@
+# RUN: llvm-mc --disassemble %s -triple=mips64el-unknown-linux -mattr +mips64r2
+
+# CHECK: daddiu t3,k0,31949
+0xcd 0x7c 0x4b 0x67
+
+# CHECK: daddu k0,at,t3
+0x2d 0xd0 0x2b 0x00
+
+# CHECK: ddiv zero,k0,s6
+0x1e 0x00 0x56 0x03
+
+# CHECK: ddivu zero,t1,t8
+0x1f 0x00 0x38 0x01
+
+# CHECK: dmfc1 v0,$f14
+0x00 0x70 0x22 0x44
+
+# CHECK: dmtc1 s7,$f5
+0x00 0x28 0xb7 0x44
+
+# CHECK: dmult t3,k0
+0x1c 0x00 0x7a 0x01
+
+# CHECK: dmultu s7,t5
+0x1d 0x00 0xed 0x02
+
+# CHECK: dsll v1,t8,0x11
+0x78 0x1c 0x18 0x00
+
+# CHECK: dsllv gp,k1,t8
+0x14 0xe0 0x1b 0x03
+
+# CHECK: dsra at,at,0x1e
+0xbb 0x0f 0x01 0x00
+
+# CHECK: dsrav at,at,s8
+0x17 0x08 0xc1 0x03
+
+# CHECK: dsrl t2,gp,0x18
+0x3a 0x56 0x1c 0x00
+
+# CHECK: dsrlv gp,t2,s7
+0x16 0xe0 0xea 0x02
+
+# CHECK: dsubu gp,k1,t8
+0x2f 0xe0 0x78 0x03
+
+# CHECK: lw k1,-15155(at)
+0xcd 0xc4 0x3b 0x8c
+
+# CHECK: lui at,0x1
+0x01 0x00 0x01 0x3c
+
+# CHECK: lwu v1,-1746(v1)
+0x2e 0xf9 0x63 0x9c
+
+# CHECK: lui ra,0x1
+0x01 0x00 0x1f 0x3c
+
+# CHECK: sw k0,-15159(at)
+0xc9 0xc4 0x3a 0xac
+
+# CHECK: ld k0,3958(zero)
+0x76 0x0f 0x1a 0xdc
+
+# CHECK: sd a2,17767(zero)
+0x67 0x45 0x06 0xfc
+
+# CHECK: dclo t1,t8
+0x25 0x48 0x09 0x73
+
+# CHECK: dclz k0,t1
+0x24 0xd0 0x3a 0x71
+
+# CHECK: dext a3,gp,0x1d,0x1f
+0x43 0xf7 0x87 0x7f
+
+# CHECK: dins s4,gp,0xf,0x1
+0xc7 0x7b 0x94 0x7f
+
+# CHECK: dsbh a3,gp
+0xa4 0x38 0x1c 0x7c
+
+# CHECK: dshd v1,t6
+0x64 0x19 0x0e 0x7c
+
+# CHECK: drotr s4,k1,0x6
+0xba 0xa1 0x3b 0x00
+
+# CHECK: drotrv t8,s7,a1
+0x56 0xc0 0xb7 0x00
diff --git a/test/MC/Disassembler/X86/intel-syntax.txt b/test/MC/Disassembler/X86/intel-syntax.txt
index 3391e45cc00..a5dbcf29b96 100644
--- a/test/MC/Disassembler/X86/intel-syntax.txt
+++ b/test/MC/Disassembler/X86/intel-syntax.txt
@@ -1,4 +1,4 @@
-# RUN: llvm-mc --disassemble %s -triple=x86_64-apple-darwin9 -x86-asm-syntax=intel | FileCheck %s
+# RUN: llvm-mc --disassemble %s -triple=x86_64-apple-darwin9 --output-asm-variant=1 | FileCheck %s
 
 # CHECK: movsb
 0xa4
diff --git a/test/MC/Mips/elf-bigendian.ll b/test/MC/Mips/elf-bigendian.ll
index 875ba3ba965..71c69bb7afa 100644
--- a/test/MC/Mips/elf-bigendian.ll
+++ b/test/MC/Mips/elf-bigendian.ll
@@ -5,18 +5,18 @@
 
 ; Make sure that a section table (text) entry is correct.
 ; CHECK:   (('sh_name', 0x{{[0]*}}5) # '.text'
-; CHECKNEXT:   ('sh_type', 0x{{[0]*}}1)
-; CHECKNEXT:   ('sh_flags', 0x{{[0]*}}6)
-; CHECKNEXT:   ('sh_addr', 0x{{{[0-9,a-f]+}})
-; CHECKNEXT:   ('sh_offset', 0x{{{[0-9,a-f]+}})
-; CHECKNEXT:   ('sh_size', 0x{{{[0-9,a-f]+}})
-; CHECKNEXT:   ('sh_link', 0x{{[0]+}})
-; CHECKNEXT:   ('sh_info', 0x{{[0]+}})
-; CHECKNEXT:   ('sh_addralign', 0x{{[0]*}}4)
-; CHECKNEXT:   ('sh_entsize', 0x{{[0]+}})
+; CHECK-NEXT:   ('sh_type', 0x{{[0]*}}1)
+; CHECK-NEXT:   ('sh_flags', 0x{{[0]*}}6)
+; CHECK-NEXT:   ('sh_addr', 0x{{[0-9,a-f]+}})
+; CHECK-NEXT:   ('sh_offset', 0x{{[0-9,a-f]+}})
+; CHECK-NEXT:   ('sh_size', 0x{{[0-9,a-f]+}})
+; CHECK-NEXT:   ('sh_link', 0x{{[0]+}})
+; CHECK-NEXT:   ('sh_info', 0x{{[0]+}})
+; CHECK-NEXT:   ('sh_addralign', 0x{{[0]*}}4)
+; CHECK-NEXT:   ('sh_entsize', 0x{{[0]+}})
 
 ; See that at least first 3 instructions are correct: GP prologue
-; CHECKNEXT:   ('_section_data', '3c1c0000 279c0000 0399e021 {{[0-9,a-f]*}}')
+; CHECK-NEXT:   ('_section_data', '3c1c0000 279c0000 0399e021 {{[0-9,a-f, ]*}}')
 
 ; ModuleID = '../br1.c'
 target datalayout = "E-p:32:32:32-i1:8:8-i8:8:32-i16:16:32-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-n32"
diff --git a/test/MC/Mips/sym-offset.ll b/test/MC/Mips/sym-offset.ll
new file mode 100644
index 00000000000..59399358394
--- /dev/null
+++ b/test/MC/Mips/sym-offset.ll
@@ -0,0 +1,22 @@
+; RUN: llc -filetype=obj -mtriple mipsel-unknown-linux %s -o - | elf-dump --dump-section-data  | FileCheck %s
+
+; FIXME: use assembler instead of llc when it becomes available.
+
+@string1 = internal global [11 x i8] c"aaaaaaaaaa\00", align 1
+@string2 = internal global [10 x i8] c"aaaa\00bbbb\00", align 1
+
+define i32 @foo1(i32 %n) nounwind readonly {
+entry:
+; check that the immediate fields of lwl and lwr are three apart.
+; 8841000e        lwl     at,14(v0)
+; 9841000b        lwr     at,11(v0)
+
+; CHECK: ('_section_data', '00001c3c 00009c27 21e09903 0000828f 0e004188 0b004198
+
+  %call = tail call i32 @memcmp(i8* getelementptr inbounds ([11 x i8]* @string1, i32 0, i32 0), i8* getelementptr inbounds ([10 x i8]* @string2, i32 0, i32 0), i32 4) nounwind readonly
+  %cmp = icmp eq i32 %call, 0
+  %conv = zext i1 %cmp to i32
+  ret i32 %conv
+}
+
+declare i32 @memcmp(i8* nocapture, i8* nocapture, i32) nounwind readonly
diff --git a/test/Transforms/BBVectorize/no-ldstr-conn.ll b/test/Transforms/BBVectorize/no-ldstr-conn.ll
new file mode 100644
index 00000000000..ada2a71e36e
--- /dev/null
+++ b/test/Transforms/BBVectorize/no-ldstr-conn.ll
@@ -0,0 +1,23 @@
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+; RUN: opt < %s -bb-vectorize -bb-vectorize-req-chain-depth=2 -instcombine -gvn -S | FileCheck %s
+
+; Make sure that things (specifically getelementptr) are not connected to loads
+; and stores via the address operand (which would be bad because the address
+; is really a scalar even after vectorization)
+define i64 @test2(i64 %a) nounwind uwtable readonly {
+entry:
+  %a1 = inttoptr i64 %a to i64*
+  %a2 = getelementptr i64* %a1, i64 1
+  %a3 = getelementptr i64* %a1, i64 2
+  %v2 = load i64* %a2, align 8
+  %v3 = load i64* %a3, align 8
+  %v2a = add i64 %v2, 5
+  %v3a = add i64 %v3, 7
+  store i64 %v2a, i64* %a2, align 8
+  store i64 %v3a, i64* %a3, align 8
+  %r = add i64 %v2, %v3
+  ret i64 %r
+; CHECK: @test2
+; CHECK-NOT: getelementptr <2 x i64*>
+}
+
diff --git a/test/Transforms/BBVectorize/simple-ldstr-ptrs.ll b/test/Transforms/BBVectorize/simple-ldstr-ptrs.ll
new file mode 100644
index 00000000000..f992d415477
--- /dev/null
+++ b/test/Transforms/BBVectorize/simple-ldstr-ptrs.ll
@@ -0,0 +1,81 @@
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+; RUN: opt < %s -bb-vectorize -bb-vectorize-req-chain-depth=3 -instcombine -gvn -S | FileCheck %s
+; RUN: opt < %s -bb-vectorize -bb-vectorize-req-chain-depth=3 -bb-vectorize-aligned-only -instcombine -gvn -S | FileCheck %s -check-prefix=CHECK-AO
+
+; Simple 3-pair chain also with loads and stores (using ptrs and gep)
+define double @test1(i64* %a, i64* %b, i64* %c) nounwind uwtable readonly {
+entry:
+  %i0 = load i64* %a, align 8
+  %i1 = load i64* %b, align 8
+  %mul = mul i64 %i0, %i1
+  %arrayidx3 = getelementptr inbounds i64* %a, i64 1
+  %i3 = load i64* %arrayidx3, align 8
+  %arrayidx4 = getelementptr inbounds i64* %b, i64 1
+  %i4 = load i64* %arrayidx4, align 8
+  %mul5 = mul i64 %i3, %i4
+  %ptr = inttoptr i64 %mul to double*
+  %ptr5 = inttoptr i64 %mul5 to double*
+  %aptr = getelementptr inbounds double* %ptr, i64 2
+  %aptr5 = getelementptr inbounds double* %ptr5, i64 3
+  %av = load double* %aptr, align 16
+  %av5 = load double* %aptr5, align 16
+  %r = fmul double %av, %av5
+  store i64 %mul, i64* %c, align 8
+  %arrayidx5 = getelementptr inbounds i64* %c, i64 1
+  store i64 %mul5, i64* %arrayidx5, align 8
+  ret double %r
+; CHECK: @test1
+; CHECK: %i0.v.i0 = bitcast i64* %a to <2 x i64>*
+; CHECK: %i1.v.i0 = bitcast i64* %b to <2 x i64>*
+; CHECK: %i0 = load <2 x i64>* %i0.v.i0, align 8
+; CHECK: %i1 = load <2 x i64>* %i1.v.i0, align 8
+; CHECK: %mul = mul <2 x i64> %i0, %i1
+; CHECK: %ptr = inttoptr <2 x i64> %mul to <2 x double*>
+; CHECK: %aptr = getelementptr inbounds <2 x double*> %ptr, <2 x i64> 
+; CHECK: %aptr.v.r1 = extractelement <2 x double*> %aptr, i32 0
+; CHECK: %aptr.v.r2 = extractelement <2 x double*> %aptr, i32 1
+; CHECK: %av = load double* %aptr.v.r1, align 16
+; CHECK: %av5 = load double* %aptr.v.r2, align 16
+; CHECK: %r = fmul double %av, %av5
+; CHECK: %0 = bitcast i64* %c to <2 x i64>*
+; CHECK: store <2 x i64> %mul, <2 x i64>* %0, align 8
+; CHECK: ret double %r
+; CHECK-AO: @test1
+; CHECK-AO-NOT: load <2 x
+}
+
+; Simple 3-pair chain with loads and stores (using ptrs and gep)
+define void @test2(i64** %a, i64** %b, i64** %c) nounwind uwtable readonly {
+entry:
+  %i0 = load i64** %a, align 8
+  %i1 = load i64** %b, align 8
+  %arrayidx3 = getelementptr inbounds i64** %a, i64 1
+  %i3 = load i64** %arrayidx3, align 8
+  %arrayidx4 = getelementptr inbounds i64** %b, i64 1
+  %i4 = load i64** %arrayidx4, align 8
+  %o1 = load i64* %i1, align 8
+  %o4 = load i64* %i4, align 8
+  %ptr0 = getelementptr inbounds i64* %i0, i64 %o1
+  %ptr3 = getelementptr inbounds i64* %i3, i64 %o4
+  store i64* %ptr0, i64** %c, align 8
+  %arrayidx5 = getelementptr inbounds i64** %c, i64 1
+  store i64* %ptr3, i64** %arrayidx5, align 8
+  ret void
+; CHECK: @test2
+; CHECK: %i0.v.i0 = bitcast i64** %a to <2 x i64*>*
+; CHECK: %i1 = load i64** %b, align 8
+; CHECK: %i0 = load <2 x i64*>* %i0.v.i0, align 8
+; CHECK: %arrayidx4 = getelementptr inbounds i64** %b, i64 1
+; CHECK: %i4 = load i64** %arrayidx4, align 8
+; CHECK: %o1 = load i64* %i1, align 8
+; CHECK: %o4 = load i64* %i4, align 8
+; CHECK: %ptr0.v.i1.1 = insertelement <2 x i64> undef, i64 %o1, i32 0
+; CHECK: %ptr0.v.i1.2 = insertelement <2 x i64> %ptr0.v.i1.1, i64 %o4, i32 1
+; CHECK: %ptr0 = getelementptr inbounds <2 x i64*> %i0, <2 x i64> %ptr0.v.i1.2
+; CHECK: %0 = bitcast i64** %c to <2 x i64*>*
+; CHECK: store <2 x i64*> %ptr0, <2 x i64*>* %0, align 8
+; CHECK: ret void
+; CHECK-AO: @test2
+; CHECK-AO-NOT: <2 x
+}
+
diff --git a/test/Transforms/BBVectorize/simple-sel.ll b/test/Transforms/BBVectorize/simple-sel.ll
new file mode 100644
index 00000000000..4daa5714fbd
--- /dev/null
+++ b/test/Transforms/BBVectorize/simple-sel.ll
@@ -0,0 +1,30 @@
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
+; RUN: opt < %s -bb-vectorize -bb-vectorize-req-chain-depth=3 -instcombine -gvn -S | FileCheck %s
+
+; Basic depth-3 chain with select
+define double @test1(double %A1, double %A2, double %B1, double %B2, i1 %C1, i1 %C2) {
+; CHECK: @test1
+; CHECK: %X1.v.i1.1 = insertelement <2 x double> undef, double %B1, i32 0
+; CHECK: %X1.v.i0.1 = insertelement <2 x double> undef, double %A1, i32 0
+; CHECK: %X1.v.i1.2 = insertelement <2 x double> %X1.v.i1.1, double %B2, i32 1
+; CHECK: %X1.v.i0.2 = insertelement <2 x double> %X1.v.i0.1, double %A2, i32 1
+	%X1 = fsub double %A1, %B1
+	%X2 = fsub double %A2, %B2
+; CHECK: %X1 = fsub <2 x double> %X1.v.i0.2, %X1.v.i1.2
+	%Y1 = fmul double %X1, %A1
+	%Y2 = fmul double %X2, %A2
+; CHECK: %Y1 = fmul <2 x double> %X1, %X1.v.i0.2
+        %Z1 = select i1 %C1, double %Y1, double %B1
+        %Z2 = select i1 %C2, double %Y2, double %B2
+; CHECK: %Z1.v.i0.1 = insertelement <2 x i1> undef, i1 %C1, i32 0
+; CHECK: %Z1.v.i0.2 = insertelement <2 x i1> %Z1.v.i0.1, i1 %C2, i32 1
+; CHECK: %Z1 = select <2 x i1> %Z1.v.i0.2, <2 x double> %Y1, <2 x double> %X1.v.i1.2
+	%R  = fmul double %Z1, %Z2
+; CHECK: %Z1.v.r1 = extractelement <2 x double> %Z1, i32 0
+; CHECK: %Z1.v.r2 = extractelement <2 x double> %Z1, i32 1
+; CHECK: %R = fmul double %Z1.v.r1, %Z1.v.r2
+	ret double %R
+; CHECK: ret double %R
+}
+
+
diff --git a/test/Transforms/GlobalOpt/constantfold-initializers.ll b/test/Transforms/GlobalOpt/constantfold-initializers.ll
index af8fa324db8..ce6e2c46d62 100644
--- a/test/Transforms/GlobalOpt/constantfold-initializers.ll
+++ b/test/Transforms/GlobalOpt/constantfold-initializers.ll
@@ -12,6 +12,11 @@ target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f3
 @xs = global [2 x i32] zeroinitializer, align 4
 ; CHECK: @xs = global [2 x i32] [i32 1, i32 1]
 
+; PR12642
+%PR12642.struct = type { i8 }
+@PR12642.s = global <{}> zeroinitializer, align 1
+@PR12642.p = constant %PR12642.struct* bitcast (i8* getelementptr (i8* bitcast (<{}>* @PR12642.s to i8*), i64 1) to %PR12642.struct*), align 8
+
 define internal void @test1() {
 entry:
   store i32 1, i32* getelementptr inbounds ([2 x i32]* @xs, i64 0, i64 0)
diff --git a/test/Transforms/InstCombine/2012-04-30-SRem.ll b/test/Transforms/InstCombine/2012-04-30-SRem.ll
new file mode 100644
index 00000000000..a285d5aea5e
--- /dev/null
+++ b/test/Transforms/InstCombine/2012-04-30-SRem.ll
@@ -0,0 +1,12 @@
+; RUN: opt -instcombine -S < %s | FileCheck %s
+; PR12541
+
+define i32 @foo(i32 %x) {
+  %y = xor i32 %x, 3
+  %z = srem i32 1656690544, %y
+  %sext = shl i32 %z, 24
+  %s = ashr exact i32 %sext, 24
+  ret i32 %s
+; CHECK-NOT: and
+; The shifts were wrongly being turned into an and with 112
+}
diff --git a/test/Transforms/InstCombine/apint-shift.ll b/test/Transforms/InstCombine/apint-shift.ll
index 55243a64918..0ea73a058c0 100644
--- a/test/Transforms/InstCombine/apint-shift.ll
+++ b/test/Transforms/InstCombine/apint-shift.ll
@@ -1,70 +1,93 @@
-; This test makes sure that shit instructions are properly eliminated
+; This test makes sure that shift instructions are properly eliminated
 ; even with arbitrary precision integers.
-; RUN: opt < %s -instcombine -S | not grep sh
-; END.
+; RUN: opt < %s -instcombine -S | FileCheck %s
 
+; CHECK: @test1
+; CHECK-NOT: sh
 define i47 @test1(i47 %A) {
 	%B = shl i47 %A, 0		;  [#uses=1]
 	ret i47 %B
 }
 
+; CHECK: @test2
+; CHECK-NOT: sh
 define i41 @test2(i7 %X) {
 	%A = zext i7 %X to i41		;  [#uses=1]
 	%B = shl i41 0, %A		;  [#uses=1]
 	ret i41 %B
 }
 
+; CHECK: @test3
+; CHECK-NOT: sh
 define i41 @test3(i41 %A) {
 	%B = ashr i41 %A, 0		;  [#uses=1]
 	ret i41 %B
 }
 
+; CHECK: @test4
+; CHECK-NOT: sh
 define i39 @test4(i7 %X) {
 	%A = zext i7 %X to i39		;  [#uses=1]
 	%B = ashr i39 0, %A		;  [#uses=1]
 	ret i39 %B
 }
 
+; CHECK: @test5
+; CHECK-NOT: sh
 define i55 @test5(i55 %A) {
 	%B = lshr i55 %A, 55		;  [#uses=1]
 	ret i55 %B
 }
 
+; CHECK: @test5a
+; CHECK-NOT: sh
 define i32 @test5a(i32 %A) {
 	%B = shl i32 %A, 32		;  [#uses=1]
 	ret i32 %B
 }
 
+; CHECK: @test6
+; CHECK-NOT: sh
 define i55 @test6(i55 %A) {
 	%B = shl i55 %A, 1		;  [#uses=1]
 	%C = mul i55 %B, 3		;  [#uses=1]
 	ret i55 %C
 }
 
+; CHECK: @test7
+; CHECK-NOT: sh
 define i29 @test7(i8 %X) {
 	%A = zext i8 %X to i29		;  [#uses=1]
 	%B = ashr i29 -1, %A		;  [#uses=1]
 	ret i29 %B
 }
 
+; CHECK: @test8
+; CHECK-NOT: sh
 define i7 @test8(i7 %A) {
 	%B = shl i7 %A, 4		;  [#uses=1]
 	%C = shl i7 %B, 3		;  [#uses=1]
 	ret i7 %C
 }
 
+; CHECK: @test9
+; CHECK-NOT: sh
 define i17 @test9(i17 %A) {
 	%B = shl i17 %A, 16		;  [#uses=1]
 	%C = lshr i17 %B, 16		;  [#uses=1]
 	ret i17 %C
 }
 
+; CHECK: @test10
+; CHECK-NOT: sh
 define i19 @test10(i19 %A) {
 	%B = lshr i19 %A, 18		;  [#uses=1]
 	%C = shl i19 %B, 18		;  [#uses=1]
 	ret i19 %C
 }
 
+; CHECK: @test11
+; CHECK-NOT: sh
 define i23 @test11(i23 %A) {
 	%a = mul i23 %A, 3		;  [#uses=1]
 	%B = lshr i23 %a, 11		;  [#uses=1]
@@ -72,12 +95,16 @@ define i23 @test11(i23 %A) {
 	ret i23 %C
 }
 
+; CHECK: @test12
+; CHECK-NOT: sh
 define i47 @test12(i47 %A) {
 	%B = ashr i47 %A, 8		;  [#uses=1]
 	%C = shl i47 %B, 8		;  [#uses=1]
 	ret i47 %C
 }
 
+; CHECK: @test13
+; CHECK-NOT: sh
 define i18 @test13(i18 %A) {
 	%a = mul i18 %A, 3		;  [#uses=1]
 	%B = ashr i18 %a, 8		;  [#uses=1]
@@ -85,6 +112,8 @@ define i18 @test13(i18 %A) {
 	ret i18 %C
 }
 
+; CHECK: @test14
+; CHECK-NOT: sh
 define i35 @test14(i35 %A) {
 	%B = lshr i35 %A, 4		;  [#uses=1]
 	%C = or i35 %B, 1234		;  [#uses=1]
@@ -92,6 +121,8 @@ define i35 @test14(i35 %A) {
 	ret i35 %D
 }
 
+; CHECK: @test14a
+; CHECK-NOT: sh
 define i79 @test14a(i79 %A) {
 	%B = shl i79 %A, 4		;  [#uses=1]
 	%C = and i79 %B, 1234		;  [#uses=1]
@@ -99,12 +130,16 @@ define i79 @test14a(i79 %A) {
 	ret i79 %D
 }
 
+; CHECK: @test15
+; CHECK-NOT: sh
 define i45 @test15(i1 %C) {
 	%A = select i1 %C, i45 3, i45 1	;  [#uses=1]
 	%V = shl i45 %A, 2		;  [#uses=1]
 	ret i45 %V
 }
 
+; CHECK: @test15a
+; CHECK-NOT: sh
 define i53 @test15a(i1 %X) {
 	%A = select i1 %X, i8 3, i8 1	;  [#uses=1]
 	%B = zext i8 %A to i53		;  [#uses=1]
@@ -112,6 +147,8 @@ define i53 @test15a(i1 %X) {
 	ret i53 %V
 }
 
+; CHECK: @test16
+; CHECK-NOT: sh
 define i1 @test16(i84 %X) {
 	%tmp.3 = ashr i84 %X, 4		;  [#uses=1]
 	%tmp.6 = and i84 %tmp.3, 1	;  [#uses=1]
@@ -119,48 +156,64 @@ define i1 @test16(i84 %X) {
 	ret i1 %tmp.7
 }
 
+; CHECK: @test17
+; CHECK-NOT: sh
 define i1 @test17(i106 %A) {
 	%B = lshr i106 %A, 3		;  [#uses=1]
 	%C = icmp eq i106 %B, 1234	;  [#uses=1]
 	ret i1 %C
 }
 
+; CHECK: @test18
+; CHECK-NOT: sh
 define i1 @test18(i11 %A) {
 	%B = lshr i11 %A, 10		;  [#uses=1]
 	%C = icmp eq i11 %B, 123	;  [#uses=1]
 	ret i1 %C
 }
 
+; CHECK: @test19
+; CHECK-NOT: sh
 define i1 @test19(i37 %A) {
 	%B = ashr i37 %A, 2		;  [#uses=1]
 	%C = icmp eq i37 %B, 0		;  [#uses=1]
 	ret i1 %C
 }
 
+; CHECK: @test19a
+; CHECK-NOT: sh
 define i1 @test19a(i39 %A) {
 	%B = ashr i39 %A, 2		;  [#uses=1]
 	%C = icmp eq i39 %B, -1		;  [#uses=1]
 	ret i1 %C
 }
 
+; CHECK: @test20
+; CHECK-NOT: sh
 define i1 @test20(i13 %A) {
 	%B = ashr i13 %A, 12		;  [#uses=1]
 	%C = icmp eq i13 %B, 123	;  [#uses=1]
 	ret i1 %C
 }
 
+; CHECK: @test21
+; CHECK-NOT: sh
 define i1 @test21(i12 %A) {
 	%B = shl i12 %A, 6		;  [#uses=1]
 	%C = icmp eq i12 %B, -128		;  [#uses=1]
 	ret i1 %C
 }
 
+; CHECK: @test22
+; CHECK-NOT: sh
 define i1 @test22(i14 %A) {
 	%B = shl i14 %A, 7		;  [#uses=1]
 	%C = icmp eq i14 %B, 0		;  [#uses=1]
 	ret i1 %C
 }
 
+; CHECK: @test23
+; CHECK-NOT: sh
 define i11 @test23(i44 %A) {
 	%B = shl i44 %A, 33		;  [#uses=1]
 	%C = ashr i44 %B, 33		;  [#uses=1]
@@ -168,6 +221,8 @@ define i11 @test23(i44 %A) {
 	ret i11 %D
 }
 
+; CHECK: @test25
+; CHECK-NOT: sh
 define i37 @test25(i37 %tmp.2, i37 %AA) {
 	%x = lshr i37 %AA, 17		;  [#uses=1]
 	%tmp.3 = lshr i37 %tmp.2, 17		;  [#uses=1]
@@ -176,6 +231,8 @@ define i37 @test25(i37 %tmp.2, i37 %AA) {
 	ret i37 %tmp.6
 }
 
+; CHECK: @test26
+; CHECK-NOT: sh
 define i40 @test26(i40 %A) {
 	%B = lshr i40 %A, 1		;  [#uses=1]
 	%C = bitcast i40 %B to i40		;  [#uses=1]
diff --git a/test/Transforms/LoopStrengthReduce/X86/2012-01-13-phielim.ll b/test/Transforms/LoopStrengthReduce/X86/2012-01-13-phielim.ll
index 2dcaab82a1a..ed32ca8659c 100644
--- a/test/Transforms/LoopStrengthReduce/X86/2012-01-13-phielim.ll
+++ b/test/Transforms/LoopStrengthReduce/X86/2012-01-13-phielim.ll
@@ -61,7 +61,7 @@ exit:                                 ; preds = %cond.true29.i, %cond.true.i
 ; CHECK: @test2
 ; CHECK: %entry
 ; CHECK-NOT: mov
-; CHECK: jne
+; CHECK: je
 define void @test2(i32 %n) nounwind uwtable {
 entry:
   br i1 undef, label %while.end, label %for.cond468
diff --git a/test/Transforms/LoopStrengthReduce/pr12691.ll b/test/Transforms/LoopStrengthReduce/pr12691.ll
new file mode 100644
index 00000000000..8399434343e
--- /dev/null
+++ b/test/Transforms/LoopStrengthReduce/pr12691.ll
@@ -0,0 +1,34 @@
+; RUN: opt < %s -loop-reduce -S | FileCheck %s
+
+@d = common global i32 0, align 4
+
+define void @fn2(i32 %x) nounwind uwtable {
+entry:
+  br label %for.cond
+
+for.cond:
+  %g.0 = phi i32 [ 0, %entry ], [ %dec, %for.cond ]
+  %tobool = icmp eq i32 %x, 0
+  %dec = add nsw i32 %g.0, -1
+  br i1 %tobool, label %for.cond, label %for.end
+
+for.end:
+; CHECK:  %tmp1 = load i32* @d, align 4
+; CHECK-NEXT:  %tmp2 = load i32* @d, align 4
+; CHECK-NEXT:  %0 = sub i32 %tmp1, %tmp2
+
+  %tmp1 = load i32* @d, align 4
+  %add = add nsw i32 %tmp1, %g.0
+  %tmp2 = load i32* @d, align 4
+  %tobool26 = icmp eq i32 %x, 0
+  br i1 %tobool26, label %for.end5, label %for.body.lr.ph
+
+for.body.lr.ph:
+  %tobool3 = icmp ne i32 %tmp2, %add
+  br label %for.end5
+
+for.end5:
+  ret void
+}
+
+
diff --git a/test/Transforms/LoopUnswitch/2012-04-30-LoopUnswitch-LPad-Crash.ll b/test/Transforms/LoopUnswitch/2012-04-30-LoopUnswitch-LPad-Crash.ll
new file mode 100644
index 00000000000..261876df2aa
--- /dev/null
+++ b/test/Transforms/LoopUnswitch/2012-04-30-LoopUnswitch-LPad-Crash.ll
@@ -0,0 +1,101 @@
+; RUN: opt < %s -basicaa -instcombine -inline -functionattrs -licm -loop-unswitch -gvn -verify
+; PR12573
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.7.0"
+
+%class.D.22.42.66.102.138.158.178.198.238.242.246.250.262.294.302.338.346.379 = type { %class.C.23.43.67.103.139.159.179.199.239.243.247.251.263.295.303.339.347.376*, %class.B.21.41.65.101.137.157.177.197.237.241.245.249.261.293.301.337.345.378 }
+%class.C.23.43.67.103.139.159.179.199.239.243.247.251.263.295.303.339.347.376 = type { %class.D.22.42.66.102.138.158.178.198.238.242.246.250.262.294.302.338.346.379* }
+%class.B.21.41.65.101.137.157.177.197.237.241.245.249.261.293.301.337.345.378 = type { %class.A.20.40.64.100.136.156.176.196.236.240.244.248.260.292.300.336.344.377* }
+%class.A.20.40.64.100.136.156.176.196.236.240.244.248.260.292.300.336.344.377 = type { i8 }
+
+define void @_Z23get_reconstruction_pathv() uwtable ssp {
+entry:
+  %c = alloca %class.D.22.42.66.102.138.158.178.198.238.242.246.250.262.294.302.338.346.379, align 8
+  br label %for.cond
+
+for.cond:                                         ; preds = %for.end, %entry
+  invoke void @_ZN1DptEv(%class.D.22.42.66.102.138.158.178.198.238.242.246.250.262.294.302.338.346.379* %c)
+          to label %invoke.cont unwind label %lpad
+
+invoke.cont:                                      ; preds = %for.cond
+  invoke void @_ZN1C3endEv()
+          to label %for.cond3 unwind label %lpad
+
+for.cond3:                                        ; preds = %invoke.cont6, %invoke.cont
+  invoke void @_ZN1DptEv(%class.D.22.42.66.102.138.158.178.198.238.242.246.250.262.294.302.338.346.379* %c)
+          to label %invoke.cont4 unwind label %lpad
+
+invoke.cont4:                                     ; preds = %for.cond3
+  invoke void @_ZN1C3endEv()
+          to label %invoke.cont6 unwind label %lpad
+
+invoke.cont6:                                     ; preds = %invoke.cont4
+  br i1 undef, label %for.cond3, label %for.end
+
+lpad:                                             ; preds = %for.end, %invoke.cont4, %for.cond3, %invoke.cont, %for.cond
+  %0 = landingpad { i8*, i32 } personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*)
+          cleanup
+  resume { i8*, i32 } undef
+
+for.end:                                          ; preds = %invoke.cont6
+  invoke void @_ZN1C13_M_insert_auxER1D()
+          to label %for.cond unwind label %lpad
+}
+
+define void @_ZN1DptEv(%class.D.22.42.66.102.138.158.178.198.238.242.246.250.262.294.302.338.346.379* %this) uwtable ssp align 2 {
+entry:
+  %this.addr = alloca %class.D.22.42.66.102.138.158.178.198.238.242.246.250.262.294.302.338.346.379*, align 8
+  store %class.D.22.42.66.102.138.158.178.198.238.242.246.250.262.294.302.338.346.379* %this, %class.D.22.42.66.102.138.158.178.198.238.242.246.250.262.294.302.338.346.379** %this.addr, align 8, !tbaa !0
+  %this1 = load %class.D.22.42.66.102.138.158.178.198.238.242.246.250.262.294.302.338.346.379** %this.addr
+  %px = getelementptr inbounds %class.D.22.42.66.102.138.158.178.198.238.242.246.250.262.294.302.338.346.379* %this1, i32 0, i32 0
+  %0 = load %class.C.23.43.67.103.139.159.179.199.239.243.247.251.263.295.303.339.347.376** %px, align 8, !tbaa !0
+  %tobool = icmp ne %class.C.23.43.67.103.139.159.179.199.239.243.247.251.263.295.303.339.347.376* %0, null
+  br i1 %tobool, label %cond.end, label %cond.false
+
+cond.false:                                       ; preds = %entry
+  call void @_Z10__assert13v() noreturn
+  unreachable
+
+cond.end:                                         ; preds = %entry
+  ret void
+}
+
+declare i32 @__gxx_personality_v0(...)
+
+declare void @_ZN1C3endEv()
+
+define void @_ZN1C13_M_insert_auxER1D() uwtable ssp align 2 {
+entry:
+  ret void
+}
+
+define void @_ZN1DD1Ev() unnamed_addr uwtable inlinehint ssp align 2 {
+entry:
+  ret void
+}
+
+define void @_ZN1DD2Ev() unnamed_addr uwtable inlinehint ssp align 2 {
+entry:
+  ret void
+}
+
+define void @_ZN1BD1Ev() unnamed_addr uwtable ssp align 2 {
+entry:
+  ret void
+}
+
+define void @_ZN1BD2Ev() unnamed_addr uwtable ssp align 2 {
+entry:
+  ret void
+}
+
+define void @_ZN1BaSERS_() uwtable ssp align 2 {
+entry:
+  unreachable
+}
+
+declare void @_Z10__assert13v() noreturn
+
+!0 = metadata !{metadata !"any pointer", metadata !1}
+!1 = metadata !{metadata !"omnipotent char", metadata !2}
+!2 = metadata !{metadata !"Simple C/C++ TBAA", null}
diff --git a/test/Transforms/ObjCARC/escape.ll b/test/Transforms/ObjCARC/escape.ll
new file mode 100644
index 00000000000..3f694cf1d5a
--- /dev/null
+++ b/test/Transforms/ObjCARC/escape.ll
@@ -0,0 +1,131 @@
+; RUN: opt -objc-arc -S < %s | FileCheck %s
+; rdar://11229925
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+
+%struct.__block_byref_weakLogNTimes = type { i8*, %struct.__block_byref_weakLogNTimes*, i32, i32, i8*, i8*, void (...)* }
+%struct.__block_descriptor = type { i64, i64 }
+
+; Don't optimize away the retainBlock, because the object's address "escapes"
+; with the objc_storeWeak call.
+
+; CHECK: define void @test0(
+; CHECK: %tmp7 = call i8* @objc_retainBlock(i8* %tmp6) nounwind, !clang.arc.copy_on_escape !0
+; CHECK: call void @objc_release(i8* %tmp7) nounwind, !clang.imprecise_release !0
+; CHECK: }
+define void @test0() nounwind {
+entry:
+  %weakLogNTimes = alloca %struct.__block_byref_weakLogNTimes, align 8
+  %block = alloca <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>, align 8
+  %byref.isa = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 0
+  store i8* null, i8** %byref.isa, align 8
+  %byref.forwarding = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 1
+  store %struct.__block_byref_weakLogNTimes* %weakLogNTimes, %struct.__block_byref_weakLogNTimes** %byref.forwarding, align 8
+  %byref.flags = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 2
+  store i32 33554432, i32* %byref.flags, align 8
+  %byref.size = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 3
+  store i32 48, i32* %byref.size, align 4
+  %tmp1 = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 4
+  store i8* bitcast (void (i8*, i8*)* @__Block_byref_object_copy_ to i8*), i8** %tmp1, align 8
+  %tmp2 = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 5
+  store i8* bitcast (void (i8*)* @__Block_byref_object_dispose_ to i8*), i8** %tmp2, align 8
+  %weakLogNTimes1 = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 6
+  %tmp3 = bitcast void (...)** %weakLogNTimes1 to i8**
+  %tmp4 = call i8* @objc_initWeak(i8** %tmp3, i8* null) nounwind
+  %block.isa = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 0
+  store i8* null, i8** %block.isa, align 8
+  %block.flags = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 1
+  store i32 1107296256, i32* %block.flags, align 8
+  %block.reserved = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 2
+  store i32 0, i32* %block.reserved, align 4
+  %block.invoke = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 3
+  store i8* bitcast (void (i8*, i32)* @__main_block_invoke_0 to i8*), i8** %block.invoke, align 8
+  %block.descriptor = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 4
+  store %struct.__block_descriptor* null, %struct.__block_descriptor** %block.descriptor, align 8
+  %block.captured = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 5
+  %tmp5 = bitcast %struct.__block_byref_weakLogNTimes* %weakLogNTimes to i8*
+  store i8* %tmp5, i8** %block.captured, align 8
+  %tmp6 = bitcast <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block to i8*
+  %tmp7 = call i8* @objc_retainBlock(i8* %tmp6) nounwind, !clang.arc.copy_on_escape !0
+  %tmp8 = load %struct.__block_byref_weakLogNTimes** %byref.forwarding, align 8
+  %weakLogNTimes3 = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %tmp8, i64 0, i32 6
+  %tmp9 = bitcast void (...)** %weakLogNTimes3 to i8**
+  %tmp10 = call i8* @objc_storeWeak(i8** %tmp9, i8* %tmp7) nounwind
+  %tmp11 = getelementptr inbounds i8* %tmp7, i64 16
+  %tmp12 = bitcast i8* %tmp11 to i8**
+  %tmp13 = load i8** %tmp12, align 8
+  %tmp14 = bitcast i8* %tmp13 to void (i8*, i32)*
+  call void %tmp14(i8* %tmp7, i32 10) nounwind, !clang.arc.no_objc_arc_exceptions !0
+  call void @objc_release(i8* %tmp7) nounwind, !clang.imprecise_release !0
+  call void @_Block_object_dispose(i8* %tmp5, i32 8) nounwind
+  call void @objc_destroyWeak(i8** %tmp3) nounwind
+  ret void
+}
+
+; Like test0, but it makes a regular call instead of a storeWeak call,
+; so the optimization is valid.
+
+; CHECK: define void @test1(
+; CHECK-NOT: @objc_retainBlock
+; CHECK: }
+define void @test1() nounwind {
+entry:
+  %weakLogNTimes = alloca %struct.__block_byref_weakLogNTimes, align 8
+  %block = alloca <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>, align 8
+  %byref.isa = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 0
+  store i8* null, i8** %byref.isa, align 8
+  %byref.forwarding = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 1
+  store %struct.__block_byref_weakLogNTimes* %weakLogNTimes, %struct.__block_byref_weakLogNTimes** %byref.forwarding, align 8
+  %byref.flags = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 2
+  store i32 33554432, i32* %byref.flags, align 8
+  %byref.size = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 3
+  store i32 48, i32* %byref.size, align 4
+  %tmp1 = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 4
+  store i8* bitcast (void (i8*, i8*)* @__Block_byref_object_copy_ to i8*), i8** %tmp1, align 8
+  %tmp2 = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 5
+  store i8* bitcast (void (i8*)* @__Block_byref_object_dispose_ to i8*), i8** %tmp2, align 8
+  %weakLogNTimes1 = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %weakLogNTimes, i64 0, i32 6
+  %tmp3 = bitcast void (...)** %weakLogNTimes1 to i8**
+  %tmp4 = call i8* @objc_initWeak(i8** %tmp3, i8* null) nounwind
+  %block.isa = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 0
+  store i8* null, i8** %block.isa, align 8
+  %block.flags = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 1
+  store i32 1107296256, i32* %block.flags, align 8
+  %block.reserved = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 2
+  store i32 0, i32* %block.reserved, align 4
+  %block.invoke = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 3
+  store i8* bitcast (void (i8*, i32)* @__main_block_invoke_0 to i8*), i8** %block.invoke, align 8
+  %block.descriptor = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 4
+  store %struct.__block_descriptor* null, %struct.__block_descriptor** %block.descriptor, align 8
+  %block.captured = getelementptr inbounds <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block, i64 0, i32 5
+  %tmp5 = bitcast %struct.__block_byref_weakLogNTimes* %weakLogNTimes to i8*
+  store i8* %tmp5, i8** %block.captured, align 8
+  %tmp6 = bitcast <{ i8*, i32, i32, i8*, %struct.__block_descriptor*, i8* }>* %block to i8*
+  %tmp7 = call i8* @objc_retainBlock(i8* %tmp6) nounwind, !clang.arc.copy_on_escape !0
+  %tmp8 = load %struct.__block_byref_weakLogNTimes** %byref.forwarding, align 8
+  %weakLogNTimes3 = getelementptr inbounds %struct.__block_byref_weakLogNTimes* %tmp8, i64 0, i32 6
+  %tmp9 = bitcast void (...)** %weakLogNTimes3 to i8**
+  %tmp10 = call i8* @not_really_objc_storeWeak(i8** %tmp9, i8* %tmp7) nounwind
+  %tmp11 = getelementptr inbounds i8* %tmp7, i64 16
+  %tmp12 = bitcast i8* %tmp11 to i8**
+  %tmp13 = load i8** %tmp12, align 8
+  %tmp14 = bitcast i8* %tmp13 to void (i8*, i32)*
+  call void %tmp14(i8* %tmp7, i32 10) nounwind, !clang.arc.no_objc_arc_exceptions !0
+  call void @objc_release(i8* %tmp7) nounwind, !clang.imprecise_release !0
+  call void @_Block_object_dispose(i8* %tmp5, i32 8) nounwind
+  call void @objc_destroyWeak(i8** %tmp3) nounwind
+  ret void
+}
+
+declare void @__Block_byref_object_copy_(i8*, i8*) nounwind
+declare void @__Block_byref_object_dispose_(i8*) nounwind
+declare void @objc_destroyWeak(i8**)
+declare i8* @objc_initWeak(i8**, i8*)
+declare void @__main_block_invoke_0(i8* nocapture, i32) nounwind ssp
+declare void @_Block_object_dispose(i8*, i32)
+declare i8* @objc_retainBlock(i8*)
+declare i8* @objc_storeWeak(i8**, i8*)
+declare i8* @not_really_objc_storeWeak(i8**, i8*)
+declare void @objc_release(i8*)
+
+!0 = metadata !{}
diff --git a/test/Transforms/Reassociate/pr12245.ll b/test/Transforms/Reassociate/pr12245.ll
new file mode 100644
index 00000000000..84098bdb006
--- /dev/null
+++ b/test/Transforms/Reassociate/pr12245.ll
@@ -0,0 +1,50 @@
+; RUN: opt < %s -basicaa -inline -instcombine -reassociate -dse -disable-output
+; PR12245
+
+@a = common global i32 0, align 4
+@d = common global i32 0, align 4
+
+define i32 @fn2() nounwind uwtable ssp {
+entry:
+  %0 = load i32* @a, align 4, !tbaa !0
+  %dec = add nsw i32 %0, -1
+  store i32 %dec, i32* @a, align 4, !tbaa !0
+  %1 = load i32* @d, align 4, !tbaa !0
+  %sub = sub nsw i32 %dec, %1
+  store i32 %sub, i32* @d, align 4, !tbaa !0
+  %2 = load i32* @a, align 4, !tbaa !0
+  %dec1 = add nsw i32 %2, -1
+  store i32 %dec1, i32* @a, align 4, !tbaa !0
+  %3 = load i32* @d, align 4, !tbaa !0
+  %sub2 = sub nsw i32 %dec1, %3
+  store i32 %sub2, i32* @d, align 4, !tbaa !0
+  %4 = load i32* @a, align 4, !tbaa !0
+  %dec3 = add nsw i32 %4, -1
+  store i32 %dec3, i32* @a, align 4, !tbaa !0
+  %5 = load i32* @d, align 4, !tbaa !0
+  %sub4 = sub nsw i32 %dec3, %5
+  store i32 %sub4, i32* @d, align 4, !tbaa !0
+  %6 = load i32* @a, align 4, !tbaa !0
+  %dec5 = add nsw i32 %6, -1
+  store i32 %dec5, i32* @a, align 4, !tbaa !0
+  %7 = load i32* @d, align 4, !tbaa !0
+  %sub6 = sub nsw i32 %dec5, %7
+  store i32 %sub6, i32* @d, align 4, !tbaa !0
+  %8 = load i32* @a, align 4, !tbaa !0
+  %dec7 = add nsw i32 %8, -1
+  store i32 %dec7, i32* @a, align 4, !tbaa !0
+  %9 = load i32* @d, align 4, !tbaa !0
+  %sub8 = sub nsw i32 %dec7, %9
+  store i32 %sub8, i32* @d, align 4, !tbaa !0
+  ret i32 0
+}
+
+define i32 @fn1() nounwind uwtable ssp {
+entry:
+  %call = call i32 @fn2()
+  ret i32 %call
+}
+
+!0 = metadata !{metadata !"int", metadata !1}
+!1 = metadata !{metadata !"omnipotent char", metadata !2}
+!2 = metadata !{metadata !"Simple C/C++ TBAA"}
diff --git a/test/Transforms/SimplifyLibCalls/floor.ll b/test/Transforms/SimplifyLibCalls/floor.ll
index 8780e32e0a0..03dcdf585f9 100644
--- a/test/Transforms/SimplifyLibCalls/floor.ll
+++ b/test/Transforms/SimplifyLibCalls/floor.ll
@@ -1,16 +1,31 @@
-; RUN: opt < %s -simplify-libcalls -S > %t
-; RUN: not grep {call.*floor(} %t
-; RUN: grep {call.*floorf(} %t
-; RUN: not grep {call.*ceil(} %t
-; RUN: grep {call.*ceilf(} %t
-; RUN: not grep {call.*nearbyint(} %t
-; RUN: grep {call.*nearbyintf(} %t
-; XFAIL: sparc
+; RUN: opt < %s -simplify-libcalls -S -mtriple "i386-pc-linux" | FileCheck -check-prefix=DO-SIMPLIFY %s
+; RUN: opt < %s -simplify-libcalls -S -mtriple "i386-pc-win32" | FileCheck -check-prefix=DONT-SIMPLIFY %s
+; RUN: opt < %s -simplify-libcalls -S -mtriple "x86_64-pc-win32" | FileCheck -check-prefix=C89-SIMPLIFY %s
+; RUN: opt < %s -simplify-libcalls -S -mtriple "i386-pc-mingw32" | FileCheck -check-prefix=DO-SIMPLIFY %s
+; RUN: opt < %s -simplify-libcalls -S -mtriple "x86_64-pc-mingw32" | FileCheck -check-prefix=DO-SIMPLIFY %s
+; RUN: opt < %s -simplify-libcalls -S -mtriple "sparc-sun-solaris" | FileCheck -check-prefix=DO-SIMPLIFY %s
+
+; DO-SIMPLIFY: call float @floorf(
+; DO-SIMPLIFY: call float @ceilf(
+; DO-SIMPLIFY: call float @roundf(
+; DO-SIMPLIFY: call float @nearbyintf(
+
+; C89-SIMPLIFY: call float @floorf(
+; C89-SIMPLIFY: call float @ceilf(
+; C89-SIMPLIFY: call double @round(
+; C89-SIMPLIFY: call double @nearbyint(
+
+; DONT-SIMPLIFY: call double @floor(
+; DONT-SIMPLIFY: call double @ceil(
+; DONT-SIMPLIFY: call double @round(
+; DONT-SIMPLIFY: call double @nearbyint(
 
 declare double @floor(double)
 
 declare double @ceil(double)
 
+declare double @round(double)
+
 declare double @nearbyint(double)
 
 define float @test_floor(float %C) {
@@ -29,8 +44,14 @@ define float @test_ceil(float %C) {
 	ret float %F
 }
 
-; PR8466
-; XFAIL: win32
+define float @test_round(float %C) {
+	%D = fpext float %C to double		;  [#uses=1]
+	; --> roundf
+        %E = call double @round( double %D )		;  [#uses=1]
+	%F = fptrunc double %E to float		;  [#uses=1]
+	ret float %F
+}
+
 define float @test_nearbyint(float %C) {
 	%D = fpext float %C to double		;  [#uses=1]
 	; --> nearbyintf
diff --git a/test/Transforms/SimplifyLibCalls/win-math.ll b/test/Transforms/SimplifyLibCalls/win-math.ll
new file mode 100644
index 00000000000..367e5b80721
--- /dev/null
+++ b/test/Transforms/SimplifyLibCalls/win-math.ll
@@ -0,0 +1,275 @@
+; RUN: opt -O2 -S -mtriple=i386-pc-win32 < %s | FileCheck %s -check-prefix=WIN32
+; RUN: opt -O2 -S -mtriple=x86_64-pc-win32 < %s | FileCheck %s -check-prefix=WIN64
+; RUN: opt -O2 -S -mtriple=i386-pc-mingw32 < %s | FileCheck %s -check-prefix=MINGW32
+; RUN: opt -O2 -S -mtriple=x86_64-pc-mingw32 < %s | FileCheck %s -check-prefix=MINGW64
+
+; x86 win32 msvcrt does not provide entry points for single-precision libm.
+; x86-64 win32 msvcrt does (except for fabsf)
+; msvcrt does not provide C99 math, but mingw32 does.
+
+declare double @acos(double %x)
+define float @float_acos(float %x) nounwind readnone {
+; WIN32: @float_acos
+; WIN32-NOT: float @acosf
+; WIN32: double @acos
+    %1 = fpext float %x to double
+    %2 = call double @acos(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @asin(double %x)
+define float @float_asin(float %x) nounwind readnone {
+; WIN32: @float_asin
+; WIN32-NOT: float @asinf
+; WIN32: double @asin
+    %1 = fpext float %x to double
+    %2 = call double @asin(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @atan(double %x)
+define float @float_atan(float %x) nounwind readnone {
+; WIN32: @float_atan
+; WIN32-NOT: float @atanf
+; WIN32: double @atan
+    %1 = fpext float %x to double
+    %2 = call double @atan(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @atan2(double %x, double %y)
+define float @float_atan2(float %x, float %y) nounwind readnone {
+; WIN32: @float_atan2
+; WIN32-NOT: float @atan2f
+; WIN32: double @atan2
+    %1 = fpext float %x to double
+    %2 = fpext float %y to double
+    %3 = call double @atan2(double %1, double %2)
+    %4 = fptrunc double %3 to float
+    ret float %4
+}
+
+declare double @ceil(double %x)
+define float @float_ceil(float %x) nounwind readnone {
+; WIN32: @float_ceil
+; WIN32-NOT: float @ceilf
+; WIN32: double @ceil
+; WIN64: @float_ceil
+; WIN64: float @ceilf
+; WIN64-NOT: double @ceil
+; MINGW32: @float_ceil
+; MINGW32: float @ceilf
+; MINGW32-NOT: double @ceil
+; MINGW64: @float_ceil
+; MINGW64: float @ceilf
+; MINGW64-NOT: double @ceil
+    %1 = fpext float %x to double
+    %2 = call double @ceil(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @_copysign(double %x)
+define float @float_copysign(float %x) nounwind readnone {
+; WIN32: @float_copysign
+; WIN32-NOT: float @copysignf
+; WIN32-NOT: float @_copysignf
+; WIN32: double @_copysign
+    %1 = fpext float %x to double
+    %2 = call double @_copysign(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @cos(double %x)
+define float @float_cos(float %x) nounwind readnone {
+; WIN32: @float_cos
+; WIN32-NOT: float @cosf
+; WIN32: double @cos
+    %1 = fpext float %x to double
+    %2 = call double @cos(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @cosh(double %x)
+define float @float_cosh(float %x) nounwind readnone {
+; WIN32: @float_cosh
+; WIN32-NOT: float @coshf
+; WIN32: double @cosh
+    %1 = fpext float %x to double
+    %2 = call double @cosh(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @exp(double %x, double %y)
+define float @float_exp(float %x, float %y) nounwind readnone {
+; WIN32: @float_exp
+; WIN32-NOT: float @expf
+; WIN32: double @exp
+    %1 = fpext float %x to double
+    %2 = fpext float %y to double
+    %3 = call double @exp(double %1, double %2)
+    %4 = fptrunc double %3 to float
+    ret float %4
+}
+
+declare double @fabs(double %x, double %y)
+define float @float_fabs(float %x, float %y) nounwind readnone {
+; WIN32: @float_fabs
+; WIN32-NOT: float @fabsf
+; WIN32: double @fabs
+; WIN64: @float_fabs
+; WIN64-NOT: float @fabsf
+; WIN64: double @fabs
+    %1 = fpext float %x to double
+    %2 = fpext float %y to double
+    %3 = call double @fabs(double %1, double %2)
+    %4 = fptrunc double %3 to float
+    ret float %4
+}
+
+declare double @floor(double %x)
+define float @float_floor(float %x) nounwind readnone {
+; WIN32: @float_floor
+; WIN32-NOT: float @floorf
+; WIN32: double @floor
+; WIN64: @float_floor
+; WIN64: float @floorf
+; WIN64-NOT: double @floor
+; MINGW32: @float_floor
+; MINGW32: float @floorf
+; MINGW32-NOT: double @floor
+; MINGW64: @float_floor
+; MINGW64: float @floorf
+; MINGW64-NOT: double @floor
+    %1 = fpext float %x to double
+    %2 = call double @floor(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @fmod(double %x, double %y)
+define float @float_fmod(float %x, float %y) nounwind readnone {
+; WIN32: @float_fmod
+; WIN32-NOT: float @fmodf
+; WIN32: double @fmod
+    %1 = fpext float %x to double
+    %2 = fpext float %y to double
+    %3 = call double @fmod(double %1, double %2)
+    %4 = fptrunc double %3 to float
+    ret float %4
+}
+
+declare double @log(double %x)
+define float @float_log(float %x) nounwind readnone {
+; WIN32: @float_log
+; WIN32-NOT: float @logf
+; WIN32: double @log
+    %1 = fpext float %x to double
+    %2 = call double @log(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @pow(double %x, double %y)
+define float @float_pow(float %x, float %y) nounwind readnone {
+; WIN32: @float_pow
+; WIN32-NOT: float @powf
+; WIN32: double @pow
+    %1 = fpext float %x to double
+    %2 = fpext float %y to double
+    %3 = call double @pow(double %1, double %2)
+    %4 = fptrunc double %3 to float
+    ret float %4
+}
+
+declare double @sin(double %x)
+define float @float_sin(float %x) nounwind readnone {
+; WIN32: @float_sin
+; WIN32-NOT: float @sinf
+; WIN32: double @sin
+    %1 = fpext float %x to double
+    %2 = call double @sin(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @sinh(double %x)
+define float @float_sinh(float %x) nounwind readnone {
+; WIN32: @float_sinh
+; WIN32-NOT: float @sinhf
+; WIN32: double @sinh
+    %1 = fpext float %x to double
+    %2 = call double @sinh(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @sqrt(double %x)
+define float @float_sqrt(float %x) nounwind readnone {
+; WIN32: @float_sqrt
+; WIN32-NOT: float @sqrtf
+; WIN32: double @sqrt
+; WIN64: @float_sqrt
+; WIN64: float @sqrtf
+; WIN64-NOT: double @sqrt
+; MINGW32: @float_sqrt
+; MINGW32: float @sqrtf
+; MINGW32-NOT: double @sqrt
+; MINGW64: @float_sqrt
+; MINGW64: float @sqrtf
+; MINGW64-NOT: double @sqrt
+    %1 = fpext float %x to double
+    %2 = call double @sqrt(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @tan(double %x)
+define float @float_tan(float %x) nounwind readnone {
+; WIN32: @float_tan
+; WIN32-NOT: float @tanf
+; WIN32: double @tan
+    %1 = fpext float %x to double
+    %2 = call double @tan(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+declare double @tanh(double %x)
+define float @float_tanh(float %x) nounwind readnone {
+; WIN32: @float_tanh
+; WIN32-NOT: float @tanhf
+; WIN32: double @tanh
+    %1 = fpext float %x to double
+    %2 = call double @tanh(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
+; win32 does not have round; mingw32 does
+declare double @round(double %x)
+define float @float_round(float %x) nounwind readnone {
+; WIN32: @float_round
+; WIN32-NOT: float @roundf
+; WIN32: double @round
+; WIN64: @float_round
+; WIN64-NOT: float @roundf
+; WIN64: double @round
+; MINGW32: @float_round
+; MINGW32: float @roundf
+; MINGW32-NOT: double @round
+; MINGW64: @float_round
+; MINGW64: float @roundf
+; MINGW64-NOT: double @round
+    %1 = fpext float %x to double
+    %2 = call double @round(double %1)
+    %3 = fptrunc double %2 to float
+    ret float %3
+}
+
diff --git a/test/Verifier/fpaccuracy.ll b/test/Verifier/fpaccuracy.ll
deleted file mode 100644
index 2fefde09f18..00000000000
--- a/test/Verifier/fpaccuracy.ll
+++ /dev/null
@@ -1,31 +0,0 @@
-; RUN: not llvm-as < %s |& FileCheck %s
-
-define void @foo(i32 %i, float %f, <2 x float> %g) {
-  %s = add i32 %i, %i, !fpaccuracy !0
-; CHECK: fpaccuracy requires a floating point result!
-  %t = fadd float %f, %f, !fpaccuracy !1
-; CHECK: fpaccuracy takes one operand!
-  %u = fadd float %f, %f, !fpaccuracy !2
-; CHECK: fpaccuracy takes one operand!
-  %v = fadd float %f, %f, !fpaccuracy !3
-; CHECK: fpaccuracy ULPs not a floating point number!
-  %w = fadd float %f, %f, !fpaccuracy !0
-; Above line is correct.
-  %w2 = fadd <2 x float> %g, %g, !fpaccuracy !0
-; Above line is correct.
-  %x = fadd float %f, %f, !fpaccuracy !4
-; CHECK: fpaccuracy ULPs is negative!
-  %y = fadd float %f, %f, !fpaccuracy !5
-; CHECK: fpaccuracy ULPs is negative!
-  %z = fadd float %f, %f, !fpaccuracy !6
-; CHECK: fpaccuracy ULPs not a normal number!
-  ret void
-}
-
-!0 = metadata !{ float 1.0 }
-!1 = metadata !{ }
-!2 = metadata !{ float 1.0, float 1.0 }
-!3 = metadata !{ i32 1 }
-!4 = metadata !{ float -1.0 }
-!5 = metadata !{ float -0.0 }
-!6 = metadata !{ float 0x7FFFFFFF00000000 }
diff --git a/test/Verifier/fpmath.ll b/test/Verifier/fpmath.ll
new file mode 100644
index 00000000000..b764a63f0a4
--- /dev/null
+++ b/test/Verifier/fpmath.ll
@@ -0,0 +1,31 @@
+; RUN: not llvm-as < %s |& FileCheck %s
+
+define void @fpmath1(i32 %i, float %f, <2 x float> %g) {
+  %s = add i32 %i, %i, !fpmath !0
+; CHECK: fpmath requires a floating point result!
+  %t = fadd float %f, %f, !fpmath !1
+; CHECK: fpmath takes one operand!
+  %u = fadd float %f, %f, !fpmath !2
+; CHECK: fpmath takes one operand!
+  %v = fadd float %f, %f, !fpmath !3
+; CHECK: invalid fpmath accuracy!
+  %w = fadd float %f, %f, !fpmath !0
+; Above line is correct.
+  %w2 = fadd <2 x float> %g, %g, !fpmath !0
+; Above line is correct.
+  %x = fadd float %f, %f, !fpmath !4
+; CHECK: fpmath accuracy not a positive number!
+  %y = fadd float %f, %f, !fpmath !5
+; CHECK: fpmath accuracy not a positive number!
+  %z = fadd float %f, %f, !fpmath !6
+; CHECK: fpmath accuracy not a positive number!
+  ret void
+}
+
+!0 = metadata !{ float 1.0 }
+!1 = metadata !{ }
+!2 = metadata !{ float 1.0, float 1.0 }
+!3 = metadata !{ i32 1 }
+!4 = metadata !{ float -1.0 }
+!5 = metadata !{ float 0.0 }
+!6 = metadata !{ float 0x7FFFFFFF00000000 }
diff --git a/test/lit.cfg b/test/lit.cfg
index c58935956a4..d74bc7bcb58 100644
--- a/test/lit.cfg
+++ b/test/lit.cfg
@@ -236,8 +236,13 @@ if loadable_module:
 
 # llc knows whether he is compiled with -DNDEBUG.
 import subprocess
-llc_cmd = subprocess.Popen([os.path.join(llvm_tools_dir, 'llc'), '-version'],
+try:
+    llc_cmd = subprocess.Popen([os.path.join(llvm_tools_dir, 'llc'), '-version'],
                            stdout = subprocess.PIPE)
+except OSError, why:
+    print "Could not find llc in " + llvm_tools_dir
+    exit(42)
+
 if re.search(r'with assertions', llc_cmd.stdout.read()):
     config.available_features.add('asserts')
 llc_cmd.wait()
diff --git a/tools/llc/llc.cpp b/tools/llc/llc.cpp
index 9e30ac198b2..ceff8a6c8cc 100644
--- a/tools/llc/llc.cpp
+++ b/tools/llc/llc.cpp
@@ -209,29 +209,6 @@ DontPlaceZerosInBSS("nozero-initialized-in-bss",
   cl::desc("Don't place zero-initialized symbols into bss section"),
   cl::init(false));
 
-static cl::opt
-EnableJITExceptionHandling("jit-enable-eh",
-  cl::desc("Emit exception handling information"),
-  cl::init(false));
-
-// In debug builds, make this default to true.
-#ifdef NDEBUG
-#define EMIT_DEBUG false
-#else
-#define EMIT_DEBUG true
-#endif
-static cl::opt
-EmitJitDebugInfo("jit-emit-debug",
-  cl::desc("Emit debug information to debugger"),
-  cl::init(EMIT_DEBUG));
-#undef EMIT_DEBUG
-
-static cl::opt
-EmitJitDebugInfoToDisk("jit-emit-debug-to-disk",
-  cl::Hidden,
-  cl::desc("Emit debug info objfiles to disk"),
-  cl::init(false));
-
 static cl::opt
 EnableGuaranteedTailCallOpt("tailcallopt",
   cl::desc("Turn fastcc calls into tail calls by (potentially) changing ABI."),
@@ -463,9 +440,6 @@ int main(int argc, char **argv) {
   if (FloatABIForCalls != FloatABI::Default)
     Options.FloatABIType = FloatABIForCalls;
   Options.NoZerosInBSS = DontPlaceZerosInBSS;
-  Options.JITExceptionHandling = EnableJITExceptionHandling;
-  Options.JITEmitDebugInfo = EmitJitDebugInfo;
-  Options.JITEmitDebugInfoToDisk = EmitJitDebugInfoToDisk;
   Options.GuaranteedTailCallOpt = EnableGuaranteedTailCallOpt;
   Options.DisableTailCalls = DisableTailCalls;
   Options.StackAlignmentOverride = OverrideStackAlignment;
diff --git a/tools/lli/lli.cpp b/tools/lli/lli.cpp
index efcc1f5870c..2e2bf7d66ec 100644
--- a/tools/lli/lli.cpp
+++ b/tools/lli/lli.cpp
@@ -141,6 +141,28 @@ namespace {
                                 "Large code model"),
                      clEnumValEnd));
 
+  cl::opt
+  EnableJITExceptionHandling("jit-enable-eh",
+    cl::desc("Emit exception handling information"),
+    cl::init(false));
+
+  cl::opt
+// In debug builds, make this default to true.
+#ifdef NDEBUG
+#define EMIT_DEBUG false
+#else
+#define EMIT_DEBUG true
+#endif
+  EmitJitDebugInfo("jit-emit-debug",
+    cl::desc("Emit debug information to debugger"),
+    cl::init(EMIT_DEBUG));
+#undef EMIT_DEBUG
+
+  static cl::opt
+  EmitJitDebugInfoToDisk("jit-emit-debug-to-disk",
+    cl::Hidden,
+    cl::desc("Emit debug info objfiles to disk"),
+    cl::init(false));
 }
 
 static ExecutionEngine *EE = 0;
@@ -229,6 +251,12 @@ int main(int argc, char **argv, char * const *envp) {
   }
   builder.setOptLevel(OLvl);
 
+  TargetOptions Options;
+  Options.JITExceptionHandling = EnableJITExceptionHandling;
+  Options.JITEmitDebugInfo = EmitJitDebugInfo;
+  Options.JITEmitDebugInfoToDisk = EmitJitDebugInfoToDisk;
+  builder.setTargetOptions(Options);
+
   EE = builder.create();
   if (!EE) {
     if (!ErrorMsg.empty())
diff --git a/tools/llvm-mc/Disassembler.cpp b/tools/llvm-mc/Disassembler.cpp
index a8cd7c1c897..5f2fdb80714 100644
--- a/tools/llvm-mc/Disassembler.cpp
+++ b/tools/llvm-mc/Disassembler.cpp
@@ -17,21 +17,18 @@
 #include "../../lib/MC/MCDisassembler/EDInst.h"
 #include "../../lib/MC/MCDisassembler/EDOperand.h"
 #include "../../lib/MC/MCDisassembler/EDToken.h"
-#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCDisassembler.h"
 #include "llvm/MC/MCInst.h"
-#include "llvm/MC/MCInstPrinter.h"
-#include "llvm/MC/MCInstrInfo.h"
-#include "llvm/MC/MCRegisterInfo.h"
+#include "llvm/MC/MCStreamer.h"
 #include "llvm/MC/MCSubtargetInfo.h"
 #include "llvm/ADT/OwningPtr.h"
 #include "llvm/ADT/Triple.h"
-#include "llvm/ADT/Twine.h"
 #include "llvm/Support/MemoryBuffer.h"
 #include "llvm/Support/MemoryObject.h"
 #include "llvm/Support/SourceMgr.h"
 #include "llvm/Support/TargetRegistry.h"
 #include "llvm/Support/raw_ostream.h"
+
 using namespace llvm;
 
 typedef std::vector > ByteArrayTy;
@@ -56,8 +53,9 @@ public:
 }
 
 static bool PrintInsts(const MCDisassembler &DisAsm,
-                       MCInstPrinter &Printer, const ByteArrayTy &Bytes,
-                       SourceMgr &SM, raw_ostream &Out) {
+                       const ByteArrayTy &Bytes,
+                       SourceMgr &SM, raw_ostream &Out,
+                       MCStreamer &Streamer) {
   // Wrap the vector in a MemoryObject.
   VectorMemoryObject memoryObject(Bytes);
 
@@ -87,8 +85,7 @@ static bool PrintInsts(const MCDisassembler &DisAsm,
       // Fall through
 
     case MCDisassembler::Success:
-      Printer.printInst(&Inst, Out, "");
-      Out << "\n";
+      Streamer.EmitInstruction(Inst);
       break;
     }
   }
@@ -145,56 +142,22 @@ static bool ByteArrayFromString(ByteArrayTy &ByteArray,
 
 int Disassembler::disassemble(const Target &T,
                               const std::string &Triple,
-                              const std::string &Cpu,
-                              const std::string &FeaturesStr,
+                              MCSubtargetInfo &STI,
+                              MCStreamer &Streamer,
                               MemoryBuffer &Buffer,
+                              SourceMgr &SM,
                               raw_ostream &Out) {
-  // Set up disassembler.
-  OwningPtr AsmInfo(T.createMCAsmInfo(Triple));
-
-  if (!AsmInfo) {
-    errs() << "error: no assembly info for target " << Triple << "\n";
-    return -1;
-  }
-
-  OwningPtr STI(T.createMCSubtargetInfo(Triple, Cpu,
-                                                               FeaturesStr));
-  if (!STI) {
-    errs() << "error: no subtarget info for target " << Triple << "\n";
-    return -1;
-  }
-
-  OwningPtr DisAsm(T.createMCDisassembler(*STI));
+  OwningPtr DisAsm(T.createMCDisassembler(STI));
   if (!DisAsm) {
     errs() << "error: no disassembler for target " << Triple << "\n";
     return -1;
   }
 
-  OwningPtr MRI(T.createMCRegInfo(Triple));
-  if (!MRI) {
-    errs() << "error: no register info for target " << Triple << "\n";
-    return -1;
-  }
-
-  OwningPtr MII(T.createMCInstrInfo());
-  if (!MII) {
-    errs() << "error: no instruction info for target " << Triple << "\n";
-    return -1;
-  }
-
-  int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
-  OwningPtr IP(T.createMCInstPrinter(AsmPrinterVariant, *AsmInfo,
-                                                    *MII, *MRI, *STI));
-  if (!IP) {
-    errs() << "error: no instruction printer for target " << Triple << '\n';
-    return -1;
-  }
+  // Set up initial section manually here
+  Streamer.InitSections();
 
   bool ErrorOccurred = false;
 
-  SourceMgr SM;
-  SM.AddNewSourceBuffer(&Buffer, SMLoc());
-
   // Convert the input to a vector for disassembly.
   ByteArrayTy ByteArray;
   StringRef Str = Buffer.getBuffer();
@@ -202,7 +165,7 @@ int Disassembler::disassemble(const Target &T,
   ErrorOccurred |= ByteArrayFromString(ByteArray, Str, SM);
 
   if (!ByteArray.empty())
-    ErrorOccurred |= PrintInsts(*DisAsm, *IP, ByteArray, SM, Out);
+    ErrorOccurred |= PrintInsts(*DisAsm, ByteArray, SM, Out, Streamer);
 
   return ErrorOccurred;
 }
@@ -236,12 +199,10 @@ static int verboseEvaluator(uint64_t *V, unsigned R, void *Arg) {
 
 int Disassembler::disassembleEnhanced(const std::string &TS,
                                       MemoryBuffer &Buffer,
+                                      SourceMgr &SM,
                                       raw_ostream &Out) {
   ByteArrayTy ByteArray;
   StringRef Str = Buffer.getBuffer();
-  SourceMgr SM;
-
-  SM.AddNewSourceBuffer(&Buffer, SMLoc());
 
   if (ByteArrayFromString(ByteArray, Str, SM)) {
     return -1;
diff --git a/tools/llvm-mc/Disassembler.h b/tools/llvm-mc/Disassembler.h
index e8cd92db030..17d622f1d9d 100644
--- a/tools/llvm-mc/Disassembler.h
+++ b/tools/llvm-mc/Disassembler.h
@@ -22,18 +22,23 @@ namespace llvm {
 class MemoryBuffer;
 class Target;
 class raw_ostream;
+class SourceMgr;
+class MCSubtargetInfo;
+class MCStreamer;
 
 class Disassembler {
 public:
-  static int disassemble(const Target &target,
-                         const std::string &tripleString,
-                         const std::string &Cpu,
-                         const std::string &FeaturesStr,
-                         MemoryBuffer &buffer,
+  static int disassemble(const Target &T,
+                         const std::string &Triple,
+                         MCSubtargetInfo &STI,
+                         MCStreamer &Streamer,
+                         MemoryBuffer &Buffer,
+                         SourceMgr &SM,
                          raw_ostream &Out);
 
   static int disassembleEnhanced(const std::string &tripleString,
                                  MemoryBuffer &buffer,
+                                 SourceMgr &SM,
                                  raw_ostream &Out);
 };
 
diff --git a/tools/llvm-mc/llvm-mc.cpp b/tools/llvm-mc/llvm-mc.cpp
index d882e01529d..36a482e908c 100644
--- a/tools/llvm-mc/llvm-mc.cpp
+++ b/tools/llvm-mc/llvm-mc.cpp
@@ -15,6 +15,7 @@
 #include "llvm/MC/MCParser/AsmLexer.h"
 #include "llvm/MC/MCParser/MCAsmLexer.h"
 #include "llvm/MC/MCAsmBackend.h"
+#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCCodeEmitter.h"
 #include "llvm/MC/MCInstPrinter.h"
@@ -242,37 +243,11 @@ static void setDwarfDebugFlags(int argc, char **argv) {
   }
 }
 
-static int AsLexInput(const char *ProgName) {
-  OwningPtr BufferPtr;
-  if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, BufferPtr)) {
-    errs() << ProgName << ": " << ec.message() << '\n';
-    return 1;
-  }
-  MemoryBuffer *Buffer = BufferPtr.take();
+static int AsLexInput(SourceMgr &SrcMgr, MCAsmInfo &MAI, tool_output_file *Out) {
 
-  SourceMgr SrcMgr;
-
-  // Tell SrcMgr about this buffer, which is what TGParser will pick up.
-  SrcMgr.AddNewSourceBuffer(Buffer, SMLoc());
-
-  // Record the location of the include directories so that the lexer can find
-  // it later.
-  SrcMgr.setIncludeDirs(IncludeDirs);
-
-  const Target *TheTarget = GetTarget(ProgName);
-  if (!TheTarget)
-    return 1;
-
-  llvm::OwningPtr MAI(TheTarget->createMCAsmInfo(TripleName));
-  assert(MAI && "Unable to create target asm info!");
-
-  AsmLexer Lexer(*MAI);
+  AsmLexer Lexer(MAI);
   Lexer.setBuffer(SrcMgr.getMemoryBuffer(0));
 
-  OwningPtr Out(GetOutputStream());
-  if (!Out)
-    return 1;
-
   bool Error = false;
   while (Lexer.Lex().isNot(AsmToken::Eof)) {
     AsmToken Tok = Lexer.getTok();
@@ -346,13 +321,49 @@ static int AsLexInput(const char *ProgName) {
     Out->os() << "\")\n";
   }
 
-  // Keep output if no errors.
-  if (Error == 0) Out->keep();
-
   return Error;
 }
 
-static int AssembleInput(const char *ProgName) {
+static int AssembleInput(const char *ProgName, const Target *TheTarget, 
+                         SourceMgr &SrcMgr, MCContext &Ctx, MCStreamer &Str,
+                         MCAsmInfo &MAI, MCSubtargetInfo &STI) {
+  OwningPtr Parser(createMCAsmParser(SrcMgr, Ctx,
+                                                  Str, MAI));
+  OwningPtr TAP(TheTarget->createMCAsmParser(STI, *Parser));
+  if (!TAP) {
+    errs() << ProgName
+           << ": error: this target does not support assembly parsing.\n";
+    return 1;
+  }
+
+  Parser->setShowParsedOperands(ShowInstOperands);
+  Parser->setTargetParser(*TAP.get());
+
+  int Res = Parser->Run(NoInitialTextSection);
+
+  return Res;
+}
+
+int main(int argc, char **argv) {
+  // Print a stack trace if we signal out.
+  sys::PrintStackTraceOnErrorSignal();
+  PrettyStackTraceProgram X(argc, argv);
+  llvm_shutdown_obj Y;  // Call llvm_shutdown() on exit.
+
+  // Initialize targets and assembly printers/parsers.
+  llvm::InitializeAllTargetInfos();
+  llvm::InitializeAllTargetMCs();
+  llvm::InitializeAllAsmParsers();
+  llvm::InitializeAllDisassemblers();
+
+  // Register the target printer for --version.
+  cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion);
+
+  cl::ParseCommandLineOptions(argc, argv, "llvm machine code playground\n");
+  TripleName = Triple::normalize(TripleName);
+  setDwarfDebugFlags(argc, argv);
+
+  const char *ProgName = argv[0];
   const Target *TheTarget = GetTarget(ProgName);
   if (!TheTarget)
     return 1;
@@ -413,7 +424,6 @@ static int AssembleInput(const char *ProgName) {
   OwningPtr
     STI(TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr));
 
-  // FIXME: There is a bit of code duplication with addPassesToEmitFile.
   if (FileType == OFT_AssemblyFile) {
     MCInstPrinter *IP =
       TheTarget->createMCInstPrinter(OutputAsmVariant, *MAI, *MCII, *MRI, *STI);
@@ -440,93 +450,24 @@ static int AssembleInput(const char *ProgName) {
                                                 NoExecStack));
   }
 
-  OwningPtr Parser(createMCAsmParser(SrcMgr, Ctx,
-                                                  *Str.get(), *MAI));
-  OwningPtr TAP(TheTarget->createMCAsmParser(*STI, *Parser));
-  if (!TAP) {
-    errs() << ProgName
-           << ": error: this target does not support assembly parsing.\n";
-    return 1;
-  }
-
-  Parser->setShowParsedOperands(ShowInstOperands);
-  Parser->setTargetParser(*TAP.get());
-
-  int Res = Parser->Run(NoInitialTextSection);
-
-  // Keep output if no errors.
-  if (Res == 0) Out->keep();
-
-  return Res;
-}
-
-static int DisassembleInput(const char *ProgName, bool Enhanced) {
-  const Target *TheTarget = GetTarget(ProgName);
-  if (!TheTarget)
-    return 0;
-
-  OwningPtr Buffer;
-  if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, Buffer)) {
-    errs() << ProgName << ": " << ec.message() << '\n';
-    return 1;
-  }
-
-  OwningPtr Out(GetOutputStream());
-  if (!Out)
-    return 1;
-
-  int Res;
-  if (Enhanced) {
-    Res =
-      Disassembler::disassembleEnhanced(TripleName, *Buffer.take(), Out->os());
-  } else {
-    // Package up features to be passed to target/subtarget
-    std::string FeaturesStr;
-    if (MAttrs.size()) {
-      SubtargetFeatures Features;
-      for (unsigned i = 0; i != MAttrs.size(); ++i)
-        Features.AddFeature(MAttrs[i]);
-      FeaturesStr = Features.getString();
-    }
-
-    Res = Disassembler::disassemble(*TheTarget, TripleName, MCPU, FeaturesStr,
-                                    *Buffer.take(), Out->os());
-  }
-
-  // Keep output if no errors.
-  if (Res == 0) Out->keep();
-
-  return Res;
-}
-
-
-int main(int argc, char **argv) {
-  // Print a stack trace if we signal out.
-  sys::PrintStackTraceOnErrorSignal();
-  PrettyStackTraceProgram X(argc, argv);
-  llvm_shutdown_obj Y;  // Call llvm_shutdown() on exit.
-
-  // Initialize targets and assembly printers/parsers.
-  llvm::InitializeAllTargetInfos();
-  llvm::InitializeAllTargetMCs();
-  llvm::InitializeAllAsmParsers();
-  llvm::InitializeAllDisassemblers();
-
-  // Register the target printer for --version.
-  cl::AddExtraVersionPrinter(TargetRegistry::printRegisteredTargetsForVersion);
-
-  cl::ParseCommandLineOptions(argc, argv, "llvm machine code playground\n");
-  TripleName = Triple::normalize(TripleName);
-  setDwarfDebugFlags(argc, argv);
-
+  int Res = 1;
   switch (Action) {
   case AC_AsLex:
-    return AsLexInput(argv[0]);
+    Res = AsLexInput(SrcMgr, *MAI, Out.get());
+    break;
   case AC_Assemble:
-    return AssembleInput(argv[0]);
+    Res = AssembleInput(ProgName, TheTarget, SrcMgr, Ctx, *Str, *MAI, *STI);
+    break;
   case AC_Disassemble:
-    return DisassembleInput(argv[0], false);
+    Res = Disassembler::disassemble(*TheTarget, TripleName, *STI, *Str,
+                                    *Buffer, SrcMgr, Out->os());
+    break;
   case AC_EDisassemble:
-    return DisassembleInput(argv[0], true);
+    Res =  Disassembler::disassembleEnhanced(TripleName, *Buffer, SrcMgr, Out->os());
+    break;
   }
+
+  // Keep output if no errors.
+  if (Res == 0) Out->keep();
+  return Res;
 }
diff --git a/tools/llvm-shlib/Makefile b/tools/llvm-shlib/Makefile
index 2d2e2c55b80..75bee07d171 100644
--- a/tools/llvm-shlib/Makefile
+++ b/tools/llvm-shlib/Makefile
@@ -67,6 +67,9 @@ ifeq ($(HOST_OS), $(filter $(HOST_OS), Linux FreeBSD OpenBSD GNU))
     # Include everything from the .a's into the shared library.
     LLVMLibsOptions := -Wl,--whole-archive $(LLVMLibsOptions) \
                        -Wl,--no-whole-archive
+endif
+
+ifeq ($(HOST_OS), $(filter $(HOST_OS), Linux FreeBSD GNU))
     # Add soname to the library.
     LLVMLibsOptions += -Wl,--soname,lib$(LIBRARYNAME)$(SHLIBEXT)
 endif
diff --git a/tools/llvm-stress/llvm-stress.cpp b/tools/llvm-stress/llvm-stress.cpp
index d284ea5e42c..fb05a589e8b 100644
--- a/tools/llvm-stress/llvm-stress.cpp
+++ b/tools/llvm-stress/llvm-stress.cpp
@@ -412,7 +412,7 @@ struct ExtractElementModifier: public Modifier {
     Value *Val0 = getRandomVectorValue();
     Value *V = ExtractElementInst::Create(Val0,
              ConstantInt::get(Type::getInt32Ty(BB->getContext()),
-             Ran->Rand() % cast(Val0->getType())->getNumElements()), 
+             Ran->Rand() % cast(Val0->getType())->getNumElements()),
              "E", BB->getTerminator());
     return PT->push_back(V);
   }
@@ -476,7 +476,7 @@ struct CastModifier: public Modifier {
       DestTy = pickVectorType(VecTy->getNumElements());
     }
 
-    // no need to casr.
+    // no need to cast.
     if (VTy == DestTy) return;
 
     // Pointers:
@@ -487,9 +487,11 @@ struct CastModifier: public Modifier {
         new BitCastInst(V, DestTy, "PC", BB->getTerminator()));
     }
 
+    unsigned VSize = VTy->getScalarType()->getPrimitiveSizeInBits();
+    unsigned DestSize = DestTy->getScalarType()->getPrimitiveSizeInBits();
+
     // Generate lots of bitcasts.
-    if ((Ran->Rand() & 1) &&
-        VTy->getPrimitiveSizeInBits() == DestTy->getPrimitiveSizeInBits()) {
+    if ((Ran->Rand() & 1) && VSize == DestSize) {
       return PT->push_back(
         new BitCastInst(V, DestTy, "BC", BB->getTerminator()));
     }
@@ -497,11 +499,11 @@ struct CastModifier: public Modifier {
     // Both types are integers:
     if (VTy->getScalarType()->isIntegerTy() &&
         DestTy->getScalarType()->isIntegerTy()) {
-      if (VTy->getScalarType()->getPrimitiveSizeInBits() >
-          DestTy->getScalarType()->getPrimitiveSizeInBits()) {
+      if (VSize > DestSize) {
         return PT->push_back(
           new TruncInst(V, DestTy, "Tr", BB->getTerminator()));
       } else {
+        assert(VSize < DestSize && "Different int types with the same size?");
         if (Ran->Rand() & 1)
           return PT->push_back(
             new ZExtInst(V, DestTy, "ZE", BB->getTerminator()));
@@ -531,14 +533,15 @@ struct CastModifier: public Modifier {
     // Both floats.
     if (VTy->getScalarType()->isFloatingPointTy() &&
         DestTy->getScalarType()->isFloatingPointTy()) {
-      if (VTy->getScalarType()->getPrimitiveSizeInBits() >
-          DestTy->getScalarType()->getPrimitiveSizeInBits()) {
+      if (VSize > DestSize) {
         return PT->push_back(
           new FPTruncInst(V, DestTy, "Tr", BB->getTerminator()));
-      } else {
+      } else if (VSize < DestSize) {
         return PT->push_back(
           new FPExtInst(V, DestTy, "ZE", BB->getTerminator()));
       }
+      // If VSize == DestSize, then the two types must be fp128 and ppc_fp128,
+      // for which there is no defined conversion. So do nothing.
     }
   }
 
diff --git a/tools/lto/LTOCodeGenerator.cpp b/tools/lto/LTOCodeGenerator.cpp
index 77c06a655b1..6382a3f219e 100644
--- a/tools/lto/LTOCodeGenerator.cpp
+++ b/tools/lto/LTOCodeGenerator.cpp
@@ -64,7 +64,7 @@ LTOCodeGenerator::LTOCodeGenerator()
   : _context(getGlobalContext()),
     _linker("LinkTimeOptimizer", "ld-temp.o", _context), _target(NULL),
     _emitDwarfDebugInfo(false), _scopeRestrictionsDone(false),
-    _runInternalizePass(false), _codeModel(LTO_CODEGEN_PIC_MODEL_DYNAMIC),
+    _codeModel(LTO_CODEGEN_PIC_MODEL_DYNAMIC),
     _nativeObjectFile(NULL) {
   InitializeAllTargets();
   InitializeAllTargetMCs();
@@ -355,8 +355,10 @@ bool LTOCodeGenerator::generateObjectFile(raw_ostream &out,
   // Add an appropriate TargetData instance for this module...
   passes.add(new TargetData(*_target->getTargetData()));
 
-  PassManagerBuilder().populateLTOPassManager(passes,
-                                              _runInternalizePass,
+  // Enabling internalize here would use its AllButMain variant. It
+  // keeps only main if it exists and does nothing for libraries. Instead
+  // we create the pass ourselves with the symbol list provided by the linker.
+  PassManagerBuilder().populateLTOPassManager(passes, /*Internalize=*/false,
                                               !DisableInline,
                                               DisableGVNLoadPRE);
 
diff --git a/tools/lto/LTOCodeGenerator.h b/tools/lto/LTOCodeGenerator.h
index bac3e6efe90..032dc379172 100644
--- a/tools/lto/LTOCodeGenerator.h
+++ b/tools/lto/LTOCodeGenerator.h
@@ -54,8 +54,6 @@ struct LTOCodeGenerator {
   const void *compile(size_t *length, std::string &errMsg);
   void setCodeGenDebugOptions(const char *opts);
 
-  void enableInternalizePass() { _runInternalizePass = true; }
-
 private:
   bool generateObjectFile(llvm::raw_ostream &out, std::string &errMsg);
   void applyScopeRestrictions();
diff --git a/tools/lto/lto.cpp b/tools/lto/lto.cpp
index addf7877c96..a7e633d14b9 100644
--- a/tools/lto/lto.cpp
+++ b/tools/lto/lto.cpp
@@ -183,12 +183,6 @@ void lto_codegen_add_must_preserve_symbol(lto_code_gen_t cg,
   cg->addMustPreserveSymbol(symbol);
 }
 
-/// lto_codegen_set_whole_program_optimization - Enable the internalize pass
-/// during LTO optimizations.
-void lto_codegen_set_whole_program_optimization(lto_code_gen_t cg) {
-  cg->enableInternalizePass();
-}
-
 /// lto_codegen_write_merged_modules - Writes a new file at the specified path
 /// that contains the merged contents of all modules added so far. Returns true
 /// on error (check lto_get_error_message() for details).
diff --git a/tools/lto/lto.exports b/tools/lto/lto.exports
index f471f1ad6a2..b900bfb594b 100644
--- a/tools/lto/lto.exports
+++ b/tools/lto/lto.exports
@@ -27,7 +27,6 @@ lto_codegen_set_assembler_args
 lto_codegen_set_assembler_path
 lto_codegen_set_cpu
 lto_codegen_compile_to_file
-lto_codegen_set_whole_program_optimization
 LLVMCreateDisasm
 LLVMDisasmDispose
 LLVMDisasmInstruction
diff --git a/tools/opt/opt.cpp b/tools/opt/opt.cpp
index 30da863b411..a5b0511fd98 100644
--- a/tools/opt/opt.cpp
+++ b/tools/opt/opt.cpp
@@ -114,6 +114,9 @@ static cl::opt
 OptLevelO3("O3",
            cl::desc("Optimization level 3. Similar to llvm-gcc -O3"));
 
+static cl::opt
+TargetTriple("mtriple", cl::desc("Override target triple for module"));
+
 static cl::opt
 UnitAtATime("funit-at-a-time",
             cl::desc("Enable IPO. This is same as llvm-gcc's -funit-at-a-time"),
@@ -512,6 +515,10 @@ int main(int argc, char **argv) {
     return 1;
   }
 
+  // If we are supposed to override the target triple, do so now.
+  if (!TargetTriple.empty())
+    M->setTargetTriple(Triple::normalize(TargetTriple));
+
   // Figure out what stream we are supposed to write to...
   OwningPtr Out;
   if (NoOutput) {
diff --git a/unittests/CMakeLists.txt b/unittests/CMakeLists.txt
index 5d691728d80..78009a86df5 100644
--- a/unittests/CMakeLists.txt
+++ b/unittests/CMakeLists.txt
@@ -165,7 +165,6 @@ add_llvm_unittest(Support
   Support/CommandLineTest.cpp
   Support/ConstantRangeTest.cpp
   Support/EndianTest.cpp
-  Support/JSONParserTest.cpp
   Support/LeakDetectorTest.cpp
   Support/MathExtrasTest.cpp
   Support/Path.cpp
diff --git a/unittests/Support/JSONParserTest.cpp b/unittests/Support/JSONParserTest.cpp
deleted file mode 100644
index e9efb817c29..00000000000
--- a/unittests/Support/JSONParserTest.cpp
+++ /dev/null
@@ -1,191 +0,0 @@
-//===- unittest/Tooling/JSONParserTest ------------------------------------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/Support/Casting.h"
-#include "llvm/Support/JSONParser.h"
-#include "llvm/ADT/Twine.h"
-#include "gtest/gtest.h"
-
-namespace llvm {
-
-// Checks that the given input gives a parse error. Makes sure that an error
-// text is available and the parse fails.
-static void ExpectParseError(StringRef Message, StringRef Input) {
-  SourceMgr SM;
-  JSONParser Parser(Input, &SM);
-  EXPECT_FALSE(Parser.validate()) << Message << ": " << Input;
-  EXPECT_TRUE(Parser.failed()) << Message << ": " << Input;
-}
-
-// Checks that the given input can be parsed without error.
-static void ExpectParseSuccess(StringRef Message, StringRef Input) {
-  SourceMgr SM;
-  JSONParser Parser(Input, &SM);
-  EXPECT_TRUE(Parser.validate()) << Message << ": " << Input;
-}
-
-TEST(JSONParser, FailsOnEmptyString) {
-  ExpectParseError("Empty JSON text", "");
-}
- 
-TEST(JSONParser, FailsIfStartsWithString) {
-  ExpectParseError("Top-level string", "\"x\"");
-}
-
-TEST(JSONParser, ParsesEmptyArray) {
-  ExpectParseSuccess("Empty array", "[]");
-}
-
-TEST(JSONParser, FailsIfNotClosingArray) {
-  ExpectParseError("Not closing array", "[");
-  ExpectParseError("Not closing array", "  [  ");
-  ExpectParseError("Not closing array", "  [x");
-}
-
-TEST(JSONParser, ParsesEmptyArrayWithWhitespace) {
-  ExpectParseSuccess("Array with spaces", "  [  ]  ");
-  ExpectParseSuccess("All whitespaces", "\t\r\n[\t\n \t\r ]\t\r \n\n");
-}
-
-TEST(JSONParser, ParsesEmptyObject) {
-  ExpectParseSuccess("Empty object", "[{}]");
-}
-
-TEST(JSONParser, ParsesObject) {
-  ExpectParseSuccess("Object with an entry", "[{\"a\":\"/b\"}]");
-}
-
-TEST(JSONParser, ParsesMultipleKeyValuePairsInObject) {
-  ExpectParseSuccess("Multiple key, value pairs",
-                     "[{\"a\":\"/b\",\"c\":\"d\",\"e\":\"f\"}]");
-}
-
-TEST(JSONParser, FailsIfNotClosingObject) {
-  ExpectParseError("Missing close on empty", "[{]");
-  ExpectParseError("Missing close after pair", "[{\"a\":\"b\"]");
-}
-
-TEST(JSONParser, FailsIfMissingColon) {
-  ExpectParseError("Missing colon between key and value", "[{\"a\"\"/b\"}]");
-  ExpectParseError("Missing colon between key and value", "[{\"a\" \"b\"}]");
-}
-
-TEST(JSONParser, FailsOnMissingQuote) {
-  ExpectParseError("Missing open quote", "[{a\":\"b\"}]");
-  ExpectParseError("Missing closing quote", "[{\"a\":\"b}]");
-}
-
-TEST(JSONParser, ParsesEscapedQuotes) {
-  ExpectParseSuccess("Parses escaped string in key and value",
-                     "[{\"a\":\"\\\"b\\\"  \\\" \\\"\"}]");
-}
-
-TEST(JSONParser, ParsesEmptyString) {
-  ExpectParseSuccess("Parses empty string in value", "[{\"a\":\"\"}]");
-}
-
-TEST(JSONParser, FailsOnMissingString) {
-  ExpectParseError("Missing value", "[{\"a\":}]");
-  ExpectParseError("Missing key", "[{:\"b\"}]");
-}
-
-TEST(JSONParser, ParsesMultipleObjects) {
-  ExpectParseSuccess(
-      "Multiple objects in array",
-      "["
-      " { \"a\" : \"b\" },"
-      " { \"a\" : \"b\" },"
-      " { \"a\" : \"b\" }"
-      "]");
-}
-
-TEST(JSONParser, FailsOnMissingComma) {
-  ExpectParseError(
-      "Missing comma",
-      "["
-      " { \"a\" : \"b\" }"
-      " { \"a\" : \"b\" }"
-      "]");
-}
-
-TEST(JSONParser, FailsOnSuperfluousComma) {
-  ExpectParseError("Superfluous comma in array", "[ { \"a\" : \"b\" }, ]");
-  ExpectParseError("Superfluous comma in object", "{ \"a\" : \"b\", }");
-}
-
-TEST(JSONParser, ParsesSpacesInBetweenTokens) {
-  ExpectParseSuccess(
-      "Various whitespace between tokens",
-      " \t \n\n \r [ \t \n\n \r"
-      " \t \n\n \r { \t \n\n \r\"a\"\t \n\n \r :"
-      " \t \n\n \r \"b\"\t \n\n \r } \t \n\n \r,\t \n\n \r"
-      " \t \n\n \r { \t \n\n \r\"a\"\t \n\n \r :"
-      " \t \n\n \r \"b\"\t \n\n \r } \t \n\n \r]\t \n\n \r");
-}
-
-TEST(JSONParser, ParsesArrayOfArrays) {
-  ExpectParseSuccess("Array of arrays", "[[]]");
-}
-
-TEST(JSONParser, HandlesEndOfFileGracefully) {
-  ExpectParseError("In string starting with EOF", "[\"");
-  ExpectParseError("In string hitting EOF", "[\"   ");
-  ExpectParseError("In string escaping EOF", "[\"  \\");
-  ExpectParseError("In array starting with EOF", "[");
-  ExpectParseError("In array element starting with EOF", "[[], ");
-  ExpectParseError("In array hitting EOF", "[[] ");
-  ExpectParseError("In array hitting EOF", "[[]");
-  ExpectParseError("In object hitting EOF", "{\"\"");
-}
-
-// Checks that the given string can be parsed into an identical string inside
-// of an array.
-static void ExpectCanParseString(StringRef String) {
-  std::string StringInArray = (llvm::Twine("[\"") + String + "\"]").str();
-  SourceMgr SM;
-  JSONParser Parser(StringInArray, &SM);
-  const JSONArray *ParsedArray = dyn_cast(Parser.parseRoot());
-  StringRef ParsedString =
-      dyn_cast(*ParsedArray->begin())->getRawText();
-  EXPECT_EQ(String, ParsedString.str());
-}
-
-// Checks that parsing the given string inside an array fails.
-static void ExpectCannotParseString(StringRef String) {
-  std::string StringInArray = (llvm::Twine("[\"") + String + "\"]").str();
-  ExpectParseError((Twine("When parsing string \"") + String + "\"").str(),
-                   StringInArray);
-}
-
-TEST(JSONParser, ParsesStrings) {
-  ExpectCanParseString("");
-  ExpectCannotParseString("\\");
-  ExpectCannotParseString("\"");
-  ExpectCanParseString(" ");
-  ExpectCanParseString("\\ ");
-  ExpectCanParseString("\\\"");
-  ExpectCannotParseString("\"\\");
-  ExpectCannotParseString(" \\");
-  ExpectCanParseString("\\\\");
-  ExpectCannotParseString("\\\\\\");
-  ExpectCanParseString("\\\\\\\\");
-  ExpectCanParseString("\\\" ");
-  ExpectCannotParseString("\\\\\" ");
-  ExpectCanParseString("\\\\\\\" ");
-  ExpectCanParseString("    \\\\  \\\"  \\\\\\\"   ");
-}
-
-TEST(JSONParser, WorksWithIteratorAlgorithms) {
-  SourceMgr SM;
-  JSONParser Parser("[\"1\", \"2\", \"3\", \"4\", \"5\", \"6\"]", &SM);
-  const JSONArray *Array = dyn_cast(Parser.parseRoot());
-  EXPECT_EQ(6, std::distance(Array->begin(), Array->end()));
-}
-
-} // end namespace llvm
diff --git a/unittests/Support/MDBuilderTest.cpp b/unittests/Support/MDBuilderTest.cpp
new file mode 100644
index 00000000000..d54c7e8e8d7
--- /dev/null
+++ b/unittests/Support/MDBuilderTest.cpp
@@ -0,0 +1,105 @@
+//===- llvm/unittests/Support/MDBuilderTest.cpp - MDBuilder unit tests ----===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "gtest/gtest.h"
+#include "llvm/Operator.h"
+#include "llvm/Support/IRBuilder.h"
+#include "llvm/Support/MDBuilder.h"
+using namespace llvm;
+
+namespace {
+
+class MDBuilderTest : public testing::Test {
+protected:
+  LLVMContext Context;
+};
+
+TEST_F(MDBuilderTest, createString) {
+  MDBuilder MDHelper(Context);
+  MDString *Str0 = MDHelper.createString("");
+  MDString *Str1 = MDHelper.createString("string");
+  EXPECT_EQ(Str0->getString(), StringRef(""));
+  EXPECT_EQ(Str1->getString(), StringRef("string"));
+}
+TEST_F(MDBuilderTest, createFPMath) {
+  MDBuilder MDHelper(Context);
+  MDNode *MD0 = MDHelper.createFPMath(0.0);
+  MDNode *MD1 = MDHelper.createFPMath(1.0);
+  EXPECT_EQ(MD0, (MDNode *)0);
+  EXPECT_NE(MD1, (MDNode *)0);
+  EXPECT_EQ(MD1->getNumOperands(), 1U);
+  Value *Op = MD1->getOperand(0);
+  EXPECT_TRUE(isa(Op));
+  EXPECT_TRUE(Op->getType()->isFloatingPointTy());
+  ConstantFP *Val = cast(Op);
+  EXPECT_TRUE(Val->isExactlyValue(1.0));
+}
+TEST_F(MDBuilderTest, createRangeMetadata) {
+  MDBuilder MDHelper(Context);
+  APInt A(8, 1), B(8, 2);
+  MDNode *R0 = MDHelper.createRange(A, A);
+  MDNode *R1 = MDHelper.createRange(A, B);
+  EXPECT_EQ(R0, (MDNode *)0);
+  EXPECT_NE(R1, (MDNode *)0);
+  EXPECT_EQ(R1->getNumOperands(), 2U);
+  EXPECT_TRUE(isa(R1->getOperand(0)));
+  EXPECT_TRUE(isa(R1->getOperand(1)));
+  ConstantInt *C0 = cast(R1->getOperand(0));
+  ConstantInt *C1 = cast(R1->getOperand(1));
+  EXPECT_EQ(C0->getValue(), A);
+  EXPECT_EQ(C1->getValue(), B);
+}
+TEST_F(MDBuilderTest, createAnonymousTBAARoot) {
+  MDBuilder MDHelper(Context);
+  MDNode *R0 = MDHelper.createAnonymousTBAARoot();
+  MDNode *R1 = MDHelper.createAnonymousTBAARoot();
+  EXPECT_NE(R0, R1);
+  EXPECT_GE(R0->getNumOperands(), 1U);
+  EXPECT_GE(R1->getNumOperands(), 1U);
+  EXPECT_EQ(R0->getOperand(0), R0);
+  EXPECT_EQ(R1->getOperand(0), R1);
+  EXPECT_TRUE(R0->getNumOperands() == 1 || R0->getOperand(1) == 0);
+  EXPECT_TRUE(R1->getNumOperands() == 1 || R1->getOperand(1) == 0);
+}
+TEST_F(MDBuilderTest, createTBAARoot) {
+  MDBuilder MDHelper(Context);
+  MDNode *R0 = MDHelper.createTBAARoot("Root");
+  MDNode *R1 = MDHelper.createTBAARoot("Root");
+  EXPECT_EQ(R0, R1);
+  EXPECT_GE(R0->getNumOperands(), 1U);
+  EXPECT_TRUE(isa(R0->getOperand(0)));
+  EXPECT_EQ(cast(R0->getOperand(0))->getString(), "Root");
+  EXPECT_TRUE(R0->getNumOperands() == 1 || R0->getOperand(1) == 0);
+}
+TEST_F(MDBuilderTest, createTBAANode) {
+  MDBuilder MDHelper(Context);
+  MDNode *R = MDHelper.createTBAARoot("Root");
+  MDNode *N0 = MDHelper.createTBAANode("Node", R);
+  MDNode *N1 = MDHelper.createTBAANode("edoN", R);
+  MDNode *N2 = MDHelper.createTBAANode("Node", R, true);
+  MDNode *N3 = MDHelper.createTBAANode("Node", R);
+  EXPECT_EQ(N0, N3);
+  EXPECT_NE(N0, N1);
+  EXPECT_NE(N0, N2);
+  EXPECT_GE(N0->getNumOperands(), 2U);
+  EXPECT_GE(N1->getNumOperands(), 2U);
+  EXPECT_GE(N2->getNumOperands(), 3U);
+  EXPECT_TRUE(isa(N0->getOperand(0)));
+  EXPECT_TRUE(isa(N1->getOperand(0)));
+  EXPECT_TRUE(isa(N2->getOperand(0)));
+  EXPECT_EQ(cast(N0->getOperand(0))->getString(), "Node");
+  EXPECT_EQ(cast(N1->getOperand(0))->getString(), "edoN");
+  EXPECT_EQ(cast(N2->getOperand(0))->getString(), "Node");
+  EXPECT_EQ(N0->getOperand(1), R);
+  EXPECT_EQ(N1->getOperand(1), R);
+  EXPECT_EQ(N2->getOperand(1), R);
+  EXPECT_TRUE(isa(N2->getOperand(2)));
+  EXPECT_EQ(cast(N2->getOperand(2))->getZExtValue(), 1U);
+}
+}
diff --git a/unittests/VMCore/InstructionsTest.cpp b/unittests/VMCore/InstructionsTest.cpp
index 218a9a08c43..d002101cd3b 100644
--- a/unittests/VMCore/InstructionsTest.cpp
+++ b/unittests/VMCore/InstructionsTest.cpp
@@ -12,8 +12,11 @@
 #include "llvm/Constants.h"
 #include "llvm/DerivedTypes.h"
 #include "llvm/LLVMContext.h"
+#include "llvm/Operator.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/Analysis/ValueTracking.h"
+#include "llvm/Support/MDBuilder.h"
+#include "llvm/Support/IRBuilder.h"
 #include "llvm/Target/TargetData.h"
 #include "gtest/gtest.h"
 
@@ -226,5 +229,19 @@ TEST(InstructionsTest, VectorGep) {
   delete PtrVecB;
 }
 
+TEST(InstructionsTest, FPMathOperator) {
+  LLVMContext &Context = getGlobalContext();
+  IRBuilder<> Builder(Context);
+  MDBuilder MDHelper(Context);
+  Instruction *I = Builder.CreatePHI(Builder.getDoubleTy(), 0);
+  MDNode *MD1 = MDHelper.createFPMath(1.0);
+  Value *V1 = Builder.CreateFAdd(I, I, "", MD1);
+  EXPECT_TRUE(isa(V1));
+  FPMathOperator *O1 = cast(V1);
+  EXPECT_EQ(O1->getFPAccuracy(), 1.0);
+  delete V1;
+  delete I;
+}
+
 }  // end anonymous namespace
 }  // end namespace llvm
diff --git a/utils/Makefile b/utils/Makefile
index b98376006ed..ecb30bed7c6 100644
--- a/utils/Makefile
+++ b/utils/Makefile
@@ -9,7 +9,7 @@
 
 LEVEL = ..
 PARALLEL_DIRS := FileCheck FileUpdate TableGen PerfectShuffle \
-	      count fpcmp llvm-lit not unittest json-bench
+	      count fpcmp llvm-lit not unittest
 
 EXTRA_DIST := check-each-file codegen-diff countloc.sh \
               DSAclean.py DSAextract.py emacs findsym.pl GenLibDeps.pl \
diff --git a/utils/TableGen/AsmMatcherEmitter.cpp b/utils/TableGen/AsmMatcherEmitter.cpp
index 39a3c25d99d..dc92a6c218e 100644
--- a/utils/TableGen/AsmMatcherEmitter.cpp
+++ b/utils/TableGen/AsmMatcherEmitter.cpp
@@ -1181,7 +1181,8 @@ void AsmMatcherInfo::BuildInfo() {
   unsigned VariantCount = Target.getAsmParserVariantCount();
   for (unsigned VC = 0; VC != VariantCount; ++VC) {
     Record *AsmVariant = Target.getAsmParserVariant(VC);
-    std::string CommentDelimiter = AsmVariant->getValueAsString("CommentDelimiter");
+    std::string CommentDelimiter =
+      AsmVariant->getValueAsString("CommentDelimiter");
     std::string RegisterPrefix = AsmVariant->getValueAsString("RegisterPrefix");
     int AsmVariantNo = AsmVariant->getValueAsInt("Variant");
 
@@ -1208,11 +1209,11 @@ void AsmMatcherInfo::BuildInfo() {
           // reject it.  We reject aliases and ignore instructions for now.
           if (OI.MINumOperands != 1) {
             // FIXME: Should reject these.  The ARM backend hits this with $lane
-            // in a bunch of instructions. It is unclear what the right answer is.
+            // in a bunch of instructions. The right answer is unclear.
             DEBUG({
                 errs() << "warning: '" << CGI.TheDef->getName() << "': "
-                       << "ignoring instruction with multi-operand tied operand '"
-                       << OI.Name << "'\n";
+                     << "ignoring instruction with multi-operand tied operand '"
+                     << OI.Name << "'\n";
               });
             continue;
           }
@@ -1248,8 +1249,8 @@ void AsmMatcherInfo::BuildInfo() {
       // If the tblgen -match-prefix option is specified (for tblgen hackers),
       // filter the set of instruction aliases we consider, based on the target
       // instruction.
-      if (!StringRef(Alias->ResultInst->TheDef->getName()).startswith(
-                                                                      MatchPrefix))
+      if (!StringRef(Alias->ResultInst->TheDef->getName())
+            .startswith( MatchPrefix))
         continue;
 
       OwningPtr II(new MatchableInfo(Alias));
@@ -1328,6 +1329,9 @@ void AsmMatcherInfo::BuildInfo() {
     Record *Rec = AllTokenAliases[i];
     ClassInfo *FromClass = getTokenClass(Rec->getValueAsString("FromToken"));
     ClassInfo *ToClass = getTokenClass(Rec->getValueAsString("ToToken"));
+    if (FromClass == ToClass)
+      throw TGError(Rec->getLoc(),
+                    "error: Destination value identical to source value.");
     FromClass->SuperClasses.push_back(ToClass);
   }
 
@@ -1868,7 +1872,8 @@ static void EmitComputeAvailableFeatures(AsmMatcherInfo &Info,
     SubtargetFeatureInfo &SFI = *it->second;
 
     OS << "  if (";
-    std::string CondStorage = SFI.TheDef->getValueAsString("AssemblerCondString");
+    std::string CondStorage =
+      SFI.TheDef->getValueAsString("AssemblerCondString");
     StringRef Conds = CondStorage;
     std::pair Comma = Conds.split(',');
     bool First = true;
diff --git a/utils/TableGen/AsmWriterEmitter.cpp b/utils/TableGen/AsmWriterEmitter.cpp
index e0b0aace33e..d079b45e8d5 100644
--- a/utils/TableGen/AsmWriterEmitter.cpp
+++ b/utils/TableGen/AsmWriterEmitter.cpp
@@ -289,7 +289,8 @@ void AsmWriterEmitter::EmitPrintInstruction(raw_ostream &O) {
   for (unsigned i = 0, e = NumberedInstructions.size(); i != e; ++i) {
     AsmWriterInst *AWI = CGIAWIMap[NumberedInstructions[i]];
     if (AWI != 0 &&
-        AWI->Operands[0].OperandType == AsmWriterOperand::isLiteralTextOperand &&
+        AWI->Operands[0].OperandType ==
+                 AsmWriterOperand::isLiteralTextOperand &&
         !AWI->Operands[0].Str.empty()) {
       std::string Str = AWI->Operands[0].Str;
       UnescapeString(Str);
@@ -589,7 +590,7 @@ class IAPrinter {
   std::map OpMap;
   std::string Result;
   std::string AsmString;
-  std::vector ReqFeatures;
+  SmallVector ReqFeatures;
 public:
   IAPrinter(std::string R, std::string AS)
     : Result(R), AsmString(AS) {}
@@ -919,7 +920,7 @@ void AsmWriterEmitter::EmitPrintAliasInstruction(raw_ostream &O) {
   O << "      }\n";
   O << "    }\n";
   O << "  }\n\n";
-  
+
   O << "  return true;\n";
   O << "}\n\n";
 
diff --git a/utils/TableGen/CodeGenDAGPatterns.cpp b/utils/TableGen/CodeGenDAGPatterns.cpp
index d2ddf232b32..d4b02fbd2f3 100644
--- a/utils/TableGen/CodeGenDAGPatterns.cpp
+++ b/utils/TableGen/CodeGenDAGPatterns.cpp
@@ -17,6 +17,7 @@
 #include "llvm/TableGen/Record.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/Twine.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
 #include 
@@ -2483,10 +2484,9 @@ static void InferFromPattern(const CodeGenInstruction &Inst,
     // If we decided that this is a store from the pattern, then the .td file
     // entry is redundant.
     if (MayStore)
-      fprintf(stderr,
-              "Warning: mayStore flag explicitly set on instruction '%s'"
-              " but flag already inferred from pattern.\n",
-              Inst.TheDef->getName().c_str());
+      PrintWarning(Inst.TheDef->getLoc(),
+                   "mayStore flag explicitly set on "
+                   "instruction, but flag already inferred from pattern.");
     MayStore = true;
   }
 
@@ -2494,24 +2494,25 @@ static void InferFromPattern(const CodeGenInstruction &Inst,
     // If we decided that this is a load from the pattern, then the .td file
     // entry is redundant.
     if (MayLoad)
-      fprintf(stderr,
-              "Warning: mayLoad flag explicitly set on instruction '%s'"
-              " but flag already inferred from pattern.\n",
-              Inst.TheDef->getName().c_str());
+      PrintWarning(Inst.TheDef->getLoc(),
+                   "mayLoad flag explicitly set on "
+                   "instruction, but flag already inferred from pattern.");
     MayLoad = true;
   }
 
   if (Inst.neverHasSideEffects) {
     if (HadPattern)
-      fprintf(stderr, "Warning: neverHasSideEffects set on instruction '%s' "
-              "which already has a pattern\n", Inst.TheDef->getName().c_str());
+      PrintWarning(Inst.TheDef->getLoc(),
+                   "neverHasSideEffects flag explicitly set on "
+                   "instruction, but flag already inferred from pattern.");
     HasSideEffects = false;
   }
 
   if (Inst.hasSideEffects) {
     if (HasSideEffects)
-      fprintf(stderr, "Warning: hasSideEffects set on instruction '%s' "
-              "which already inferred this.\n", Inst.TheDef->getName().c_str());
+      PrintWarning(Inst.TheDef->getLoc(),
+                   "hasSideEffects flag explicitly set on "
+                   "instruction, but flag already inferred from pattern.");
     HasSideEffects = true;
   }
 
diff --git a/utils/TableGen/CodeGenRegisters.cpp b/utils/TableGen/CodeGenRegisters.cpp
index 7ce4f878a3e..45c5bb8b3a5 100644
--- a/utils/TableGen/CodeGenRegisters.cpp
+++ b/utils/TableGen/CodeGenRegisters.cpp
@@ -19,6 +19,7 @@
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/StringExtras.h"
+#include "llvm/ADT/Twine.h"
 
 using namespace llvm;
 
@@ -896,11 +897,10 @@ void CodeGenRegBank::computeComposites() {
           if (i1d->second == Reg3) {
             // Conflicting composition? Emit a warning but allow it.
             if (CodeGenSubRegIndex *Prev = Idx1->addComposite(Idx2, i1d->first))
-              errs() << "Warning: SubRegIndex " << Idx1->getQualifiedName()
-                     << " and " << Idx2->getQualifiedName()
-                     << " compose ambiguously as "
-                     << Prev->getQualifiedName() << " or "
-                     << i1d->first->getQualifiedName() << "\n";
+              PrintWarning(Twine("SubRegIndex") + Idx1->getQualifiedName() +
+                     " and " + Idx2->getQualifiedName() +
+                     " compose ambiguously as " + Prev->getQualifiedName() +
+                     " or " + i1d->first->getQualifiedName());
           }
         }
       }
diff --git a/utils/TableGen/EDEmitter.cpp b/utils/TableGen/EDEmitter.cpp
index 3809a4576a5..fe484ca8cf1 100644
--- a/utils/TableGen/EDEmitter.cpp
+++ b/utils/TableGen/EDEmitter.cpp
@@ -594,6 +594,7 @@ static int ARMFlagFromOpName(LiteralConstantEmitter *type,
   IMM("jtblock_operand");
   IMM("nohash_imm");
   IMM("p_imm");
+  IMM("pf_imm");
   IMM("c_imm");
   IMM("coproc_option_imm");
   IMM("imod_op");
diff --git a/utils/TableGen/RegisterInfoEmitter.cpp b/utils/TableGen/RegisterInfoEmitter.cpp
index a2478a7330e..97fcca3a850 100644
--- a/utils/TableGen/RegisterInfoEmitter.cpp
+++ b/utils/TableGen/RegisterInfoEmitter.cpp
@@ -17,22 +17,23 @@
 #include "CodeGenTarget.h"
 #include "CodeGenRegisters.h"
 #include "SequenceToOffsetTable.h"
+#include "llvm/TableGen/Error.h"
 #include "llvm/TableGen/Record.h"
 #include "llvm/ADT/BitVector.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/Twine.h"
 #include "llvm/Support/Format.h"
 #include 
 #include 
 using namespace llvm;
 
 // runEnums - Print out enum values for all of the registers.
-void
-RegisterInfoEmitter::runEnums(raw_ostream &OS,
-                              CodeGenTarget &Target, CodeGenRegBank &Bank) {
+void RegisterInfoEmitter::runEnums(raw_ostream &OS,
+                                   CodeGenTarget &Target, CodeGenRegBank &Bank) {
   const std::vector &Registers = Bank.getRegisters();
 
-  // Register enums are stored as uint16_t in the tables. Make sure we'll fit
+  // Register enums are stored as uint16_t in the tables. Make sure we'll fit.
   assert(Registers.size() <= 0xffff && "Too many regs to fit in tables");
 
   std::string Namespace = Registers[0]->TheDef->getValueAsString("Namespace");
@@ -208,8 +209,8 @@ RegisterInfoEmitter::EmitRegMappingTables(raw_ostream &OS,
     std::vector RegNums = Reg->getValueAsListOfInts("DwarfNumbers");
     maxLength = std::max((size_t)maxLength, RegNums.size());
     if (DwarfRegNums.count(Reg))
-      errs() << "Warning: DWARF numbers for register " << getQualifiedName(Reg)
-             << "specified multiple times\n";
+      PrintWarning(Reg->getLoc(), Twine("DWARF numbers for register ") +
+                   getQualifiedName(Reg) + "specified multiple times");
     DwarfRegNums[Reg] = RegNums;
   }
 
diff --git a/utils/buildit/build_llvm b/utils/buildit/build_llvm
index 88a26d30955..994fb0696ee 100755
--- a/utils/buildit/build_llvm
+++ b/utils/buildit/build_llvm
@@ -228,19 +228,6 @@ fi
 cd $DEST_DIR$DEST_ROOT
 rm -f bin/.dir etc/llvm/.dir lib/.dir
 
-# Remove PPC64 fat slices.
-cd $DEST_DIR$DEST_ROOT/bin
-if [ $MACOSX_DEPLOYMENT_TARGET = "10.4" ]; then
-    find . -perm 755 -type f \! \( -name '*gccas' -o -name '*gccld' -o -name llvm-config \) \
-        -exec lipo -extract ppc -extract i386 {} -output {} \;
-elif [ $MACOSX_DEPLOYMENT_TARGET = "10.5" ]; then
-    find . -perm 755 -type f \! \( -name '*gccas' -o -name '*gccld' -o -name llvm-config \) \
-        -exec lipo -extract ppc7400 -extract i386 {} -output {} \;
-else
-    find . -perm 755 -type f \! \( -name '*gccas' -o -name '*gccld' -o -name llvm-config \) \
-        -exec lipo -extract i386 -extract x86_64 {} -output {} \;
-fi
-
 # The Hello dylib is an example of how to build a pass.
 # The BugpointPasses module is only used to test bugpoint.
 # These unversioned dylibs cause verification failures, so do not install them.
diff --git a/utils/json-bench/CMakeLists.txt b/utils/json-bench/CMakeLists.txt
deleted file mode 100644
index 03ac51ce64a..00000000000
--- a/utils/json-bench/CMakeLists.txt
+++ /dev/null
@@ -1,5 +0,0 @@
-add_llvm_utility(json-bench
-  JSONBench.cpp
-  )
-
-target_link_libraries(json-bench LLVMSupport)
diff --git a/utils/json-bench/JSONBench.cpp b/utils/json-bench/JSONBench.cpp
deleted file mode 100644
index ca8a36a03ab..00000000000
--- a/utils/json-bench/JSONBench.cpp
+++ /dev/null
@@ -1,85 +0,0 @@
-//===- JSONBench - Benchmark the JSONParser implementation ----------------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// This program executes the JSONParser on differntly sized JSON texts and
-// outputs the run time.
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/ADT/Twine.h"
-#include "llvm/Support/CommandLine.h"
-#include "llvm/Support/JSONParser.h"
-#include "llvm/Support/Timer.h"
-#include "llvm/Support/raw_ostream.h"
-
-static llvm::cl::opt
-Verify("verify", llvm::cl::desc(
-         "Run a quick verification useful for regression testing"),
-       llvm::cl::init(false));
-
-static llvm::cl::opt
-MemoryLimitMB("memory-limit", llvm::cl::desc(
-                "Do not use more megabytes of memory"),
-	          llvm::cl::init(1000));
-
-void benchmark(llvm::TimerGroup &Group, llvm::StringRef Name,
-               llvm::StringRef JSONText) {
-  llvm::Timer BaseLine((Name + ": Loop").str(), Group);
-  BaseLine.startTimer();
-  char C = 0;
-  for (llvm::StringRef::iterator I = JSONText.begin(),
-                                 E = JSONText.end();
-       I != E; ++I) { C += *I; }
-  BaseLine.stopTimer();
-  volatile char DontOptimizeOut = C; (void)DontOptimizeOut;
-
-  llvm::Timer Parsing((Name + ": Parsing").str(), Group);
-  Parsing.startTimer();
-  llvm::SourceMgr SM;
-  llvm::JSONParser Parser(JSONText, &SM);
-  if (!Parser.validate()) {
-    llvm::errs() << "Parsing error in JSON parser benchmark.\n";
-    exit(1);
-  }
-  Parsing.stopTimer();
-}
-
-std::string createJSONText(size_t MemoryMB, unsigned ValueSize) {
-  std::string JSONText;
-  llvm::raw_string_ostream Stream(JSONText);
-  Stream << "[\n";
-  size_t MemoryBytes = MemoryMB * 1024 * 1024;
-  while (JSONText.size() < MemoryBytes) {
-    Stream << " {\n"
-           << "  \"key1\": \"" << std::string(ValueSize, '*') << "\",\n"
-           << "  \"key2\": \"" << std::string(ValueSize, '*') << "\",\n"
-           << "  \"key3\": \"" << std::string(ValueSize, '*') << "\"\n"
-           << " }";
-    Stream.flush();
-    if (JSONText.size() < MemoryBytes) Stream << ",";
-    Stream << "\n";
-  }
-  Stream << "]\n";
-  Stream.flush();
-  return JSONText;
-}
-
-int main(int argc, char **argv) {
-  llvm::cl::ParseCommandLineOptions(argc, argv);
-  llvm::TimerGroup Group("JSON parser benchmark");
-  if (Verify) {
-    benchmark(Group, "Fast", createJSONText(10, 500));
-  } else {
-    benchmark(Group, "Small Values", createJSONText(MemoryLimitMB, 5));
-    benchmark(Group, "Medium Values", createJSONText(MemoryLimitMB, 500));
-    benchmark(Group, "Large Values", createJSONText(MemoryLimitMB, 50000));
-  }
-  return 0;
-}
-
diff --git a/utils/release/findRegressions-nightly.py b/utils/release/findRegressions-nightly.py
index e801dab4aba..ddf89835e2c 100755
--- a/utils/release/findRegressions-nightly.py
+++ b/utils/release/findRegressions-nightly.py
@@ -1,4 +1,4 @@
-#!/usr/bin/python
+#!/usr/bin/env python
 import re, string, sys, os, time
 
 DEBUG = 0
diff --git a/utils/release/findRegressions-simple.py b/utils/release/findRegressions-simple.py
index 758623199ff..8d3b4cfca29 100755
--- a/utils/release/findRegressions-simple.py
+++ b/utils/release/findRegressions-simple.py
@@ -1,4 +1,4 @@
-#!/usr/bin/python
+#!/usr/bin/env python
 import re, string, sys, os, time, math
 
 DEBUG = 0
diff --git a/utils/release/tag.sh b/utils/release/tag.sh
index 80da47a4db0..399d5c5a7fa 100755
--- a/utils/release/tag.sh
+++ b/utils/release/tag.sh
@@ -16,33 +16,40 @@ set -e
 
 release=""
 rc=""
+rebranch="no"
 
 base_url="https://llvm.org/svn/llvm-project"
 
 function usage() {
-    echo "usage: `basename $0` -release "
+    echo "usage: `basename $0` -release  [-rebranch]"
     echo "usage: `basename $0` -release  -rc "
     echo " "
     echo "  -release   The version number of the release"
     echo "  -rc        The release candidate number"
+    echo "  -rebranch       Remove existing branch, if present, before branching"
     echo "  -final          Tag final release candidate"
 }
 
 function tag_version() {
     set -x
-    for proj in llvm cfe dragonegg test-suite compiler-rt libcxx libcxxabi ; do
-        if ! svn ls $base_url/$proj/branches/release_$release > /dev/null 2>&1 ; then
-            svn copy -m "Creating release_$release branch" \
-                $base_url/$proj/trunk \
+    for proj in llvm cfe dragonegg test-suite compiler-rt ; do
+        if svn ls $base_url/$proj/branches/release_$release > /dev/null 2>&1 ; then
+            if [ $rebranch = "no" ]; then
+                continue
+            fi
+            svn remove -m "Removing old release_$release branch for rebranching." \
                 $base_url/$proj/branches/release_$release
         fi
+        svn copy -m "Creating release_$release branch" \
+            $base_url/$proj/trunk \
+            $base_url/$proj/branches/release_$release
     done
     set +x
 }
 
 function tag_release_candidate() {
     set -x
-    for proj in llvm cfe dragonegg test-suite compiler-rt libcxx libcxxabi ; do
+    for proj in llvm cfe dragonegg test-suite compiler-rt ; do
         if ! svn ls $base_url/$proj/tags/RELEASE_$release > /dev/null 2>&1 ; then
             svn mkdir -m "Creating release directory for release_$release." $base_url/$proj/tags/RELEASE_$release
         fi
@@ -65,6 +72,9 @@ while [ $# -gt 0 ]; do
             shift
             rc="rc$1"
             ;;
+        -rebranch | --rebranch )
+            rebranch="yes"
+            ;;
         -final | --final )
             rc="final"
             ;;
diff --git a/utils/release/test-release.sh b/utils/release/test-release.sh
index ad1af5fef05..d2946d8762a 100755
--- a/utils/release/test-release.sh
+++ b/utils/release/test-release.sh
@@ -18,7 +18,7 @@ else
     MAKE=make
 fi
 
-projects="llvm cfe dragonegg test-suite"
+projects="llvm cfe dragonegg compiler-rt test-suite"
 
 # Base SVN URL for the sources.
 Base_url="http://llvm.org/svn/llvm-project"
@@ -214,6 +214,9 @@ function export_sources() {
     if [ ! -h llvm-test ]; then
         ln -s ../../test-suite.src llvm-test
     fi
+    if [ ! -h compiler-rt ]; then
+        ln -s ../../compiler-rt.src compiler-rt
+    fi
     cd $BuildDir
 }