oota-llvm.git
12 years agoadd a new @llvm.donothing intrinsic that, well, does nothing, and teach CodeGen to...
Nuno Lopes [Thu, 28 Jun 2012 22:30:12 +0000 (22:30 +0000)]
add a new @llvm.donothing intrinsic that, well, does nothing, and teach CodeGen to ignore calls to it

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159383 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago'Promote' vector [su]int_to_fp should widen elements.
Jim Grosbach [Thu, 28 Jun 2012 21:03:44 +0000 (21:03 +0000)]
'Promote' vector [su]int_to_fp should widen elements.

Teach vector legalization how to honor Promote for int to float
conversions. The code checking whether to promote the operation knew
to look at the operand, but the actual promotion code didn't. This
fixes that. The operand is promoted up via [zs]ext.

rdar://11762659

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159378 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoChanged the formatting sequence of a curly brace to
Jack Carter [Thu, 28 Jun 2012 20:46:26 +0000 (20:46 +0000)]
Changed the formatting sequence of a curly brace to
the comment per code review feedback.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159376 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRegenerate.
Eric Christopher [Thu, 28 Jun 2012 20:35:00 +0000 (20:35 +0000)]
Regenerate.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159374 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove layering violation #include.
Bill Wendling [Thu, 28 Jun 2012 20:17:05 +0000 (20:17 +0000)]
Remove layering violation #include.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159372 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix hexagon gcc builtin names to use '_' instead of '.'.
Benjamin Kramer [Thu, 28 Jun 2012 20:08:47 +0000 (20:08 +0000)]
Fix hexagon gcc builtin names to use '_' instead of '.'.

This way the generated GCC builtin to LLVM intrinsic converter
actually works.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159370 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoEnable automatic GCC<->LLVM intrinsic translation for mips.
Benjamin Kramer [Thu, 28 Jun 2012 19:09:53 +0000 (19:09 +0000)]
Enable automatic GCC<->LLVM intrinsic translation for mips.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159367 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDefine MIPS DSP Rev1 intrinsics. That allows frontend to emit a correct IR.
Simon Atanasyan [Thu, 28 Jun 2012 18:20:28 +0000 (18:20 +0000)]
Define MIPS DSP Rev1 intrinsics. That allows frontend to emit a correct IR.
This patch was reviewed in the llvm-commits list by Jim Grosbach.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159364 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMemoryBuiltins:
Nuno Lopes [Thu, 28 Jun 2012 16:34:03 +0000 (16:34 +0000)]
MemoryBuiltins:
 - recognize C++ new(std::nothrow) friends
 - ignore ExtractElement and ExtractValue instructions in size/offset analysis (all easy cases are probably folded away before we get here)
 - also recognize realloc as noalias

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159356 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agomake LazyValueInfo analyze the default case of switch statements (we know that in...
Nuno Lopes [Thu, 28 Jun 2012 16:13:37 +0000 (16:13 +0000)]
make LazyValueInfo analyze the default case of switch statements (we know that in the default branch the value cannot be any of the switch cases)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159353 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoadd ConstantRange::difference (to perform set difference/relative complement)
Nuno Lopes [Thu, 28 Jun 2012 16:10:13 +0000 (16:10 +0000)]
add ConstantRange::difference (to perform set difference/relative complement)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159352 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDevirtualize DIScope and subclasses.
Benjamin Kramer [Thu, 28 Jun 2012 14:25:45 +0000 (14:25 +0000)]
Devirtualize DIScope and subclasses.

Nothing in here makes use of the virtuality.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159349 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[asan] set a hard limit on the number of instructions instrumented pear each BB....
Kostya Serebryany [Thu, 28 Jun 2012 09:34:41 +0000 (09:34 +0000)]
[asan] set a hard limit on the number of instructions instrumented pear each BB. This is (hopefully temporary) workaround for PR13225

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159344 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSphinxify the bitcode format document.
Bill Wendling [Thu, 28 Jun 2012 08:43:12 +0000 (08:43 +0000)]
Sphinxify the bitcode format document.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159340 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix Windows build after r159281: s/iterator/const_iterator
Alexey Samsonov [Thu, 28 Jun 2012 07:47:50 +0000 (07:47 +0000)]
Fix Windows build after r159281: s/iterator/const_iterator

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159334 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove the setup for variables that are expanded in the lit.site.cfg into
Chandler Carruth [Thu, 28 Jun 2012 06:36:24 +0000 (06:36 +0000)]
Move the setup for variables that are expanded in the lit.site.cfg into
a dedicated helper function. This will enable re-using the same logic
for Clang's lit setup, etc.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159333 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoPrecompute SCEV pointer analysis prior to instruction fusion in BBVectorize.
Hal Finkel [Thu, 28 Jun 2012 05:42:45 +0000 (05:42 +0000)]
Precompute SCEV pointer analysis prior to instruction fusion in BBVectorize.

When both a load/store and its address computation are being vectorized, it can
happen that the address-computation vectorization destroys SCEV's ability
to analyize the relative pointer offsets. As a result (like with the aliasing
analysis info), we need to precompute the necessary information prior to
instruction fusing.

This was found during stress testing (running through the test suite with a very
low required chain length); unfortunately, I don't have a small test case.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159332 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove a useless check in BBVectorize.
Hal Finkel [Thu, 28 Jun 2012 05:42:43 +0000 (05:42 +0000)]
Remove a useless check in BBVectorize.

A shuffle mask will always be a constant, but I did not realize that
when I originally wrote the code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159331 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAllow BBVectorize to form non-2^n-length vectors.
Hal Finkel [Thu, 28 Jun 2012 05:42:42 +0000 (05:42 +0000)]
Allow BBVectorize to form non-2^n-length vectors.

The original algorithm only used recursive pair fusion of equal-length
types. This is now extended to allow pairing of any types that share
the same underlying scalar type. Because we would still generally
prefer the 2^n-length types, those are formed first. Then a second
set of iterations form the non-2^n-length types.

Also, a call to SimplifyInstructionsInBlock has been added after each
pairing iteration. This takes care of DCE (and a few other things)
that make the following iterations execute somewhat faster. For the
same reason, some of the simple shuffle-combination cases are now
handled internally.

There is some additional refactoring work to be done, but I've had
many requests for this feature, so additional refactoring will come
soon in future commits (as will additional test cases).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159330 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRefactor operation equivalence checking in BBVectorize by extending Instruction:...
Hal Finkel [Thu, 28 Jun 2012 05:42:26 +0000 (05:42 +0000)]
Refactor operation equivalence checking in BBVectorize by extending Instruction::isSameOperationAs.

Maintaining this kind of checking in different places is dangerous, extending
Instruction::isSameOperationAs consolidates this logic into one place. Here
I've added an optional flags parameter and two flags that are important for
vectorization: CompareIgnoringAlignment and CompareUsingScalarTypes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159329 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoOnly print out the tag if it's there.
Bill Wendling [Thu, 28 Jun 2012 02:17:58 +0000 (02:17 +0000)]
Only print out the tag if it's there.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159328 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDon't output an empty string.
Bill Wendling [Thu, 28 Jun 2012 02:12:20 +0000 (02:12 +0000)]
Don't output an empty string.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159327 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThe Mips specific inline asm operand modifier 'z' has the
Jack Carter [Thu, 28 Jun 2012 01:33:40 +0000 (01:33 +0000)]
The Mips specific inline asm operand modifier 'z' has the
following description in the gnu sources:

    Print $0 if operand is zero otherwise print the op normally.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159324 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agomake LVI::getEdgeValue() always intersect the constraints of the edge with the range...
Nuno Lopes [Thu, 28 Jun 2012 01:16:18 +0000 (01:16 +0000)]
make LVI::getEdgeValue() always intersect the constraints of the edge with the range of the block. Previously it was only performing the intersection for a few cases, thus losing precision

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159320 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agofix a off-by-one bug in intersectWith(), and add a bunch of tests
Nuno Lopes [Thu, 28 Jun 2012 00:59:33 +0000 (00:59 +0000)]
fix a off-by-one bug in intersectWith(), and add a bunch of tests

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159319 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUse the interface through DIDescriptor to get the tag/version for a debug info
Bill Wendling [Thu, 28 Jun 2012 00:41:44 +0000 (00:41 +0000)]
Use the interface through DIDescriptor to get the tag/version for a debug info
MDNode.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159317 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSilence unused variable warning.
Richard Trieu [Thu, 28 Jun 2012 00:41:11 +0000 (00:41 +0000)]
Silence unused variable warning.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159316 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix cmake failure from moving files around.
Bill Wendling [Thu, 28 Jun 2012 00:18:12 +0000 (00:18 +0000)]
Fix cmake failure from moving files around.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159314 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove 'site.exp' building from both CMake and configure+make.
Chandler Carruth [Thu, 28 Jun 2012 00:16:51 +0000 (00:16 +0000)]
Remove 'site.exp' building from both CMake and configure+make.

This is another vestige of the DejaGNU roots. There were FIXMEs in the
lit setup to add a 'lit.site.cfg', which has been around for quite some
time now, so I've properly switched the handling of the 4 things
actually used in site.exp to go through lit.site.cfg now. No more
parsing of the .exp file, one fewer configure-style generated file,
etc., etc.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159313 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove lib/Analysis/DebugInfo.cpp to lib/VMCore/DebugInfo.cpp and
Bill Wendling [Thu, 28 Jun 2012 00:05:13 +0000 (00:05 +0000)]
Move lib/Analysis/DebugInfo.cpp to lib/VMCore/DebugInfo.cpp and
include/llvm/Analysis/DebugInfo.h to include/llvm/DebugInfo.h.

The reasoning is because the DebugInfo module is simply an interface to the
debug info MDNodes and has nothing to do with analysis.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159312 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove the last vestiges of the '-lit' and '-dg' test runner split by
Chandler Carruth [Thu, 28 Jun 2012 00:03:15 +0000 (00:03 +0000)]
Remove the last vestiges of the '-lit' and '-dg' test runner split by
removing '-lit' qualifiers from make rules. I've left a legacy
'check-local-lit' rule in case build scripts have this encoded
somewhere.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159311 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove 'check-dg', a wrapper around 'check-local-dg' which was just
Chandler Carruth [Thu, 28 Jun 2012 00:03:13 +0000 (00:03 +0000)]
Remove 'check-dg', a wrapper around 'check-local-dg' which was just
nuked.

Add a comment that the 'check-lit' rule is really just a legacy of
having two test runners.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159310 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove autoconf support for runtest and tclsh, some of the last vestigal
Chandler Carruth [Wed, 27 Jun 2012 23:53:41 +0000 (23:53 +0000)]
Remove autoconf support for runtest and tclsh, some of the last vestigal
bits of DejaGNU.

Eric, you may want to remove the TCLSH bits from aclocal.m4 and
regenerate... I didn't want to touch the m4 file lest something
exploded.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159308 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRip out legacy DejaGNU support from our Makefiles. This hasn't been the
Chandler Carruth [Wed, 27 Jun 2012 23:48:39 +0000 (23:48 +0000)]
Rip out legacy DejaGNU support from our Makefiles. This hasn't been the
default in forever, and hasn't even worked since most of the .exp files
were removed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159307 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoLLVM-GCC is dead. Really. I promise. ;]
Chandler Carruth [Wed, 27 Jun 2012 23:34:25 +0000 (23:34 +0000)]
LLVM-GCC is dead. Really. I promise. ;]

More importantly, these files don't even have the variable that these
lines purport to substite.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159304 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThis allows hello world to be compiled for Mips 64 direct object.
Jack Carter [Wed, 27 Jun 2012 23:13:42 +0000 (23:13 +0000)]
This allows hello world to be compiled for Mips 64 direct object.

It takes advantage of r159299 which introduces relocation support for N64.
elf-dump needed to be upgraded to support N64 relocations as well.

This passes make check.

Jack

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159302 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThis allows hello world to be compiled for Mips 64 direct object.
Jack Carter [Wed, 27 Jun 2012 22:48:25 +0000 (22:48 +0000)]
This allows hello world to be compiled for Mips 64 direct object.

It takes advantage of r159299 which introduces relocation support for N64.
elf-dump needed to be upgraded to support N64 relocations as well.

This passes make check.

Jack

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159301 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoWhitespace.
Chad Rosier [Wed, 27 Jun 2012 22:34:28 +0000 (22:34 +0000)]
Whitespace.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159300 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThe ELF relocation record format is different for N64
Jack Carter [Wed, 27 Jun 2012 22:28:30 +0000 (22:28 +0000)]
The ELF relocation record format is different for N64
which many Mips 64 ABIs use than for O64 which many
if not all other target ABIs use.

Most architectures have the following 64 bit relocation record format:

  typedef struct
  {
    Elf64_Addr   r_offset; /* Address of reference */
    Elf64_Xword  r_info;   /* Symbol index and type of relocation */
  } Elf64_Rel;

  typedef struct
  {
    Elf64_Addr    r_offset;
    Elf64_Xword   r_info;
    Elf64_Sxword  r_addend;
  } Elf64_Rela;

Whereas N64 has the following format:

  typedef struct
  {
    Elf64_Addr    r_offset;/* Address of reference */
    Elf64_Word  r_sym;     /* Symbol index */
    Elf64_Byte  r_ssym;    /* Special symbol */
    Elf64_Byte  r_type3;   /* Relocation type */
    Elf64_Byte  r_type2;   /* Relocation type */
    Elf64_Byte  r_type;    /* Relocation type */
  } Elf64_Rel;

  typedef struct
  {
    Elf64_Addr    r_offset;/* Address of reference */
    Elf64_Word  r_sym;     /* Symbol index */
    Elf64_Byte  r_ssym;    /* Special symbol */
    Elf64_Byte  r_type3;   /* Relocation type */
    Elf64_Byte  r_type2;   /* Relocation type */
    Elf64_Byte  r_type;    /* Relocation type */
    Elf64_Sxword  r_addend;
  } Elf64_Rela;

The structure is the same size, but the r_info data element
is now 5 separate elements. Besides the content aspects,
endian byte reordering will be different for the area with
each element being endianized separately.

I treat this as generic and continue to pass r_type as
an integer masking and unmasking the byte sized N64
values for N64 mode. I've implemented this and it causes no
affect on other current targets.

This passes make check.

Jack

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159299 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRefactor and speed up DFA generator.
Anshuman Dasgupta [Wed, 27 Jun 2012 19:38:29 +0000 (19:38 +0000)]
Refactor and speed up DFA generator.

Patch by Ivan Llopard!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159281 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRevert r159136 due to PR13124.
Matt Beaumont-Gay [Wed, 27 Jun 2012 17:10:33 +0000 (17:10 +0000)]
Revert r159136 due to PR13124.

Original commit message:

If a constant or a function has linkonce_odr linkage and unnamed_addr, mark it
hidden. Being linkonce_odr guarantees that it is available in every dso that
needs it. Being a constant/function with unnamed_addr guarantees that the
copies don't have to be merged.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159272 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoWhen users ask for -mcpu=help or -mattr=help, just output the help without
Duncan Sands [Wed, 27 Jun 2012 16:23:48 +0000 (16:23 +0000)]
When users ask for -mcpu=help or -mattr=help, just output the help without
requiring a module.  Original patch by Sunay Ismail, simplified by Arnaud
de Grandmaison, then complicated by me (if a triple was specified on the
command line, output help for that triple, not for the default).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159268 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSome reassociate optimizations create new instructions, which they insert just
Duncan Sands [Wed, 27 Jun 2012 14:19:00 +0000 (14:19 +0000)]
Some reassociate optimizations create new instructions, which they insert just
before the expression root.  Any existing operators that are changed to use one
of them needs to be moved between it and the expression root, and recursively
for the operators using that one.  When I rewrote RewriteExprTree I accidentally
inverted the logic, resulting in the compacting going down from operators to
operands rather than up from operands to the operators using them, oops.  Fix
this, resolving PR12963.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159265 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTeach assembler to handle capitalised operation values for DSB instructions
Richard Barton [Wed, 27 Jun 2012 09:48:23 +0000 (09:48 +0000)]
Teach assembler to handle capitalised operation values for DSB instructions

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159259 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoClean up the 'check' CMake build rule a bit, notable renaming it to
Chandler Carruth [Wed, 27 Jun 2012 09:44:16 +0000 (09:44 +0000)]
Clean up the 'check' CMake build rule a bit, notable renaming it to
'check-llvm'.

Don't worry! 'check' still works! =] To rationalize the names of targets
used to run tests, the vague plan is the following:

make check-llvm  # run LLVM reg/unit tests  (currently 'check')
make check-clang # run Clang reg/unit tests (currently 'clang-test')
make check-rt    # run CompilerRT reg/unit tests
make check-asan  # run ASan reg/unit tests (subset of -rt)
make check-tsan  # run TSan reg/unit tests (subset of -rt)
make check-all   # run as much of the above as is available

The last one respects what projects are checked out and built for
a given tree. Personally, I would like to eventually make 'check' be an
alias for 'check-all'. For now however, it is an alias for 'check-llvm',
and thus no behavior has changed.

While this patch and my plan only really apply to CMake, I think it
might be good to similarly rationalize the naming scheme for the Make
builds.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159258 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoPrevent ARM Assembler crashing on unrecognised assembly format for DSB instruction
Richard Barton [Wed, 27 Jun 2012 09:36:19 +0000 (09:36 +0000)]
Prevent ARM Assembler crashing on unrecognised assembly format for DSB instruction

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159257 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSphinxify the exception handling doc.
Bill Wendling [Wed, 27 Jun 2012 07:20:57 +0000 (07:20 +0000)]
Sphinxify the exception handling doc.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159254 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSilence uninitialized variable warning in MipsISelDAGToDAG.cpp.
Akira Hatanaka [Wed, 27 Jun 2012 00:49:46 +0000 (00:49 +0000)]
Silence uninitialized variable warning in MipsISelDAGToDAG.cpp.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159243 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTest case for r159240.
Akira Hatanaka [Wed, 27 Jun 2012 00:40:34 +0000 (00:40 +0000)]
Test case for r159240.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159242 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoExclude both libcxx and compiler-rt until we get their CMake builds
Chandler Carruth [Wed, 27 Jun 2012 00:30:08 +0000 (00:30 +0000)]
Exclude both libcxx and compiler-rt until we get their CMake builds
suitable for building as a whole-project.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159241 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix bug in computation of stack size in MipsFrameLowering.cpp.
Akira Hatanaka [Wed, 27 Jun 2012 00:20:39 +0000 (00:20 +0000)]
Fix bug in computation of stack size in MipsFrameLowering.cpp.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159240 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoReduce indentation in function. Rearrange some methods. No functionality change.
Bill Wendling [Tue, 26 Jun 2012 23:22:18 +0000 (23:22 +0000)]
Reduce indentation in function. Rearrange some methods. No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159239 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTableGen: AsmMatcher diagnostics preference detail.
Jim Grosbach [Tue, 26 Jun 2012 22:58:01 +0000 (22:58 +0000)]
TableGen: AsmMatcher diagnostics preference detail.

Don't override a custom diagnostic w/ a generic InvalidOperand, all else
being equal.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159238 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRevamp how debugging information is emitted for debug info objects.
Bill Wendling [Tue, 26 Jun 2012 22:57:33 +0000 (22:57 +0000)]
Revamp how debugging information is emitted for debug info objects.

It's not necessary for each DI class to have its own copy of `print' and
`dump'. Instead, just give DIDescriptor those methods and have it call the
appropriate debugging printing routine based on the type of the debug
information.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159237 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd a missing check to avoid dereference null. No sensible test case possible. Sorry...
Evan Cheng [Tue, 26 Jun 2012 22:54:59 +0000 (22:54 +0000)]
Add a missing check to avoid dereference null. No sensible test case possible. Sorry. rdar://11745134

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159236 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove a instcombine transform that (no longer?) makes sense:
Evan Cheng [Tue, 26 Jun 2012 22:03:13 +0000 (22:03 +0000)]
Remove a instcombine transform that (no longer?) makes sense:
    // C - zext(bool) -> bool ? C - 1 : C
    if (ZExtInst *ZI = dyn_cast<ZExtInst>(Op1))
      if (ZI->getSrcTy()->isIntegerTy(1))
        return SelectInst::Create(ZI->getOperand(0), SubOne(C), C);

This ends up forming sext i1 instructions that codegen to terrible code. e.g.
int blah(_Bool x, _Bool y) {
  return (x - y) + 1;
}
=>
        movzbl  %dil, %eax
        movzbl  %sil, %ecx
        shll    $31, %ecx
        sarl    $31, %ecx
        leal    1(%rax,%rcx), %eax
        ret

Without the rule, llvm now generates:
        movzbl  %sil, %ecx
        movzbl  %dil, %eax
        incl    %eax
        subl    %ecx, %eax
        ret

It also helps with ARM (and pretty much any target that doesn't have a sext i1 :-).

The transformation was done as part of Eli's r75531. He has given the ok to
remove it.

rdar://11748024

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159230 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoImplement getHostCPUName for ARM/linux. This will be used to implement -march=native...
Benjamin Kramer [Tue, 26 Jun 2012 21:36:32 +0000 (21:36 +0000)]
Implement getHostCPUName for ARM/linux. This will be used to implement -march=native in clang.

The cpuid registers are only available in privileged mode so we don't have
an OS-independent way of implementing this. ARM doesn't provide a list of
processor IDs so the list is somewhat incomplete.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159228 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix llc's -print-before=pass and -print-after=pass.
Rafael Espindola [Tue, 26 Jun 2012 21:33:36 +0000 (21:33 +0000)]
Fix llc's -print-before=pass and -print-after=pass.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159227 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoX86: add GATHER intrinsics (AVX2) in LLVM
Manman Ren [Tue, 26 Jun 2012 19:47:59 +0000 (19:47 +0000)]
X86: add GATHER intrinsics (AVX2) in LLVM

Support the following intrinsics:
llvm.x86.avx2.gather.d.pd, llvm.x86.avx2.gather.q.pd
llvm.x86.avx2.gather.d.pd.256, llvm.x86.avx2.gather.q.pd.256
llvm.x86.avx2.gather.d.ps, llvm.x86.avx2.gather.q.ps
llvm.x86.avx2.gather.d.ps.256, llvm.x86.avx2.gather.q.ps.256

Modified Disassembler to handle VSIB addressing mode.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159221 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTeach TableGen to put chains on more instructions
Tim Northover [Tue, 26 Jun 2012 18:46:28 +0000 (18:46 +0000)]
Teach TableGen to put chains on more instructions

When generating selection tables for Pat instances, TableGen relied on
an output Instruction's Pattern field being set to infer whether a
chain should be added.

This patch adds additional logic to check various flag fields so that
correct code can be generated even if Pattern is unset.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159217 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix ThreadLocalImpl::getInstance for --disable-threads.
Argyrios Kyrtzidis [Tue, 26 Jun 2012 17:13:58 +0000 (17:13 +0000)]
Fix ThreadLocalImpl::getInstance for --disable-threads.

PR13114.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159210 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAllow targets to inject passes before the virtual register rewriter.
Jakob Stoklund Olesen [Tue, 26 Jun 2012 17:09:29 +0000 (17:09 +0000)]
Allow targets to inject passes before the virtual register rewriter.

Such passes can be used to tweak the register assignments in a
target-dependent way, for example to avoid write-after-write
dependencies.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159209 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoIntegersSubsetTest: Due to compilation failure with -std=c11, replaced -1UL with...
Stepan Dyatkovskiy [Tue, 26 Jun 2012 16:08:20 +0000 (16:08 +0000)]
IntegersSubsetTest: Due to compilation failure with -std=c11, replaced -1UL with NOT_A_NUMBER constant (0xffff).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159207 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThere are a number of generic inline asm operand modifiers that
Jack Carter [Tue, 26 Jun 2012 13:49:27 +0000 (13:49 +0000)]
There are a number of generic inline asm operand modifiers that
up to r158925 were handled as processor specific. Making them
generic and putting tests for these modifiers in the CodeGen/Generic
directory caused a number of targets to fail.

This commit addresses that problem by having the targets call
the generic routine for generic modifiers that they don't currently
have explicit code for.

For now only generic print operands 'c' and 'n' are supported.vi

Affected files:

    test/CodeGen/Generic/asm-large-immediate.ll
    lib/Target/PowerPC/PPCAsmPrinter.cpp
    lib/Target/NVPTX/NVPTXAsmPrinter.cpp
    lib/Target/ARM/ARMAsmPrinter.cpp
    lib/Target/XCore/XCoreAsmPrinter.cpp
    lib/Target/X86/X86AsmPrinter.cpp
    lib/Target/Hexagon/HexagonAsmPrinter.cpp
    lib/Target/CellSPU/SPUAsmPrinter.cpp
    lib/Target/Sparc/SparcAsmPrinter.cpp
    lib/Target/MBlaze/MBlazeAsmPrinter.cpp
    lib/Target/Mips/MipsAsmPrinter.cpp

MSP430 isn't represented because it did not even run with
the long existing 'c' modifier and it was not apparent what
needs to be done to get it inline asm ready.

Contributer: Jack Carter

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159203 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoReplacing zero-sized alloca's with a null pointer is too aggressive, instead
Duncan Sands [Tue, 26 Jun 2012 13:39:21 +0000 (13:39 +0000)]
Replacing zero-sized alloca's with a null pointer is too aggressive, instead
merge all zero-sized alloca's into one, fixing c43204g from the Ada ACATS
conformance testsuite.  What happened there was that a variable sized object
was being allocated on the stack, "alloca i8, i32 %size".  It was then being
passed to another function, which tested that the address was not null (raising
an exception if it was) then manipulated %size bytes in it (load and/or store).
The optimizers cleverly managed to deduce that %size was zero (congratulations
to them, as it isn't at all obvious), which made the alloca zero size, causing
the optimizers to replace it with null, which then caused the check mentioned
above to fail, and the exception to be raised, wrongly.  Note that no loads
and stores were actually being done to the alloca (the loop that does them is
executed %size times, i.e. is not executed), only the not-null address check.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159202 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoIntegersSubsetMapping: implemented "diff" operation. Operation allows at the same...
Stepan Dyatkovskiy [Tue, 26 Jun 2012 11:57:43 +0000 (11:57 +0000)]
IntegersSubsetMapping: implemented "diff" operation. Operation allows at the same time perform up to three operations:
- LHS exclude RHS
- LHS intersect RHS (LHS successors will keeped)
- RHS exclude LHS
The complexity is N+M, where
  N is size of LHS
  M is size of RHS.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159201 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoIntegersSubsetMapping: removed exclude operation, it will replaced with more universa...
Stepan Dyatkovskiy [Tue, 26 Jun 2012 11:41:47 +0000 (11:41 +0000)]
IntegersSubsetMapping: removed exclude operation, it will replaced with more universal "diff" operation in next commit.
Changes was separated onto two commits for better readability.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159200 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSphyinxify the Bugpoint document.
Bill Wendling [Tue, 26 Jun 2012 11:37:00 +0000 (11:37 +0000)]
Sphyinxify the Bugpoint document.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159199 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemoved unused variable
Elena Demikhovsky [Tue, 26 Jun 2012 10:50:07 +0000 (10:50 +0000)]
Removed unused variable

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159197 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRename to match other X86_64* names.
Bill Wendling [Tue, 26 Jun 2012 10:05:06 +0000 (10:05 +0000)]
Rename to match other X86_64* names.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159196 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoShuffle optimization for AVX/AVX2.
Elena Demikhovsky [Tue, 26 Jun 2012 08:04:10 +0000 (08:04 +0000)]
Shuffle optimization for AVX/AVX2.
The current patch optimizes frequently used shuffle patterns and gives these instruction sequence reduction.
Before:
      vshufps $-35, %xmm1, %xmm0, %xmm2 ## xmm2 = xmm0[1,3],xmm1[1,3]
       vpermilps       $-40, %xmm2, %xmm2 ## xmm2 = xmm2[0,2,1,3]
       vextractf128    $1, %ymm1, %xmm1
       vextractf128    $1, %ymm0, %xmm0
       vshufps $-35, %xmm1, %xmm0, %xmm0 ## xmm0 = xmm0[1,3],xmm1[1,3]
       vpermilps       $-40, %xmm0, %xmm0 ## xmm0 = xmm0[0,2,1,3]
       vinsertf128     $1, %xmm0, %ymm2, %ymm0
After:
      vshufps $13, %ymm0, %ymm1, %ymm1 ## ymm1 = ymm1[1,3],ymm0[0,0],ymm1[5,7],ymm0[4,4]
      vshufps $13, %ymm0, %ymm0, %ymm0 ## ymm0 = ymm0[1,3,0,0,5,7,4,4]
      vunpcklps       %ymm1, %ymm0, %ymm0 ## ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159188 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUpdate a bunch of stale comments that dated from when this folled the
Chandler Carruth [Tue, 26 Jun 2012 05:16:37 +0000 (05:16 +0000)]
Update a bunch of stale comments that dated from when this folled the
very first (and worst) placement algorithm. These should now more
accurately reflect the reality of the pass.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159185 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove some duplicate instructions that exist only to given different mnemonics for...
Craig Topper [Tue, 26 Jun 2012 04:12:49 +0000 (04:12 +0000)]
Remove some duplicate instructions that exist only to given different mnemonics for the assembler. Use InstAlias instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159184 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoEnable the new LoopInfo algorithm by default.
Andrew Trick [Tue, 26 Jun 2012 04:11:38 +0000 (04:11 +0000)]
Enable the new LoopInfo algorithm by default.

The primary advantage is that loop optimizations will be applied in a
stable order. This helps debugging and unit test creation. It is also
a better overall implementation without pathologically bad performance
on deep functions.

On large functions (llvm-stress --size=200000 | opt -loops)
Before: 0.1263s
After:  0.0225s

On deep functions (after tweaking llvm-stress, thanks Nadav):
Before: 0.2281s
After:  0.0227s

See r158790 for more comments.

The loop tree is now consistently generated in forward order, but loop
passes are applied in reverse order over the program. If we have a
loop optimization that prefers forward order, that can easily be
achieved by adding a different type of LoopPassManager.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159183 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove unnecessary FIXME
Andrew Trick [Tue, 26 Jun 2012 04:11:34 +0000 (04:11 +0000)]
Remove unnecessary FIXME

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159182 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMake sure type is not extended or untyped before create a constant of the type. No...
Evan Cheng [Tue, 26 Jun 2012 01:19:33 +0000 (01:19 +0000)]
Make sure type is not extended or untyped before create a constant of the type. No test case. Found by inspection.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159179 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTypo.
Eric Christopher [Tue, 26 Jun 2012 00:28:15 +0000 (00:28 +0000)]
Typo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159178 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMake some ugly hacks for inline asm operands which name a specific register a bit...
Eli Friedman [Mon, 25 Jun 2012 23:42:33 +0000 (23:42 +0000)]
Make some ugly hacks for inline asm operands which name a specific register a bit more thorough.  PR13196.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159176 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agorevert my previous commit (r159173), since as Eli pointed out, it's perfectly ok...
Nuno Lopes [Mon, 25 Jun 2012 23:26:10 +0000 (23:26 +0000)]
revert my previous commit (r159173), since as Eli pointed out, it's perfectly ok to mark realloc as noalias

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159175 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agodo not set realloc() as NotAlias, since it can return the same pointer. This whole...
Nuno Lopes [Mon, 25 Jun 2012 22:55:50 +0000 (22:55 +0000)]
do not set realloc() as NotAlias, since it can return the same pointer. This whole thing should be upgraded to use the MemoryBuiltin interface anyway..

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159173 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoARM: update peephole optimization.
Manman Ren [Mon, 25 Jun 2012 21:49:38 +0000 (21:49 +0000)]
ARM: update peephole optimization.

More condition codes are included when deciding whether to remove cmp after
a sub instruction. Specifically, we extend from GE|LT|GT|LE to
GE|LT|GT|LE|HS|LS|HI|LO|EQ|NE. If we have "sub a, b; cmp b, a; movhs", we
should be able to replace with "sub a, b; movls".

rdar: 11725965

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159166 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDefine DAGOperand, an empty base class for RegisterClass and Operand. This allows...
Owen Anderson [Mon, 25 Jun 2012 21:25:16 +0000 (21:25 +0000)]
Define DAGOperand, an empty base class for RegisterClass and Operand.  This allows one to write multiclasses that are polymorphic over both registers and non-register operands.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159162 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix the objc_autoreleasedReturnValue optimization code to locate
Dan Gohman [Mon, 25 Jun 2012 19:47:37 +0000 (19:47 +0000)]
Fix the objc_autoreleasedReturnValue optimization code to locate
the call correctly even in the case where it is an invoke. This
fixes rdar://11714057.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159157 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoEnforce stricter liveness rules for PHIs.
Jakob Stoklund Olesen [Mon, 25 Jun 2012 18:18:27 +0000 (18:18 +0000)]
Enforce stricter liveness rules for PHIs.

Verify that all paths from the entry block to a virtual register read
pass through a def. Enable this check even when MRI->isSSA() is false.

Verify that the live range of a virtual register is live out of all
predecessor blocks, even for PHI-values.

This requires that PHIElimination sometimes inserts IMPLICIT_DEF
instruction in predecessor blocks.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159150 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRun ProcessImplicitDefs on SSA form where it can be much simpler.
Jakob Stoklund Olesen [Mon, 25 Jun 2012 18:12:18 +0000 (18:12 +0000)]
Run ProcessImplicitDefs on SSA form where it can be much simpler.

Implicitly defined virtual registers can simply have the <undef> bit set
on all uses, and copies can be turned into implicit defs recursively.

Physical registers are a bit trickier. We handle the common case where a
physreg def is used by a nearby instruction in the same basic block. For
more complicated cases, just leave the IMPLICIT_DEF instruction in.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159149 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoimprove optimization of invoke instructions:
Nuno Lopes [Mon, 25 Jun 2012 17:11:47 +0000 (17:11 +0000)]
improve optimization of invoke instructions:
 - simplifycfg:  invoke undef/null -> unreachable
 - instcombine:  invoke new  -> invoke expect(0, 0)  (an arbitrary NOOP intrinsic;  only done if the allocated memory is unused, of course)
 - verifier:  allow invoke of intrinsics  (to make the previous step work)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159146 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agocheck for the NoAlias attribute through CallSite
Nuno Lopes [Mon, 25 Jun 2012 16:17:54 +0000 (16:17 +0000)]
check for the NoAlias attribute through CallSite

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159145 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoadd CallSite/CallInst/InvokeInst::hasFnAttr()
Nuno Lopes [Mon, 25 Jun 2012 16:16:58 +0000 (16:16 +0000)]
add CallSite/CallInst/InvokeInst::hasFnAttr()

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159144 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoPR13013: ELF Type identification fails for MSB type ELF files.
Meador Inge [Mon, 25 Jun 2012 14:48:43 +0000 (14:48 +0000)]
PR13013: ELF Type identification fails for MSB type ELF files.

Fix 'sys::IdentifyFileType' to work with big and little endian byte orderings
when reading the ELF object file type.

Initial patch by Stefan Hepp.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159138 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoIf a constant or a function has linkonce_odr linkage and unnamed_addr, mark it
Rafael Espindola [Mon, 25 Jun 2012 14:30:31 +0000 (14:30 +0000)]
If a constant or a function has linkonce_odr linkage and unnamed_addr, mark it
hidden. Being linkonce_odr guarantees that it is available in every dso that
needs it. Being a constant/function with unnamed_addr guarantees that the
copies don't have to be merged.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159136 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThe name (and comment describing) of llvm::GetFirstDebuigLocInBasicBlock no longer...
Eli Bendersky [Mon, 25 Jun 2012 10:13:14 +0000 (10:13 +0000)]
The name (and comment describing) of llvm::GetFirstDebuigLocInBasicBlock no longer represents what the function does. Therefore, the function is removed and its functionality is folded into the only place in the code-base where it was being used.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159133 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd SSE2 predicate to CVTPS2PD instructions. Doesn't matter much because there are...
Craig Topper [Mon, 25 Jun 2012 06:51:42 +0000 (06:51 +0000)]
Add SSE2 predicate to CVTPS2PD instructions. Doesn't matter much because there are no patterns in the instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159127 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove codegen only instruction in favor of one that has the same definition. Make...
Craig Topper [Mon, 25 Jun 2012 06:16:00 +0000 (06:16 +0000)]
Remove codegen only instruction in favor of one that has the same definition. Make some pattern operands more explicit about types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159126 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoJust remove generic support for C++11 alignas -- GCC is already
Chandler Carruth [Mon, 25 Jun 2012 05:20:13 +0000 (05:20 +0000)]
Just remove generic support for C++11 alignas -- GCC is already
advertising complete support w/o alignas implemented, and its
implementation of alignas in the latest versions is so convoluted as to
be unusable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159125 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTeach PHIElimination to handle <undef> operands.
Jakob Stoklund Olesen [Mon, 25 Jun 2012 03:36:12 +0000 (03:36 +0000)]
Teach PHIElimination to handle <undef> operands.

When a PHI use is <undef>, don't emit a copy in the predecessor block,
but insert an IMPLICIT_DEF instruction instead. This ensures that
virtual register uses are always jointly dominated by defs, even if some
of them are IMPLICIT_DEF.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159121 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoHandle <undef> operands in TwoAddressInstructionPass.
Jakob Stoklund Olesen [Mon, 25 Jun 2012 03:27:12 +0000 (03:27 +0000)]
Handle <undef> operands in TwoAddressInstructionPass.

When the source register to a 2-addr instruction is undefined, there is
no need to attempt any transformations - simply replace the source
register with the destination register.

This also comes up when lowering IMPLICIT_DEF instructions - make sure
the <undef> flag is moved to the new partial register def operand:

  %vreg8<def> = INSERT_SUBREG %vreg9<undef>, %vreg0<kill>, sub_16bit
rewrite undef:
  %vreg8<def> = INSERT_SUBREG %vreg8<undef>, %vreg0<kill>, sub_16bit
convert to:
  %vreg8:sub_16bit<def,read-undef> = COPY %vreg0<kill>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159120 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago%RCX is not a function live-out in eh.return functions.
Jakob Stoklund Olesen [Sun, 24 Jun 2012 15:53:01 +0000 (15:53 +0000)]
%RCX is not a function live-out in eh.return functions.

The function live-out registers must be live at all function returns,
and %RCX is only used by eh.return. When a function also has a normal
return, only %RAX holds a return value.

This fixes PR13188.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159116 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agollvm/lib: [CMake] Add explicit dependency to intrinsics_gen.
NAKAMURA Takumi [Sun, 24 Jun 2012 13:32:01 +0000 (13:32 +0000)]
llvm/lib: [CMake] Add explicit dependency to intrinsics_gen.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159112 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAllow controlling vectorization of boolean values separately from other integer types.
Hal Finkel [Sun, 24 Jun 2012 13:28:01 +0000 (13:28 +0000)]
Allow controlling vectorization of boolean values separately from other integer types.

These are used as the result of comparisons, and often handled differently from larger integer types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159111 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove dyn_cast + dereference pattern by replacing it with a cast and changing
Nick Lewycky [Sun, 24 Jun 2012 10:15:42 +0000 (10:15 +0000)]
Remove dyn_cast + dereference pattern by replacing it with a cast and changing
the safety check to look for the same type we're going to actually cast to.
Fixes PR13180!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@159110 91177308-0d34-0410-b5e6-96231b3b80d8