oota-llvm.git
11 years agoClaim ownership.
Chad Rosier [Wed, 14 Nov 2012 18:43:21 +0000 (18:43 +0000)]
Claim ownership.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167943 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoFix the largest offender of determinism in BBVectorize
Hal Finkel [Wed, 14 Nov 2012 18:38:11 +0000 (18:38 +0000)]
Fix the largest offender of determinism in BBVectorize

Iterating over the children of each node in the potential vectorization
plan must happen in a deterministic order (because it affects which children
are erased when two children conflict). There was no need for this data
structure to be a map in the first place, so replacing it with a vector
is a small change.

I believe that this was the last remaining instance if iterating over the
elements of a Dense* container where the iteration order could matter.
There are some remaining iterations over std::*map containers where the order
might matter, but so long as the Value* for instructions in a block increase
with the order of the instructions in the block (or decrease) monotonically,
then this will appear to be deterministic.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167942 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoSort the code owner list alphabetically. Add myself as lib/DebugInfo owner.
Benjamin Kramer [Wed, 14 Nov 2012 18:17:45 +0000 (18:17 +0000)]
Sort the code owner list alphabetically. Add myself as lib/DebugInfo owner.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167940 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoX86: Better diagnostics for 32-bit vs. 64-bit mode mismatches.
Jim Grosbach [Wed, 14 Nov 2012 18:04:47 +0000 (18:04 +0000)]
X86: Better diagnostics for 32-bit vs. 64-bit mode mismatches.

When an instruction as written requires 32-bit mode and we're assembling
in 64-bit mode, or vice-versa, issue a more specific diagnostic about
what's wrong.

rdar://12700702

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167937 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agos/assert/llvm_unreachable/
Matt Beaumont-Gay [Wed, 14 Nov 2012 17:58:11 +0000 (17:58 +0000)]
s/assert/llvm_unreachable/

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167936 91177308-0d34-0410-b5e6-96231b3b80d8

11 years agoGrab debug information for code ownership.
Eric Christopher [Wed, 14 Nov 2012 17:33:40 +0000 (17:33 +0000)]
Grab debug information for code ownership.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167933 91177308-0d34-0410-b5e6-96231b3b80d8

11 years ago[TSan] fix indentation
Alexey Samsonov [Wed, 14 Nov 2012 14:33:59 +0000 (14:33 +0000)]
[TSan] fix indentation

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167928 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoEmit relocations from .debug_aranges to .debug_info for asm files
Alexey Samsonov [Wed, 14 Nov 2012 09:55:38 +0000 (09:55 +0000)]
Emit relocations from .debug_aranges to .debug_info for asm files

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167926 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRevert some redundant parts of r142605.
Patrik Hägglund [Wed, 14 Nov 2012 09:04:56 +0000 (09:04 +0000)]
Revert some redundant parts of r142605.

This seems like redundant leftovers from r142288 - exposing
TargetData::parseSpecifier to LLParser - which got reverted. Removes
redunant td != NULL checks in parseSpecifier, and simplifies the
interface to parseSpecifier and init.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167924 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSet FFLOOR of vectors to expand to keep intruction selection from failing.
Craig Topper [Wed, 14 Nov 2012 08:11:25 +0000 (08:11 +0000)]
Set FFLOOR of vectors to expand to keep intruction selection from failing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167922 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMark myself as owner of BBVectorize and PowerPC
Hal Finkel [Wed, 14 Nov 2012 07:59:10 +0000 (07:59 +0000)]
Mark myself as owner of BBVectorize and PowerPC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167921 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAccepting ownership of scheduling.
Andrew Trick [Wed, 14 Nov 2012 07:04:23 +0000 (07:04 +0000)]
Accepting ownership of scheduling.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167917 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFactor out an overly replicated typecast. No functional change.
Craig Topper [Wed, 14 Nov 2012 06:41:09 +0000 (06:41 +0000)]
Factor out an overly replicated typecast. No functional change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167916 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSet FFLOOR for vectors to expand on CellSPU to keep instruction selection from failin...
Craig Topper [Wed, 14 Nov 2012 05:56:30 +0000 (05:56 +0000)]
Set FFLOOR for vectors to expand on CellSPU to keep instruction selection from failing on llvm.floor of a vector.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167914 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd newlines to end of debug messages.
Craig Topper [Wed, 14 Nov 2012 05:20:09 +0000 (05:20 +0000)]
Add newlines to end of debug messages.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167913 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoHandle DAG CSE adding new uses during ReplaceAllUsesWith. Fixes PR14333.
Rafael Espindola [Wed, 14 Nov 2012 05:08:56 +0000 (05:08 +0000)]
Handle DAG CSE adding new uses during ReplaceAllUsesWith. Fixes PR14333.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167912 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agodocs: Fix Sphinx toctree warning.
Sean Silva [Wed, 14 Nov 2012 01:49:23 +0000 (01:49 +0000)]
docs: Fix Sphinx toctree warning.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167905 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUse TARGET2 relocation for TType references on ARM.
Anton Korobeynikov [Wed, 14 Nov 2012 01:47:00 +0000 (01:47 +0000)]
Use TARGET2 relocation for TType references on ARM.
Do some cleanup of the code while here.

Inspired by patch by Logan Chien!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167904 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agodocs: chmod -x HowToUseInstrMappings.rst
Sean Silva [Wed, 14 Nov 2012 01:41:58 +0000 (01:41 +0000)]
docs: chmod -x HowToUseInstrMappings.rst

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167903 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix broken asserts. Also, spell 'indices' correctly.
Matt Beaumont-Gay [Wed, 14 Nov 2012 00:21:27 +0000 (00:21 +0000)]
Fix broken asserts. Also, spell 'indices' correctly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167894 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[Object] Fix endianess bug by refactoring Archive::Symbol::getMember.
Michael J. Spencer [Wed, 14 Nov 2012 00:04:13 +0000 (00:04 +0000)]
[Object] Fix endianess bug by refactoring Archive::Symbol::getMember.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167893 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agouse the getSplat API. Patch by Paul Redmond.
Nadav Rotem [Wed, 14 Nov 2012 00:02:13 +0000 (00:02 +0000)]
use the getSplat API. Patch by Paul Redmond.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167892 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRevert "Use the 'count' attribute instead of the 'upper_bound' attribute."
Eric Christopher [Tue, 13 Nov 2012 23:30:43 +0000 (23:30 +0000)]
Revert "Use the 'count' attribute instead of the 'upper_bound' attribute."
temporarily as it is breaking the gdb bots.

This reverts commit r167806/e7ff4c14b157746b3e0228d2dce9f70712d1c126.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167886 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAcknowledge code ownership of MC.
Jim Grosbach [Tue, 13 Nov 2012 23:01:35 +0000 (23:01 +0000)]
Acknowledge code ownership of MC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167882 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAssignment of Bitcode ownership to Joe Abbey, after announcing proposal on
Joe Abbey [Tue, 13 Nov 2012 22:32:15 +0000 (22:32 +0000)]
Assignment of Bitcode ownership to Joe Abbey, after announcing proposal on
LLVMdev and not hearing any major objections.  Although it did spark a nice
discussion regarding what it means to own something in LLVM.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167881 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRevert -join-splitedges to a boolean cmd line option.
Andrew Trick [Tue, 13 Nov 2012 22:19:48 +0000 (22:19 +0000)]
Revert -join-splitedges to a boolean cmd line option.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167880 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThe MachineScheduler does not currently require JoinSplitEdges.
Andrew Trick [Tue, 13 Nov 2012 22:15:40 +0000 (22:15 +0000)]
The MachineScheduler does not currently require JoinSplitEdges.

This option will eventually either be enabled unconditionally or
replaced by a more general live range splitting optimization.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167879 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[MC][COFF] Emit weak symbols to the correct section. Patch by Dmitry Puzirev!
Michael J. Spencer [Tue, 13 Nov 2012 22:04:09 +0000 (22:04 +0000)]
[MC][COFF] Emit weak symbols to the correct section. Patch by Dmitry Puzirev!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167877 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRevert r167836, "llvm/test/Other/close-stderr.ll: Mark it as XFAIL:mingw32 for now...
NAKAMURA Takumi [Tue, 13 Nov 2012 21:57:42 +0000 (21:57 +0000)]
Revert r167836, "llvm/test/Other/close-stderr.ll: Mark it as XFAIL:mingw32 for now.", corresponding to r167849.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167876 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd test case to verify correct relocs being generated for
Ulrich Weigand [Tue, 13 Nov 2012 21:53:43 +0000 (21:53 +0000)]
Add test case to verify correct relocs being generated for
TLS symbols on PowerPC using the integrated assembler.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167875 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agonumerically sort the symbols, so that the testcase result is uniform
Shankar Easwaran [Tue, 13 Nov 2012 21:01:11 +0000 (21:01 +0000)]
numerically sort the symbols, so that the testcase result is uniform

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167872 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agollvm-nm: Make sort more stable when symbol names are equal.
Daniel Dunbar [Tue, 13 Nov 2012 19:39:55 +0000 (19:39 +0000)]
llvm-nm: Make sort more stable when symbol names are equal.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167866 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd (some) PowerPC TLS relocation types to ELF.h and
Ulrich Weigand [Tue, 13 Nov 2012 19:24:36 +0000 (19:24 +0000)]
Add (some) PowerPC TLS relocation types to ELF.h and
generate them from PPCELFObjectWriter::getRelocTypeInner
as appropriate.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167864 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix wrong PowerPC instruction opcodes for:
Ulrich Weigand [Tue, 13 Nov 2012 19:21:31 +0000 (19:21 +0000)]
Fix wrong PowerPC instruction opcodes for:
 - lwaux
 - lhzux
 - stbu

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167863 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix wrong PowerPC instruction encodings due to
Ulrich Weigand [Tue, 13 Nov 2012 19:19:46 +0000 (19:19 +0000)]
Fix wrong PowerPC instruction encodings due to
operand field name mismatches in:
 - AForm_3  (fmul, fmuls)
 - XFXForm_5 (mtcrf)
 - XFLForm (mtfsf)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167862 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix instruction encoding for "bd(n)z" on PowerPC,
Ulrich Weigand [Tue, 13 Nov 2012 19:15:52 +0000 (19:15 +0000)]
Fix instruction encoding for "bd(n)z" on PowerPC,
by using a new instruction format BForm_1.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167861 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix instruction encoding for "isel" on PowerPC,
Ulrich Weigand [Tue, 13 Nov 2012 19:14:19 +0000 (19:14 +0000)]
Fix instruction encoding for "isel" on PowerPC,
using a new instruction format AForm_4.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167860 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoX86: when constructing VZEXT_LOAD from other loads, makes sure its output
Manman Ren [Tue, 13 Nov 2012 19:13:05 +0000 (19:13 +0000)]
X86: when constructing VZEXT_LOAD from other loads, makes sure its output
chain is correctly setup.

As an example, if the original load must happen before later stores, we need
to make sure the constructed VZEXT_LOAD is constrained to be before the stores.

rdar://12684358

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167859 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDo not consider a machine instruction that uses and defines the same
Ulrich Weigand [Tue, 13 Nov 2012 18:40:58 +0000 (18:40 +0000)]
Do not consider a machine instruction that uses and defines the same
physical register as candidate for common subexpression elimination
in MachineCSE.

This fixes a bug on PowerPC in MultiSource/Applications/oggenc/oggenc
caused by MachineCSE invalidly merging two separate DYNALLOC insns.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167855 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdding changes to support GNU style archive library reading
Shankar Easwaran [Tue, 13 Nov 2012 18:38:42 +0000 (18:38 +0000)]
Adding changes to support GNU style archive library reading

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167853 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdding convenience function to ELF Header
Shankar Easwaran [Tue, 13 Nov 2012 18:26:22 +0000 (18:26 +0000)]
Adding convenience function to ELF Header

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167852 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix -join-splitedges: my previous "cleanup" broke it.
Andrew Trick [Tue, 13 Nov 2012 17:37:46 +0000 (17:37 +0000)]
Fix -join-splitedges: my previous "cleanup" broke it.

Working on reducing unit tests.
This won't be enabled unless a subtarget enables misched.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167851 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRevert 167755/167760. We don't want to emit crash diagnostics on command-line syntax...
Chad Rosier [Tue, 13 Nov 2012 16:42:19 +0000 (16:42 +0000)]
Revert 167755/167760. We don't want to emit crash diagnostics on command-line syntax errors.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167849 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agollvm/test/Other/close-stderr.ll: Mark it as XFAIL:mingw32 for now.
NAKAMURA Takumi [Tue, 13 Nov 2012 15:03:33 +0000 (15:03 +0000)]
llvm/test/Other/close-stderr.ll: Mark it as XFAIL:mingw32 for now.

On MSYS, 70 is not seen, but 1.

r127726 should be reworked. Candidate options are;

  1) Use not exit(70), but _exit(70), in report_fatal_error().
  2) Return with _exit(70) in ~raw_ostream().

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167836 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCodegen support for arbitrary vector getelementptrs.
Duncan Sands [Tue, 13 Nov 2012 13:01:58 +0000 (13:01 +0000)]
Codegen support for arbitrary vector getelementptrs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167830 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix the instcombine GEP index widening transform to work correctly for vector
Duncan Sands [Tue, 13 Nov 2012 13:01:00 +0000 (13:01 +0000)]
Fix the instcombine GEP index widening transform to work correctly for vector
getelementptrs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167829 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRelax the restrictions on vector of pointer types, and vector getelementptr.
Duncan Sands [Tue, 13 Nov 2012 12:59:33 +0000 (12:59 +0000)]
Relax the restrictions on vector of pointer types, and vector getelementptr.
Previously in a vector of pointers, the pointer couldn't be any pointer type,
it had to be a pointer to an integer or floating point type.  This is a hassle
for dragonegg because the GCC vectorizer happily produces vectors of pointers
where the pointer is a pointer to a struct or whatever.  Vector getelementptr
was restricted to just one index, but now that vectors of pointers can have
any pointer type it is more natural to allow arbitrary vector getelementptrs.
There is however the issue of struct GEPs, where if each lane chose different
struct fields then from that point on each lane will be working down into
unrelated types.  This seems like too much pain for too little gain, so when
you have a vector struct index all the elements are required to be the same.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167828 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDependenceAnalysis: Print all dependency pairs when dumping. Update all testcases.
Benjamin Kramer [Tue, 13 Nov 2012 12:12:02 +0000 (12:12 +0000)]
DependenceAnalysis: Print all dependency pairs when dumping. Update all testcases.

Part of a patch by Preston Briggs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167827 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agomisched: Allow subtargets to enable misched and dependent options.
Andrew Trick [Tue, 13 Nov 2012 08:47:29 +0000 (08:47 +0000)]
misched: Allow subtargets to enable misched and dependent options.

This allows me to begin enabling (or backing out) misched by default
for one subtarget at a time. To run misched we typically want to:
- Disable SelectionDAG scheduling (use the source order scheduler)
- Enable more aggressive coalescing (until we decide to always run the coalescer this way)
- Enable MachineScheduler pass itself.

Disabling PostRA sched may follow for some subtargets.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167826 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdded RegisterCoalescer support for joining global copies first.
Andrew Trick [Tue, 13 Nov 2012 08:47:25 +0000 (08:47 +0000)]
Added RegisterCoalescer support for joining global copies first.

This adds the -join-globalcopies option which can be enabled by
default once misched is also enabled.

Ideally, the register coalescer would be able to split local live
ranges in a way that produces copies that can be easily resolved by
the scheduler. Until then, this heuristic should be good enough to at
least allow the scheduler to run after coalescing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167825 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFigure out <size> argument of llvm.lifetime intrinsics at the moment they are created...
Alexey Samsonov [Tue, 13 Nov 2012 07:15:32 +0000 (07:15 +0000)]
Figure out <size> argument of llvm.lifetime intrinsics at the moment they are created (during function inlining)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167821 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTest commit.
Jyotsna Verma [Tue, 13 Nov 2012 06:31:55 +0000 (06:31 +0000)]
Test commit.
Add a blank line.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167819 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoBBVectorize: Remove temporary assert used for debugging
Hal Finkel [Tue, 13 Nov 2012 05:54:54 +0000 (05:54 +0000)]
BBVectorize: Remove temporary assert used for debugging

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167817 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoinstcombine: Migrate math library call simplifications
Meador Inge [Tue, 13 Nov 2012 04:16:17 +0000 (04:16 +0000)]
instcombine: Migrate math library call simplifications

This patch migrates the math library call simplifications from the
simplify-libcalls pass into the instcombine library call simplifier.

I have typically migrated just one simplifier at a time, but the math
simplifiers are interdependent because:

   1. CosOpt, PowOpt, and Exp2Opt all depend on UnaryDoubleFPOpt.
   2. CosOpt, PowOpt, Exp2Opt, and UnaryDoubleFPOpt all depend on
      the option -enable-double-float-shrink.

These two factors made migrating each of these simplifiers individually
more of a pain than it would be worth.  So, I migrated them all together.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167815 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove virtual keyword for two routines that should never be overridden.
Evan Cheng [Tue, 13 Nov 2012 03:14:16 +0000 (03:14 +0000)]
Remove virtual keyword for two routines that should never be overridden.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167812 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoBBVectorize: Don't vectorize vector-manipulation chains
Hal Finkel [Tue, 13 Nov 2012 03:12:40 +0000 (03:12 +0000)]
BBVectorize: Don't vectorize vector-manipulation chains

Don't choose a vectorization plan containing only shuffles and
vector inserts/extracts. Due to inperfections in the cost model,
these can lead to infinite recusion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167811 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRevert r167759. Ben is right this isn't likely to help much.
Evan Cheng [Tue, 13 Nov 2012 02:56:38 +0000 (02:56 +0000)]
Revert r167759. Ben is right this isn't likely to help much.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167809 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agomisched: Don't consider artificial edges weak edges.
Andrew Trick [Tue, 13 Nov 2012 02:35:06 +0000 (02:35 +0000)]
misched: Don't consider artificial edges weak edges.

For now be more conservative in case other out-of-tree schedulers rely
on the old behavior of artificial edges.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167808 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUse the 'count' attribute instead of the 'upper_bound' attribute.
Bill Wendling [Tue, 13 Nov 2012 02:31:47 +0000 (02:31 +0000)]
Use the 'count' attribute instead of the 'upper_bound' attribute.

If we have a type 'int a[1]' and a type 'int b[0]', the generated DWARF is the
same for both of them because we use the 'upper_bound' attribute. Instead use
the 'count' attrbute, which gives the correct number of elements in the array.
<rdar://problem/12566646>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167806 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCleanup the main RegisterCoalescer loop.
Andrew Trick [Tue, 13 Nov 2012 00:34:44 +0000 (00:34 +0000)]
Cleanup the main RegisterCoalescer loop.

Block priorities still apply outside loops.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167793 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agorevert r167740
Shuxin Yang [Tue, 13 Nov 2012 00:08:49 +0000 (00:08 +0000)]
revert r167740

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167787 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCleanup -join-splitedges. Make the loop more obvious.
Andrew Trick [Mon, 12 Nov 2012 23:59:48 +0000 (23:59 +0000)]
Cleanup -join-splitedges. Make the loop more obvious.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167785 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoBBVectorize: Only some insert element operand pairs are free.
Hal Finkel [Mon, 12 Nov 2012 23:55:36 +0000 (23:55 +0000)]
BBVectorize: Only some insert element operand pairs are free.

This fixes another infinite recursion case when using target costs.
We can only replace insert element input chains that are pure (end
with inserting into an undef).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167784 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUpdate link to external document
Jean-Daniel Dupas [Mon, 12 Nov 2012 23:43:34 +0000 (23:43 +0000)]
Update link to external document

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167781 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix test case added in patch fixing PR14314
Michael Liao [Mon, 12 Nov 2012 22:33:18 +0000 (22:33 +0000)]
Fix test case added in patch fixing PR14314

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167769 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd an option to enable prototype "fission" capabilities and debug changes.
Eric Christopher [Mon, 12 Nov 2012 22:22:20 +0000 (22:22 +0000)]
Add an option to enable prototype "fission" capabilities and debug changes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167765 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUpdate test case for r167754/r167755.
Chad Rosier [Mon, 12 Nov 2012 21:51:08 +0000 (21:51 +0000)]
Update test case for r167754/r167755.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167760 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCache size of PassVector to speed up getNumContainedPasses().
Evan Cheng [Mon, 12 Nov 2012 21:42:53 +0000 (21:42 +0000)]
Cache size of PassVector to speed up getNumContainedPasses().
getNumContainedPasses() used to compute the size of the vector on demand. It is
called repeated in loops (such as runOnFunction()) and it can be updated while
inside the loop.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167759 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdded a temporary option to avoid critical edges splitting.
Andrew Trick [Mon, 12 Nov 2012 21:42:40 +0000 (21:42 +0000)]
Added a temporary option to avoid critical edges splitting.

This teaches the register coalescer to be less prone to split critical
edges. I am currently benchmarking this with the new (post-coalescer)
scheduler. I plan to enable this by default and remove the option as
soon as misched is enabled.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167758 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRewrite DIContext interface to take an object. Update all callers.
Eric Christopher [Mon, 12 Nov 2012 21:40:38 +0000 (21:40 +0000)]
Rewrite DIContext interface to take an object. Update all callers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167757 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRevert r167620; this can be implemented using an existing CL option.
Chad Rosier [Mon, 12 Nov 2012 21:32:44 +0000 (21:32 +0000)]
Revert r167620; this can be implemented using an existing CL option.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167755 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agomisched: rename interfaceto avoid gcc warnings
Andrew Trick [Mon, 12 Nov 2012 21:28:10 +0000 (21:28 +0000)]
misched: rename interfaceto avoid gcc warnings

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167753 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoBBVectorize: Use a more sophisticated check for input cost
Hal Finkel [Mon, 12 Nov 2012 21:21:02 +0000 (21:21 +0000)]
BBVectorize: Use a more sophisticated check for input cost

The old checking code, which assumed that input shuffles and insert-elements
could always be folded (and thus were free) is too simple.
This can only happen in special circumstances.
Using the simple check caused infinite recursion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167750 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agomisched: Target-independent support for MacroFusion.
Andrew Trick [Mon, 12 Nov 2012 19:52:20 +0000 (19:52 +0000)]
misched: Target-independent support for MacroFusion.

Uses the infrastructure from r167742 to support clustering instructure
that the target processor can "fuse". e.g. cmp+jmp.

Next step: target hook implementations with test cases, and enable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167744 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoBBVectorize: Check the types of compare instructions
Hal Finkel [Mon, 12 Nov 2012 19:41:38 +0000 (19:41 +0000)]
BBVectorize: Check the types of compare instructions

The pass would previously assert when trying to compute the cost of
compare instructions with illegal vector types (like struct pointers).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167743 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agomisched: Target-independent support for load/store clustering.
Andrew Trick [Mon, 12 Nov 2012 19:40:10 +0000 (19:40 +0000)]
misched: Target-independent support for load/store clustering.

This infrastructure is generally useful for any target that wants to
strongly prefer two instructions to be adjacent after scheduling.

A following checkin will add target-specific hooks with unit
tests. Then this feature will be enabled by default with misched.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167742 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThis change is to fix rdar://12571717 which is about assertion in Reassociate pass.
Shuxin Yang [Mon, 12 Nov 2012 19:34:11 +0000 (19:34 +0000)]
This change is to fix rdar://12571717 which is about assertion in Reassociate pass.

The assertion is trigged when the Reassociater tries to transform expression
     ... + 2 * n * 3 + 2 * m + ...
  into:
     ... + 2 * (n*3 + m).

In the process of the transformation, a helper routine folds the constant 2*3 into 6,
confusing optimizer which is trying the to eliminate the common factor 2, and cannot
find 2 any more.

Review is pending. But I'd like commit first in order to help those who are waiting
for this fix.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167740 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agomisched: Infrastructure for weak DAG edges.
Andrew Trick [Mon, 12 Nov 2012 19:28:57 +0000 (19:28 +0000)]
misched: Infrastructure for weak DAG edges.

This adds support for weak DAG edges to the general scheduling
infrastructure in preparation for MachineScheduler support for
heuristics based on weak edges.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167738 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMake TOC order deterministic by using MapVector instead of DenseMap.
Ulrich Weigand [Mon, 12 Nov 2012 19:13:24 +0000 (19:13 +0000)]
Make TOC order deterministic by using MapVector instead of DenseMap.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167737 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agofix a spelling mistake
Nadav Rotem [Mon, 12 Nov 2012 18:45:12 +0000 (18:45 +0000)]
fix a spelling mistake

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167734 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoBBVectorize: Check the input types of shuffles for legality
Hal Finkel [Mon, 12 Nov 2012 14:50:59 +0000 (14:50 +0000)]
BBVectorize: Check the input types of shuffles for legality

This fixes a bug where shuffles were being fused such that the
resulting input types were not legal on the target. This would
occur only when both inputs and dependencies were also foldable
operations (such as other shuffles) and there were other connected
pairs in the same block.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167731 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDon't use __cxa_demangle under MSVC (which doesn't have it)
Alexander Potapenko [Mon, 12 Nov 2012 14:49:58 +0000 (14:49 +0000)]
Don't use __cxa_demangle under MSVC (which doesn't have it)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167730 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[ASan] fixup for r167725: Don't fetch name of StructType if it is literal
Alexey Samsonov [Mon, 12 Nov 2012 14:47:00 +0000 (14:47 +0000)]
[ASan] fixup for r167725: Don't fetch name of StructType if it is literal

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167729 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFixup for r167558: Store raw pointer (instead of reference) to RelocMap in DIContext...
Alexey Samsonov [Mon, 12 Nov 2012 14:25:36 +0000 (14:25 +0000)]
Fixup for r167558: Store raw pointer (instead of reference) to RelocMap in DIContext. This is needed to prevent crashes because of dangling reference if the clients don't provide RelocMap to DIContext constructor.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167728 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoNormalize memcmp constant folding results.
Meador Inge [Mon, 12 Nov 2012 14:00:45 +0000 (14:00 +0000)]
Normalize memcmp constant folding results.

The library call simplifier folds memcmp calls with all constant arguments
to a constant.  For example:

  memcmp("foo", "foo", 3) ->  0
  memcmp("hel", "foo", 3) ->  1
  memcmp("foo", "hel", 3) -> -1

The folding is implemented in terms of the system memcmp that LLVM gets
linked with.  It currently just blindly uses the value returned from
the system memcmp as the folded constant.

This patch normalizes the values returned from the system memcmp to
(-1, 0, 1) so that we get consistent results across multiple platforms.
The test cases were adjusted accordingly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167726 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[ASan]: Add minimalistic support for turning off initialization-order checking for...
Alexey Samsonov [Mon, 12 Nov 2012 14:00:01 +0000 (14:00 +0000)]
[ASan]: Add minimalistic support for turning off initialization-order checking for globals of specified types. Tests for this behavior will go to ASan test suite in compiler-rt.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167725 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agodo not play preprocessor tricks with 'private', use public interfaces instead; this...
Gabor Greif [Mon, 12 Nov 2012 13:34:59 +0000 (13:34 +0000)]
do not play preprocessor tricks with 'private', use public interfaces instead; this appeases the VC++ buildbots

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167724 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[ASan] Add llvm-symbolizer from to tools/
Alexander Potapenko [Mon, 12 Nov 2012 11:33:29 +0000 (11:33 +0000)]
[ASan] Add llvm-symbolizer from to tools/
This is the second and last (2/2) part of a change that moves llvm-symbolizer to llvm/tools/, which will allow to build it
with both cmake and configure+make.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167723 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoadd unit test for waymarking algorithm (Use::getUser)
Gabor Greif [Mon, 12 Nov 2012 10:01:17 +0000 (10:01 +0000)]
add unit test for waymarking algorithm (Use::getUser)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167720 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove unused field.
Eric Christopher [Mon, 12 Nov 2012 07:35:12 +0000 (07:35 +0000)]
Remove unused field.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167719 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix PR14314
Michael Liao [Mon, 12 Nov 2012 06:49:17 +0000 (06:49 +0000)]
Fix PR14314

- Fix operand order for atomic sub, where the minuend is the value
  loaded from memory and the subtrahend is the parameter specified.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167718 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd --enable-werror and --enable-cxx11 to projects/sample/
Craig Topper [Mon, 12 Nov 2012 06:11:12 +0000 (06:11 +0000)]
Add --enable-werror and --enable-cxx11 to projects/sample/

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167716 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[NVPTX] Add more precise PTX/SM target attributes
Justin Holewinski [Mon, 12 Nov 2012 03:16:43 +0000 (03:16 +0000)]
[NVPTX] Add more precise PTX/SM target attributes

Each SM and PTX version is modeled as a subtarget feature/CPU. Additionally,
PTX 3.1 is added as the default PTX version to be out-of-the-box compatible
with CUDA 5.0.

Available CPUs for this target:

  sm_10 - Select the sm_10 processor.
  sm_11 - Select the sm_11 processor.
  sm_12 - Select the sm_12 processor.
  sm_13 - Select the sm_13 processor.
  sm_20 - Select the sm_20 processor.
  sm_21 - Select the sm_21 processor.
  sm_30 - Select the sm_30 processor.
  sm_35 - Select the sm_35 processor.

Available features for this target:

  ptx30 - Use PTX version 3.0.
  ptx31 - Use PTX version 3.1.
  sm_10 - Target SM 1.0.
  sm_11 - Target SM 1.1.
  sm_12 - Target SM 1.2.
  sm_13 - Target SM 1.3.
  sm_20 - Target SM 2.0.
  sm_21 - Target SM 2.1.
  sm_30 - Target SM 3.0.
  sm_35 - Target SM 3.5.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167699 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDelete a stale comment. No functional change.
Meador Inge [Mon, 12 Nov 2012 00:28:15 +0000 (00:28 +0000)]
Delete a stale comment.  No functional change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167698 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove some helper methods to being static functions in the implementation file.
Craig Topper [Sun, 11 Nov 2012 22:45:02 +0000 (22:45 +0000)]
Move some helper methods to being static functions in the implementation file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167696 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove hard-coded constant in Transforms/InstCombine/memcmp-1.ll
Meador Inge [Sun, 11 Nov 2012 07:10:25 +0000 (07:10 +0000)]
Remove hard-coded constant in Transforms/InstCombine/memcmp-1.ll

Transforms/InstCombine/memcmp-1.ll has a test case that looks like:

  @foo = constant [4 x i8] c"foo\00"
  @hel = constant [4 x i8] c"hel\00"

  ...

  %mem1 = getelementptr [4 x i8]* @hel, i32 0, i32 0
  %mem2 = getelementptr [4 x i8]* @foo, i32 0, i32 0
  %ret = call i32 @memcmp(i8* %mem1, i8* %mem2, i32 3)
  ret i32 %ret
  ; CHECK: ret i32 2

The folded return value (2 above) is computed using the system memcmp
that the compiler is linked with.  This can return different values on
different systems.  The test was originally written on an OS X 10.7.5
x86-64 box and passed.  However, it failed on one of the x86-64 FreeBSD
buildbots because the system memcpy on that machine returned a different
value (1 instead of 2).

I fixed the test by checking the folding constants with regexes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167691 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoinstcombine: Migrate memset optimizations
Meador Inge [Sun, 11 Nov 2012 06:49:03 +0000 (06:49 +0000)]
instcombine: Migrate memset optimizations

This patch migrates the memset optimizations from the simplify-libcalls
pass into the instcombine library call simplifier.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167689 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUpdate the vectorizer docs.
Nadav Rotem [Sun, 11 Nov 2012 06:47:51 +0000 (06:47 +0000)]
Update the vectorizer docs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167688 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoinstcombine: Migrate memmove optimizations
Meador Inge [Sun, 11 Nov 2012 06:22:40 +0000 (06:22 +0000)]
instcombine: Migrate memmove optimizations

This patch migrates the memmove optimizations from the simplify-libcalls
pass into the instcombine library call simplifier.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167687 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoinstcombine: Migrate memcpy optimizations
Meador Inge [Sun, 11 Nov 2012 05:54:34 +0000 (05:54 +0000)]
instcombine: Migrate memcpy optimizations

This patch migrates the memcpy optimizations from the simplify-libcalls
pass into the instcombine library call simplifier.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167686 91177308-0d34-0410-b5e6-96231b3b80d8