oota-llvm.git
10 years ago[PM] [cleanup] Rearrange the public and private sections of this class
Chandler Carruth [Mon, 2 Dec 2013 12:35:56 +0000 (12:35 +0000)]
[PM] [cleanup] Rearrange the public and private sections of this class
to be a bit more sensible. The public interface now is first followed by
the implementation details.

This also resolves a FIXME to make something private -- it was already
possible as the one special caller was already a friend.

No functionality changed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196095 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoXCoreFrameLowering.cpp: Use [in,out] instead of [in] [out]. [-Wdocumentation]
NAKAMURA Takumi [Mon, 2 Dec 2013 11:31:25 +0000 (11:31 +0000)]
XCoreFrameLowering.cpp: Use [in,out] instead of [in] [out]. [-Wdocumentation]

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196094 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CMake] add_lit_target: Tests should be excluded from "Build Solution".
NAKAMURA Takumi [Mon, 2 Dec 2013 11:31:19 +0000 (11:31 +0000)]
[CMake] add_lit_target: Tests should be excluded from "Build Solution".

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196093 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoXCore target: Make handling of large frames not dependent upon an FP.
Robert Lytton [Mon, 2 Dec 2013 11:05:28 +0000 (11:05 +0000)]
XCore target: Make handling of large frames not dependent upon an FP.

eliminateFrameIndex() has been reworked to handle both small & large frames
with either a FP or SP.
An additional Slot is required for Scavenging spills when not using FP for large frames.
Reworked the handling of Register Scavenging.

Whether we are using an FP or not, whether it is a large frame or not,
and whether we are using a large code model or not are now independent.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196091 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM: add pseudo-instructions for lit-pool global materialisation
Tim Northover [Mon, 2 Dec 2013 10:35:41 +0000 (10:35 +0000)]
ARM: add pseudo-instructions for lit-pool global materialisation

These are used by MachO only at the moment, and (much like the existing
MOVW/MOVT set) work around the fact that the labels used in the actual
instructions often contain PC-dependent components, which means that repeatedly
materialising the same global can't be CSEed.

With small modifications, it could be adapted to how ELF finds the address of
_GLOBAL_OFFSET_TABLE_, which would give similar benefits in PIC mode there.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196090 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoXCore: Unbreak C++11 build.
Benjamin Kramer [Mon, 2 Dec 2013 10:29:26 +0000 (10:29 +0000)]
XCore: Unbreak C++11 build.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196089 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoXCore target: fix large code model 'select' indirect address handling.
Robert Lytton [Mon, 2 Dec 2013 10:18:37 +0000 (10:18 +0000)]
XCore target: fix large code model 'select' indirect address handling.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196088 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoXCore target: Add large code model
Robert Lytton [Mon, 2 Dec 2013 10:18:31 +0000 (10:18 +0000)]
XCore target: Add large code model

When using large code model:
Global objects larger than 'CodeModelLargeSize' bytes are placed in sections named with a trailing ".large"
The folded global address of such objects are lowered into the const pool.

During inspection it was noted that LowerConstantPool() was using a default offset of zero.
A fix was made, but due to only offsets of zero being generated, testing only verifies the change is not detrimental.

Correct the flags emitted for explicitly specified sections.

We assume the size of the object queried by getSectionForConstant() is never greater than CodeModelLargeSize.
To handle greater than CodeModelLargeSize, changes to AsmPrinter would be required.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196087 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoXCore target: extend tests in preparation
Robert Lytton [Mon, 2 Dec 2013 10:18:24 +0000 (10:18 +0000)]
XCore target: extend tests in preparation

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196086 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoXCore target: Fix eliminateFrameIndex() to handle large frames
Robert Lytton [Mon, 2 Dec 2013 10:18:19 +0000 (10:18 +0000)]
XCore target: Fix eliminateFrameIndex() to handle large frames

Large frame offsets are loaded from the ConstantPool.
Where possible, offsets are encoded using the smaller MKMSK instruction.
Large frame offsets can only be used when there is a frame-pointer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196085 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoXCore target: Enable frames larger than 65535 to be lowered
Robert Lytton [Mon, 2 Dec 2013 10:18:14 +0000 (10:18 +0000)]
XCore target: Enable frames larger than 65535 to be lowered

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196084 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove an empty directory left after r194304, which incompletely reverted r194218
Alexander Kornienko [Mon, 2 Dec 2013 09:28:49 +0000 (09:28 +0000)]
Remove an empty directory left after r194304, which incompletely reverted r194218

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196081 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[tsan] fix instrumentation of vector vptr updates (https://code.google.com/p/thread...
Kostya Serebryany [Mon, 2 Dec 2013 08:07:15 +0000 (08:07 +0000)]
[tsan] fix instrumentation of vector vptr updates (https://code.google.com/p/thread-sanitizer/issues/detail?id=43)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196079 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUpdate the LTO GoldPlugin documentation
Alp Toker [Mon, 2 Dec 2013 07:15:33 +0000 (07:15 +0000)]
Update the LTO GoldPlugin documentation

 * Update build instructions to reflect the current source tree layout.
 * Don't inflict CVS on readers; there's a perfectly good git mirror.
 * configure with --disable-werror making it possible to build using clang.
 * ar and nm-new now support the -plugin option.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196069 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove leftovers from a non-MC asm printer.
Rafael Espindola [Mon, 2 Dec 2013 05:42:16 +0000 (05:42 +0000)]
Remove leftovers from a non-MC asm printer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196068 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove #if 0 declarations.
Rafael Espindola [Mon, 2 Dec 2013 05:24:28 +0000 (05:24 +0000)]
Remove #if 0 declarations.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196067 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove dead code.
Rafael Espindola [Mon, 2 Dec 2013 05:10:04 +0000 (05:10 +0000)]
Remove dead code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196066 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoChange the default of AsmWriterClassName and isMCAsmWriter.
Rafael Espindola [Mon, 2 Dec 2013 04:55:42 +0000 (04:55 +0000)]
Change the default of AsmWriterClassName and isMCAsmWriter.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196065 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRename test with misspelt filename
Alp Toker [Mon, 2 Dec 2013 04:31:36 +0000 (04:31 +0000)]
Rename test with misspelt filename

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196064 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove dead declarations.
Rafael Espindola [Mon, 2 Dec 2013 04:18:19 +0000 (04:18 +0000)]
Remove dead declarations.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196063 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRefactor for clarity and efficiency.
Rafael Espindola [Mon, 2 Dec 2013 03:26:43 +0000 (03:26 +0000)]
Refactor for clarity and efficiency.

The PPC GetSymbolFromOperand already prefixed stubs of MO_ExternalSymbol, so
this should be a nop.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196059 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAlso test the created stubs on 32 bits.
Rafael Espindola [Sun, 1 Dec 2013 21:24:30 +0000 (21:24 +0000)]
Also test the created stubs on 32 bits.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196052 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd -mcpu to stackmap.ll
Andrew Trick [Sun, 1 Dec 2013 18:17:05 +0000 (18:17 +0000)]
Add -mcpu to stackmap.ll

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196051 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM: fix bug in -Oz stack adjustment folding
Tim Northover [Sun, 1 Dec 2013 14:16:24 +0000 (14:16 +0000)]
ARM: fix bug in -Oz stack adjustment folding

Previously, we clobbered callee-saved registers when folding an "add
sp, #N" into a "pop {rD, ...}" instruction. This change checks whether
a register we're going to add to the "pop" could actually be live
outside the function before doing so and should fix the issue.

This should fix PR18081.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196046 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRevamp error checking in the ms inline asm parser.
Benjamin Kramer [Sun, 1 Dec 2013 11:47:42 +0000 (11:47 +0000)]
Revamp error checking in the ms inline asm parser.

- Actually abort when an error occurred.
- Check that the frontend lookup worked when parsing length/size/type operators.

Tested by a clang test. PR18096.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196044 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoEnsure bitcode encoding of linkage types stays stable. Patch by Boaz Ouriel
Michael Kuperstein [Sun, 1 Dec 2013 10:16:35 +0000 (10:16 +0000)]
Ensure bitcode encoding of linkage types stays stable. Patch by Boaz Ouriel

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196042 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse accessor methods instead.
Bill Wendling [Sun, 1 Dec 2013 03:40:42 +0000 (03:40 +0000)]
Use accessor methods instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196006 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse 'unsigned char' to get this past gcc error message:
Bill Wendling [Sun, 1 Dec 2013 03:36:07 +0000 (03:36 +0000)]
Use 'unsigned char' to get this past gcc error message:

  error: invalid conversion from 'unsigned char' to '{anonymous}::Sequence'

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196004 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd a scheduling model (with itinerary) for the PPC POWER7
Hal Finkel [Sat, 30 Nov 2013 20:55:12 +0000 (20:55 +0000)]
Add a scheduling model (with itinerary) for the PPC POWER7

This adds a scheduling model for the POWER7 (P7) core, and enables the
machine-instruction scheduler when targeting the P7. Scheduling for the P7,
like earlier ooo PPC cores, requires considering both dispatch group hazards,
and functional unit resources and latencies. These are both modeled in a
combined itinerary. Dispatch group formation is still handled by the post-RA
scheduler (which still needs to be updated for the P7, but nevertheless does a
pretty good job).

One interesting aspect of this change is that I've also enabled to use of AA
duing CodeGen for the P7 (just as it is for the embedded cores). The benchmark
results seem to support this decision (see below), and while this is normally
useful for in-order cores, and not for ooo cores like the P7, I think that the
dispatch slot hazards are enough like in-order resources to make the AA useful.

Test suite significant performance differences (where negative is a speedup,
and positive is a regression) vs. the current situation:

MultiSource/Benchmarks/BitBench/drop3/drop3
  with AA: N/A
  without AA: -28.7614% +/- 19.8356%
(significantly against AA)

MultiSource/Benchmarks/FreeBench/neural/neural
  with AA: -17.7406% +/- 11.2712%
  without AA: N/A
(significantly in favor of AA)

MultiSource/Benchmarks/SciMark2-C/scimark2
  with AA: -11.2079% +/- 1.80543%
  without AA: -11.3263% +/- 2.79651%

MultiSource/Benchmarks/TSVC/Symbolics-flt/Symbolics-flt
  with AA: -41.8649% +/- 17.0053%
  without AA: -34.5256% +/- 23.7072%

MultiSource/Benchmarks/mafft/pairlocalalign
  with AA: 25.3016% +/- 17.8614%
  without AA: 38.6629% +/- 14.9391%
(significantly in favor of AA)

MultiSource/Benchmarks/sim/sim
  with AA: N/A
  without AA: 13.4844% +/- 7.18195%
(significantly in favor of AA)

SingleSource/Benchmarks/BenchmarkGame/Large/fasta
  with AA: 15.0664% +/- 6.70216%
  without AA: 12.7747% +/- 8.43043%

SingleSource/Benchmarks/BenchmarkGame/puzzle
  with AA: 82.2713% +/- 26.3567%
  without AA: 75.7525% +/- 41.1842%

SingleSource/Benchmarks/Misc/flops-2
  with AA: -37.1621% +/- 20.7964%
  without AA: -35.2342% +/- 20.2999%
(significantly in favor of AA)

These are 99.5% confidence intervals from 5 runs per configuration. Regarding
the choice to turn on AA during CodeGen, of these results, four seem
significantly in favor of using AA, and one seems significantly against. I'm
not making this decision based on these numbers alone, but these results
seem consistent with results I have from other tests, and so I think that, on
balance, using AA is a win.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195981 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSplit some PPC itinerary classes
Hal Finkel [Sat, 30 Nov 2013 20:41:13 +0000 (20:41 +0000)]
Split some PPC itinerary classes

In preparation for adding scheduling definitions for the POWER7, split some PPC
itinerary classes so that the P7's latencies and hazards can be better
described. For the most part, this means differentiating indexed from non-index
pre-increment loads and stores. Also, differentiate single from
double-precision sqrt.

No functionality change intended (except for a more-specific latency for
single-precision sqrt on the A2).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195980 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoConvert a PPC test from grep to FileCheck
Hal Finkel [Sat, 30 Nov 2013 20:04:33 +0000 (20:04 +0000)]
Convert a PPC test from grep to FileCheck

Convert this test to FileCheck, and improve it to check for the instructions it
is trying to exclude instead of checking for register use (especially because
grepping for r1 can be thrown off, for example, by a use of r12).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195979 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDesensitize a couple of PPC regression tests
Hal Finkel [Sat, 30 Nov 2013 19:52:28 +0000 (19:52 +0000)]
Desensitize a couple of PPC regression tests

Use CHECK-DAG to make these regression tests more resilient against changes in
instruction scheduling.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195978 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUpdate the cpu specified on some PPC regression tests
Hal Finkel [Sat, 30 Nov 2013 19:39:27 +0000 (19:39 +0000)]
Update the cpu specified on some PPC regression tests

Some of these tests did not specify a cpu but were also sensitive to
instruction scheduling and/or register assignment choices. A few others
similarly-sensitive tests specified a cpu (often the POWER7), and while the P7
currently uses the default model for PPC64, this will soon change. For those
tests which should not really be cpu-dependent anyway, the cpu is set to the
generic 'ppc64'.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195977 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTest case for issue with microMIPS long branch.
Zoran Jovanovic [Sat, 30 Nov 2013 19:13:15 +0000 (19:13 +0000)]
Test case for issue with microMIPS long branch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195976 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFixed issue with microMIPS long branch.
Zoran Jovanovic [Sat, 30 Nov 2013 19:12:28 +0000 (19:12 +0000)]
Fixed issue with microMIPS long branch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195975 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][msa] MSA loads and stores have a 10-bit offset. Account for this when lowering...
Daniel Sanders [Sat, 30 Nov 2013 13:47:57 +0000 (13:47 +0000)]
[mips][msa] MSA loads and stores have a 10-bit offset. Account for this when lowering FrameIndex.

This prevents the compiler from emitting invalid ld.[bhwd]'s and st.[bhwd]'s
when the stack frame is between 512 and 32,768 bytes in size.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195973 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][msa] A small refactor to reduce patch noise in my next commit
Daniel Sanders [Sat, 30 Nov 2013 13:15:21 +0000 (13:15 +0000)]
[mips][msa] A small refactor to reduce patch noise in my next commit

No functional change. An if-statement has been split into two nested if-statements.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195972 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoForce CPU type to unbreak unit tests on Haswell machines.
Juergen Ributzka [Sat, 30 Nov 2013 03:07:16 +0000 (03:07 +0000)]
Force CPU type to unbreak unit tests on Haswell machines.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195971 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoReverse the order of eviction checks for possible compile time savings. No functionality.
Andrew Trick [Fri, 29 Nov 2013 23:49:38 +0000 (23:49 +0000)]
Reverse the order of eviction checks for possible compile time savings. No functionality.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195969 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoPart 1 of 3 patches that completes very long conditional branches
Reed Kotler [Fri, 29 Nov 2013 22:32:56 +0000 (22:32 +0000)]
Part 1 of 3 patches that completes very long conditional branches
in constant islands for Mips16. We introdcuce JalB16 as a synomnym
for Jal16. It makes it easier to read and is also necessary because
Jal16 is a call instruction but JalB16 is being used as a branch.
Various parts of LLVM will not work properly even in this late stage of
the backend if we use what was declared as a call instruction to function
as a branch. For one, basic block labels may not get emitted in some
situations.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195968 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRevert revision 195965.
Zoran Jovanovic [Fri, 29 Nov 2013 22:10:02 +0000 (22:10 +0000)]
Revert revision 195965.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195967 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agomips: XFAIL llvm-cov test
Petar Jovanovic [Fri, 29 Nov 2013 21:59:09 +0000 (21:59 +0000)]
mips: XFAIL llvm-cov test

XFAIL llvm-cov.test for MIPS until big-endian issues are fixed for llvm-cov.
The test does pass on MIPS little-endian.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195966 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFixed issue with microMIPS long branch.
Zoran Jovanovic [Fri, 29 Nov 2013 21:41:24 +0000 (21:41 +0000)]
Fixed issue with microMIPS long branch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195965 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdjust PPC A2 input operand latencies
Hal Finkel [Fri, 29 Nov 2013 07:04:59 +0000 (07:04 +0000)]
Adjust PPC A2 input operand latencies

On the PPC A2, instructions are only issued after their input operands are
ready. Model this by specifying that input operands are read at dispatch (0
cycles after issue). This changes all input operand latencies from 1 to 0.

Significant test-suite performance changes (these are 99.5% confidence
intervals on 6 runs for both before and after):

speedups:
MultiSource/Benchmarks/sim/sim
-1.21915% +/- 0.175063%
MultiSource/Benchmarks/TSVC/LinearDependence-flt/LinearDependence-flt
-1.23946% +/- 1.05133%
SingleSource/Benchmarks/Misc/flops-2
-1.24237% +/- 0.681362%
MultiSource/Applications/JM/lencod/lencod
-1.33992% +/- 0.757498%
MultiSource/Benchmarks/TSVC/InductionVariable-flt/InductionVariable-flt
-1.51802% +/- 1.21468%
MultiSource/Benchmarks/TSVC/GlobalDataFlow-flt/GlobalDataFlow-flt
-2.18818% +/- 1.28605%
MultiSource/Benchmarks/TSVC/Packing-flt/Packing-flt
-2.21977% +/- 1.19499%
SingleSource/Benchmarks/BenchmarkGame/spectral-norm
-2.29822% +/- 0.671871%
MultiSource/Benchmarks/TSVC/Packing-dbl/Packing-dbl
-2.40975% +/- 0.355931%
SingleSource/Benchmarks/Misc/fp-convert
-2.41899% +/- 1.04751%
MultiSource/Benchmarks/TSVC/Searching-dbl/Searching-dbl
-2.50349% +/- 0.126765%
SingleSource/Benchmarks/Misc/flops-3
-3.00214% +/- 0.700795%
MultiSource/Benchmarks/TSVC/LoopRestructuring-flt/LoopRestructuring-flt
-3.56995% +/- 3.2929%
MultiSource/Applications/sgefa/sgefa
-4.24908% +/- 2.00413%
MultiSource/Benchmarks/ASC_Sequoia/IRSmk/IRSmk
-18.1294% +/- 3.96489%

regressions:
MultiSource/Benchmarks/TSVC/Reductions-dbl/Reductions-dbl
1.03249% +/- 0.178547%
MultiSource/Applications/hexxagon/hexxagon
1.16597% +/- 0.285235%
MultiSource/Benchmarks/TSVC/IndirectAddressing-flt/IndirectAddressing-flt
1.39576% +/- 1.07855%
SingleSource/Benchmarks/Misc-C++/stepanov_v1p2
1.71539% +/- 0.173182%
MultiSource/Benchmarks/Fhourstones-3.1/fhourstones3.1
1.90013% +/- 0.866472%
MultiSource/Benchmarks/TSVC/Recurrences-dbl/Recurrences-dbl
2.39854% +/- 1.05914%
MultiSource/Benchmarks/TSVC/ControlFlow-dbl/ControlFlow-dbl
2.4402% +/- 0.817904%
MultiSource/Benchmarks/TSVC/LoopRestructuring-dbl/LoopRestructuring-dbl
5.87997% +/- 3.3172%
MultiSource/Benchmarks/Trimaran/netbench-crc/netbench-crc
9.02643% +/- 5.79591%
MultiSource/Benchmarks/VersaBench/bmm/bmm
10.3517% +/- 1.227%

Obviously, there are data points on both sides of this; but I think, overall,
this supports making the change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195951 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTeach LocalStackSlotAllocation that stackmaps/patchpoints don't have range
Lang Hames [Fri, 29 Nov 2013 06:35:30 +0000 (06:35 +0000)]
Teach LocalStackSlotAllocation that stackmaps/patchpoints don't have range
constraints on their frame offsets.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195950 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoCreate a PPC440 SchedMachineModel
Hal Finkel [Fri, 29 Nov 2013 06:32:17 +0000 (06:32 +0000)]
Create a PPC440 SchedMachineModel

Some of the older PPC processor definitions don't have associated
SchedMachineModels; correct this for the PPC440.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195949 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFixup PPC440 load/store operand latencies
Hal Finkel [Fri, 29 Nov 2013 06:19:43 +0000 (06:19 +0000)]
Fixup PPC440 load/store operand latencies

The operand latencies for loads and stores in the PPC440 itinerary were wrong
(the store operands are all inputs, and the "with update" (pre-increment)
instructions need a latency for the additional output).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195948 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdjust PPC440 operand latencies
Hal Finkel [Fri, 29 Nov 2013 05:59:00 +0000 (05:59 +0000)]
Adjust PPC440 operand latencies

The operand latencies for the PPC440 should be specified relative to dispatch,
not relative to the initial fetch-and-decode stages. Because most instructions
(ignoring bypass) wait in dispatch until their operands are ready, this is
modeled as reading input operands "at dispatch" (0 cycles after issue), and so
every input and output operand has 4 cycles subtracted from it.

This could alter scheduling slightly, but I don't expect a large effect.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195947 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDon't model the fetch and decode units for the PPC440
Hal Finkel [Fri, 29 Nov 2013 05:58:38 +0000 (05:58 +0000)]
Don't model the fetch and decode units for the PPC440

Modeling the fetch and decode units in the PPC440 itinerary does not add
anything to the hazard detection capability (and so modeling them just wastes
compile time).

No functionality change intended.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195946 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove unused variable from r195944.
Lang Hames [Fri, 29 Nov 2013 03:36:53 +0000 (03:36 +0000)]
Remove unused variable from r195944.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195945 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRefactor a lot of patchpoint/stackmap related code to simplify and make it
Lang Hames [Fri, 29 Nov 2013 03:07:54 +0000 (03:07 +0000)]
Refactor a lot of patchpoint/stackmap related code to simplify and make it
target independent.

Most of the x86 specific stackmap/patchpoint handling was necessitated by the
use of the native address-mode format for frame index operands. PEI has now
been modified to treat stackmap/patchpoint similarly to DEBUG_INFO, allowing
us to use a simple, platform independent register/offset pair for frame
indexes on stackmap/patchpoints.

Notes:
  - Folding is now platform independent and automatically supported.
  - Emiting patchpoints with direct memory references now just involves calling
    the TargetLoweringBase::emitPatchPoint utility method from the target's
    XXXTargetLowering::EmitInstrWithCustomInserter method. (See
    X86TargetLowering for an example).
  - No more ugly platform-specific operand parsers.

This patch shouldn't change the generated output for X86.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195944 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAArch64: The pattern match should check the range of the immediate value.
Hao Liu [Fri, 29 Nov 2013 02:11:22 +0000 (02:11 +0000)]
AArch64: The pattern match should check the range of the immediate value.
Or we can generate some illegal instructions.
E.g. shrn2 v0.4s, v1.2d, #35. The legal range should be in [1, 16].

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195941 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd missing test case for bsl_f64 support of AArch64 NEON.
Jiangning Liu [Fri, 29 Nov 2013 01:38:08 +0000 (01:38 +0000)]
Add missing test case for bsl_f64 support of AArch64 NEON.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195939 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd missing pattern for supporting intrinsic function vbsl_f64 with
Jiangning Liu [Fri, 29 Nov 2013 01:37:15 +0000 (01:37 +0000)]
Add missing pattern for supporting intrinsic function vbsl_f64 with
argument double floating point.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195938 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[AArch64 NEON]Fix a assertion failure when disassemble SHLL instruction.
Kevin Qin [Fri, 29 Nov 2013 01:29:16 +0000 (01:29 +0000)]
[AArch64 NEON]Fix a assertion failure when disassemble SHLL instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195936 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRein in overzealous InstCombine of fptrunc(OP(fpextend, fpextend)).
Stephen Canon [Thu, 28 Nov 2013 21:38:05 +0000 (21:38 +0000)]
Rein in overzealous InstCombine of fptrunc(OP(fpextend, fpextend)).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195934 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRefactor to remove a bit of duplication. No functionality change.
Rafael Espindola [Thu, 28 Nov 2013 20:12:44 +0000 (20:12 +0000)]
Refactor to remove a bit of duplication. No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195933 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSilence sign-compare warning and reduce nesting.
Benjamin Kramer [Thu, 28 Nov 2013 19:58:56 +0000 (19:58 +0000)]
Silence sign-compare warning and reduce nesting.

No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195932 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove an always true parameter.
Rafael Espindola [Thu, 28 Nov 2013 19:35:07 +0000 (19:35 +0000)]
Remove an always true parameter.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195931 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CMake] Let add_public_tablegen_target() provide intrinsics_gen, too.
NAKAMURA Takumi [Thu, 28 Nov 2013 17:04:31 +0000 (17:04 +0000)]
[CMake] Let add_public_tablegen_target() provide intrinsics_gen, too.

I think, in principle, intrinsics_gen may be added explicitly.
That said, it can be added incidentally, since each target already has dependencies to llvm-tblgen.
Almost all source files depend on both CommonTaleGen and intrinsics_gen.

Explicit add_dependencies() have been pruned under lib/Target.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195929 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CMake] Also OptionTests can be free from add_dependencies() with add_public_tablegen...
NAKAMURA Takumi [Thu, 28 Nov 2013 17:04:13 +0000 (17:04 +0000)]
[CMake] Also OptionTests can be free from add_dependencies() with add_public_tablegen_target().

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195928 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CMake] Let add_public_tablegen_target responsible to provide dependency to CommonTab...
NAKAMURA Takumi [Thu, 28 Nov 2013 17:04:04 +0000 (17:04 +0000)]
[CMake] Let add_public_tablegen_target responsible to provide dependency to CommonTableGen.

add_public_tablegen_target adds *CommonTableGen to LLVM_COMMON_DEPENDS.
LLVM_COMMON_DEPENDS affects add_llvm_library (and other add_target stuff) within its scope.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195927 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoThe global prefix is always one char. Don't use a string for it.
Rafael Espindola [Thu, 28 Nov 2013 17:00:49 +0000 (17:00 +0000)]
The global prefix is always one char. Don't use a string for it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195926 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CMake] Prune include_directories() in llvm/lib/Target, take #2.
NAKAMURA Takumi [Thu, 28 Nov 2013 15:30:37 +0000 (15:30 +0000)]
[CMake] Prune include_directories() in llvm/lib/Target, take #2.

I forgot to commit them. They were staging in my local repo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195924 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips] Revert test commit r195922.
Daniel Sanders [Thu, 28 Nov 2013 15:26:33 +0000 (15:26 +0000)]
[mips] Revert test commit r195922.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195923 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips] A test commit to test my Herald and Audit workflow
Daniel Sanders [Thu, 28 Nov 2013 15:25:43 +0000 (15:25 +0000)]
[mips] A test commit to test my Herald and Audit workflow

Will be reverted in the next commit

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195922 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CMake] Prune include_directories() in llvm/lib/Target. add_llvm_target() sets them.
NAKAMURA Takumi [Thu, 28 Nov 2013 14:53:30 +0000 (14:53 +0000)]
[CMake] Prune include_directories() in llvm/lib/Target. add_llvm_target() sets them.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195921 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd newline at eof.
NAKAMURA Takumi [Thu, 28 Nov 2013 14:52:52 +0000 (14:52 +0000)]
Add newline at eof.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195920 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAs myself as code-owner of the MIPS backend (lib/Target/Mips/*)
Daniel Sanders [Thu, 28 Nov 2013 09:36:44 +0000 (09:36 +0000)]
As myself as code-owner of the MIPS backend (lib/Target/Mips/*)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195915 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[OCaml] Add a slash accidentally omitted from Makefile
Peter Zotov [Thu, 28 Nov 2013 09:03:28 +0000 (09:03 +0000)]
[OCaml] Add a slash accidentally omitted from Makefile

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195912 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse the mangler consistently instead of using getGlobalPrefix directly.
Rafael Espindola [Thu, 28 Nov 2013 08:59:52 +0000 (08:59 +0000)]
Use the mangler consistently instead of using getGlobalPrefix directly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195911 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDon't share functional units among the PPC itineraries
Hal Finkel [Thu, 28 Nov 2013 06:05:59 +0000 (06:05 +0000)]
Don't share functional units among the PPC itineraries

Instead of sharing functional unit names between the various PPC itineraries,
give each core its own unit names prefixed with the core name.  This follows
the convention used by other backends (such as ARM), and removes a non-obvious
ordering dependency between the various PPCSchedule*.td files.

No functionality change intended.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195908 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove the variable only used by assert to avoid the build failure
Jiangning Liu [Thu, 28 Nov 2013 01:34:55 +0000 (01:34 +0000)]
Remove the variable only used by assert to avoid the build failure
caused by build options [-Werror,-Wunused-variable].

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195905 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAArch64: Fix a bug about disassembling post-index load single element to 4 vectors
Hao Liu [Thu, 28 Nov 2013 01:07:45 +0000 (01:07 +0000)]
AArch64: Fix a bug about disassembling post-index load single element to 4 vectors

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195903 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoCheck in conditional branches for constant islands. Still need to finish
Reed Kotler [Thu, 28 Nov 2013 00:56:37 +0000 (00:56 +0000)]
Check in conditional branches for constant islands. Still need to finish
conditional branches for very large targets. That will be the next small
patch. Everything now should in principle work as good (functionality
wise) as without constant islands so we decided at Mips/Imagination to
make constant islands the default for Mips16 now so that it will get
excercised a lot and this port is still experimentatl though hopefully soon
we will change the status. Some more cleanup and code review is in order
but things are converging fast.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195902 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips] Redefine TAILCALL as a pseudo instruction.
Akira Hatanaka [Wed, 27 Nov 2013 23:58:32 +0000 (23:58 +0000)]
[mips] Redefine TAILCALL as a pseudo instruction.

No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195896 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDebugInfo: Do not include variables only referenced by templates in aranges.
David Blaikie [Wed, 27 Nov 2013 23:53:52 +0000 (23:53 +0000)]
DebugInfo: Do not include variables only referenced by templates in aranges.

ARanges included even extern variables referenced by pointer non-type
template parameters even though that variable isn't part of this
compilation unit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195895 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd MipsOptimizePICCall.cpp to CMakeLists.txt.
Akira Hatanaka [Wed, 27 Nov 2013 23:47:25 +0000 (23:47 +0000)]
Add MipsOptimizePICCall.cpp to CMakeLists.txt.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195894 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips] Implement the following optimizations using dominance information to
Akira Hatanaka [Wed, 27 Nov 2013 23:38:42 +0000 (23:38 +0000)]
[mips] Implement the following optimizations using dominance information to
make PIC calls a little more efficient:

1. Remove instructions setting up $gp if it is known that a function has been
   called at least once.
2. Save the address of a called function in a register instead of loading
   it from the GOT at every call site.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195892 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd IIC_ prefix to PPC instruction-class names
Hal Finkel [Wed, 27 Nov 2013 23:26:09 +0000 (23:26 +0000)]
Add IIC_ prefix to PPC instruction-class names

This adds the IIC_ prefix to the instruction itinerary class names, giving the
PPC backend a naming convention for itinerary classes that is more consistent
with that used by the X86 and ARM backends.

Instruction scheduling in the PPC backend needs a bunch of cleanup and
improvement (especially for the ooo cores). This is just a preliminary step.

No functionality change intended.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195890 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDon't set GlobalPrefix to the default value.
Rafael Espindola [Wed, 27 Nov 2013 21:57:54 +0000 (21:57 +0000)]
Don't set GlobalPrefix to the default value.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195884 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoThe R600 has its own asm printer which doesn't use GlobalPrefix. Drop it.
Rafael Espindola [Wed, 27 Nov 2013 21:52:37 +0000 (21:52 +0000)]
The R600 has its own asm printer which doesn't use GlobalPrefix. Drop it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195883 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Expand vector FABS
Tom Stellard [Wed, 27 Nov 2013 21:23:39 +0000 (21:23 +0000)]
R600: Expand vector FABS

NOTE: This is a candidate for the 3.4 branch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195881 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Implement spilling of SGPRs v5
Tom Stellard [Wed, 27 Nov 2013 21:23:35 +0000 (21:23 +0000)]
R600/SI: Implement spilling of SGPRs v5

SGPRs are spilled into VGPRs using the {READ,WRITE}LANE_B32 instructions.

v2:
  - Fix encoding of Lane Mask
  - Use correct register flags, so we don't overwrite the low dword
    when restoring multi-dword registers.

v3:
  - Register spilling seems to hang the GPU, so replace all shaders
    that need spilling with a dummy shader.

v4:
  - Fix *LANE definitions
  - Change destination reg class for 32-bit SMRD instructions

v5:
  - Remove small optimization that was crashing Serious Sam 3.

https://bugs.freedesktop.org/show_bug.cgi?id=68224
https://bugs.freedesktop.org/show_bug.cgi?id=71285

NOTE: This is a candidate for the 3.4 branch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195880 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Use SGPR_32 register class for 32-bit SMRD outputs
Tom Stellard [Wed, 27 Nov 2013 21:23:29 +0000 (21:23 +0000)]
R600/SI: Use SGPR_32 register class for 32-bit SMRD outputs

Writing to the M0 register from an SMRD instruction hangs the GPU, so
we need to use the SGPR_32 register class, which does not include M0.

NOTE: This is a candidate for the 3.4 branch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195879 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Add support for ISD::FROUND
Tom Stellard [Wed, 27 Nov 2013 21:23:20 +0000 (21:23 +0000)]
R600: Add support for ISD::FROUND

NOTE: This is a candidate for the 3.4 branch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195878 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoShow stackmap entry encodings in stackmap debug logs. This makes it easier to
Lang Hames [Wed, 27 Nov 2013 20:10:16 +0000 (20:10 +0000)]
Show stackmap entry encodings in stackmap debug logs. This makes it easier to
cross-reference debug output with encoded stack-maps, and to create stackmap
test-cases.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195874 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse FileCheck and expand the test a bit.
Rafael Espindola [Wed, 27 Nov 2013 19:22:14 +0000 (19:22 +0000)]
Use FileCheck and expand the test a bit.

In particular, check the name of the symbol we are putting in the constant pool.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195865 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove dead code.
Rafael Espindola [Wed, 27 Nov 2013 18:38:14 +0000 (18:38 +0000)]
Remove dead code.

MO_ExternalSymbol and MO_JumpTableIndex don't show up in inline asm.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195861 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoConvert two if sequences to switches.
Rafael Espindola [Wed, 27 Nov 2013 18:26:51 +0000 (18:26 +0000)]
Convert two if sequences to switches.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195859 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse a switch.
Rafael Espindola [Wed, 27 Nov 2013 18:18:24 +0000 (18:18 +0000)]
Use a switch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195857 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse the same tls section name as msvc.
Rafael Espindola [Wed, 27 Nov 2013 15:52:11 +0000 (15:52 +0000)]
Use the same tls section name as msvc.

We currently error in clang with:
"error: thread-local storage is unsupported for the current target", but we
can start to get the llvm level ready.

When compiling

template<typename T>
struct foo {
  static __declspec(thread) int bar;
};
template<typename T>
__declspec(therad) int foo<T>::bar;
template struct foo<int>;

msvc produces

SECTION HEADER #3
   .tls$ name
       0 physical address
       0 virtual address
       4 size of raw data
     12F file pointer to raw data (0000012F to 00000132)
       0 file pointer to relocation table
       0 file pointer to line numbers
       0 number of relocations
       0 number of line numbers
C0301040 flags
         Initialized Data
         COMDAT; sym= "public: static int foo<int>::bar" (?bar@?$foo@H@@2HA)
         4 byte align
         Read Write

gcc produces a ".data$__emutls_v.<symbol>" for the testcase with
__declspec(thread) replaced with thread_local.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195849 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove more dead code now that this is only used for inline asm.
Rafael Espindola [Wed, 27 Nov 2013 15:13:06 +0000 (15:13 +0000)]
Remove more dead code now that this is only used for inline asm.

MO_ConstantPoolIndex is handled in printLeaMemReference.
MO_JumpTableIndex and MO_ExternalSymbol don't show up in inline asm.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195847 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix the AArch64 NEON bug exposed by checking constant integer argument range of ACLE...
Jiangning Liu [Wed, 27 Nov 2013 14:02:25 +0000 (14:02 +0000)]
Fix the AArch64 NEON bug exposed by checking constant integer argument range of ACLE intrinsics.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195843 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[OCaml] Embed rpath into stub libraries and native executables
Peter Zotov [Wed, 27 Nov 2013 11:03:18 +0000 (11:03 +0000)]
[OCaml] Embed rpath into stub libraries and native executables

This commit embeds a set of linker flags with hardcoded paths to
the LLVM shared library on --enable-shared builds into .cmxa files
and stub dynamic libraries. This solution closely follows existing
rules for rpath in the LLVM tools, which had to be modified because
of differences in toolchain.

Without this patch, OCaml tests as well as opam bindings broke,
as neither of those updates LD_LIBRARY_PATH to include
the $prefix/lib directory.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195834 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoConvert more methods in static helpers.
Rafael Espindola [Wed, 27 Nov 2013 07:34:09 +0000 (07:34 +0000)]
Convert more methods in static helpers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195826 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoConvert these methods into static functions.
Rafael Espindola [Wed, 27 Nov 2013 07:14:26 +0000 (07:14 +0000)]
Convert these methods into static functions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195825 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoCleanup and test X86AsmPrinter::printPCRelImm.
Rafael Espindola [Wed, 27 Nov 2013 06:53:13 +0000 (06:53 +0000)]
Cleanup and test X86AsmPrinter::printPCRelImm.

It is only used for asm printing.

On X86 we put basic block addresses on register before passing them to inline
asm, so the MO_MachineBasicBlock case was dead.

MO_ExternalSymbol was dead since any symbol being passed to inline asm
is represented as MO_GlobalAddress.

The MO_GlobalAddress and MO_Register cases were not tested.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195824 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[docs] Mention gotcha regarding implicit BB numbering
Sean Silva [Wed, 27 Nov 2013 04:55:23 +0000 (04:55 +0000)]
[docs] Mention gotcha regarding implicit BB numbering

Impetus for the clarification by Mikael Lyngvig.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195812 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix comment in PPCA2Model
Hal Finkel [Wed, 27 Nov 2013 03:12:56 +0000 (03:12 +0000)]
Fix comment in PPCA2Model

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@195807 91177308-0d34-0410-b5e6-96231b3b80d8