oota-llvm.git
14 years agoRefactor code.
Evan Cheng [Sun, 8 Nov 2009 00:15:23 +0000 (00:15 +0000)]
Refactor code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86423 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix CMake reporting of target triple.
Daniel Dunbar [Sat, 7 Nov 2009 23:52:27 +0000 (23:52 +0000)]
Fix CMake reporting of target triple.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86419 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoStop running get_target_triple more than we need to.
Daniel Dunbar [Sat, 7 Nov 2009 23:52:20 +0000 (23:52 +0000)]
Stop running get_target_triple more than we need to.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86418 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix MSVC warning ( | with bool and unsigned int).
Daniel Dunbar [Sat, 7 Nov 2009 23:51:55 +0000 (23:51 +0000)]
Fix MSVC warning ( | with bool and unsigned int).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86417 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix class -> struct tag.
Daniel Dunbar [Sat, 7 Nov 2009 23:21:30 +0000 (23:21 +0000)]
Fix class -> struct tag.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86416 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agox86 vector shuffle cleanup/fixes:
Nate Begeman [Sat, 7 Nov 2009 23:17:15 +0000 (23:17 +0000)]
x86 vector shuffle cleanup/fixes:

1. rename the movhp patfrag to movlhps, since thats what it actually matches
2. eliminate the bogus movhps load and store patterns, they were incorrect.  The load transforms are already handled (correctly) by shufps/unpack.
3. revert a recent test change to its correct form.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86415 91177308-0d34-0410-b5e6-96231b3b80d8

14 years ago80-column cleanup of file header comments
Jim Grosbach [Sat, 7 Nov 2009 22:00:39 +0000 (22:00 +0000)]
80-column cleanup of file header comments

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86408 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoSupport alignment specifier for NEON vld/vst instructions
Jim Grosbach [Sat, 7 Nov 2009 21:25:39 +0000 (21:25 +0000)]
Support alignment specifier for NEON vld/vst instructions

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86404 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoImprove tail call elimination to handle the switch statement.
Nick Lewycky [Sat, 7 Nov 2009 21:10:15 +0000 (21:10 +0000)]
Improve tail call elimination to handle the switch statement.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86403 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agot2ldrpci_pic can be used for blockaddress as well.
Evan Cheng [Sat, 7 Nov 2009 19:40:04 +0000 (19:40 +0000)]
t2ldrpci_pic can be used for blockaddress as well.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86400 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agotemporarily remove these tests, as they are breaking in the buildbot,
Chris Lattner [Sat, 7 Nov 2009 19:13:17 +0000 (19:13 +0000)]
temporarily remove these tests, as they are breaking in the buildbot,
Eric, please investigate.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86399 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agomake instcombine only rewrite a chain of computation
Chris Lattner [Sat, 7 Nov 2009 19:11:46 +0000 (19:11 +0000)]
make instcombine only rewrite a chain of computation
(eliminating some extends) if the new type of the
computation is legal or if both the source and dest
are illegal.  This prevents instcombine from changing big
chains of computation into i64 on 32-bit targets for
example.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86398 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoindicate what the native integer types for the target are.
Chris Lattner [Sat, 7 Nov 2009 19:07:32 +0000 (19:07 +0000)]
indicate what the native integer types for the target are.
Please verify.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86397 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoall targets should be required to declare legal integer types. My plan to
Chris Lattner [Sat, 7 Nov 2009 18:53:00 +0000 (18:53 +0000)]
all targets should be required to declare legal integer types.  My plan to
make it optional doesn't work out.  If you don't want to specify this, don't
specify a TD string at all.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86394 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoremove empty files.
Chris Lattner [Sat, 7 Nov 2009 18:03:32 +0000 (18:03 +0000)]
remove empty files.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86392 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRevert r86359, it is breaking the self host on the
Chris Lattner [Sat, 7 Nov 2009 17:59:32 +0000 (17:59 +0000)]
Revert r86359, it is breaking the self host on the
llvm-gcc-i386-darwin9 build bot.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86391 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFirst try of the post-inc operands handling... Not fully worked, though :(
Anton Korobeynikov [Sat, 7 Nov 2009 17:15:25 +0000 (17:15 +0000)]
First try of the post-inc operands handling... Not fully worked, though :(

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86386 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd some dummy support for post-incremented loads
Anton Korobeynikov [Sat, 7 Nov 2009 17:15:06 +0000 (17:15 +0000)]
Add some dummy support for post-incremented loads

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86385 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd 8 bit libcalls and make use of them for msp430
Anton Korobeynikov [Sat, 7 Nov 2009 17:14:39 +0000 (17:14 +0000)]
Add 8 bit libcalls and make use of them for msp430

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86384 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd few pseudo-source-values
Anton Korobeynikov [Sat, 7 Nov 2009 17:13:57 +0000 (17:13 +0000)]
Add few pseudo-source-values

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86383 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoInitial support for addrmode handling. Tests by Brian Lucas!
Anton Korobeynikov [Sat, 7 Nov 2009 17:13:35 +0000 (17:13 +0000)]
Initial support for addrmode handling. Tests by Brian Lucas!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86382 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoSome preliminary variable asmprinting
Anton Korobeynikov [Sat, 7 Nov 2009 17:12:58 +0000 (17:12 +0000)]
Some preliminary variable asmprinting

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86381 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoUse '.L' for global private prefix (as mspgcc)
Anton Korobeynikov [Sat, 7 Nov 2009 17:12:38 +0000 (17:12 +0000)]
Use '.L' for global private prefix (as mspgcc)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86380 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoDrop old asmprinter stuff
Anton Korobeynikov [Sat, 7 Nov 2009 17:12:21 +0000 (17:12 +0000)]
Drop old asmprinter stuff

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86379 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoIt turns out that the testcase in question uncovered subreg-handling bug.
Anton Korobeynikov [Sat, 7 Nov 2009 15:20:32 +0000 (15:20 +0000)]
It turns out that the testcase in question uncovered subreg-handling bug.
Add assert in asmprinter to catch such cases and xfail the tests.
PR is to be filled.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86375 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoadd the ability for TargetData to return information about legal integer
Chris Lattner [Sat, 7 Nov 2009 09:35:34 +0000 (09:35 +0000)]
add the ability for TargetData to return information about legal integer
datatypes on a given CPU.  This is intended to allow instcombine and other
transformations to avoid converting big sequences of operations to an
inconvenient width, and will help clean up after SRoA.  See also "Adding
legal integer sizes to TargetData" on Feb 1, 2009 on llvmdev, and PR3451.

Comments welcome.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86370 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agomore cleanup.
Chris Lattner [Sat, 7 Nov 2009 09:23:04 +0000 (09:23 +0000)]
more cleanup.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86369 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoadd some missing #includes
Chris Lattner [Sat, 7 Nov 2009 09:20:54 +0000 (09:20 +0000)]
add some missing #includes

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86367 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agorewrite TargetData to use StringRef/raw_ostream instead of thrashing std::strings.
Chris Lattner [Sat, 7 Nov 2009 09:13:23 +0000 (09:13 +0000)]
rewrite TargetData to use StringRef/raw_ostream instead of thrashing std::strings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86366 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoprune #include / layering violation
Chris Lattner [Sat, 7 Nov 2009 09:07:01 +0000 (09:07 +0000)]
prune #include / layering violation

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86365 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoMake the need-stub variables accurate and consistent. In the case of
Jeffrey Yasskin [Sat, 7 Nov 2009 08:51:52 +0000 (08:51 +0000)]
Make the need-stub variables accurate and consistent.  In the case of
MachineRelocations, "stub" always refers to a far-call stub or a
load-a-faraway-global stub, so this patch adds "Far" to the term. (Other stubs
are used for lazy compilation and dlsym address replacement.) The variable was
also inconsistent between the positive and negative sense, and the positive
sense ("NeedStub") was more demanding than is accurate (since a nearby-enough
function can be called directly even if the platform often requires a stub).
Since the negative sense causes double-negatives, I switched to
"MayNeedFarStub" globally.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86363 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix a couple of shuffle patterns to use movhlps instead
Eric Christopher [Sat, 7 Nov 2009 08:45:53 +0000 (08:45 +0000)]
Fix a couple of shuffle patterns to use movhlps instead
of movhps as the constraint.  Changes optimizations so
update testcases as appropriate as well.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86360 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoTeach dead store elimination that certain intrinsics write to memory just like
Nick Lewycky [Sat, 7 Nov 2009 08:34:40 +0000 (08:34 +0000)]
Teach dead store elimination that certain intrinsics write to memory just like
a store.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86359 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoremove the win32 tree, it's stale and confusing.
Chris Lattner [Sat, 7 Nov 2009 08:31:52 +0000 (08:31 +0000)]
remove the win32 tree, it's stale and confusing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86358 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoreapply 86289, 86278, 86270, 86267, 86266 & 86264 plus a fix
Chris Lattner [Sat, 7 Nov 2009 08:05:03 +0000 (08:05 +0000)]
reapply 86289, 86278, 86270, 86267, 86266 & 86264 plus a fix
(making pred factoring only happen if threading is guaranteed
to be successful).

This now survives an X86-64 bootstrap of llvm-gcc.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86355 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix PR5421 by APInt'izing switch lowering.
Chris Lattner [Sat, 7 Nov 2009 07:50:34 +0000 (07:50 +0000)]
Fix PR5421 by APInt'izing switch lowering.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86354 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoOops, FunctionContainsEscapingAllocas is really used to mean two different
Nick Lewycky [Sat, 7 Nov 2009 07:42:38 +0000 (07:42 +0000)]
Oops, FunctionContainsEscapingAllocas is really used to mean two different
things. Back out part of r86349 for a moment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86353 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoDust off tail recursion elimination. Fix a fixme by applying CaptureTracking
Nick Lewycky [Sat, 7 Nov 2009 07:10:01 +0000 (07:10 +0000)]
Dust off tail recursion elimination. Fix a fixme by applying CaptureTracking
and add a .ll to demo the new capability.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86349 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agollvmc: Add a '-time' option.
Mikhail Glushenkov [Sat, 7 Nov 2009 06:33:58 +0000 (06:33 +0000)]
llvmc: Add a '-time' option.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86348 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoTrailing whitespace.
Mikhail Glushenkov [Sat, 7 Nov 2009 06:33:12 +0000 (06:33 +0000)]
Trailing whitespace.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86347 91177308-0d34-0410-b5e6-96231b3b80d8

14 years ago80-col violation.
Mikhail Glushenkov [Sat, 7 Nov 2009 06:33:01 +0000 (06:33 +0000)]
80-col violation.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86346 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agomerge cmp1 into cmp0 and filecheckize.
Chris Lattner [Sat, 7 Nov 2009 06:19:20 +0000 (06:19 +0000)]
merge cmp1 into cmp0 and filecheckize.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86345 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoUpdate some globals to use ManagedStatic.
Lang Hames [Sat, 7 Nov 2009 05:50:28 +0000 (05:50 +0000)]
Update some globals to use ManagedStatic.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86342 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix memoizing of CvtRndSatSDNode
Mon P Wang [Sat, 7 Nov 2009 04:46:25 +0000 (04:46 +0000)]
Fix memoizing of CvtRndSatSDNode

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86340 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFixed Overload table bug noticed by Jakob
Mon P Wang [Sat, 7 Nov 2009 04:07:33 +0000 (04:07 +0000)]
Fixed Overload table bug noticed by Jakob

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86332 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoMissed this.
Evan Cheng [Sat, 7 Nov 2009 04:07:30 +0000 (04:07 +0000)]
Missed this.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86331 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRefactor code. Fix a potential missing check. Teach isIdentical() about tLDRpci_pic.
Evan Cheng [Sat, 7 Nov 2009 04:04:34 +0000 (04:04 +0000)]
Refactor code. Fix a potential missing check. Teach isIdentical() about tLDRpci_pic.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86330 91177308-0d34-0410-b5e6-96231b3b80d8

14 years ago- Add TargetInstrInfo::isIdentical(). It's similar to MachineInstr::isIdentical
Evan Cheng [Sat, 7 Nov 2009 03:52:02 +0000 (03:52 +0000)]
- Add TargetInstrInfo::isIdentical(). It's similar to MachineInstr::isIdentical
  except it doesn't care if the definitions' virtual registers differ. This is
  used by machine LICM and other MI passes to perform CSE.
- Teach Thumb2InstrInfo::isIdentical() to check two t2LDRpci_pic are identical.
  Since pc relative constantpool entries are always different, this requires it
  it check if the values can actually the same.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86328 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoUpdate CMake file.
Ted Kremenek [Sat, 7 Nov 2009 03:26:59 +0000 (03:26 +0000)]
Update CMake file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86325 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd code to check at SelectionDAGISel::LowerArguments time to see if return values...
Kenneth Uildriks [Sat, 7 Nov 2009 02:11:54 +0000 (02:11 +0000)]
Add code to check at SelectionDAGISel::LowerArguments time to see if return values can be lowered to registers.  Coming soon, code to perform sret-demotion if return values cannot be lowered to registers

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86324 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix inverted conflict test in -early-coalesce.
Jakob Stoklund Olesen [Sat, 7 Nov 2009 01:58:40 +0000 (01:58 +0000)]
Fix inverted conflict test in -early-coalesce.

A non-identity copy cannot be coalesced when the phi join destination register
is live at the copy site.

Also verify the condition that the PHI join source register is only used in
the PHI join. Otherwise the coalescing is invalid.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86322 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRevert following patches to fix llvmgcc bootstrap.
Devang Patel [Sat, 7 Nov 2009 01:32:59 +0000 (01:32 +0000)]
Revert following patches to fix llvmgcc bootstrap.
86289, 86278, 86270, 86267, 86266 & 86264
Chris, please take a look.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86321 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoMy previous patch (r84124) for setting the encoding bits 4 and 7 of DPSoRegFrm
Johnny Chen [Sat, 7 Nov 2009 00:54:36 +0000 (00:54 +0000)]
My previous patch (r84124) for setting the encoding bits 4 and 7 of DPSoRegFrm
was wrong and too aggressive in the sense that DPSoRegFrm includes both constant
shifts (with Inst{4} = 0) and register controlled shifts (with Inst{4} = 1 and
Inst{7} = 0).  The 'rr' fragment of the multiclass definitions actually means
register/register with no shift, see A8-11.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86319 91177308-0d34-0410-b5e6-96231b3b80d8

14 years ago- new SROA mallocs should have the mallocs running-or'ed, not the malloc's bitcast
Victor Hernandez [Sat, 7 Nov 2009 00:41:19 +0000 (00:41 +0000)]
- new SROA mallocs should have the mallocs running-or'ed, not the malloc's bitcast
- fix ProcessInternalGlobal() debug output

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86317 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFit in 80 columns
Victor Hernandez [Sat, 7 Nov 2009 00:36:50 +0000 (00:36 +0000)]
Fit in 80 columns

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86316 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAvoid "ambiguous 'else'" warning from gcc.
Jeffrey Yasskin [Sat, 7 Nov 2009 00:26:47 +0000 (00:26 +0000)]
Avoid "ambiguous 'else'" warning from gcc.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86314 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRe-commit r86077 now that r86290 fixes the 179.art and 175.vpr ARM regressions.
Victor Hernandez [Sat, 7 Nov 2009 00:16:28 +0000 (00:16 +0000)]
Re-commit r86077 now that r86290 fixes the 179.art and 175.vpr ARM regressions.

Here is the original commit message:

This commit updates malloc optimizations to operate on malloc calls that have constant int size arguments.

Update CreateMalloc so that its callers specify the size to allocate:
MallocInst-autoupgrade users use non-TargetData-computed allocation sizes.
Optimization uses use TargetData to compute the allocation size.

Now that malloc calls can have constant sizes, update isArrayMallocHelper() to use TargetData to determine the size of the malloced type and the size of malloced arrays.
Extend getMallocType() to support malloc calls that have non-bitcast uses.

Update OptimizeGlobalAddressOfMalloc() to optimize malloc calls that have non-bitcast uses.  The bitcast use of a malloc call has to be treated specially here because the uses of the bitcast need to be replaced and the bitcast needs to be erased (just like the malloc call) for OptimizeGlobalAddressOfMalloc() to work correctly.

Update PerformHeapAllocSRoA() to optimize malloc calls that have non-bitcast uses.  The bitcast use of the malloc is not handled specially here because ReplaceUsesOfMallocWithGlobal replaces through the bitcast use.

Update OptimizeOnceStoredGlobal() to not care about the malloc calls' bitcast use.

Update all globalopt malloc tests to not rely on autoupgraded-MallocInsts, but instead use explicit malloc calls with correct allocation sizes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86311 91177308-0d34-0410-b5e6-96231b3b80d8

14 years ago80-columns
Jim Grosbach [Sat, 7 Nov 2009 00:13:30 +0000 (00:13 +0000)]
80-columns

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86310 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoGive the JITResolver a direct pointer to its JITEmitter, and use that instead
Jeffrey Yasskin [Sat, 7 Nov 2009 00:00:10 +0000 (00:00 +0000)]
Give the JITResolver a direct pointer to its JITEmitter, and use that instead
of going through the global TheJIT variable.  This makes it easier to use
features of JITEmitter that aren't in JITCodeEmitter for fixing PR5201.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86305 91177308-0d34-0410-b5e6-96231b3b80d8

14 years ago- Add pseudo instructions tLDRpci_pic and t2LDRpci_pic which does a pc-relative
Evan Cheng [Fri, 6 Nov 2009 23:52:48 +0000 (23:52 +0000)]
- Add pseudo instructions tLDRpci_pic and t2LDRpci_pic which does a pc-relative
  load of a GV from constantpool and then add pc. It allows the code sequence to
  be rematerializable so it would be hoisted by machine licm.
- Add a late pass to break these pseudo instructions into a number of real
  instructions. Also move the code in Thumb2 IT pass that breaks up t2MOVi32imm
  to this pass. This is done before post regalloc scheduling to allow the
  scheduler to proper schedule these instructions. It also allow them to be
  if-converted and shrunk by later passes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86304 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoHonour subreg machine operands during asmprinting
Anton Korobeynikov [Fri, 6 Nov 2009 23:45:15 +0000 (23:45 +0000)]
Honour subreg machine operands during asmprinting

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86303 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoPrint VMOV (immediate) operands as hexadecimal values. Apple's assembler
Bob Wilson [Fri, 6 Nov 2009 23:33:28 +0000 (23:33 +0000)]
Print VMOV (immediate) operands as hexadecimal values.  Apple's assembler
will not accept negative values for these.  LLVM's default operand printing
sign extends values, so that valid unsigned values appear as negative
immediates.  Print all VMOV immediate operands as hex values to resolve this.
Radar 7372576.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86301 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix a bug where we'd call SplitBlockPredecessors with a pred in the
Chris Lattner [Fri, 6 Nov 2009 23:19:58 +0000 (23:19 +0000)]
Fix a bug where we'd call SplitBlockPredecessors with a pred in the
set only once even if it has multiple edges to BB.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86299 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix a broken test.
Bob Wilson [Fri, 6 Nov 2009 23:06:42 +0000 (23:06 +0000)]
Fix a broken test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86298 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix comment typos.
Bob Wilson [Fri, 6 Nov 2009 22:38:38 +0000 (22:38 +0000)]
Fix comment typos.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86295 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRemove ARMPCLabelIndex from ARMISelLowering. Use ARMFunctionInfo::createConstPoolEntr...
Evan Cheng [Fri, 6 Nov 2009 22:24:13 +0000 (22:24 +0000)]
Remove ARMPCLabelIndex from ARMISelLowering. Use ARMFunctionInfo::createConstPoolEntryUId() instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86294 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoCallInst::CreateMalloc() and CallInst::CreateFree() need to create calls with correct...
Victor Hernandez [Fri, 6 Nov 2009 21:43:21 +0000 (21:43 +0000)]
CallInst::CreateMalloc() and CallInst::CreateFree() need to create calls with correct calling convention

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86290 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRemove function left over from other jump threading cleanup.
Eli Friedman [Fri, 6 Nov 2009 21:24:57 +0000 (21:24 +0000)]
Remove function left over from other jump threading cleanup.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86289 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agofix typo
Gabor Greif [Fri, 6 Nov 2009 20:10:46 +0000 (20:10 +0000)]
fix typo

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86281 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix a problem discovered on self host.
Chris Lattner [Fri, 6 Nov 2009 19:21:48 +0000 (19:21 +0000)]
Fix a problem discovered on self host.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86278 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoremove more code subsumed by r86264
Chris Lattner [Fri, 6 Nov 2009 18:24:32 +0000 (18:24 +0000)]
remove more code subsumed by r86264

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86270 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoTolerate invalid derived type.
Devang Patel [Fri, 6 Nov 2009 18:24:05 +0000 (18:24 +0000)]
Tolerate invalid derived type.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86269 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoeliminate some more code subsumed by r86264
Chris Lattner [Fri, 6 Nov 2009 18:22:54 +0000 (18:22 +0000)]
eliminate some more code subsumed by r86264

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86267 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoremove now redundant code, r86264 handles this case.
Chris Lattner [Fri, 6 Nov 2009 18:20:58 +0000 (18:20 +0000)]
remove now redundant code, r86264 handles this case.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86266 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoExtend jump threading to support much more general threading
Chris Lattner [Fri, 6 Nov 2009 18:15:14 +0000 (18:15 +0000)]
Extend jump threading to support much more general threading
predicates.  This allows us to jump thread things like:

_ZN12StringSwitchI5ColorE4CaseILj7EEERS1_RAT__KcRKS0_.exit119:
  %tmp1.i24166 = phi i8 [ 1, %bb5.i117 ], [ %tmp1.i24165, %_Z....exit ], [ %tmp1.i24165, %bb4.i114 ]
  %toBoolnot.i87 = icmp eq i8 %tmp1.i24166, 0     ; <i1> [#uses=1]
  %tmp4.i90 = icmp eq i32 %tmp2.i, 6              ; <i1> [#uses=1]
  %or.cond173 = and i1 %toBoolnot.i87, %tmp4.i90  ; <i1> [#uses=1]
  br i1 %or.cond173, label %bb4.i96, label %_ZN12...

Where it is "obvious" that when coming from %bb5.i117 that the 'and' is always
false.  This triggers a surprisingly high number of times in the testsuite,
and gets us closer to generating good code for doug's strswitch testcase.

This also make a bunch of other code in jump threading redundant, I'll rip
out in the next patch.  This survived an enable-checking llvm-gcc bootstrap.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86264 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoUse WriteAsOperand to print GlobalAddress MachineOperands. This
Dan Gohman [Fri, 6 Nov 2009 18:03:10 +0000 (18:03 +0000)]
Use WriteAsOperand to print GlobalAddress MachineOperands. This
prints them with the leading '@'.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86261 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoDo not bother to emit debug info for nameless global variable.
Devang Patel [Fri, 6 Nov 2009 17:58:12 +0000 (17:58 +0000)]
Do not bother to emit debug info for nameless global variable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86259 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoPass StringRef by value.
Daniel Dunbar [Fri, 6 Nov 2009 10:58:06 +0000 (10:58 +0000)]
Pass StringRef by value.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86251 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoclang++ points out that this is pointless.
Chris Lattner [Fri, 6 Nov 2009 06:33:01 +0000 (06:33 +0000)]
clang++ points out that this is pointless.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86239 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoremove some more Context arguments.
Chris Lattner [Fri, 6 Nov 2009 05:59:53 +0000 (05:59 +0000)]
remove some more Context arguments.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86235 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoremove a bunch of extraneous LLVMContext arguments
Chris Lattner [Fri, 6 Nov 2009 04:27:31 +0000 (04:27 +0000)]
remove a bunch of extraneous LLVMContext arguments
from various APIs, addressing PR5325.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86231 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoNewNighlytTest: Fix timestamp format to actually make sense (it was missing the hour).
Daniel Dunbar [Fri, 6 Nov 2009 04:12:13 +0000 (04:12 +0000)]
NewNighlytTest: Fix timestamp format to actually make sense (it was missing the hour).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86229 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoNewNightlyTest: Add -noclean option, which doesn't run 'make clean' before building...
Daniel Dunbar [Fri, 6 Nov 2009 04:12:07 +0000 (04:12 +0000)]
NewNightlyTest: Add -noclean option, which doesn't run 'make clean' before building LLVM (for testing).

Also, switch to always running 'make clean' in the test-suite directories.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86228 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoNewNightlyTest: Unbreak passing the build directory via a positional argument.
Daniel Dunbar [Fri, 6 Nov 2009 04:12:02 +0000 (04:12 +0000)]
NewNightlyTest: Unbreak passing the build directory via a positional argument.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86227 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoNewNightlyTest: Add -llvmgccdir as alternative to environment variable.
Daniel Dunbar [Fri, 6 Nov 2009 04:11:29 +0000 (04:11 +0000)]
NewNightlyTest: Add -llvmgccdir as alternative to environment variable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86226 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoRevert r86077 because it caused crashes in 179.art and 175.vpr on ARM
Victor Hernandez [Fri, 6 Nov 2009 01:33:24 +0000 (01:33 +0000)]
Revert r86077 because it caused crashes in 179.art and 175.vpr on ARM

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86213 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoDo not try to emit debug info entry for dead global variable.
Devang Patel [Fri, 6 Nov 2009 01:30:04 +0000 (01:30 +0000)]
Do not try to emit debug info entry for dead global variable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86212 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoDon't print a redundant tab for inline asm, and do use the new printKill.
Dan Gohman [Fri, 6 Nov 2009 00:19:43 +0000 (00:19 +0000)]
Don't print a redundant tab for inline asm, and do use the new printKill.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86206 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdd a bunch of missing "template" keywords to disambiguate dependent template names...
Douglas Gregor [Fri, 6 Nov 2009 00:12:53 +0000 (00:12 +0000)]
Add a bunch of missing "template" keywords to disambiguate dependent template names. GCC eats this ill-formed code, Clang does not. I already filed PR5404 to improve recovery in this case

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86204 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix PR5315, original patch by Nicolas Capens!
Eric Christopher [Fri, 6 Nov 2009 00:11:57 +0000 (00:11 +0000)]
Fix PR5315, original patch by Nicolas Capens!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86203 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFactor out the printing of the leading tab into printInlineAsm.
Dan Gohman [Fri, 6 Nov 2009 00:04:54 +0000 (00:04 +0000)]
Factor out the printing of the leading tab into printInlineAsm.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86199 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoMake printImplicitDef and printKill non-virtual, since they don't
Dan Gohman [Fri, 6 Nov 2009 00:04:05 +0000 (00:04 +0000)]
Make printImplicitDef and printKill non-virtual, since they don't
need to be overridden.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86198 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoUse SUBREG_TO_REG instead of INSERT_SUBREG to model x86-64's
Dan Gohman [Thu, 5 Nov 2009 23:53:08 +0000 (23:53 +0000)]
Use SUBREG_TO_REG instead of INSERT_SUBREG to model x86-64's
implicit zero-extend.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86196 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoTeach LSR to avoid calling SplitCriticalEdge on edges with indirectbr.
Dan Gohman [Thu, 5 Nov 2009 23:34:59 +0000 (23:34 +0000)]
Teach LSR to avoid calling SplitCriticalEdge on edges with indirectbr.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86193 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoUpdate these tests for the new label names.
Dan Gohman [Thu, 5 Nov 2009 23:31:40 +0000 (23:31 +0000)]
Update these tests for the new label names.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86192 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoFix the label name generation for address-taken labels to avoid potential
Dan Gohman [Thu, 5 Nov 2009 23:14:35 +0000 (23:14 +0000)]
Fix the label name generation for address-taken labels to avoid potential
problems with name collisions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86189 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoMake a few more LLVM headers parsable as standalone headers.
Douglas Gregor [Thu, 5 Nov 2009 23:01:30 +0000 (23:01 +0000)]
Make a few more LLVM headers parsable as standalone headers.

Fix some problems with the hidden copy constructors for
ImmutableMap/ImmutableSet found by Clang++.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86186 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoTeach lit's SyntaxCheckTest two new tricks:
Douglas Gregor [Thu, 5 Nov 2009 22:58:04 +0000 (22:58 +0000)]
Teach lit's SyntaxCheckTest two new tricks:
  - skip .svn directories
  - add a set of excluded filenames so we can easily skip tests

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86185 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAdded support for renumbering existing index list elements. Removed some junk from...
Lang Hames [Thu, 5 Nov 2009 22:20:57 +0000 (22:20 +0000)]
Added support for renumbering existing index list elements. Removed some junk from the initial numbering code in runOnMachineFunction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86184 91177308-0d34-0410-b5e6-96231b3b80d8

14 years agoAvoid calling getUniqueExitBlocks from within LoopSimplify, as it depends
Dan Gohman [Thu, 5 Nov 2009 21:48:32 +0000 (21:48 +0000)]
Avoid calling getUniqueExitBlocks from within LoopSimplify, as it depends
on loops having dedicated exits, which LoopSimplify can no longer always
guarantee.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@86181 91177308-0d34-0410-b5e6-96231b3b80d8