oota-llvm.git
12 years agoAdd missing VEX predicates to VMOVSDto64rr/VMOVSDto64mr. This fixes a few
Chad Rosier [Tue, 10 Jan 2012 22:14:06 +0000 (22:14 +0000)]
Add missing VEX predicates to VMOVSDto64rr/VMOVSDto64mr.  This fixes a few
failing test cases on our internal AVX nightly tester.
rdar://10663637

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147881 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoLet asm parser query asm syntax dialect.
Devang Patel [Tue, 10 Jan 2012 21:49:42 +0000 (21:49 +0000)]
Let asm parser query asm syntax dialect.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147880 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoThis is the matching change for the data structure name changes for the
Kevin Enderby [Tue, 10 Jan 2012 21:12:34 +0000 (21:12 +0000)]
This is the matching change for the data structure name changes for the
functional change in r147860 to use DW_TAG_label's instead TAG_subprogram's.
This only changes names and updates comments.  No functional change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147877 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoARM updating VST2 pseudo-lowering fixed vs. register update.
Jim Grosbach [Tue, 10 Jan 2012 21:11:12 +0000 (21:11 +0000)]
ARM updating VST2 pseudo-lowering fixed vs. register update.

rdar://10663487

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147876 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix some leftover control reaches end of non-void function warnings.
Benjamin Kramer [Tue, 10 Jan 2012 20:47:20 +0000 (20:47 +0000)]
Fix some leftover control reaches end of non-void function warnings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147874 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTeach the triple library about the androideabi environment.
Chandler Carruth [Tue, 10 Jan 2012 19:46:00 +0000 (19:46 +0000)]
Teach the triple library about the androideabi environment.

Patch by Evgeniy Stepanov.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147871 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove default case for covered enum outside of switch.
Richard Smith [Tue, 10 Jan 2012 19:43:09 +0000 (19:43 +0000)]
Move default case for covered enum outside of switch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147870 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFor i386, don't use the generic code.
Bill Wendling [Tue, 10 Jan 2012 19:41:30 +0000 (19:41 +0000)]
For i386, don't use the generic code.

As the comment around 7746 says, it's better to use the x87 extended precision
here than SSE. And the generic code doesn't know how to do that. It also regains
the speed lost for the uint64_to_float.c testcase.
<rdar://problem/10669858>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147869 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix a -Wreturn-type warning in g++.
Richard Smith [Tue, 10 Jan 2012 19:10:22 +0000 (19:10 +0000)]
Fix a -Wreturn-type warning in g++.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147867 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCleanup these asserts to follow common LLVM style and coding
Chandler Carruth [Tue, 10 Jan 2012 18:18:52 +0000 (18:18 +0000)]
Cleanup these asserts to follow common LLVM style and coding
conventions. Also, clarify the grouping of one of the asserts to silence
-Wparentheses.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147863 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd 'llvm_unreachable' to passify GCC's understanding of the constraints
Chandler Carruth [Tue, 10 Jan 2012 18:08:01 +0000 (18:08 +0000)]
Add 'llvm_unreachable' to passify GCC's understanding of the constraints
of several newly un-defaulted switches. This also helps optimizers
(including LLVM's) recognize that every case is covered, and we should
assume as much.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147861 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoVarious crash reporting tools have a problem with the dwarf generated for
Kevin Enderby [Tue, 10 Jan 2012 17:52:29 +0000 (17:52 +0000)]
Various crash reporting tools have a problem with the dwarf generated for
assembly source when it generates the TAG_subprogram dwarf debug info for
the labels that have nothing between them as in this bit of assembly source:

% cat ZeroLength.s
_func1:
_func2:
 nop

One solution would be to not emit the subsequent labels with the same address
and use the next label with a different address or the end of the section for
the AT_high_pc value of the TAG_subprogram.

Turns out in llvm-mc it is not possible in all cases to determine of two
symbols have the same value at the point we put out the TAG_subprogram dwarf
debug info.

So we will have llvm-mc instead of putting out TAG_subprogram's put out
DW_TAG_label's.  And the DW_TAG_label does not have a AT_high_pc value which
avoids the problem.

This commit is only the functional change to make the diffs clear as to what is
really being changed.  The next commit will be to clean up the names of such
things like MCGenDwarfSubprogramEntry to something like MCGenDwarfLabelEntry.

rdar://10666925

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147860 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd definition for intel asm variant.
Devang Patel [Tue, 10 Jan 2012 17:51:54 +0000 (17:51 +0000)]
Add definition for intel asm variant.

Right now, this just adds additional entries in match table. The parser does not use them yet.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147859 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRecord asm variant id in MatchEntry and check it while matching instruction.
Devang Patel [Tue, 10 Jan 2012 17:50:43 +0000 (17:50 +0000)]
Record asm variant id in MatchEntry and check it while matching instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147858 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove unnecessary default cases in switches that cover all enum values.
David Blaikie [Tue, 10 Jan 2012 16:47:17 +0000 (16:47 +0000)]
Remove unnecessary default cases in switches that cover all enum values.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147855 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix a bug in the legalization of shuffle vectors. When we emulate shuffles using...
Nadav Rotem [Tue, 10 Jan 2012 14:28:46 +0000 (14:28 +0000)]
Fix a bug in the legalization of shuffle vectors. When we emulate shuffles using BUILD_VECTORS we may be using a BV of different type. Make sure to cast it back.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147851 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd definitions for AMD's bobcat (aka btver1)
Benjamin Kramer [Tue, 10 Jan 2012 11:50:02 +0000 (11:50 +0000)]
Add definitions for AMD's bobcat (aka btver1)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147846 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix a crash in AVX2 when trying to broadcast a double into a 128-bit vector. There...
Craig Topper [Tue, 10 Jan 2012 08:23:59 +0000 (08:23 +0000)]
Fix a crash in AVX2 when trying to broadcast a double into a 128-bit vector. There is no vbroadcastsd xmm, but we do need to support 64-bit integers broadcasted into xmm. Also factor the AVX check into the isVectorBroadcast function. This makes more sense since the AVX2 check was already inside.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147844 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove hasXMM/hasXMMInt functions. Move callers to hasSSE1/hasSSE2. This is the final...
Craig Topper [Tue, 10 Jan 2012 06:54:16 +0000 (06:54 +0000)]
Remove hasXMM/hasXMMInt functions. Move callers to hasSSE1/hasSSE2. This is the final piece to remove the AVX hack that disabled SSE.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147843 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove hasSSE*orAVX functions and change all callers to use just hasSSE*. AVX is...
Craig Topper [Tue, 10 Jan 2012 06:37:29 +0000 (06:37 +0000)]
Remove hasSSE*orAVX functions and change all callers to use just hasSSE*. AVX is now an SSE level and no longer disables SSE checks.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147842 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoInstruction selection priority fixes to remove the XMM/XMMInt/orAVX predicates. Anoth...
Craig Topper [Tue, 10 Jan 2012 06:30:56 +0000 (06:30 +0000)]
Instruction selection priority fixes to remove the XMM/XMMInt/orAVX predicates. Another commit will remove orAVX functions from X86SubTarget.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147841 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAllow machine-cse to look across MBB boundary when cse'ing instructions that
Evan Cheng [Tue, 10 Jan 2012 02:02:58 +0000 (02:02 +0000)]
Allow machine-cse to look across MBB boundary when cse'ing instructions that
define physical registers. It's currently very restrictive, only catching
cases where the CE is in an immediate (and only) predecessor. But it catches
a surprising large number of cases.

rdar://10660865

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147827 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoEnable LSR IV Chains with sufficient heuristics.
Andrew Trick [Tue, 10 Jan 2012 01:45:08 +0000 (01:45 +0000)]
Enable LSR IV Chains with sufficient heuristics.

These heuristics are sufficient for enabling IV chains by
default. Performance analysis has been done for i386, x86_64, and
thumbv7. The optimization is rarely important, but can significantly
speed up certain cases by eliminating spill code within the
loop. Unrolled loops are prime candidates for IV chains. In many
cases, the final code could still be improved with more target
specific optimization following LSR. The goal of this feature is for
LSR to make the best choice of induction variables.

Instruction selection may not completely take advantage of this
feature yet. As a result, there could be cases of slight code size
increase.

Code size can be worse on x86 because it doesn't support postincrement
addressing. In fact, when chains are formed, you may see redundant
address plus stride addition in the addressing mode. GenerateIVChains
tries to compensate for the common cases.

On ARM, code size increase can be mitigated by using postincrement
addressing, but downstream codegen currently misses some opportunities.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147826 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAccurately model hardware alignment rounding.
Jakob Stoklund Olesen [Tue, 10 Jan 2012 01:34:59 +0000 (01:34 +0000)]
Accurately model hardware alignment rounding.

On Thumb, the displacement computation hardware uses the address of the
current instruction rouned down to a multiple of 4.  Include this
rounding in the UserOffset we compute for each instruction.

When inline asm is present, the instruction alignment may not be known.
Constrain the maximum displacement instead in that case.

This makes it possible for CreateNewWater() and OffsetIsInRange() to
agree about the valid displacements.  When they disagree, infinite
looping happens.

As always, test cases for this stuff are insane.

<rdar://problem/10660175>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147825 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove the logging streamer.
Rafael Espindola [Tue, 10 Jan 2012 00:40:39 +0000 (00:40 +0000)]
Remove the logging streamer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147820 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCatch runaway ARMConstantIslandPass even in -Asserts builds.
Jakob Stoklund Olesen [Mon, 9 Jan 2012 22:16:24 +0000 (22:16 +0000)]
Catch runaway ARMConstantIslandPass even in -Asserts builds.

The pass is prone to looping, and it is better to crash than loop
forever, even in a -Asserts build.

<rdar://problem/10660175>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147806 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix asm string wrt variants.
Devang Patel [Mon, 9 Jan 2012 21:32:02 +0000 (21:32 +0000)]
Fix asm string wrt variants.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147805 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUse descriptive variable name and remove incorrect operand number check.
Devang Patel [Mon, 9 Jan 2012 21:30:46 +0000 (21:30 +0000)]
Use descriptive variable name and remove incorrect operand number check.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147802 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdding IV chain generation to LSR.
Andrew Trick [Mon, 9 Jan 2012 21:18:52 +0000 (21:18 +0000)]
Adding IV chain generation to LSR.

After collecting chains, check if any should be materialized. If so,
hide the chained IV users from the LSR solver. LSR will only solve for
the head of the chain. GenerateIVChains will then materialize the
chained IV users by computing the IV relative to its previous value in
the chain.

In theory, chained IV users could be exposed to LSR's solver. This
would be considerably complicated to implement and I'm not aware of a
case where we need it. In practice it's more important to
intelligently prune the search space of nontrivial loops before
running the solver, otherwise the solver is often forced to prune the
most optimal solutions. Hiding the chained users does this well, so
that LSR is more likely to find the best IV for the chain as a whole.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147801 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdding collection of IV chains to LSR.
Andrew Trick [Mon, 9 Jan 2012 19:50:34 +0000 (19:50 +0000)]
Adding collection of IV chains to LSR.

This collects a set of IV uses within the loop whose values can be
computed relative to each other in a sequence. Following checkins will
make use of this information.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147797 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSplit AsmParser into two components - AsmParser and AsmParserVariant
Devang Patel [Mon, 9 Jan 2012 19:13:28 +0000 (19:13 +0000)]
Split AsmParser into two components - AsmParser and AsmParserVariant

AsmParser holds info specific to target parser.
AsmParserVariant holds info specific to asm variants supported by the target.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147787 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago"Minor LSR debugging stuff"
Andrew Trick [Mon, 9 Jan 2012 18:58:16 +0000 (18:58 +0000)]
"Minor LSR debugging stuff"

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147785 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUpdate language check. Do not ignore DW_LANG_Python.
Devang Patel [Mon, 9 Jan 2012 17:49:47 +0000 (17:49 +0000)]
Update language check. Do not ignore DW_LANG_Python.
Patch by  Joe Groff!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147781 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove assert to the right place.
Benjamin Kramer [Mon, 9 Jan 2012 17:36:29 +0000 (17:36 +0000)]
Move assert to the right place.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147779 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoInstCombine: Teach foldLogOpOfMaskedICmpsHelper that sign bit tests are bit tests.
Benjamin Kramer [Mon, 9 Jan 2012 17:23:27 +0000 (17:23 +0000)]
InstCombine: Teach foldLogOpOfMaskedICmpsHelper that sign bit tests are bit tests.

This subsumes several other transforms while enabling us to catch more cases.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147777 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDon't rely on the fact that shift values are never very large, and thus
Chandler Carruth [Mon, 9 Jan 2012 09:47:25 +0000 (09:47 +0000)]
Don't rely on the fact that shift values are never very large, and thus
this substraction will result in small negative numbers at worst which
become very large positive numbers on assignment and are thus caught by
the <=4 check on the next line. The >0 check clearly intended to catch
these as negative numbers.

Spotted by inspection, and impossible to trigger given the shift widths
that can be used.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147773 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCleanup and FileCheck-ize a test.
Chandler Carruth [Mon, 9 Jan 2012 09:44:26 +0000 (09:44 +0000)]
Cleanup and FileCheck-ize a test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147772 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove AVX hack in X86Subtarget. AVX/AVX2 are now treated as an SSE level. Predicate...
Craig Topper [Mon, 9 Jan 2012 09:02:13 +0000 (09:02 +0000)]
Remove AVX hack in X86Subtarget. AVX/AVX2 are now treated as an SSE level. Predicate functions have been altered to maintain previous names and behavior.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147770 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd HasAVX predicate to some of the AVX patterns.
Craig Topper [Mon, 9 Jan 2012 08:34:00 +0000 (08:34 +0000)]
Add HasAVX predicate to some of the AVX patterns.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147769 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoReorder a bunch of patterns to put the AVX version first thus giving it priority...
Craig Topper [Mon, 9 Jan 2012 08:10:38 +0000 (08:10 +0000)]
Reorder a bunch of patterns to put the AVX version first thus giving it priority over the SSE version. Another step towards trying to remove the AVX hack that disables SSE from X86Subtarget.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147768 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoClean up patterns for MOVNT*. Not sure why there were floating point types on MOVNTPS...
Craig Topper [Mon, 9 Jan 2012 06:52:46 +0000 (06:52 +0000)]
Clean up patterns for MOVNT*. Not sure why there were floating point types on MOVNTPS and MOVNTDQ. And v4i64 was completely missing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147767 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMark MOVNTI as being supported in SSE2 OR AVX mode. This instruction has no AVX equiv...
Craig Topper [Mon, 9 Jan 2012 06:38:55 +0000 (06:38 +0000)]
Mark MOVNTI as being supported in SSE2 OR AVX mode. This instruction has no AVX equivalent so we should use the SSE version.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147766 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMove SSE2 logical operations PAND/POR/PXOR/PANDN above SSE1 logical operations ANDPS...
Craig Topper [Mon, 9 Jan 2012 05:07:01 +0000 (05:07 +0000)]
Move SSE2 logical operations PAND/POR/PXOR/PANDN above SSE1 logical operations ANDPS/ORPS/XORPS/ANDNPS. This fixes a pattern ordering issue that meant that the SSE2 instructions could never be directly selected since the SSE1 patterns would always match first. This is largely moot with the ExeDepsFix pass, but I'm trying to audit for all such ordering issues.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147765 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoChange some places that were checking for AVX OR SSE1/2 to use hasXMM/hasXMMInt inste...
Craig Topper [Mon, 9 Jan 2012 02:28:15 +0000 (02:28 +0000)]
Change some places that were checking for AVX OR SSE1/2 to use hasXMM/hasXMMInt instead. Also fix one place that checked SSE3, but accidentally excluded AVX to use hasSSE3orAVX. This is a step towards removing the AVX hack from the X86Subtarget.h

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147764 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDon't print an unused label before .cfi_endproc.
Rafael Espindola [Mon, 9 Jan 2012 00:17:29 +0000 (00:17 +0000)]
Don't print an unused label before .cfi_endproc.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147763 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDon't disable MMX support when AVX is enabled. Fix predicates for MMX instructions...
Craig Topper [Mon, 9 Jan 2012 00:11:29 +0000 (00:11 +0000)]
Don't disable MMX support when AVX is enabled. Fix predicates for MMX instructions that were added along with SSE instructions to check for AVX in addition to SSE level.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147762 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoEnable FISTTP* instructions when AVX is enabled.
Craig Topper [Sun, 8 Jan 2012 23:04:21 +0000 (23:04 +0000)]
Enable FISTTP* instructions when AVX is enabled.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147758 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTweak my last commit to be less conservative about uses.
Benjamin Kramer [Sun, 8 Jan 2012 21:12:51 +0000 (21:12 +0000)]
Tweak my last commit to be less conservative about uses.

We still save an instruction when just the "and" part is replaced.
Also change the code to match comments more closely.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147753 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDon't forget to transfer implicit uses of return instruction.
Evan Cheng [Sun, 8 Jan 2012 20:41:16 +0000 (20:41 +0000)]
Don't forget to transfer implicit uses of return instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147752 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAvoid eraseing copies from a reserved register unless the definition can be
Evan Cheng [Sun, 8 Jan 2012 19:52:28 +0000 (19:52 +0000)]
Avoid eraseing copies from a reserved register unless the definition can be
safely proven not to have been clobbered. No small test case possible.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147751 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoInstCombine: If we have a bit test and a sign test anded/ored together, merge the...
Benjamin Kramer [Sun, 8 Jan 2012 18:32:24 +0000 (18:32 +0000)]
InstCombine: If we have a bit test and a sign test anded/ored together, merge the sign bit into the bit test.

This is common in bit field code, e.g. checking if the first or the last bit of a bit field is set.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147749 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoReverted commit #147601 upon Evan's request.
Victor Umansky [Sun, 8 Jan 2012 17:20:33 +0000 (17:20 +0000)]
Reverted commit #147601 upon Evan's request.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147748 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove MCELFStreamer.h.
Rafael Espindola [Sat, 7 Jan 2012 23:18:39 +0000 (23:18 +0000)]
Remove MCELFStreamer.h.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147745 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoDon't print a label before .cfi_startproc when we don't need to. This makes
Rafael Espindola [Sat, 7 Jan 2012 22:42:19 +0000 (22:42 +0000)]
Don't print a label before .cfi_startproc when we don't need to. This makes
the produce assembly when using CFI just a bit more readable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147743 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMake clever use of alignment and padding to shrink GlobalValue.
Benjamin Kramer [Sat, 7 Jan 2012 21:17:16 +0000 (21:17 +0000)]
Make clever use of alignment and padding to shrink GlobalValue.

-8 bytes on x86_64, no change on x86.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147742 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMatch SelectionDAG logic for enabling movt.
Jakob Stoklund Olesen [Sat, 7 Jan 2012 20:49:15 +0000 (20:49 +0000)]
Match SelectionDAG logic for enabling movt.

Darwin doesn't do static, and ELF targets only support static.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147740 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix typo in the X86 backend readme. Patch from Jaeden Amero.
Craig Topper [Sat, 7 Jan 2012 20:35:21 +0000 (20:35 +0000)]
Fix typo in the X86 backend readme. Patch from Jaeden Amero.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147739 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove VectorExtras. This unused helper was written for a type of API that is discour...
Benjamin Kramer [Sat, 7 Jan 2012 19:42:13 +0000 (19:42 +0000)]
Remove VectorExtras. This unused helper was written for a type of API that is discouraged now.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147738 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRemove unnecessary check of hasAVX(). It's already included in hasXMM().
Craig Topper [Sat, 7 Jan 2012 18:48:43 +0000 (18:48 +0000)]
Remove unnecessary check of hasAVX(). It's already included in hasXMM().

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147734 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoReplace some uses of hasNUsesOfValue(0, X) with !hasAnyUseOfValue(X)
Craig Topper [Sat, 7 Jan 2012 18:31:09 +0000 (18:31 +0000)]
Replace some uses of hasNUsesOfValue(0, X) with !hasAnyUseOfValue(X)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147733 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoPort the trick to skip the check for empty buckets from StringMap to DenseMap.
Benjamin Kramer [Sat, 7 Jan 2012 13:12:07 +0000 (13:12 +0000)]
Port the trick to skip the check for empty buckets from StringMap to DenseMap.

This should fix the odd behavior that find() is slower than lookup().

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147731 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd some DAG combines for SUBC/SUBE. If nothing uses the carry/borrow out of subc...
Craig Topper [Sat, 7 Jan 2012 09:06:39 +0000 (09:06 +0000)]
Add some DAG combines for SUBC/SUBE. If nothing uses the carry/borrow out of subc, turn it into a sub. Turn (subc x, x) into 0 with no borrow. Turn (subc x, 0) into x with no borrow. Turn (subc -1, x) into (xor x, -1) with no borrow. Turn sube with no borrow in into subc.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147728 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix TableGen so that it will emit the correct signature for FastEmit_f:
Cameron Zwarich [Sat, 7 Jan 2012 08:18:37 +0000 (08:18 +0000)]
Fix TableGen so that it will emit the correct signature for FastEmit_f:

  /// FastEmit_f - This method is called by target-independent code
  /// to request that an instruction with the given type, opcode, and
  /// floating-point immediate operand be emitted.
  virtual unsigned FastEmit_f(MVT VT,
                              MVT RetVT,
                              unsigned Opcode,
                              const ConstantFP *FPImm);

Currently, it emits an accidentally overloaded version without the const on the
ConstantFP*. This doesn't affect anything in the tree, since nothing causes that
method to be autogenerated, but I have been playing with some ARM TableGen
refactorings that hit this problem.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147727 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoOptimize reserved register coalescing.
Jakob Stoklund Olesen [Sat, 7 Jan 2012 07:39:50 +0000 (07:39 +0000)]
Optimize reserved register coalescing.

Reserved registers don't have proper live ranges, their LiveInterval
simply has a snippet of liveness for each def.  Virtual registers with a
single value that is a copy of a reserved register (typically %esp) can
be coalesced with the reserved register if the live range doesn't
overlap any reserved register defs.

When coalescing with a reserved register, don't modify the reserved
register live range.  Just leave it as a bunch of dead defs.  This
eliminates quadratic coalescer behavior in i386 functions with many
function calls.

PR11699

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147726 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUse the 'regalloc' debug tag for most register allocator tracing.
Jakob Stoklund Olesen [Sat, 7 Jan 2012 07:39:47 +0000 (07:39 +0000)]
Use the 'regalloc' debug tag for most register allocator tracing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147725 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoEnable redundant phi elimination after LSR.
Andrew Trick [Sat, 7 Jan 2012 07:08:17 +0000 (07:08 +0000)]
Enable redundant phi elimination after LSR.

This will be more important as we extend the LSR pass in ways that don't rely on the formula solver. In particular, we need it for constructing IV chains.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147724 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix dead link
Eli Bendersky [Sat, 7 Jan 2012 04:11:27 +0000 (04:11 +0000)]
Fix dead link

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147721 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUse getRegForValue() to materialize the address of ARM globals.
Jakob Stoklund Olesen [Sat, 7 Jan 2012 04:07:22 +0000 (04:07 +0000)]
Use getRegForValue() to materialize the address of ARM globals.

This enables basic local CSE, giving us 20% smaller code for
consumer-typeset in -O0 builds.

<rdar://problem/10658692>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147720 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRevert part of r147716. Looks like x87 instructions kill markers are all messed
Evan Cheng [Sat, 7 Jan 2012 03:35:48 +0000 (03:35 +0000)]
Revert part of r147716. Looks like x87 instructions kill markers are all messed
up so branch folding pass can't use the scavenger. :-(  This doesn't breaks
anything currently. It just means targets which do not carefully update kill
markers cannot run post-ra scheduler (not new, it has always been the case).

We should fix this at some point since it's really hacky.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147719 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoLSR: Don't optimize loops if an outer loop has no preheader.
Andrew Trick [Sat, 7 Jan 2012 03:16:50 +0000 (03:16 +0000)]
LSR: Don't optimize loops if an outer loop has no preheader.

LoopSimplify may not run on some outer loops, e.g. because of indirect
branches. SCEVExpander simply cannot handle outer loops with no preheaders.
Fixes rdar://10655343 SCEVExpander segfault.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147718 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSplit Finish into Finish and FinishImpl to have a common place to do end of
Rafael Espindola [Sat, 7 Jan 2012 03:13:18 +0000 (03:13 +0000)]
Split Finish into Finish and FinishImpl to have a common place to do end of
file error checking. Use that to error on an unfinished cfi_startproc.

The error is not nice, but is already better than a segmentation fault.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147717 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdded a late machine instruction copy propagation pass. This catches
Evan Cheng [Sat, 7 Jan 2012 03:02:36 +0000 (03:02 +0000)]
Added a late machine instruction copy propagation pass. This catches
opportunities that only present themselves after late optimizations
such as tail duplication .e.g.
## BB#1:
        movl    %eax, %ecx
        movl    %ecx, %eax
        ret

The register allocator also leaves some of them around (due to false
dep between copies from phi-elimination, etc.)

This required some changes in codegen passes. Post-ra scheduler and the
pseudo-instruction expansion passes have been moved after branch folding
and tail merging. They were before branch folding before because it did
not always update block livein's. That's fixed now. The pass change makes
independently since we want to properly schedule instructions after
branch folding / tail duplication.

rdar://10428165
rdar://10640363

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147716 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCopy implicit defs (e.g. r0) when changing tBX_RET to tPOP_RET. This bug is
Evan Cheng [Sat, 7 Jan 2012 02:55:54 +0000 (02:55 +0000)]
Copy implicit defs (e.g. r0) when changing tBX_RET to tPOP_RET. This bug is
exposed with an upcoming change will would delete the copy to return register
because there is no use! It's amazing anything works.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147715 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUse movw+movt in ARMFastISel::ARMMaterializeGV.
Jakob Stoklund Olesen [Sat, 7 Jan 2012 01:47:05 +0000 (01:47 +0000)]
Use movw+movt in ARMFastISel::ARMMaterializeGV.

This eliminates a lot of constant pool entries for -O0 builds of code
with many global variable accesses.

This speeds up -O0 codegen of consumer-typeset by 2x because the
constant island pass no longer has to look at thousands of constant pool
entries.

<rdar://problem/10629774>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147712 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoLSR: run DeleteDeadPhis before replaceCongruentPhis.
Andrew Trick [Sat, 7 Jan 2012 01:36:44 +0000 (01:36 +0000)]
LSR: run DeleteDeadPhis before replaceCongruentPhis.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147711 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoRefactor.
Devang Patel [Sat, 7 Jan 2012 01:33:34 +0000 (01:33 +0000)]
Refactor.

Store AsmParser info locally. A small step towards emitting match entries for multiple asm variants.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147710 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoCleanup comments and argument types related to my previous replaceCongruentPhis checkin.
Andrew Trick [Sat, 7 Jan 2012 01:29:21 +0000 (01:29 +0000)]
Cleanup comments and argument types related to my previous replaceCongruentPhis checkin.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147709 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoEliminate an error check that may not work with all asm syntax variants.
Devang Patel [Sat, 7 Jan 2012 01:22:23 +0000 (01:22 +0000)]
Eliminate an error check that may not work with all asm syntax variants.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147708 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoExtended replaceCongruentPhis to handle mixed phi types.
Andrew Trick [Sat, 7 Jan 2012 01:12:09 +0000 (01:12 +0000)]
Extended replaceCongruentPhis to handle mixed phi types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147707 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMake the 'x' constraint work for AVX registers as well.
Eric Christopher [Sat, 7 Jan 2012 01:02:09 +0000 (01:02 +0000)]
Make the 'x' constraint work for AVX registers as well.

Fixes rdar://10614894

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147704 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoMissing raw_ostream.h breaks MSVC build.
Andrew Trick [Sat, 7 Jan 2012 00:54:28 +0000 (00:54 +0000)]
Missing raw_ostream.h breaks MSVC build.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147703 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agocomment typo
Andrew Trick [Sat, 7 Jan 2012 00:29:20 +0000 (00:29 +0000)]
comment typo

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147701 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoExpose isNonConstantNegative to users of ScalarEvolution.
Andrew Trick [Sat, 7 Jan 2012 00:27:31 +0000 (00:27 +0000)]
Expose isNonConstantNegative to users of ScalarEvolution.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147700 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd comment.
Chad Rosier [Fri, 6 Jan 2012 23:45:47 +0000 (23:45 +0000)]
Add comment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147696 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAdd a comment and ensure that anyone else looking at this code doesn't start
Eric Christopher [Fri, 6 Jan 2012 23:03:37 +0000 (23:03 +0000)]
Add a comment and ensure that anyone else looking at this code doesn't start
to bleed from the eyes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147695 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUse const vector references instead of a vector copy. Spotted by Devang.
Eric Christopher [Fri, 6 Jan 2012 23:03:34 +0000 (23:03 +0000)]
Use const vector references instead of a vector copy. Spotted by Devang.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147694 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoUse -> instead of (*iter).
Eric Christopher [Fri, 6 Jan 2012 23:03:27 +0000 (23:03 +0000)]
Use -> instead of (*iter).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147693 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoEnable aligned NEON spilling by default.
Jakob Stoklund Olesen [Fri, 6 Jan 2012 22:19:37 +0000 (22:19 +0000)]
Enable aligned NEON spilling by default.

Experiments show this to be a small speedup for modern ARM cores.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147689 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoPut all IVUsers in the processed set. Allow querying IVUsers with isIVUserOrOperand.
Andrew Trick [Fri, 6 Jan 2012 21:41:55 +0000 (21:41 +0000)]
Put all IVUsers in the processed set. Allow querying IVUsers with isIVUserOrOperand.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147686 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAbort AdjustBBOffsetsAfter early when possible.
Jakob Stoklund Olesen [Fri, 6 Jan 2012 21:40:15 +0000 (21:40 +0000)]
Abort AdjustBBOffsetsAfter early when possible.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147685 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agolit/lit.TestRunner: Add an extra_substitutions argument for executeShTest --
Daniel Dunbar [Fri, 6 Jan 2012 21:39:06 +0000 (21:39 +0000)]
lit/lit.TestRunner: Add an extra_substitutions argument for executeShTest --
useful for test suites which want to piggyback onto the "shtest" format style.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147684 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoSCEVExpander: hoistStep should check strict dominance.
Andrew Trick [Fri, 6 Jan 2012 21:23:43 +0000 (21:23 +0000)]
SCEVExpander: hoistStep should check strict dominance.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147683 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoTracing to help investigate issues with SjLj spill code.
Andrew Trick [Fri, 6 Jan 2012 21:16:27 +0000 (21:16 +0000)]
Tracing to help investigate issues with SjLj spill code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147682 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoInitializing to false makes better sense. Thanks, David.
Chad Rosier [Fri, 6 Jan 2012 20:11:59 +0000 (20:11 +0000)]
Initializing to false makes better sense.  Thanks, David.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147679 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix uninitialized variable warning.
Chad Rosier [Fri, 6 Jan 2012 20:02:49 +0000 (20:02 +0000)]
Fix uninitialized variable warning.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147676 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix uninitialized variable warning.
Chad Rosier [Fri, 6 Jan 2012 19:59:58 +0000 (19:59 +0000)]
Fix uninitialized variable warning.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147675 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix a leak I noticed while reviewing the accelerator table changes. Passes
Eric Christopher [Fri, 6 Jan 2012 19:35:04 +0000 (19:35 +0000)]
Fix a leak I noticed while reviewing the accelerator table changes. Passes
lldb testsuite.

rdar://10652330

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147673 91177308-0d34-0410-b5e6-96231b3b80d8

12 years ago[asan] cleanup: remove the SIGILL-related code (compiler part)
Kostya Serebryany [Fri, 6 Jan 2012 18:09:21 +0000 (18:09 +0000)]
[asan] cleanup: remove the SIGILL-related code (compiler part)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147667 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoFix typo in string
Eli Bendersky [Fri, 6 Jan 2012 07:49:17 +0000 (07:49 +0000)]
Fix typo in string

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147654 91177308-0d34-0410-b5e6-96231b3b80d8

12 years agoAs part of the ongoing work in finalizing the accelerator tables, extend
Eric Christopher [Fri, 6 Jan 2012 04:35:23 +0000 (04:35 +0000)]
As part of the ongoing work in finalizing the accelerator tables, extend
the debug type accelerator tables to contain the tag and a flag
stating whether or not a compound type is a complete type.

rdar://10652330

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@147651 91177308-0d34-0410-b5e6-96231b3b80d8