oota-llvm.git
13 years agoprune #includes.
Chris Lattner [Thu, 16 Jun 2011 21:27:52 +0000 (21:27 +0000)]
prune #includes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133194 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agomove the address space into the subclass data field, saving a word on PointerType.
Chris Lattner [Thu, 16 Jun 2011 21:17:17 +0000 (21:17 +0000)]
move the address space into the subclass data field, saving a word on PointerType.

This limits the # address spaces to 2^23, which should be good enough.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133192 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agotidy up some comments, store the 'isvararg' bit for FunctionType in
Chris Lattner [Thu, 16 Jun 2011 21:08:21 +0000 (21:08 +0000)]
tidy up some comments, store the 'isvararg' bit for FunctionType in
the SubclassData field, saving a word.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133191 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoremove Type::getVAArgsPromotedType, which is dead, and tidy up a bit.
Chris Lattner [Thu, 16 Jun 2011 21:00:43 +0000 (21:00 +0000)]
remove Type::getVAArgsPromotedType, which is dead, and tidy up a bit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133190 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoThere's no need to be so picky about the particular register.
Nick Lewycky [Thu, 16 Jun 2011 21:00:00 +0000 (21:00 +0000)]
There's no need to be so picky about the particular register.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133189 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoFix ARCOpt to insert releases on both successors of an invoke rather
Dan Gohman [Thu, 16 Jun 2011 20:57:14 +0000 (20:57 +0000)]
Fix ARCOpt to insert releases on both successors of an invoke rather
than trying to insert them immediately after the invoke.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133188 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoMove PBQP off allocation_order_begin. No functional change intended.
Jakob Stoklund Olesen [Thu, 16 Jun 2011 20:37:45 +0000 (20:37 +0000)]
Move PBQP off allocation_order_begin. No functional change intended.

I think PBQP could use RegisterClassInfo, but it didn't fit neatly with
the external interfaces that PBQP uses, so I'll leave that to Lang.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133186 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoIntroduce MachineBranchProbabilityInfo class, which has similar API to
Jakub Staszak [Thu, 16 Jun 2011 20:22:37 +0000 (20:22 +0000)]
Introduce MachineBranchProbabilityInfo class, which has similar API to
BranchProbabilityInfo (expect setEdgeWeight which is not available here).
Branch Weights are kept in MachineBasicBlocks. To turn off this analysis
set -use-mbpi=false.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133184 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoChange the REG_SEQUENCE SDNode to take an explict register class ID as its first...
Owen Anderson [Thu, 16 Jun 2011 18:17:13 +0000 (18:17 +0000)]
Change the REG_SEQUENCE SDNode to take an explict register class ID as its first operand.  This operand is lowered away by the time we reach MachineInstrs, so the actual register-allocation handling of them doesn't need to change.
This is intended to support using REG_SEQUENCE SDNode's with type MVT::untyped, and is part of the long road to eliminating some of the hacks we currently use to support register pairs and other strange constraints, particularly on ARM NEON.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133178 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoSwitch linear scan to using RegisterClassInfo.
Jakob Stoklund Olesen [Thu, 16 Jun 2011 18:17:00 +0000 (18:17 +0000)]
Switch linear scan to using RegisterClassInfo.

This avoids the manual filtering of reserved registers and removes the
dependency on allocation_order_begin().

Palliative care...

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133177 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoMove test for appropriate directory.
Galina Kistanova [Thu, 16 Jun 2011 18:13:57 +0000 (18:13 +0000)]
Move test for appropriate directory.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133176 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoMark ldrexd/strexd w/ volatile memory by default
Bruno Cardoso Lopes [Thu, 16 Jun 2011 18:11:32 +0000 (18:11 +0000)]
Mark ldrexd/strexd w/ volatile memory by default

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133175 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoTest commit.
Jakub Staszak [Thu, 16 Jun 2011 18:01:17 +0000 (18:01 +0000)]
Test commit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133174 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoPTX: Finish new calling convention implementation
Justin Holewinski [Thu, 16 Jun 2011 17:50:00 +0000 (17:50 +0000)]
PTX: Finish new calling convention implementation

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133172 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoPTX: Rename register classes for readability and combine int and fp registers
Justin Holewinski [Thu, 16 Jun 2011 17:49:58 +0000 (17:49 +0000)]
PTX: Rename register classes for readability and combine int and fp registers

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133171 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd TargetRegisterInfo::getRawAllocationOrder().
Jakob Stoklund Olesen [Thu, 16 Jun 2011 17:42:25 +0000 (17:42 +0000)]
Add TargetRegisterInfo::getRawAllocationOrder().

This virtual function will replace allocation_order_begin/end as the one
to override when implementing custom allocation orders. It is simpler to
have one function return an ArrayRef than having two virtual functions
computing different ends of the same array.

Use getRawAllocationOrder() in place of allocation_order_begin() where
it makes sense, but leave some clients that look like they really want
the filtered allocation orders from RegisterClassInfo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133170 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd testcase for r133050 which added support for printing and parsing escaped
Nick Lewycky [Thu, 16 Jun 2011 17:14:38 +0000 (17:14 +0000)]
Add testcase for r133050 which added support for printing and parsing escaped
names for named metadata nodes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133166 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoFix formatting.
Owen Anderson [Thu, 16 Jun 2011 16:52:24 +0000 (16:52 +0000)]
Fix formatting.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133164 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoDocument nonlazybind.
Dan Gohman [Thu, 16 Jun 2011 16:03:13 +0000 (16:03 +0000)]
Document nonlazybind.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133160 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd a comment describing why transforming (shl x, 1) to (add x, x) is to be
Dan Gohman [Thu, 16 Jun 2011 15:55:48 +0000 (15:55 +0000)]
Add a comment describing why transforming (shl x, 1) to (add x, x) is to be
considered safe enough in this context.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133159 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoPTX: Fix whitespace errors
Justin Holewinski [Thu, 16 Jun 2011 15:17:11 +0000 (15:17 +0000)]
PTX: Fix whitespace errors

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133158 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd AVX suport for fpextend.
Bruno Cardoso Lopes [Thu, 16 Jun 2011 07:03:21 +0000 (07:03 +0000)]
Add AVX suport for fpextend.

Original patch by Syoyo Fujita with more comments by me.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133153 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoPrempt some obnoxious compiler from complaing about signed/unsigned
Jakob Stoklund Olesen [Thu, 16 Jun 2011 03:07:40 +0000 (03:07 +0000)]
Prempt some obnoxious compiler from complaing about signed/unsigned
compares.

2^30 is actually the limit on the number of physical registers per
TargetRegisterInfo.h.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133142 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoMake sure to pass an unsigned to a printf format that is always %u.
Jakob Stoklund Olesen [Thu, 16 Jun 2011 02:55:56 +0000 (02:55 +0000)]
Make sure to pass an unsigned to a printf format that is always %u.

This should unbreak the native ARM testers.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133141 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoFileCheck-ize test, and make it work on EABI hosts, like clang-native-arm-cortex-a9.
Eli Friedman [Thu, 16 Jun 2011 02:36:32 +0000 (02:36 +0000)]
FileCheck-ize test, and make it work on EABI hosts, like clang-native-arm-cortex-a9.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133139 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoForce a triple here so this test doesn't fail on EABI hosts (like clang-native-arm...
Eli Friedman [Thu, 16 Jun 2011 01:49:31 +0000 (01:49 +0000)]
Force a triple here so this test doesn't fail on EABI hosts (like clang-native-arm-cortex-a9).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133134 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoCommit the right set of tests for r133124. Sorry 'bout that!
Nick Lewycky [Thu, 16 Jun 2011 01:35:45 +0000 (01:35 +0000)]
Commit the right set of tests for r133124. Sorry 'bout that!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133133 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoReenabling this test with REQUIRES: Asserts
Andrew Trick [Thu, 16 Jun 2011 01:34:41 +0000 (01:34 +0000)]
Reenabling this test with REQUIRES: Asserts

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133132 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd support to lit for build mode requirements. e.g.
Andrew Trick [Thu, 16 Jun 2011 01:33:35 +0000 (01:33 +0000)]
Add support to lit for build mode requirements. e.g.
REQUIRES: Asserts
REQUIRES: Debug

This required chaining test configuration properties. It seems like a
generally good thing to do.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133131 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoTypos.
Chad Rosier [Thu, 16 Jun 2011 01:24:24 +0000 (01:24 +0000)]
Typos.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133128 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoRevision r128665 added an optimization to make use of NEON multiplier
Chad Rosier [Thu, 16 Jun 2011 01:21:54 +0000 (01:21 +0000)]
Revision r128665 added an optimization to make use of NEON multiplier
accumulator forwarding.  Specifically (from SVN log entry):

Distribute (A + B) * C to (A * C) + (B * C) to make use of NEON multiplier
accumulator forwarding:
vadd d3, d0, d1
vmul d3, d3, d2
=>
vmul d3, d0, d2
vmla d3, d1, d2

Make sure it catches cases where operand 1 is add/fadd/sub/fsub, which was
intended in the original revision.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133127 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd a DAGCombine for (ext (binop (load x), cst)).
Nick Lewycky [Thu, 16 Jun 2011 01:15:49 +0000 (01:15 +0000)]
Add a DAGCombine for (ext (binop (load x), cst)).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133124 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoSilence warnings in non assert builds. Patch by David Blaikie
Bruno Cardoso Lopes [Thu, 16 Jun 2011 00:40:02 +0000 (00:40 +0000)]
Silence warnings in non assert builds. Patch by David Blaikie

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133118 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoRename the test. Thanks Cameron! Use shorter/generic names.
Anna Zaks [Thu, 16 Jun 2011 00:34:10 +0000 (00:34 +0000)]
Rename the test. Thanks Cameron! Use shorter/generic names.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133115 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoFunction::getNumBlockIDs() should be used instead of Function::size() to set the...
Anna Zaks [Thu, 16 Jun 2011 00:03:21 +0000 (00:03 +0000)]
Function::getNumBlockIDs() should be used instead of Function::size() to set the upper limit on the block IDs since basic blocks might get removed (simplified away) after being initially numbered. Plus the test case, in which SelectionDAGBuilder::visitBr() calls llvm::MachineFunction::removeFromMBBNumbering(), which introduces the hole in numbering leading to an assert in llc (prior to the fix).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133113 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd a limit to the number of instructions memdep will scan in a single block. This...
Eli Friedman [Wed, 15 Jun 2011 23:59:25 +0000 (23:59 +0000)]
Add a limit to the number of instructions memdep will scan in a single block.  This prevents (at least in some cases) O(N^2) runtime in passes like DSE.

The limit in this patch is probably too high, but it is enough to stop DSE from going completely insane on a testcase I have (which has a single block with around 50,000 non-aliasing stores in it).

rdar://9471075

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133111 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoThe ARC language-specific optimizer. Credit to Dan Gohman.
John McCall [Wed, 15 Jun 2011 23:37:01 +0000 (23:37 +0000)]
The ARC language-specific optimizer.  Credit to Dan Gohman.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133108 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agounittests: add test for APInt::toString()
Dylan Noblesmith [Wed, 15 Jun 2011 23:36:34 +0000 (23:36 +0000)]
unittests: add test for APInt::toString()

Follow up to r133032.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133107 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd a new MVT::untyped. This will be used in future work for modelling ISA features...
Owen Anderson [Wed, 15 Jun 2011 23:35:18 +0000 (23:35 +0000)]
Add a new MVT::untyped.  This will be used in future work for modelling ISA features like register pairs and lists with "interesting" constraints (such as ARM NEON contiguous register lists or even-odd paired registers).  We need to be able to generate these instructions (often from intrinsics), but don't want to have to assign a legal type to them.  Instead, we'll use an "untyped" edge to bypass the type-checking and simply ensure that the register classes match.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133106 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoUse set operations instead of plain lists to enumerate register classes.
Jakob Stoklund Olesen [Wed, 15 Jun 2011 23:28:14 +0000 (23:28 +0000)]
Use set operations instead of plain lists to enumerate register classes.

This simplifies many of the target description files since it is common
for register classes to be related or contain sequences of numbered
registers.

I have verified that this doesn't change the files generated by TableGen
for ARM and X86. It alters the allocation order of MBlaze GPR and Mips
FGR32 registers, but I believe the change is benign.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133105 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoUpdate the Clang diagnostic emitter to emit IDs for diagnostic categories.
John McCall [Wed, 15 Jun 2011 21:43:52 +0000 (21:43 +0000)]
Update the Clang diagnostic emitter to emit IDs for diagnostic categories.

Patch by Argyrios Kyrtzidis.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133093 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoTestcase for previous commit.
Rafael Espindola [Wed, 15 Jun 2011 21:18:51 +0000 (21:18 +0000)]
Testcase for previous commit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133089 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoSimplify; no significant functionality change.
Eli Friedman [Wed, 15 Jun 2011 21:08:25 +0000 (21:08 +0000)]
Simplify; no significant functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133086 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoFix cmake build.
Rafael Espindola [Wed, 15 Jun 2011 21:03:04 +0000 (21:03 +0000)]
Fix cmake build.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133085 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoHandle jump tables. Test to follow soon.
Rafael Espindola [Wed, 15 Jun 2011 21:00:28 +0000 (21:00 +0000)]
Handle jump tables. Test to follow soon.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133083 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoTry fixing http://google1.osuosl.org:8011/builders/clang-i686-freebsd/builds/3548
Argyrios Kyrtzidis [Wed, 15 Jun 2011 20:39:05 +0000 (20:39 +0000)]
Try fixing google1.osuosl.org:8011/builders/clang-i686-freebsd/builds/3548

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133081 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd a new function attribute, nonlazybind, which inhibits lazy-loading
John McCall [Wed, 15 Jun 2011 20:36:13 +0000 (20:36 +0000)]
Add a new function attribute, nonlazybind, which inhibits lazy-loading
optimizations when emitting calls to the function;  instead those calls may
use faster relocations which require the function to be immediately resolved
upon loading the dynamic object featuring the call.  This is useful when it
is known that the function will be called frequently and pervasively and
therefore there is no merit in delaying binding of the function.

Currently only implemented for x86-64, where it turns into a call through
the global offset table.

Patch by Dan Gohman, who assures me that he's going to add LangRef documentation
for this once it's committed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133080 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoRemove unused code.
Eli Friedman [Wed, 15 Jun 2011 19:58:09 +0000 (19:58 +0000)]
Remove unused code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133078 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoDocument ADT/PackedVector.h in "Programmer's Manual" doc.
Argyrios Kyrtzidis [Wed, 15 Jun 2011 19:56:01 +0000 (19:56 +0000)]
Document ADT/PackedVector.h in "Programmer's Manual" doc.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133077 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd unit tests for ADT/PackedVector
Argyrios Kyrtzidis [Wed, 15 Jun 2011 19:19:09 +0000 (19:19 +0000)]
Add unit tests for ADT/PackedVector

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133075 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoConstify PackedVector::reference::operator T().
Argyrios Kyrtzidis [Wed, 15 Jun 2011 19:18:13 +0000 (19:18 +0000)]
Constify PackedVector::reference::operator T().

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133074 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoDiagnostic for undefined assembler local symbols.
Jim Grosbach [Wed, 15 Jun 2011 18:33:28 +0000 (18:33 +0000)]
Diagnostic for undefined assembler local symbols.

Re-apply 133010, with fixes for inline assembler.

Original commit message:
"When an assembler local symbol is used but not defined in a module, a
Darwin assembler wants to issue a diagnostic to that effect."

Added fix to only perform the check when finalizing, as otherwise we're not
done and undefined symbols may simply not have been encountered yet.

Passes "make check" and a self-host check on Darwin.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133071 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoDisabling this test until I can figure out the right lit flags.
Andrew Trick [Wed, 15 Jun 2011 18:25:38 +0000 (18:25 +0000)]
Disabling this test until I can figure out the right lit flags.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133068 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoRemove custom allocation orders in SystemZ.
Jakob Stoklund Olesen [Wed, 15 Jun 2011 18:02:56 +0000 (18:02 +0000)]
Remove custom allocation orders in SystemZ.

Note that this actually changes code generation, and someone who
understands this target better should check the changes.

- R12Q is now allocatable. I think it was omitted from the allocation
  order by mistake since it isn't reserved. It as apparently used as a
  GOT pointer sometimes, and it should probably be reserved if that is
  the case.

- The GR64 registers are allocated in a different order now. The
  register allocator will automatically put the CSRs last. There were
  other changes to the order that may have been significant.

The test fix is because r0 and r1 swapped places in the allocation order.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133067 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAnother revsh pattern. rdar://9609059
Evan Cheng [Wed, 15 Jun 2011 17:17:48 +0000 (17:17 +0000)]
Another revsh pattern. rdar://9609059

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133064 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdded -stress-sched flag in the Asserts build.
Andrew Trick [Wed, 15 Jun 2011 17:16:12 +0000 (17:16 +0000)]
Added -stress-sched flag in the Asserts build.
Added a test case for handling physreg aliases during pre-RA-sched.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133063 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoTargetLoweringOpt is a struct used by DAGCombine, not a pass.
Chad Rosier [Wed, 15 Jun 2011 16:48:02 +0000 (16:48 +0000)]
TargetLoweringOpt is a struct used by DAGCombine, not a pass.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133062 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoMake PPC64CompilationCallback compilable no non-darwin platforms.
Roman Divacky [Wed, 15 Jun 2011 15:29:47 +0000 (15:29 +0000)]
Make PPC64CompilationCallback compilable no non-darwin platforms.

Patch by Nathan Whitehorn!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133059 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agogetZeroExtendInReg needs to get a scalar type
Nadav Rotem [Wed, 15 Jun 2011 14:37:18 +0000 (14:37 +0000)]
getZeroExtendInReg needs to get a scalar type

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133057 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoThis test was failing on X86 machines which do not have SSE4. Fixed the test by
Nadav Rotem [Wed, 15 Jun 2011 12:26:53 +0000 (12:26 +0000)]
This test was failing on X86 machines which do not have SSE4. Fixed the test by
specifying that the target CPU is corei7.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133053 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoEnable the simplification of truncating-store after fixing the usage of
Nadav Rotem [Wed, 15 Jun 2011 11:19:12 +0000 (11:19 +0000)]
Enable the simplification of truncating-store after fixing the usage of
GetDemandBits (which must operate on the vector element type).

Fix the a usage of getZeroExtendInReg which must also be done on scalar types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133052 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoReplace the statically generated hashtables for checking register relationships with...
Owen Anderson [Wed, 15 Jun 2011 06:53:50 +0000 (06:53 +0000)]
Replace the statically generated hashtables for checking register relationships with just scanning the (typically tiny) static lists.

At the time I wrote this code (circa 2007), TargetRegisterInfo was using a std::set to perform these queries.  Switching to the static hashtables was an obvious improvement, but in reality there's no reason to do anything other than scan.
With this change, total LLC time on a whole-program 403.gcc is reduced by approximately 1.5%, almost all of which comes from a 15% reduction in LiveVariables time.  It also reduces the binary size of LLC by 86KB, thanks to eliminating a bunch of very large static tables.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133051 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoTeach the .ll parser to handle named metadata with non-simple names.
Nick Lewycky [Wed, 15 Jun 2011 06:37:58 +0000 (06:37 +0000)]
Teach the .ll parser to handle named metadata with non-simple names.
Unfortunately we can't follow what the rest of the language does (wrapping it
in double-quotes) because that would cause an ambiguity with metadata strings,
so instead we escape any unusual characters with \xx escaping.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133050 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoA minor simplification: no functional change.
Bob Wilson [Wed, 15 Jun 2011 06:04:34 +0000 (06:04 +0000)]
A minor simplification: no functional change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133047 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoUse a SetTheory instance to expand register lists in register classes.
Jakob Stoklund Olesen [Wed, 15 Jun 2011 05:09:20 +0000 (05:09 +0000)]
Use a SetTheory instance to expand register lists in register classes.

This prepares tablegen to compute register lists from set theoretic dag
expressions. This doesn't really make any difference as long as
Target.td still declares RegisterClass::MemberList as [Register].

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133043 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoRemove the LTO stuff from the profile_rt library's Makefile.
Bill Wendling [Wed, 15 Jun 2011 04:56:13 +0000 (04:56 +0000)]
Remove the LTO stuff from the profile_rt library's Makefile.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133042 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoGive CodeGenRegisterClass a real sorted member set.
Jakob Stoklund Olesen [Wed, 15 Jun 2011 04:50:36 +0000 (04:50 +0000)]
Give CodeGenRegisterClass a real sorted member set.

Make the Elements vector private and expose an ArrayRef through
getOrder() instead. getOrder will eventually provide multiple
user-specified allocation orders.

Use the sorted member set for member and subclass tests. Clean up a lot
of ad hoc searches.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133040 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoImprove the heuristic to emit the alias if the number of hard-coded registers
Bill Wendling [Wed, 15 Jun 2011 04:31:19 +0000 (04:31 +0000)]
Improve the heuristic to emit the alias if the number of hard-coded registers
are also greater than the alias.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133038 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoStop using memdep for a check that didn't really make sense with memdep. In terms...
Eli Friedman [Wed, 15 Jun 2011 01:25:56 +0000 (01:25 +0000)]
Stop using memdep for a check that didn't really make sense with memdep. In terms of specific issues, using memdep here checks irrelevant instructions and won't work properly once we start returning "unknown" more aggressively from memdep.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133035 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoPerformBFICombine - (bfi A, (and B, Mask1), Mask2) -> (bfi A, B, Mask2) iff
Evan Cheng [Wed, 15 Jun 2011 01:12:31 +0000 (01:12 +0000)]
PerformBFICombine - (bfi A, (and B, Mask1), Mask2) -> (bfi A, B, Mask2) iff
the bits being cleared by the AND are not demanded by the BFI.

The previous BFI dag combine rule was actually incorrect (or used to be
correct until BFI representation changed).

rdar://9609030

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133034 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoadd option for literal formatting to APInt::toString()
Ted Kremenek [Wed, 15 Jun 2011 00:51:55 +0000 (00:51 +0000)]
add option for literal formatting to APInt::toString()

toString() now takes an optional bool argument that,
depending on the radix, adds the appropriate prefix
to the integer's string representation that makes it into a
meaningful C literal, e.g.:

hexademical: '-f' becomes '-0xf'
octal: '77' becomes '077'
binary: '110' becomes '0b110'

Patch by nobled@dreamwidth.org!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133032 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd "unknown" results for memdep, which mean "I don't know whether a dependence for...
Eli Friedman [Wed, 15 Jun 2011 00:47:34 +0000 (00:47 +0000)]
Add "unknown" results for memdep, which mean "I don't know whether a dependence for the given instruction exists in the given block".  This cleans up all the existing hacks in memdep which represent this concept by returning clobber with various unrelated instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133031 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoMove the list of register classes into CodeGenRegBank as well.
Jakob Stoklund Olesen [Wed, 15 Jun 2011 00:20:40 +0000 (00:20 +0000)]
Move the list of register classes into CodeGenRegBank as well.

No functional change intended.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133029 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd an optimization that looks for a specific pair-wise add pattern and generates...
Tanya Lattner [Tue, 14 Jun 2011 23:48:48 +0000 (23:48 +0000)]
Add an optimization that looks for a specific pair-wise add pattern and generates a vpaddl instruction instead of scalarizing the add.
Includes a test case.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133027 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd triple.
Rafael Espindola [Tue, 14 Jun 2011 23:47:36 +0000 (23:47 +0000)]
Add triple.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133026 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAnna's test commit (#2).
Anna Zaks [Tue, 14 Jun 2011 22:40:29 +0000 (22:40 +0000)]
Anna's test commit (#2).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133023 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoWhen pattern matching during instruction selection make sure shl x,1 is not
Chad Rosier [Tue, 14 Jun 2011 22:29:10 +0000 (22:29 +0000)]
When pattern matching during instruction selection make sure shl x,1 is not
converted to add x,x if x is a undef.  add undef, undef does not guarantee
that the resulting low order bit is zero.
Fixes <rdar://problem/9453156> and <rdar://problem/9487392>.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133022 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoCheck the llc output.
Rafael Espindola [Tue, 14 Jun 2011 22:24:32 +0000 (22:24 +0000)]
Check the llc output.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133021 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoPR10136: fix PPCTargetLowering::LowerCall_SVR4 so that a necessary CopyToReg doesn...
Eli Friedman [Tue, 14 Jun 2011 22:16:20 +0000 (22:16 +0000)]
PR10136: fix PPCTargetLowering::LowerCall_SVR4 so that a necessary CopyToReg doesn't appear to be dead.

Roman, since you're writing tests for other PPC-SVR4 vararg-related stuff, would you mind writing a test for this?

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133018 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAnna's test commit.
Anna Zaks [Tue, 14 Jun 2011 22:10:12 +0000 (22:10 +0000)]
Anna's test commit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133017 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoTest case for x86 MMX inline asm. rdar://problem/8886707
Stuart Hastings [Tue, 14 Jun 2011 21:51:38 +0000 (21:51 +0000)]
Test case for x86 MMX inline asm.  rdar://problem/8886707

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133014 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoRevert 133010. Self-hosted buildbot unhappy.
Jim Grosbach [Tue, 14 Jun 2011 21:51:20 +0000 (21:51 +0000)]
Revert 133010. Self-hosted buildbot unhappy.

Apparently llvm itself generates undefined assembler local labels, causing
self-hosting problems with this patch. Reverting until that's sorted out.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133013 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoDiagnostic for undefined assembler local symbols.
Jim Grosbach [Tue, 14 Jun 2011 21:13:25 +0000 (21:13 +0000)]
Diagnostic for undefined assembler local symbols.

When an assembler local symbol is used but not defined in a module, a
Darwin assembler wants to issue a diagnostic to that effect.

rdar://9559714

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133010 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd a test for the recent regression.
Rafael Espindola [Tue, 14 Jun 2011 20:38:50 +0000 (20:38 +0000)]
Add a test for the recent regression.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133009 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoRevert r133004 ; it's breaking nightly tests.
Eli Friedman [Tue, 14 Jun 2011 19:30:33 +0000 (19:30 +0000)]
Revert r133004 ; it's breaking nightly tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133007 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoPartial revert of 132882.
Rafael Espindola [Tue, 14 Jun 2011 18:12:31 +0000 (18:12 +0000)]
Partial revert of 132882.
Dan noted that this would work on the case shown on the commit message. I think
the case that was failing was a bb ending with a redundant conditional jump:

...
jne foo

foo:
...

I was unable to find any such case in the tests or in a debug build of clang,
so I will revert this part of the patch and watch the bots.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133004 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAlso recognize ARM v4t and v5e variants.
Evan Cheng [Tue, 14 Jun 2011 18:08:33 +0000 (18:08 +0000)]
Also recognize ARM v4t and v5e variants.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133002 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoThis test is still failing. Delete the rest of it.
Dan Gohman [Tue, 14 Jun 2011 18:07:36 +0000 (18:07 +0000)]
This test is still failing. Delete the rest of it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133001 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoRevert r132991. This test is failing on the
Dan Gohman [Tue, 14 Jun 2011 18:03:11 +0000 (18:03 +0000)]
Revert r132991. This test is failing on the
llvm-gcc-x86_64-linux-selfhost buildbot and others.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133000 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoFix a compile time regression caused by too small hash tables.
Jakob Stoklund Olesen [Tue, 14 Jun 2011 16:58:16 +0000 (16:58 +0000)]
Fix a compile time regression caused by too small hash tables.

Measure the worst case number of probes for a miss instead of the less
conservative number of probes required for an insertion.

Lower the limit to < 6 probes worst case.

This doubles the size of the ARM and X86 hash tables, other targets are
unaffected. LiveVariables runs 12% faster with this change.

<rdar://problem/9598545>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132999 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd 132986 back, but avoid non-determinism if a bb address gets reused.
Rafael Espindola [Tue, 14 Jun 2011 15:31:54 +0000 (15:31 +0000)]
Add 132986 back, but avoid non-determinism if a bb address gets reused.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132995 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd a testcase for #9623
Nadav Rotem [Tue, 14 Jun 2011 13:23:10 +0000 (13:23 +0000)]
Add a testcase for #9623

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132991 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agorevert 132986 to see if the bots go green.
Rafael Espindola [Tue, 14 Jun 2011 12:48:26 +0000 (12:48 +0000)]
revert 132986 to see if the bots go green.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132988 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoThis testcase cause a failure on some bots. Remove the failing test until
Nadav Rotem [Tue, 14 Jun 2011 09:10:37 +0000 (09:10 +0000)]
This testcase cause a failure on some bots. Remove the failing test until
further investigation.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132986 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd a testcase for checking the integer-promotion of many different vector
Nadav Rotem [Tue, 14 Jun 2011 08:11:52 +0000 (08:11 +0000)]
Add a testcase for checking the integer-promotion of many different vector
types (with power of two types such as 8,16,32 .. 512).

Fix a bug in the integer promotion of bitcast nodes. Enable integer expanding
only if the target of the conversion is an integer (when the type action is
scalarize).

Add handling to the legalization of vector load/store in cases where the saved
vector is integer-promoted.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132985 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoDisable trunc-store simplification on vectors.
Nadav Rotem [Tue, 14 Jun 2011 07:18:26 +0000 (07:18 +0000)]
Disable trunc-store simplification on vectors.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132984 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoBe more obvious about what is being tested.
Cameron Zwarich [Tue, 14 Jun 2011 06:33:51 +0000 (06:33 +0000)]
Be more obvious about what is being tested.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132982 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoImplement Jakob's suggestion on how to detect fall thought without calling
Rafael Espindola [Tue, 14 Jun 2011 06:08:32 +0000 (06:08 +0000)]
Implement Jakob's suggestion on how to detect fall thought without calling
AnalyzeBranch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132981 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoSince ARM's prefetch implementation predicted the presence of a instruction
Bruno Cardoso Lopes [Tue, 14 Jun 2011 05:11:46 +0000 (05:11 +0000)]
Since ARM's prefetch implementation predicted the presence of a instruction
cache prefetch and now that the info from "prefetch" to "ARMPreload" is present,
only add a testcase for PLI.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132978 91177308-0d34-0410-b5e6-96231b3b80d8

13 years agoAdd one more argument to the prefetch intrinsic to indicate whether it's a data
Bruno Cardoso Lopes [Tue, 14 Jun 2011 04:58:37 +0000 (04:58 +0000)]
Add one more argument to the prefetch intrinsic to indicate whether it's a data
or instruction cache access. Update the targets to match it and also teach
autoupgrade.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132976 91177308-0d34-0410-b5e6-96231b3b80d8