oota-llvm.git
9 years ago[SimplifyLibCalls] Enable double to float shrinking for copysign
Matthias Braun [Wed, 3 Dec 2014 21:46:29 +0000 (21:46 +0000)]
[SimplifyLibCalls] Enable double to float shrinking for copysign

rdar://19049359

Differential Revision: http://reviews.llvm.org/D6495

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223269 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Hexagon] Converting member InstrDesc to static variable.
Colin LeMahieu [Wed, 3 Dec 2014 21:40:25 +0000 (21:40 +0000)]
[Hexagon] Converting member InstrDesc to static variable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223268 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix grammar-o.
Nico Weber [Wed, 3 Dec 2014 20:58:23 +0000 (20:58 +0000)]
Fix grammar-o.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223265 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Hexagon] Converting subclass members to an implicit operand.
Colin LeMahieu [Wed, 3 Dec 2014 20:23:22 +0000 (20:23 +0000)]
[Hexagon] Converting subclass members to an implicit operand.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223264 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agorange-for some things
David Blaikie [Wed, 3 Dec 2014 19:58:49 +0000 (19:58 +0000)]
range-for some things

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223263 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSimplify CodeGenRegBank::inferMatchingSuperRegClass & its caller by passing an iterat...
David Blaikie [Wed, 3 Dec 2014 19:58:48 +0000 (19:58 +0000)]
Simplify CodeGenRegBank::inferMatchingSuperRegClass & its caller by passing an iterator rather than index

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223262 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSimplify ownership of RegClasses by using list<CodeGenRegisterClass> instead of vecto...
David Blaikie [Wed, 3 Dec 2014 19:58:45 +0000 (19:58 +0000)]
Simplify ownership of RegClasses by using list<CodeGenRegisterClass> instead of vector<CodeGenRegisterClass*>

This complicates a few algorithms due to not having random access, but
not by a huge degree I don't think (open to debate/design
discussion/etc).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223261 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRange-for some stuff related to RegClasses, and comment cases where range-for isn...
David Blaikie [Wed, 3 Dec 2014 19:58:41 +0000 (19:58 +0000)]
Range-for some stuff related to RegClasses, and comment cases where range-for isn't suitable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223260 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMake the Verifier more strict about gc.statepoints
Philip Reames [Wed, 3 Dec 2014 19:53:15 +0000 (19:53 +0000)]
Make the Verifier more strict about gc.statepoints

The recently added documentation for statepoints claimed that we checked the parameters of the various intrinsics for validity.  This patch adds the code to actually do so.  I also removed a couple of redundant checks for conditions which are checked elsewhere in the Verifier and simplified the logic using the helper functions from Statepoint.h.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223259 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd TableGen info for Power8.
Will Schmidt [Wed, 3 Dec 2014 18:46:30 +0000 (18:46 +0000)]
Add TableGen info for Power8.
This is based on the Power7 version, with units added and renamed to match P8.

Differential Revision: http://reviews.llvm.org/D6358

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223257 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoChange the name to be in style.
Roman Divacky [Wed, 3 Dec 2014 18:39:44 +0000 (18:39 +0000)]
Change the name to be in style.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223255 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix sphinx error from Statepoints.rst
Matt Arsenault [Wed, 3 Dec 2014 18:35:11 +0000 (18:35 +0000)]
Fix sphinx error from Statepoints.rst

It was complaining it wasn't included in any toctree

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223254 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Move SIInsertWaits into AMDGPUPassConfig::addPreSched2()
Tom Stellard [Wed, 3 Dec 2014 18:27:08 +0000 (18:27 +0000)]
R600/SI: Move SIInsertWaits into AMDGPUPassConfig::addPreSched2()

This pass needs to be run after PrologEpilogInserter, because
that pass may inserter spill code which reads or writes memory.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223253 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Don't run SI passes on R600 subtargets
Tom Stellard [Wed, 3 Dec 2014 18:27:05 +0000 (18:27 +0000)]
R600/SI: Don't run SI passes on R600 subtargets

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223252 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAArch64: fix wrong-endian parameter passing.
Tim Northover [Wed, 3 Dec 2014 17:49:26 +0000 (17:49 +0000)]
AArch64: fix wrong-endian parameter passing.

The blocked arguments code didn't take account of the hacks needed to support
it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223247 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[NFC] Fixing pendantic warning extra semicolons.
Colin LeMahieu [Wed, 3 Dec 2014 17:36:39 +0000 (17:36 +0000)]
[NFC] Fixing pendantic warning extra semicolons.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223246 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Hexagon] [NFC] Moving function implementations out of header. Clang-formatting...
Colin LeMahieu [Wed, 3 Dec 2014 17:35:39 +0000 (17:35 +0000)]
[Hexagon] [NFC] Moving function implementations out of header.  Clang-formatting files.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223245 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix test to use the right metadata node (reapply r223239 plus a fix) and also to...
Nick Lewycky [Wed, 3 Dec 2014 17:32:44 +0000 (17:32 +0000)]
Fix test to use the right metadata node (reapply r223239 plus a fix) and also to use the correct path to the GCNO file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223244 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Hexagon] [NFC] Renaming *packetStart to *packetBegin
Colin LeMahieu [Wed, 3 Dec 2014 17:31:43 +0000 (17:31 +0000)]
[Hexagon] [NFC] Renaming *packetStart to *packetBegin

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223243 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert r223239, which broke some bots.
Alexander Potapenko [Wed, 3 Dec 2014 16:03:08 +0000 (16:03 +0000)]
Revert r223239, which broke some bots.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223240 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix the metadata number used by llvm.gcov to match the number of the inserted metadat...
Alexander Potapenko [Wed, 3 Dec 2014 15:15:58 +0000 (15:15 +0000)]
Fix the metadata number used by llvm.gcov to match the number of the inserted metadata node.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223239 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSilencing several "multiple copy constructors" warnings from MSVC; NFC.
Aaron Ballman [Wed, 3 Dec 2014 14:44:16 +0000 (14:44 +0000)]
Silencing several "multiple copy constructors" warnings from MSVC; NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223238 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSilencing a 32-bit implicit conversion warning in MSVC; NFC.
Aaron Ballman [Wed, 3 Dec 2014 14:39:58 +0000 (14:39 +0000)]
Silencing a 32-bit implicit conversion warning in MSVC; NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223237 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agomsan] Add compile-time checks for missing origins.
Evgeniy Stepanov [Wed, 3 Dec 2014 14:15:53 +0000 (14:15 +0000)]
msan] Add compile-time checks for missing origins.

This change makes MemorySanitizer instrumentation a bit more strict
about instructions that have no origin id assigned to them.

This would have caught the bug that was fixed in r222918.

This is re-commit of r222997, reverted in r223211, with 3 more
missing origins added.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223236 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoInstCombine: simplify signed range checks
Erik Eckstein [Wed, 3 Dec 2014 10:39:15 +0000 (10:39 +0000)]
InstCombine: simplify signed range checks

Try to convert two compares of a signed range check into a single unsigned compare.
Examples:
(icmp sge x, 0) & (icmp slt x, n) --> icmp ult x, n
(icmp slt x, 0) | (icmp sgt x, n) --> icmp ugt x, n

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223224 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[PowerPC] Print all inline-asm consts as signed numbers
Hal Finkel [Wed, 3 Dec 2014 09:37:50 +0000 (09:37 +0000)]
[PowerPC] Print all inline-asm consts as signed numbers

Almost all immediates in PowerPC assembly (both 32-bit and 64-bit) are signed
numbers, and it is important that we print them as such. To make sure that
happens, we change PPCTargetLowering::LowerAsmOperandForConstraint so that it
does all intermediate checks on a signed-extended int64_t value, and then
creates the resulting target constant using MVT::i64. This will ensure that all
negative values are printed as negative values (mirroring what is done in other
backends to achieve the same sign-extension effect).

This came up in the context of inline assembly like this:
  "add%I2   %0,%0,%2", ..., "Ir"(-1ll)
where we used to print:
  addi   3,3,4294967295
and gcc would print:
  addi   3,3,-1
and gas accepts both forms, but our builtin assembler (correctly) does not. Now
we print -1 like gcc does.

While here, I replaced a bunch of custom integer checks with isInt<16> and
friends from MathExtras.h.

Thanks to Paul Hargrove for the bug report.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223220 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoEmit ABI_FP_rounding attribute.
Charlie Turner [Wed, 3 Dec 2014 08:12:26 +0000 (08:12 +0000)]
Emit ABI_FP_rounding attribute.

LLVM understands a -enable-sign-dependent-rounding-fp-math codegen option. When
the user has specified this option, the Tag_ABI_FP_rounding attribute should be
emitted with value 1. This option currently does not appear to disable
transformations and optimizations that assume default floating point rounding
behavior, AFAICT, but the intention should be recorded in the build attributes,
regardless of what the compiler actually does with the intention.

Change-Id: If838578df3dc652b6f2796b8d152545674bcb30e

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223218 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd tests for default value of Tag_ABI_FP_rounding.
Charlie Turner [Wed, 3 Dec 2014 07:59:50 +0000 (07:59 +0000)]
Add tests for default value of Tag_ABI_FP_rounding.

Change-Id: I051866d073fc6ce87ce3e693a3762da6d81f4393

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223217 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix a typo in the documentation of LTO
Benjamin Poulain [Wed, 3 Dec 2014 07:32:36 +0000 (07:32 +0000)]
Fix a typo in the documentation of LTO

Fix defininitions->definitions.

Reviewed by David Blaikie.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223216 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAsk the module for its the identified types.
Rafael Espindola [Wed, 3 Dec 2014 07:18:23 +0000 (07:18 +0000)]
Ask the module for its the identified types.

When lazy reading a module, the types used in a function will not be visible to
a TypeFinder until the body is read.

This patch fixes that by asking the module for its identified struct types.
If a materializer is present, the module asks it. If not, it uses a TypeFinder.

This fixes pr21374.

I will be the first to say that this is ugly, but it was the best I could find.

Some of the options I looked at:

* Asking the LLVMContext. This could be made to work for gold, but not currently
  for ld64. ld64 will load multiple modules into a single context before merging
  them. This causes us to see types from future merges. Unfortunately,
  MappedTypes is not just a cache when it comes to opaque types. Once the
  mapping has been made, we have to remember it for as long as the key may
  be used. This would mean moving MappedTypes to the Linker class and having
  to drop the Linker::LinkModules static methods, which are visible from C.

* Adding an option to ignore function bodies in the TypeFinder. This would
  fix the PR by picking the worst result. It would work, but unfortunately
  we are currently quite dependent on the upfront type merging. I will
  try to reduce our dependency, but it is not clear that we will be able
  to get rid of it for now.

The only clean solution I could think of is making the Module own the types.
This would have other advantages, but it is a much bigger change. I will
propose it, but it is nice to have this fixed while that is discussed.

With the gold plugin, this patch takes the number of types in the LTO clang
binary from 52817 to 49669.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223215 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoADT: Rename argument in emplace_back_impl
Duncan P. N. Exon Smith [Wed, 3 Dec 2014 05:53:24 +0000 (05:53 +0000)]
ADT: Rename argument in emplace_back_impl

Rename a functor argument in r223201 from `emplace` to `construct` to
reduce confusion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223212 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRevert r222997. The newly added compile-time checks are finding missing origins,...
Nick Lewycky [Wed, 3 Dec 2014 05:47:00 +0000 (05:47 +0000)]
Revert r222997. The newly added compile-time checks are finding missing origins, testcase is being reduced and a PR will be posted shortly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223211 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoLoopVectorize: Remove unnecessary RAUW
Duncan P. N. Exon Smith [Wed, 3 Dec 2014 05:41:20 +0000 (05:41 +0000)]
LoopVectorize: Remove unnecessary RAUW

Remove an unnecessary `MDNode::replaceAllUsesWith()`.  In the preceding
line, `TheLoop->setLoopID()` visits all backedges and sets the new loop
ID.  This sufficiently updates the loop metadata.

Metadata RAUW is going away as part of PR21532.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223210 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Fix SIFixSGPRCopies for copies to physical registers
Matt Arsenault [Wed, 3 Dec 2014 05:22:39 +0000 (05:22 +0000)]
R600/SI: Fix SIFixSGPRCopies for copies to physical registers

This shows up when operands required to be passed in VCC are copied
to.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223208 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Remove incorrect assertion
Matt Arsenault [Wed, 3 Dec 2014 05:22:38 +0000 (05:22 +0000)]
R600/SI: Remove incorrect assertion

This can be a COPY to a physical register, such as VCC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223207 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Remove i1 pseudo VALU ops
Matt Arsenault [Wed, 3 Dec 2014 05:22:35 +0000 (05:22 +0000)]
R600/SI: Remove i1 pseudo VALU ops

Select i1 logical ops directly to 64-bit SALU instructions.
Vector i1 values are always really in SGPRs, with each
bit for each item in the wave. This saves about 4 instructions
when and/or/xoring any condition, and also helps write conditions
that need to be passed in vcc.

This should work correctly now that the SGPR live range
fixing pass works. More work is needed to eliminate the VReg_1
pseudo regclass and possibly the entire SILowerI1Copies pass.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223206 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Fix suspicious indexing
Matt Arsenault [Wed, 3 Dec 2014 05:22:32 +0000 (05:22 +0000)]
R600/SI: Fix suspicious indexing

The loop is over the operands of an instruction, and checks the
register with the sub reg index of the dest register. This probably
meant to be checking the sub reg index of the same operand.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223205 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Fix running SILowerI1Copies a second time
Matt Arsenault [Wed, 3 Dec 2014 05:22:30 +0000 (05:22 +0000)]
R600/SI: Fix running SILowerI1Copies a second time

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223204 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Fix live range error hidden by SIFoldOperands
Matt Arsenault [Wed, 3 Dec 2014 05:22:29 +0000 (05:22 +0000)]
R600/SI: Fix live range error hidden by SIFoldOperands

m0 is treated as a virtual register class with a single register
rather than the physical register it really is. This was updating
the live range of the used virtual copy of m0 from the first ds_read
instruction, and leaving the unused copy unchanged. This resulted in a
"Live segment doesn't end at a valid instruction" verifier error because
the erased instructions. Update the live range of the second copy (which
should be dead).

No test since I'm not sure how to trigger this with SIFoldOperands
enabled.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223203 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoADT: Add SmallVector<>::emplace_back(): fixup
Duncan P. N. Exon Smith [Wed, 3 Dec 2014 04:49:16 +0000 (04:49 +0000)]
ADT: Add SmallVector<>::emplace_back(): fixup

Add missing `void` return type from `!LLVM_HAS_VARIADIC_TEMPLATES` case
in r223201.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223202 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoADT: Add SmallVector<>::emplace_back()
Duncan P. N. Exon Smith [Wed, 3 Dec 2014 04:45:09 +0000 (04:45 +0000)]
ADT: Add SmallVector<>::emplace_back()

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223201 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoStructurizeCFG: Use LoopInfo analysis for better loop detection
Tom Stellard [Wed, 3 Dec 2014 04:28:32 +0000 (04:28 +0000)]
StructurizeCFG: Use LoopInfo analysis for better loop detection

We were assuming that each back-edge in a region represented a unique
loop, which is not always the case.  We need to use LoopInfo to
correctly determine which back-edges are loops.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223199 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoNVPTX: Delete dead code
Duncan P. N. Exon Smith [Wed, 3 Dec 2014 04:13:23 +0000 (04:13 +0000)]
NVPTX: Delete dead code

`MDNode` does not inherit from `User`, and it never has a name.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223198 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Enable inline assembly
Tom Stellard [Wed, 3 Dec 2014 04:08:00 +0000 (04:08 +0000)]
R600/SI: Enable inline assembly

We just needed to remove the assertion in
AMDGPURegisterInfo::getFrameRegister(), which is called when
initializing the parser for inline assembly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223197 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[OCaml] [cmake] Disable OCaml bindings if ctypes >=0.3 is not found.
Peter Zotov [Wed, 3 Dec 2014 03:39:01 +0000 (03:39 +0000)]
[OCaml] [cmake] Disable OCaml bindings if ctypes >=0.3 is not found.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223195 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Change mubuf offsets to print as decimal
Matt Arsenault [Wed, 3 Dec 2014 03:12:13 +0000 (03:12 +0000)]
R600/SI: Change mubuf offsets to print as decimal

This matches SC's behavior.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223194 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoEmit the entry block first and the exit block second, then all the blocks in between...
Nick Lewycky [Wed, 3 Dec 2014 02:45:01 +0000 (02:45 +0000)]
Emit the entry block first and the exit block second, then all the blocks in between afterwards. This is what gcc always does, and some out of tree tools depend on that.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223193 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoGCRelocateOperands: Try to appease msc17.
NAKAMURA Takumi [Wed, 3 Dec 2014 02:40:24 +0000 (02:40 +0000)]
GCRelocateOperands: Try to appease msc17.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223192 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoPrologue support
Peter Collingbourne [Wed, 3 Dec 2014 02:08:38 +0000 (02:08 +0000)]
Prologue support

Patch by Ben Gamari!

This redefines the `prefix` attribute introduced previously and
introduces a `prologue` attribute.  There are a two primary usecases
that these attributes aim to serve,

  1. Function prologue sigils

  2. Function hot-patching: Enable the user to insert `nop` operations
     at the beginning of the function which can later be safely replaced
     with a call to some instrumentation facility

  3. Runtime metadata: Allow a compiler to insert data for use by the
     runtime during execution. GHC is one example of a compiler that
     needs this functionality for its tables-next-to-code functionality.

Previously `prefix` served cases (1) and (2) quite well by allowing the user
to introduce arbitrary data at the entrypoint but before the function
body. Case (3), however, was poorly handled by this approach as it
required that prefix data was valid executable code.

Here we redefine the notion of prefix data to instead be data which
occurs immediately before the function entrypoint (i.e. the symbol
address). Since prefix data now occurs before the function entrypoint,
there is no need for the data to be valid code.

The previous notion of prefix data now goes under the name "prologue
data" to emphasize its duality with the function epilogue.

The intention here is to handle cases (1) and (2) with prologue data and
case (3) with prefix data.

References
----------

This idea arose out of discussions[1] with Reid Kleckner in response to a
proposal to introduce the notion of symbol offsets to enable handling of
case (3).

[1] http://lists.cs.uiuc.edu/pipermail/llvmdev/2014-May/073235.html

Test Plan: testsuite

Differential Revision: http://reviews.llvm.org/D6454

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223189 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoExceptionDemo: Let setMCJITMemoryManager() take unique_ptr, since r223183.
NAKAMURA Takumi [Wed, 3 Dec 2014 02:05:51 +0000 (02:05 +0000)]
ExceptionDemo: Let setMCJITMemoryManager() take unique_ptr, since r223183.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223188 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[X86][MC] Intel syntax: accept implicit memory operand sizes larger than 80.
Ahmed Bougacha [Wed, 3 Dec 2014 02:03:26 +0000 (02:03 +0000)]
[X86][MC] Intel syntax: accept implicit memory operand sizes larger than 80.

The X86AsmParser intel handling was refactored in r216481, making it
try each different memory operand size to see which one matches.
Operand sizes larger than 80 ("[xyz]mmword ptr") were forgotten, which
led to an "invalid operand" error for code such as:
  movdqa [rax], xmm0

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223187 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[MCJIT] Unique-ptrify the RTDyldMemoryManager member of MCJIT. NFC.
Lang Hames [Wed, 3 Dec 2014 00:51:19 +0000 (00:51 +0000)]
[MCJIT] Unique-ptrify the RTDyldMemoryManager member of MCJIT. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223183 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[PowerPC] Fix readcyclecounter to be custom expanded for all 32-bit targets
Hal Finkel [Wed, 3 Dec 2014 00:19:17 +0000 (00:19 +0000)]
[PowerPC] Fix readcyclecounter to be custom expanded for all 32-bit targets

We need to use the custom expansion of readcyclecounter on all 32-bit targets
(even those with 64-bit registers). This should fix the ppc64 buildbot.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223182 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAArch64: strengthen Darwin ABI alignment assumptions
Tim Northover [Tue, 2 Dec 2014 23:53:43 +0000 (23:53 +0000)]
AArch64: strengthen Darwin ABI alignment assumptions

A global variable without an explicit alignment specified should be assumed to
be ABI-aligned according to its type, like on other platforms. This allows us
to use better memory operations when accessing it.

rdar://18533701

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223180 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoUse a typed enum instead of 'unsigned char' for packed field. NFC.
Pete Cooper [Tue, 2 Dec 2014 23:34:23 +0000 (23:34 +0000)]
Use a typed enum instead of 'unsigned char' for packed field.  NFC.

This makes it easier to debug Twine as the 'Kind' fields now show their enum values in lldb and not escaped characters.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223178 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAArch64: don't be too greedy when folding :lo12: accesses into mem ops.
Tim Northover [Tue, 2 Dec 2014 23:13:39 +0000 (23:13 +0000)]
AArch64: don't be too greedy when folding :lo12: accesses into mem ops.

This frequently leads to cases like:
   ldr xD, [xN, :lo12:var]
   add xA, xN, :lo12:var
   ldr xD, [xA, #8]

where the ADD would have been needed anyway, and the two distinct addressing
modes can prevent the formation of an ldp. Because of how we handle ADRP
(aggressively forming an ADRP/ADD pseudo-inst at ISel time), this pattern also
results in duplicated ADRP instructions (one on its own to cover the ldr, and
one combined with the add).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223172 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoPR21302. Vectorize only bottom-tested loops.
Michael Zolotukhin [Tue, 2 Dec 2014 22:59:06 +0000 (22:59 +0000)]
PR21302. Vectorize only bottom-tested loops.

rdar://problem/18886083

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223171 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoApply loop-rotate to several vectorizer tests.
Michael Zolotukhin [Tue, 2 Dec 2014 22:59:02 +0000 (22:59 +0000)]
Apply loop-rotate to several vectorizer tests.

Such loops shouldn't be vectorized due to the loops form.
After applying loop-rotate (+simplifycfg) the tests again start to check
what they are intended to check.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223170 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[X86][SSE] Keep 4i32 vector insertions in integer domain on SSE4.1 targets
Simon Pilgrim [Tue, 2 Dec 2014 22:31:23 +0000 (22:31 +0000)]
[X86][SSE] Keep 4i32 vector insertions in integer domain on SSE4.1 targets

4i32 shuffles for single insertions into zero vectors lowers to X86vzmovl which was using (v)blendps - causing domain switch stalls. This patch fixes this by using (v)pblendw instead.

The updated tests on test/CodeGen/X86/sse41.ll still contain a domain stall due to the use of insertps - I'm looking at fixing this in a future patch.

Differential Revision: http://reviews.llvm.org/D6458

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223165 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoGive lit a --xunit-xml-output option for saving results in xunit format
Chris Matthews [Tue, 2 Dec 2014 22:19:21 +0000 (22:19 +0000)]
Give lit a --xunit-xml-output option for saving results in xunit format

  --xunit-xml-output saves test results to disk in JUnit's xml format. This will allow Jenkins to report the details of a lit run.

  Based on a patch by David Chisnall.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223163 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[PowerPC] Implement readcyclecounter for PPC32
Hal Finkel [Tue, 2 Dec 2014 22:01:00 +0000 (22:01 +0000)]
[PowerPC] Implement readcyclecounter for PPC32

We've long supported readcyclecounter on PPC64, but it is easier there (the
read of the 64-bit time-base register can be accomplished via a single
instruction). This now provides an implementation for PPC32 as well. On PPC32,
the time-base register is still 64 bits, but can only be read 32 bits at a time
via two separate SPRs. The ISA manual explains how to do this properly (it
involves re-reading the upper bits and looping if the counter has wrapped while
being read).

This requires PPC to implement a custom integer splitting legalization for the
READCYCLECOUNTER node, turning it into a target-specific SDAG node, which then
gets turned into a pseudo-instruction, which is then expanded to the necessary
sequence (which has three SPR reads, the comparison and the branch).

Thanks to Paul Hargrove for pointing out to me that this was still unimplemented.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223161 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Emit amd_kernel_code_t header for AMDGPU environment
Tom Stellard [Tue, 2 Dec 2014 22:00:07 +0000 (22:00 +0000)]
R600/SI: Emit amd_kernel_code_t header for AMDGPU environment

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223160 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoMake sure that the TargetOptions operator== is checking the
Eric Christopher [Tue, 2 Dec 2014 21:57:15 +0000 (21:57 +0000)]
Make sure that the TargetOptions operator== is checking the
full contents of the class.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223159 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[AArch64][Stackmaps] Optimize stackmap shadows on AArch64.
Lang Hames [Tue, 2 Dec 2014 21:36:24 +0000 (21:36 +0000)]
[AArch64][Stackmaps] Optimize stackmap shadows on AArch64.

Reduce the number of nops emitted for stackmap shadows on AArch64 by counting
non-stackmap instructions up to the next branch target towards the requested
shadow.

<rdar://problem/14959522>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223156 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Move more information into SIProgramInfo struct
Tom Stellard [Tue, 2 Dec 2014 21:28:53 +0000 (21:28 +0000)]
R600/SI: Move more information into SIProgramInfo struct

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223154 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd bindings for the rest of the MCJIT options that we previously
Eric Christopher [Tue, 2 Dec 2014 21:09:01 +0000 (21:09 +0000)]
Add bindings for the rest of the MCJIT options that we previously
had support for. We're still missing a binding for an MCJIT
memory manager.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223153 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600: Cleanup some tests and add missing testcases
Matt Arsenault [Tue, 2 Dec 2014 21:02:20 +0000 (21:02 +0000)]
R600: Cleanup some tests and add missing testcases

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223151 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRestructure some assertion checking based on post commit feedback by Aaron and Tom.
Philip Reames [Tue, 2 Dec 2014 21:01:48 +0000 (21:01 +0000)]
Restructure some assertion checking based on post commit feedback by Aaron and Tom.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223150 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[mips] Fix passing of small structures for big-endian O32.
Daniel Sanders [Tue, 2 Dec 2014 20:40:27 +0000 (20:40 +0000)]
[mips] Fix passing of small structures for big-endian O32.

Summary:
Like N32/N64, they must be passed in the upper bits of the register.

The new code could be merged with the existing if-statements but I've
refrained from doing this since it will make porting the O32 implementation
to tablegen harder later.

Reviewers: vmedic

Reviewed By: vmedic

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D6463

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223148 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoIntroduce CPUStringIsValid() into MCSubtargetInfo and use it for ARM .cpu parsing.
Roman Divacky [Tue, 2 Dec 2014 20:03:22 +0000 (20:03 +0000)]
Introduce CPUStringIsValid() into MCSubtargetInfo and use it for ARM .cpu parsing.

Previously .cpu directive in ARM assembler didnt switch to the new CPU and
therefore acted as a nop. This implemented real action for .cpu and eg.
allows to assembler FreeBSD kernel with -integrated-as.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223147 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Refactor AMDGPUAsmPrinter::EmitProgramInfoSI()
Tom Stellard [Tue, 2 Dec 2014 19:45:05 +0000 (19:45 +0000)]
R600/SI: Refactor AMDGPUAsmPrinter::EmitProgramInfoSI()

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223144 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Statepoints 4/4] Statepoint infrastructure for garbage collection: Documentation
Philip Reames [Tue, 2 Dec 2014 19:37:00 +0000 (19:37 +0000)]
[Statepoints 4/4] Statepoint infrastructure for garbage collection: Documentation

This is the fourth and final patch in the statepoint series.  It contains the documentation for the statepoint intrinsics and their usage.

There's definitely still room to improve the documentation here, but I wanted to get this landed so it was available for others.  There will likely be a series of small cleanup changes over the next few weeks as we work to clarify and revise the documentation.  If you have comments or questions, please feel free to discuss them either in this commit thread, the original review thread, or on llvmdev.  Comments are more than welcome.

Reviewed by: atrick, ributzka
Differential Revision: http://reviews.llvm.org/D5683

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223143 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAppease a build bot complaining about an unused variable that's used in an assertion.
Philip Reames [Tue, 2 Dec 2014 19:28:57 +0000 (19:28 +0000)]
Appease a build bot complaining about an unused variable that's used in an assertion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223142 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agocmake: Remove MAXPATHLEN define as autoconf does not provide it
Reid Kleckner [Tue, 2 Dec 2014 18:59:08 +0000 (18:59 +0000)]
cmake: Remove MAXPATHLEN define as autoconf does not provide it

Presumably it was added to the CMake system when MAXPATHLEN was still
used by code built for Windows. Currently only lib/Support/Path.inc uses
MAXPATHLEN, and it should be available on all Unices.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223139 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRemove '#undef const' from config.h.cmake to sync with autoconf
Reid Kleckner [Tue, 2 Dec 2014 18:58:38 +0000 (18:58 +0000)]
Remove '#undef const' from config.h.cmake to sync with autoconf

This define was removed from config.h.in when Rafael removed our use of
libtool.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223138 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Statepoints 3/4] Statepoint infrastructure for garbage collection: SelectionDAGBuilder
Philip Reames [Tue, 2 Dec 2014 18:50:36 +0000 (18:50 +0000)]
[Statepoints 3/4] Statepoint infrastructure for garbage collection: SelectionDAGBuilder

This is the third patch in a small series.  It contains the CodeGen support for lowering the gc.statepoint intrinsic sequences (223078) to the STATEPOINT pseudo machine instruction (223085).  The change also includes the set of helper routines and classes for working with gc.statepoints, gc.relocates, and gc.results since the lowering code uses them.

With this change, gc.statepoints should be functionally complete.  The documentation will follow in the fourth change, and there will likely be some cleanup changes, but interested parties can start experimenting now.

I'm not particularly happy with the amount of code or complexity involved with the lowering step, but at least it's fairly well isolated.  The statepoint lowering code is split into it's own files and anyone not working on the statepoint support itself should be able to ignore it.

During the lowering process, we currently spill aggressively to stack. This is not entirely ideal (and we have plans to do better), but it's functional, relatively straight forward, and matches closely the implementations of the patchpoint intrinsics.  Most of the complexity comes from trying to keep relocated copies of values in the same stack slots across statepoints.  Doing so avoids the insertion of pointless load and store instructions to reshuffle the stack.  The current implementation isn't as effective as I'd like, but it is functional and 'good enough' for many common use cases.

In the long term, I'd like to figure out how to integrate the statepoint lowering with the register allocator.  In principal, we shouldn't need to eagerly spill at all.  The register allocator should do any spilling required and the statepoint should simply record that fact.  Depending on how challenging that turns out to be, we may invest in a smarter global stack slot assignment mechanism as a stop gap measure.

Reviewed by: atrick, ributzka

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223137 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[SwitchLowering] Handle destinations on multiple phi instructions
Bruno Cardoso Lopes [Tue, 2 Dec 2014 18:31:53 +0000 (18:31 +0000)]
[SwitchLowering] Handle destinations on multiple phi instructions

Follow up from r222926. Also handle multiple destinations from merged
cases on multiple and subsequent phi instructions.

rdar://problem/19106978

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223135 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[MachineCSE] Clear kill-flag on registers imp-def'd by the CSE'd instruction.
Ahmed Bougacha [Tue, 2 Dec 2014 18:09:51 +0000 (18:09 +0000)]
[MachineCSE] Clear kill-flag on registers imp-def'd by the CSE'd instruction.

Go through implicit defs of CSMI and MI, and clear the kill flags on
their uses in all the instructions between CSMI and MI.
We might have made some of the kill flags redundant, consider:
  subs  ... %NZCV<imp-def>        <- CSMI
  csinc ... %NZCV<imp-use,kill>   <- this kill flag isn't valid anymore
  subs  ... %NZCV<imp-def>        <- MI, to be eliminated
  csinc ... %NZCV<imp-use,kill>
Since we eliminated MI, and reused a register imp-def'd by CSMI
(here %NZCV), that register, if it was killed before MI, should have
that kill flag removed, because it's lifetime was extended.

Also, add an exhaustive testcase for the motivating example.

Reviewed by: Juergen Ributzka <juergen@apple.com>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223133 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRemove unneccessary code introduced with 223101.
Philip Reames [Tue, 2 Dec 2014 18:06:10 +0000 (18:06 +0000)]
Remove unneccessary code introduced with 223101.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223132 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Set correct number of user sgprs for HSA runtime
Tom Stellard [Tue, 2 Dec 2014 17:41:43 +0000 (17:41 +0000)]
R600/SI: Set correct number of user sgprs for HSA runtime

We don't support scratch buffers yet with HSA.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223130 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[OCaml] Add Llvm.mdnull.
Peter Zotov [Tue, 2 Dec 2014 17:35:26 +0000 (17:35 +0000)]
[OCaml] Add Llvm.mdnull.

Patch by Gideon Smeding <gideon.smeding@3ds.com>.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223129 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agofix typo in comment
Sanjay Patel [Tue, 2 Dec 2014 17:25:27 +0000 (17:25 +0000)]
fix typo in comment

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223127 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAArch64: make register block rules apply to vector types too.
Tim Northover [Tue, 2 Dec 2014 17:15:22 +0000 (17:15 +0000)]
AArch64: make register block rules apply to vector types too.

The blocking code originated in ARM, which is more aggressive about casting
types to a canonical representative before doing anything else, so I missed out
most vector HFAs and broke the ABI. This should fix it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223126 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoR600/SI: Set the ATC bit on all resource descriptors for the HSA runtime
Tom Stellard [Tue, 2 Dec 2014 17:05:41 +0000 (17:05 +0000)]
R600/SI: Set the ATC bit on all resource descriptors for the HSA runtime

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223125 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoTriple: Add AMDHSA operating system type
Tom Stellard [Tue, 2 Dec 2014 16:45:47 +0000 (16:45 +0000)]
Triple: Add AMDHSA operating system type

This operating system type represents the AMD HSA runtime,
and will be required by the R600 backend in order to generate
correct code for this runtime.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223124 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[LICM] Avoind store sinking if no preheader is available
Bruno Cardoso Lopes [Tue, 2 Dec 2014 14:22:34 +0000 (14:22 +0000)]
[LICM] Avoind store sinking if no preheader is available

Load instructions are inserted into loop preheaders when sinking stores
and later removed if not used by the SSA updater. Avoid sinking if the
loop has no preheader and avoid crashes. This fixes one more side effect
of not handling indirectbr instructions properly on LoopSimplify.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223119 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoRemove unused function.
Asiri Rathnayake [Tue, 2 Dec 2014 12:09:55 +0000 (12:09 +0000)]
Remove unused function.

Removing an unused function which is causing one of the build bots to fail.
This was introduced in the commit r223113. A proper cleanup of the so_imm
tblgen defintion (made redundant by the mod_imm definition) needs to happen
soon.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223115 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd support for ARM modified-immediate assembly syntax.
Asiri Rathnayake [Tue, 2 Dec 2014 10:53:20 +0000 (10:53 +0000)]
Add support for ARM modified-immediate assembly syntax.

Certain ARM instructions accept 32-bit immediate operands encoded as a 8-bit
integer value (0-255) and a 4-bit rotation (0-30, even). Current ARM assembly
syntax support in LLVM allows the decoded (32-bit) immediate to be specified
as a single immediate operand for such instructions:

mov r0, #4278190080

The ARMARM defines an extended assembly syntax allowing the encoding to be made
more explicit, as in:

mov r0, #255, #8 ; (same 32-bit value as above)

The behaviour of the two instructions can be different w.r.t flags, which is
documented under "Modified immediate constants" in ARMARM. This patch enables
support for this extended syntax at the MC layer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223113 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoAdd ARM relocations to ELFYAML
Will Newton [Tue, 2 Dec 2014 09:49:09 +0000 (09:49 +0000)]
Add ARM relocations to ELFYAML

Tested with check-all with no regressions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223112 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoEmit Tag_ABI_FP_denormal correctly in fast-math mode.
Charlie Turner [Tue, 2 Dec 2014 08:22:29 +0000 (08:22 +0000)]
Emit Tag_ABI_FP_denormal correctly in fast-math mode.

The default ARM floating-point mode does not support IEEE 754 mode exactly. Of
relevance to this patch is that input denormals are flushed to zero. The way in
which they're flushed to zero depends on the architecture,

  * For VFPv2, it is implementation defined as to whether the sign of zero is
    preserved.
  * For VFPv3 and above, the sign of zero is always preserved when a denormal
    is flushed to zero.

When FP support has been disabled, the strategy taken by this patch is to
assume the software support will mirror the behaviour of the hardware support
for the target *if it existed*. That is, for architectures which can only have
VFPv2, it is assumed the software will flush to positive zero. For later
architectures it is assumed the software will flush to zero preserving sign.

Change-Id: Icc5928633ba222a4ba3ca8c0df44a440445865fd

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223110 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[signext.ll] Removal Of Duplicate Test Cases
Sonam Kumari [Tue, 2 Dec 2014 05:29:47 +0000 (05:29 +0000)]
[signext.ll] Removal Of Duplicate Test Cases

Removed the duplicate test case existing in signext.ll file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223109 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix variable used only in assertion.
Nick Lewycky [Tue, 2 Dec 2014 01:09:56 +0000 (01:09 +0000)]
Fix variable used only in assertion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223101 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoFix several bugs in r221220's new program finding code.
Chandler Carruth [Tue, 2 Dec 2014 00:52:01 +0000 (00:52 +0000)]
Fix several bugs in r221220's new program finding code.

In both the Unix and Windows variants, std::getenv was called and the
result passed directly to a function accepting a StringRef. This isn't
OK because it might return a null pointer and that causes the StringRef
constructor to assert (and generally produces crash-prone code if
asserts are disabled). Fix this by independently testing the result as
non-null prior to splitting things.

This in turn uncovered another bug in the Unix variant where it would
infinitely recurse if PATH="", or after this fix if PATH isn't set.
There is no need to recurse at all. Slightly re-arrange the code to make
it clear that we can just fixup the Paths argument based on the
environment if we find anything.

I don't know of a particularly useful way to test these routines in
LLVM. I'll commit a test to Clang that ensures that its driver correctly
handles various settings of PATH. However, I have no idea how to
correctly write a Windows test for the PATHEXT change. Any Windows
developers who could provide such a test, please have at. =D

Many thanks to Nick Lewycky and others for helping debug this. =/ It was
quite nasty for us to track down.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223099 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoSimplify pointer comparisons involving memory allocation functions
Hal Finkel [Mon, 1 Dec 2014 23:38:06 +0000 (23:38 +0000)]
Simplify pointer comparisons involving memory allocation functions

System memory allocation functions, which are identified at the IR level by the
noalias attribute on the return value, must return a pointer into a memory region
disjoint from any other memory accessible to the caller. We can use this
property to simplify pointer comparisons between allocated memory and local
stack addresses and the addresses of global variables. Neither the stack nor
global variables can overlap with the region used by the memory allocator.

Fixes PR21556.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223093 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoTry to fix a bot failure due to a variable used only in an assert.
Philip Reames [Mon, 1 Dec 2014 23:27:45 +0000 (23:27 +0000)]
Try to fix a bot failure due to a variable used only in an assert.

Specifically, bot lld-x86_64-darwin13.  Resulting from change 223085.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223092 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Statepoints 2/4] Statepoint infrastructure for garbage collection: MI & x86-64 Backend
Philip Reames [Mon, 1 Dec 2014 22:52:56 +0000 (22:52 +0000)]
[Statepoints 2/4] Statepoint infrastructure for garbage collection: MI & x86-64 Backend

This is the second patch in a small series.  This patch contains the MachineInstruction and x86-64 backend pieces required to lower Statepoints.  It does not include the code to actually generate the STATEPOINT machine instruction and as a result, the entire patch is currently dead code.  I will be submitting the SelectionDAG parts within the next 24-48 hours.  Since those pieces are by far the most complicated, I wanted to minimize the size of that patch.  That patch will include the tests which exercise the functionality in this patch.  The entire series can be seen as one combined whole in http://reviews.llvm.org/D5683.

The STATEPOINT psuedo node is generated after all gc values are explicitly spilled to stack slots.  The purpose of this node is to wrap an actual call instruction while recording the spill locations of the meta arguments used for garbage collection and other purposes.  The STATEPOINT is modeled as modifing all of those locations to prevent backend optimizations from forwarding the value from before the STATEPOINT to after the STATEPOINT.  (Doing so would break relocation semantics for collectors which wish to relocate roots.)

The implementation of STATEPOINT is closely modeled on PATCHPOINT.  Eventually, much of the code in this patch will be removed.  The long term plan is to merge the functionality provided by statepoints and patchpoints.  Merging their implementations in the backend is likely to be a good starting point.

Reviewed by: atrick, ributzka

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223085 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[Statepoints 1/4] Statepoint infrastructure for garbage collection: IR Intrinsics
Philip Reames [Mon, 1 Dec 2014 21:18:12 +0000 (21:18 +0000)]
[Statepoints 1/4] Statepoint infrastructure for garbage collection: IR Intrinsics

The statepoint intrinsics are intended to enable precise root tracking through the compiler as to support garbage collectors of all types. The addition of the statepoint intrinsics to LLVM should have no impact on the compilation of any program which does not contain them. There are no side tables created, no extra metadata, and no inhibited optimizations.

A statepoint works by transforming a call site (or safepoint poll site) into an explicit relocation operation. It is the frontend's responsibility (or eventually the safepoint insertion pass we've developed, but that's not part of this patch series) to ensure that any live pointer to a GC object is correctly added to the statepoint and explicitly relocated. The relocated value is just a normal SSA value (as seen by the optimizer), so merges of relocated and unrelocated values are just normal phis. The explicit relocation operation, the fact the statepoint is assumed to clobber all memory, and the optimizers standard semantics ensure that the relocations flow through IR optimizations correctly.

This is the first patch in a small series.  This patch contains only the IR parts; the documentation and backend support will be following separately.  The entire series can be seen as one combined whole in http://reviews.llvm.org/D5683.

Reviewed by: atrick, ributzka

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223078 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[NVPTX] Do not emit .weak symbols for NVPTX
Jingyue Wu [Mon, 1 Dec 2014 21:16:17 +0000 (21:16 +0000)]
[NVPTX] Do not emit .weak symbols for NVPTX

Summary:
".weak" symbols cannot be consumed by ptxas (PR21685). This patch makes the
weak directive in MCAsmPrinter customizable, and disables emitting ".weak"
symbols for NVPTX.

Test Plan: weak-linkage.ll

Reviewers: jholewinski

Reviewed By: jholewinski

Subscribers: majnemer, jholewinski, llvm-commits

Differential Revision: http://reviews.llvm.org/D6455

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223077 91177308-0d34-0410-b5e6-96231b3b80d8

9 years agoParse 'ghccc' in .ll files as the GHC convention (cc 10)
Reid Kleckner [Mon, 1 Dec 2014 21:04:44 +0000 (21:04 +0000)]
Parse 'ghccc' in .ll files as the GHC convention (cc 10)

Previously we just used "cc 10" in the .ll files, but that isn't very
human readable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223076 91177308-0d34-0410-b5e6-96231b3b80d8

9 years ago[AArch64] Don't combine "select (setcc i1 LHS, RHS), vL, vR".
Ahmed Bougacha [Mon, 1 Dec 2014 20:59:00 +0000 (20:59 +0000)]
[AArch64] Don't combine "select (setcc i1 LHS, RHS), vL, vR".

r208210 introduced an optimization that improves the vector select
codegen by doing the setcc on vectors directly.
This is a problem they the setcc operands are i1s, because the
optimization would create vectors of i1, which aren't legal.

Part of PR21549.

Differential Revision: http://reviews.llvm.org/D6308

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223075 91177308-0d34-0410-b5e6-96231b3b80d8