oota-llvm.git
10 years agoTeach combineMetadata how to merge 'nonnull' metadata.
Philip Reames [Tue, 21 Oct 2014 21:02:19 +0000 (21:02 +0000)]
Teach combineMetadata how to merge 'nonnull' metadata.

combineMetadata is used when merging two instructions into one.  This change teaches it how to merge 'nonnull' - i.e. only preserve it on the new instruction if it's set on both sources.  This isn't actually used yet since I haven't adjusted any of the call sites to pass in nonnull as a 'known metadata'.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220325 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoPreserve 'nonnull' when changing type of the load.
Philip Reames [Tue, 21 Oct 2014 21:00:03 +0000 (21:00 +0000)]
Preserve 'nonnull' when changing type of the load.

When changing the type of a load in Chandler's recent InstCombine changes, we can preserve the new 'nonnull' metadata.

I considered adding an assert since 'nonnull' is only valid on pointer types, but casting a pointer to a non-pointer would involve more than a bitcast anyways.  If someone extends this transform to handle more than bitcasts, the verifier will report the malformed IR, so a separate assertion isn't needed.  Also, the fpmath flags would have the same problem.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220324 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoExtend the verifier to check usage of 'nonnull' metadata.
Philip Reames [Tue, 21 Oct 2014 20:56:29 +0000 (20:56 +0000)]
Extend the verifier to check usage of 'nonnull' metadata.

The recently added !nonnull metadata is only valid on loads of pointer type.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220323 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PBQP] Teach PassConfig to tell if the default register allocator is used.
Arnaud A. de Grandmaison [Tue, 21 Oct 2014 20:47:22 +0000 (20:47 +0000)]
[PBQP] Teach PassConfig to tell if the default register allocator is used.

This enables targets to adapt their pass pipeline to the register
allocator in use. For example, with the AArch64 backend, using PBQP
with the cortex-a57, the FPLoadBalancing pass is no longer necessary.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220321 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMove code a bit to avoid a few declarations. NFC.
Rafael Espindola [Tue, 21 Oct 2014 20:34:57 +0000 (20:34 +0000)]
Move code a bit to avoid a few declarations. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220317 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PBQP] Add a testcase for r220302: Fix coalescing benefits
Arnaud A. de Grandmaison [Tue, 21 Oct 2014 20:10:21 +0000 (20:10 +0000)]
[PBQP] Add a testcase for r220302: Fix coalescing benefits

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220316 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoInstCombine: Simplify FoldICmpCstShrCst
David Majnemer [Tue, 21 Oct 2014 19:51:55 +0000 (19:51 +0000)]
InstCombine: Simplify FoldICmpCstShrCst

This function was complicated by the fact that it tried to perform
canonicalizations that were already preformed by InstSimplify.  Remove
this extra code and move the tests over to InstSimplify.  Add asserts to
make sure our preconditions hold before we make any assumptions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220314 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDrop support for an old version of ld64 (from darwin 9).
Rafael Espindola [Tue, 21 Oct 2014 18:31:09 +0000 (18:31 +0000)]
Drop support for an old version of ld64 (from darwin 9).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220310 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoremove function names from comments; NFC
Sanjay Patel [Tue, 21 Oct 2014 18:26:57 +0000 (18:26 +0000)]
remove function names from comments; NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220309 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoConvert two tests to use llvm-readobj.
Rafael Espindola [Tue, 21 Oct 2014 18:24:31 +0000 (18:24 +0000)]
Convert two tests to use llvm-readobj.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220308 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Add pattern for bswap
Matt Arsenault [Tue, 21 Oct 2014 16:25:08 +0000 (16:25 +0000)]
R600/SI: Add pattern for bswap

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220304 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PBQP] Check for out of bound access in DEBUG builds
Arnaud A. de Grandmaison [Tue, 21 Oct 2014 16:24:21 +0000 (16:24 +0000)]
[PBQP] Check for out of bound access in DEBUG builds

It is just too easy to use a virtual register intead of a NodeId without a
compiler warning. This does not fix the fundamental problem, i.e. both
have the same underlying types, but increases the likelyhood to detect it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220303 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PBQP] Fix coalescing benefits
Arnaud A. de Grandmaison [Tue, 21 Oct 2014 16:24:15 +0000 (16:24 +0000)]
[PBQP] Fix coalescing benefits

As coalescing registers is a benefit, the cost should be improved (i.e. made smaller) when coalescing is possible.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220302 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoX86AsmInstrumentation.cpp: Dissolve initializer-ranged-for. MSC17 disliked it.
NAKAMURA Takumi [Tue, 21 Oct 2014 16:22:52 +0000 (16:22 +0000)]
X86AsmInstrumentation.cpp: Dissolve initializer-ranged-for. MSC17 disliked it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220301 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSilence a -Wcast-qual warning; NFC.
Aaron Ballman [Tue, 21 Oct 2014 16:12:37 +0000 (16:12 +0000)]
Silence a -Wcast-qual warning; NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220300 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTest commit
Colin LeMahieu [Tue, 21 Oct 2014 16:03:10 +0000 (16:03 +0000)]
Test commit
Fixing brief comment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220299 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoComment cleanup. NFC.
Rafael Espindola [Tue, 21 Oct 2014 15:49:46 +0000 (15:49 +0000)]
Comment cleanup. NFC.

Don't duplicate names in comments and remove useless ones. Hopefully anyone
reading this knows what main is.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220298 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd support for addmod to mri scripts.
Rafael Espindola [Tue, 21 Oct 2014 14:46:17 +0000 (14:46 +0000)]
Add support for addmod to mri scripts.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220294 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PowerPC] Avoid VSX FMA mutate when killed product reg = addend reg
Bill Schmidt [Tue, 21 Oct 2014 13:02:37 +0000 (13:02 +0000)]
[PowerPC] Avoid VSX FMA mutate when killed product reg = addend reg

With VSX enabled, test/CodeGen/PowerPC/recipest.ll exposes a bug in
the FMA mutation pass.  If we have a situation where a killed product
register is the same register as the FMA target, such as:

   %vreg5<def,tied1> = XSNMSUBADP %vreg5<tied0>, %vreg11, %vreg5,
                       %RM<imp-use>; VSFRC:%vreg5 F8RC:%vreg11

then the substitution makes no sense.  We end up getting a crash when
we try to extend the interval associated with the killed product
register, as there is already a live range for %vreg5 there.  This
patch just disables the mutation under those circumstances.

Since recipest.ll generates different code with VMX enabled, I've
modified that test to use -mattr=-vsx.  I've borrowed the code from
that test that exposed the bug and placed it in fma-mutate.ll, where
it tests several mutation opportunities including the "bad" one.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220290 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM] NEON 32-bit scalar moves are also available in VFPv2
Oliver Stannard [Tue, 21 Oct 2014 11:49:14 +0000 (11:49 +0000)]
[ARM] NEON 32-bit scalar moves are also available in VFPv2

The 32-bit variants of the NEON scalar<->GPR move instructions are
also available in VFPv2. The 8- and 16-bit variants do require NEON.

Note that the checks in the test file are all -DAG because they are
checking a mixture of stdout and stderr, and the ordering is not
guaranteed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220288 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[asan-asm-instrumentation] Fixed memory accesses with rbp as a base or an index register.
Yuri Gorshenin [Tue, 21 Oct 2014 10:22:27 +0000 (10:22 +0000)]
[asan-asm-instrumentation] Fixed memory accesses with rbp as a base or an index register.

Summary: Fixed memory accesses with rbp as a base or an index register.

Reviewers: eugenis

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D5819

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220283 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Thumb2] LDRS?[BH] cannot load to the PC
Oliver Stannard [Tue, 21 Oct 2014 09:14:15 +0000 (09:14 +0000)]
[Thumb2] LDRS?[BH] cannot load to the PC

The Thumb2 LDRS?[BH] instructions are not valid when the destination
register is the PC (these encodings are used for preload hints).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220278 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTeach the load analysis to allow finding available values which require
Chandler Carruth [Tue, 21 Oct 2014 09:00:40 +0000 (09:00 +0000)]
Teach the load analysis to allow finding available values which require
inttoptr or ptrtoint cast provided there is datalayout available.
Eventually, the datalayout can just be required but in practice it will
always be there today.

To go with the ability to expose available values requiring a ptrtoint
or inttoptr cast, helpers are added to perform one of these three casts.

These smarts are necessary to finish canonicalizing loads and stores to
the operational type requirements without regressing fundamental
combines.

I've added some test cases. These should actually improve as the load
combining and store combining improves, but they may fundamentally be
highlighting some missing combines for select in addition to exercising
the specific added logic to load analysis.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220277 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][microMIPS] Implement ADDU16 and SUBU16 instructions
Zoran Jovanovic [Tue, 21 Oct 2014 08:44:58 +0000 (08:44 +0000)]
[mips][microMIPS] Implement ADDU16 and SUBU16 instructions
Differential Revision: http://reviews.llvm.org/D5118

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220276 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][microMIPS] Implement AND16, NOT16, OR16 and XOR16 instructions
Zoran Jovanovic [Tue, 21 Oct 2014 08:32:40 +0000 (08:32 +0000)]
[mips][microMIPS] Implement AND16, NOT16, OR16 and XOR16 instructions
Differential Revision: http://reviews.llvm.org/D5117

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220275 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[mips][microMIPS] Implement microMIPS 16-bit instructions registers
Zoran Jovanovic [Tue, 21 Oct 2014 08:23:11 +0000 (08:23 +0000)]
[mips][microMIPS] Implement microMIPS 16-bit instructions registers
Differential Revision: http://reviews.llvm.org/D5116

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220273 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix a bit of confusion about .set and produce more readable assembly.
Rafael Espindola [Tue, 21 Oct 2014 01:17:30 +0000 (01:17 +0000)]
Fix a bit of confusion about .set and produce more readable assembly.

Every target we support has support for assembly that looks like

a = b - c
.long a

What is special about MachO is that the above combination suppresses the
production of a relocation.

With this change we avoid producing the intermediary labels when they don't
add any value.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220256 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDo not attribute static allocas to the call site's DebugLoc.
Paul Robinson [Tue, 21 Oct 2014 01:00:55 +0000 (01:00 +0000)]
Do not attribute static allocas to the call site's DebugLoc.

When functions are inlined, instructions without debug information are
attributed to the call site's DebugLoc. After inlining, inlined static
allocas are moved to the caller's entry block, adjacent to the caller's
original static alloca instructions. By retaining the call site's
DebugLoc, these instructions could cause instructions that were
subsequently inserted at the entry block to pick up the same DebugLoc.

Patch by Wolfgang Pieb!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220255 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMake this test a bit more strict.
Rafael Espindola [Tue, 21 Oct 2014 00:47:49 +0000 (00:47 +0000)]
Make this test a bit more strict.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220253 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTeach lit to filter the host LDFLAGS down from the build system and into
Chandler Carruth [Tue, 21 Oct 2014 00:36:28 +0000 (00:36 +0000)]
Teach lit to filter the host LDFLAGS down from the build system and into
the CGO build environment. This lets things like -rpath propagate down
to the C++ code that is built along side the Go bindings when testing
them.

Patch by Peter Collingbourne, and verified that it works by me.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220252 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoPR21202: Memory leak in Windows RWMutexImpl when using SRWLOCK
David Blaikie [Tue, 21 Oct 2014 00:34:39 +0000 (00:34 +0000)]
PR21202: Memory leak in Windows RWMutexImpl when using SRWLOCK

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220251 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMake AsmPrinter::EmitLabelOffsetDifference a static helper and simplify.
Rafael Espindola [Tue, 21 Oct 2014 00:25:49 +0000 (00:25 +0000)]
Make AsmPrinter::EmitLabelOffsetDifference a static helper and simplify.

It had exactly one caller in a position where we know hasSetDirective is true.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220250 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[MCJIT] Temporarily revert r220245 - it broke several bots.
Lang Hames [Tue, 21 Oct 2014 00:24:02 +0000 (00:24 +0000)]
[MCJIT] Temporarily revert r220245 - it broke several bots.

(See e.g. http://bb.pgr.jp/builders/cmake-llvm-x86_64-linux/builds/17653)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220249 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoIntroduce enum values for previously defined metadata types. (NFC)
Philip Reames [Tue, 21 Oct 2014 00:13:20 +0000 (00:13 +0000)]
Introduce enum values for previously defined metadata types. (NFC)

Our metadata scheme lazily assigns IDs to string metadata, but we have a mechanism to preassign them as well.  Using a preassigned ID is helpful since we get compile time type checking, and avoid some (minimal) string construction and comparison.  This change adds enum value for three existing metadata types:
+    MD_nontemporal = 9, // "nontemporal"
+    MD_mem_parallel_loop_access = 10, // "llvm.mem.parallel_loop_access"
+    MD_nonnull = 11 // "nonnull"

I went through an updated various uses as well.  I made no attempt to get all uses; I focused on the ones which were easily grepable and easily to translate.  For example, there were several items in LoopInfo.cpp I chose not to update.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220248 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoExtend the verifier to validate range metadata on calls and invokes.
Philip Reames [Mon, 20 Oct 2014 23:52:07 +0000 (23:52 +0000)]
Extend the verifier to validate range metadata on calls and invokes.

Range metadata applies to loads, call, and invokes.  We were validating that metadata applied to loads was correct according to the LangRef, but we were not validating metadata applied to calls or invokes.  This change extracts the checking functionality to a common location, reuses it for all valid locations, and adds a simple test to ensure a misused range on a call gets reported.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220246 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[MCJIT] Make MCJIT honor symbol visibility settings when populating the global
Lang Hames [Mon, 20 Oct 2014 23:39:54 +0000 (23:39 +0000)]
[MCJIT] Make MCJIT honor symbol visibility settings when populating the global
symbol table.

Patch by Anthony Pesch. Thanks Anthony!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220245 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[X86] Fix a bug in the lowering of the mask of VSELECT.
Quentin Colombet [Mon, 20 Oct 2014 23:13:30 +0000 (23:13 +0000)]
[X86] Fix a bug in the lowering of the mask of VSELECT.
X86 code to lower VSELECT messed a bit with the bits set in the mask of VSELECT
when it knows it can be lowered into BLEND. Indeed, only the high bits need to be
set for those and it optimizes those accordingly.
However, when the mask is a compile time constant, the lowering will be handled
by the generic optimizer and those modifications will generate bad code in the
generic optimizer.

This patch fixes that by preventing the optimization if the VSELECT will be
handled by the generic optimizer.

<rdar://problem/18675020>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220242 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoIntroduce a 'nonnull' metadata on Load instructions.
Philip Reames [Mon, 20 Oct 2014 22:40:55 +0000 (22:40 +0000)]
Introduce a 'nonnull' metadata on Load instructions.

The newly introduced 'nonnull' metadata is analogous to existing 'nonnull' attributes, but applies to load instructions rather than call arguments or returns.  Long term, it would be nice to combine these into a single construct.   The value of the load is allowed to vary between successive loads, but null is not a valid value to be loaded by any load marked nonnull.

Reviewed by: Hal Finkel
Differential Revision:  http://reviews.llvm.org/D5220

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220240 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[X86] Memory folding for commutative instructions (updated)
Simon Pilgrim [Mon, 20 Oct 2014 22:14:22 +0000 (22:14 +0000)]
[X86] Memory folding for commutative instructions (updated)

This patch improves support for commutative instructions in the x86 memory folding implementation by attempting to fold a commuted version of the instruction if the original folding fails - if that folding fails as well the instruction is 're-commuted' back to its original order before returning.

Updated version of r219584 (reverted in r219595) - the commutation attempt now explicitly ensures that neither of the commuted source operands are tied to the destination operand / register, which was the source of all the regressions that occurred with the original patch attempt.

Added additional regression test case provided by Joerg Sonnenberger.

Differential Revision: http://reviews.llvm.org/D5818

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220239 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoExplain why we don't always use --gc-sections.
Rafael Espindola [Mon, 20 Oct 2014 21:37:38 +0000 (21:37 +0000)]
Explain why we don't always use --gc-sections.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220237 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM: rework Thumb1 frame index rewriting
Tim Northover [Mon, 20 Oct 2014 21:28:41 +0000 (21:28 +0000)]
ARM: rework Thumb1 frame index rewriting

The previous code had a few problems, motivating the choices here.

1. It could create instructions clobbering CPSR, but the incoming MachineInstr
   didn't reflect this. A potential source of corruption. This is why the patch
   has a new PseudoInst for before lowering.
2. Similarly, there was some code to handle the incoming instruction not being
   ARMCC::AL, but this would have caused massive problems if it was actually
   invoked when a complex offset needing more than one instruction was requested.
3. It wasn't designed to handle unaligned pointers (or offsets). These should
   probably be minimised anyway, but the code needs to deal with them properly
   regardless.
4. It had some rather dubious ad-hoc code to avoid calling
   emitThumbRegPlusImmediate, a function which should be designed to do precisely
   this job.

We seem to cover the common cases correctly now, and hopefully can enhance
emitThumbRegPlusImmediate to handle any extra optimisations we need to add in
future.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220236 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTry to fix GCC error about invalid use of const_cast in const version of ErrorOr...
Alexey Samsonov [Mon, 20 Oct 2014 20:41:21 +0000 (20:41 +0000)]
Try to fix GCC error about invalid use of const_cast in const version of ErrorOr::get()

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220233 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoConstify getELFDynamicSymbolIterators standalone function. NFC.
Alexey Samsonov [Mon, 20 Oct 2014 20:33:20 +0000 (20:33 +0000)]
Constify getELFDynamicSymbolIterators standalone function. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220232 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd const version of OwningBinary::getBinary
Alexey Samsonov [Mon, 20 Oct 2014 20:32:47 +0000 (20:32 +0000)]
Add const version of OwningBinary::getBinary

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220231 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoBe more specific about return type of MachOUniversalBinary::getObjectForArch
Alexey Samsonov [Mon, 20 Oct 2014 20:30:57 +0000 (20:30 +0000)]
Be more specific about return type of MachOUniversalBinary::getObjectForArch

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220230 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoConstify input argument of RelocVisitor and DWARFContext constructors. NFC.
Alexey Samsonov [Mon, 20 Oct 2014 20:28:51 +0000 (20:28 +0000)]
Constify input argument of RelocVisitor and DWARFContext constructors. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220228 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTeach Lit to catch OSError exceptions when creating a process during the
Dan Liew [Mon, 20 Oct 2014 20:14:28 +0000 (20:14 +0000)]
Teach Lit to catch OSError exceptions when creating a process during the
execution of a shell command. This can happen for example if the
``RUN:`` line calls a python script which can work correctly under
Linux/OSX but will not work under Windows. A more useful error message
is now shown rather than an unhelpful backtrace.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220227 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMoved out IIT_V64 from common values section.
Robert Khasanov [Mon, 20 Oct 2014 19:25:05 +0000 (19:25 +0000)]
Moved out IIT_V64 from common values section.
Thanks Juergen Ributzka for notice.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220224 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[AArch64] test case for compfail fixed by r219748
Gerolf Hoflehner [Mon, 20 Oct 2014 16:08:33 +0000 (16:08 +0000)]
[AArch64] test case for compfail fixed by r219748

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220206 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix Intrinsic::getType not working with vararg
Steven Wu [Mon, 20 Oct 2014 15:47:24 +0000 (15:47 +0000)]
Fix Intrinsic::getType not working with vararg

VarArg Intrinsic functions are encoded with "void" type as the last
argument. Now Intrinsic::getType can correctly return all the intrinsic
function type.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220205 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Thumb2] RFE, SRS and "SUBS pc, lr" are undefined on v7M
Oliver Stannard [Mon, 20 Oct 2014 15:37:35 +0000 (15:37 +0000)]
[Thumb2] RFE, SRS and "SUBS pc, lr" are undefined on v7M

These instructions are related to the v7[AR] exception model, and are
not defined on v7M.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220204 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove unnecessary else.
Sid Manning [Mon, 20 Oct 2014 13:08:19 +0000 (13:08 +0000)]
Remove unnecessary else.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220200 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRevert r220174, "Always use -Wl,-gc-sections on our build."
NAKAMURA Takumi [Mon, 20 Oct 2014 12:12:21 +0000 (12:12 +0000)]
Revert r220174, "Always use -Wl,-gc-sections on our build."

It dropped required functions for plugins with gnu ld 2.20 and 2.21.

  Failing Tests (1):
      LLVM :: Feature/load_module.ll

  Hello: bin/opt: symbol lookup error: lib/LLVMHello.so: undefined symbol: _ZN4llvm11raw_ostream13write_escapedENS_9StringRefEb

  Failing Tests (1):
      Clang :: Frontend/plugins.c

  error: unable to load plugin 'lib/PrintFunctionNames.so': 'lib/PrintFunctionNames.so: undefined symbol: _ZN5clang15PluginASTAction6anchorEv'

I think we should inspect linker's version or behavior to introduce --gc-sections for --export-dynamic.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220198 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ARM] Do not select SMULW[BT] or SMLAW[BT]
Oliver Stannard [Mon, 20 Oct 2014 11:30:35 +0000 (11:30 +0000)]
[ARM] Do not select SMULW[BT] or SMLAW[BT]

The current instruction selection patterns for SMULW[BT] and SMLAW[BT]
are incorrect. These instructions multiply a 32-bit and a 16-bit value
(both signed) and return the top 32 bits of the 48-bit result. This
preserves the 16 bits of overflow, whereas the patterns they currently
match truncate the result to 16 bits then sign extend.

To select these instructions, we would need to match an ISD::SMUL_LOHI,
a sign extend, two shifts and an or. There is no way to match SMUL_LOHI
in an instruction pattern as it defines multiple values, so this would
have to be done in C++. I have raised
http://llvm.org/bugs/show_bug.cgi?id=21297 to cover allowing correct
selection of these instructions.

This fixes http://llvm.org/bugs/show_bug.cgi?id=19396

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220196 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Thumb] Fix crash in Thumb1RegisterInfo::rewriteFrameIndex
Oliver Stannard [Mon, 20 Oct 2014 11:00:18 +0000 (11:00 +0000)]
[Thumb] Fix crash in Thumb1RegisterInfo::rewriteFrameIndex

This function can, for some offsets from the SP, split one instruction
into two. Since it re-uses the original instruction as the first
instruction of the result, we need ensure its result register is not
marked as dead before we use it in the second instruction.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220194 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSwitch the default DataLayout to be little endian, and make the variable
Chandler Carruth [Mon, 20 Oct 2014 10:41:29 +0000 (10:41 +0000)]
Switch the default DataLayout to be little endian, and make the variable
be BigEndian so the default can continue to be zero-initialized.

This is one of the prerequisites to making DataLayout a constant and
always available part of every module.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220193 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove some completely superfluous trailing comments and clang-format
Chandler Carruth [Mon, 20 Oct 2014 10:35:11 +0000 (10:35 +0000)]
Remove some completely superfluous trailing comments and clang-format
this header to remove numerous formatting inconsistencies that impede
making simple changes here without large diffs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220192 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoClean up the comments and doxygen for DataLayout.
Chandler Carruth [Mon, 20 Oct 2014 10:27:53 +0000 (10:27 +0000)]
Clean up the comments and doxygen for DataLayout.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220191 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix a miscompile introduced in r220178.
Chandler Carruth [Mon, 20 Oct 2014 10:03:01 +0000 (10:03 +0000)]
Fix a miscompile introduced in r220178.

The original code had an implicit assumption that if the test for
allocas or globals was reached, the two pointers were not equal. With my
changes to make the pointer analysis more powerful here, I also had to
guard against circumstances where the results weren't useful. That in
turn violated the assumption and gave rise to a circumstance in which we
could have a store with both the queried pointer and stored pointer
rooted at *the same* alloca. Clearly, we cannot ignore such a store.
There are other things we might do in this code to better handle the
case of both pointers ending up at the same alloca or global, but it
seems best to at least make the test explicit in what it intends to
check.

I've added tests for both the alloca and global case here.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220190 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoIR: Replace DataLayout::RoundUpAlignment with RoundUpToAlignment
David Majnemer [Mon, 20 Oct 2014 06:13:33 +0000 (06:13 +0000)]
IR: Replace DataLayout::RoundUpAlignment with RoundUpToAlignment

No functional change intended, just cleaning up some code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220187 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix a somewhat subtle pair of issues with JumpThreading I introduced in
Chandler Carruth [Mon, 20 Oct 2014 05:34:36 +0000 (05:34 +0000)]
Fix a somewhat subtle pair of issues with JumpThreading I introduced in
r220178. First, the creation routine doesn't insert prior to the
terminator of the basic block provided, but really at the end of the
basic block. Instead, get the terminator and insert before that. The
next issue was that we need to ensure multiple PHI node entries for
a single predecessor re-use the same cast instruction rather than
creating new ones.

All of the logic here was without tests previously. I've reduced and
added a test case from the test suite that crashed without both of these
fixes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220186 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PBQP] Use DenseSet rather than std::set for PBQP's PoolCostAllocator
Lang Hames [Mon, 20 Oct 2014 04:26:23 +0000 (04:26 +0000)]
[PBQP] Use DenseSet rather than std::set for PBQP's PoolCostAllocator
implementation.

This is good for a ~6% reduction in total compile time on the nightly test suite
when running with -regalloc=pbqp.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220183 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTeach the load analysis driving core instcombine logic and other bits of
Chandler Carruth [Mon, 20 Oct 2014 00:24:14 +0000 (00:24 +0000)]
Teach the load analysis driving core instcombine logic and other bits of
logic to look through pointer casts, making them trivially stronger in
the face of loads and stores with intervening pointer casts.

I've included a few test cases that demonstrate the kind of folding
instcombine can do without pointer casts and then variations which
obfuscate the logic through bitcasts. Without this patch, the variations
all fail to optimize fully.

This is more important now than it has been in the past as I've started
moving the load canonicialization to more closely follow the value type
requirements rather than the pointer type requirements and thus this
needs to be prepared for more pointer casts. When I made the same change
to stores several test cases regressed without logic along these lines
so I wanted to systematically improve matters first.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220178 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd a datalayout string to this test so that it exercises the full gamut
Chandler Carruth [Mon, 20 Oct 2014 00:11:31 +0000 (00:11 +0000)]
Add a datalayout string to this test so that it exercises the full gamut
of InstCombine rather than just the bits enabled when datalayout is
optional.

The primary fixes here are because now things are little endian.

In good news, silliness like this seems like it will be going away as
we've got pretty stong consensus on dropping optional datalayout
entirely.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220176 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAlways use -Wl,-gc-sections on our build.
Rafael Espindola [Sun, 19 Oct 2014 23:24:46 +0000 (23:24 +0000)]
Always use -Wl,-gc-sections on our build.

Both bfd ld and gold correctly handle --export-dynamic, so gc-sections is safe even for binaries
that support plugins.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220174 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PowerPC] Clean up -mattr=+vsx tests to always specify -mcpu
Bill Schmidt [Sun, 19 Oct 2014 21:29:21 +0000 (21:29 +0000)]
[PowerPC] Clean up -mattr=+vsx tests to always specify -mcpu

We recently discovered an issue that reinforces what a good idea it is
to always specify -mcpu in our code generation tests, particularly for
-mattr=+vsx.  This patch ensures that all tests that specify
-mattr=+vsx also specify -mcpu=pwr7 or -mcpu=pwr8, as appropriate.

Some of the uses of -mattr=+vsx added recently don't make much sense
(when specified for -mtriple=powerpc-apple-darwin8 or -march=ppc32,
for example).  For cases like this I've just removed the extra VSX
test commands; there's enough coverage without them.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220173 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PowerPC] Temporarily disable VSX for PowerPC fast-isel tests
Bill Schmidt [Sun, 19 Oct 2014 20:48:47 +0000 (20:48 +0000)]
[PowerPC] Temporarily disable VSX for PowerPC fast-isel tests

Patch by Bill Seurer; some comment formatting changes by me.

There are a few PowerPC test cases for FastISel support that currently
fail with VSX support enabled.  The temporary workaround under
discussion in http://reviews.llvm.org/D5362 helps, but the tests still
fail because they specify -fast-isel-abort, and the VSX workaround
punts back to SelectionDAG.  We have plans to fix FastISel permanently
for VSX, but until that's in place these tests are preventing us from
enabling VSX by default.  Therefore we are adding -mattr=-vsx to these
tests until the full support is ready.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220172 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PowerPC] Re-enable VSX test line for fma.ll with -mcpu=pwr7
Bill Schmidt [Sun, 19 Oct 2014 20:27:56 +0000 (20:27 +0000)]
[PowerPC] Re-enable VSX test line for fma.ll with -mcpu=pwr7

The VSX testing variant in test/CodeGen/PowerPC/fma.ll had to be
disabled because of unexpected behavior on many of the builders.  I
tracked this down to a situation that occurs when the VSX attribute is
enabled for a target that disables the MI early scheduling pass.  This
patch adds -mcpu=pwr7 to make this predictable.  The other issue will
be addressed separately.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220171 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[ADT] Add a 'find_as' operation to DenseSet.
Lang Hames [Sun, 19 Oct 2014 19:36:33 +0000 (19:36 +0000)]
[ADT] Add a 'find_as' operation to DenseSet.

This operation is analogous to its counterpart in DenseMap: It allows lookup
via cheap-to-construct keys (provided that getHashValue and isEqual are
implemented for the cheap key-type in the DenseMapInfo specialization).

Thanks to Chandler for the review.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220168 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDo a better and more complete job of preserving metadata when combining
Chandler Carruth [Sun, 19 Oct 2014 10:46:46 +0000 (10:46 +0000)]
Do a better and more complete job of preserving metadata when combining
loads.

This handles many more cases than just the AA metadata, some of them
suggested by Hal in his review of the AA metadata handling patch. I've
tried to test this behavior where tractable to do so.

I'll point out that I have specifically *not* included a test for
debuginfo because it was going to require 2 or 3 times as much work to
craft some input which would survive the "helpful" stripping of debug
info metadata that doesn't match the desired schema. This is another
good example of why the current state of write-ability for our debug
info metadata is unacceptable. I spent over 30 minutes trying to conjure
some test case that would survive, even copying from other debug info
tests, but it always failed to survive with no explanation of why or how
I might fix it. =[

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220165 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMove previously dead code to handle computing the known bits of an alias
Chandler Carruth [Sun, 19 Oct 2014 09:06:56 +0000 (09:06 +0000)]
Move previously dead code to handle computing the known bits of an alias
up to where it actually works as intended. The problem is that
a GlobalAlias isa GlobalValue and so the prior block handled all of the
cases.

This allows us to constant fold based on the actual constant expression
in the global alias. As an example, see the last function in the newly
added test case which explicitly aligns an unaligned pointer using
constant expression math. Without this change, we fail to see that and
fold an alignment test to zero.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220164 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoInstCombine: (sub (or A B) (xor A B)) --> (and A B)
David Majnemer [Sun, 19 Oct 2014 08:32:32 +0000 (08:32 +0000)]
InstCombine: (sub (or A B) (xor A B)) --> (and A B)

The following implements the transformation:
(sub (or A B) (xor A B)) --> (and A B).

Patch by Ankur Garg!

Differential Revision: http://reviews.llvm.org/D5719

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220163 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoInstCombine: Optimize icmp eq/ne (shl Const2, A), Const1
David Majnemer [Sun, 19 Oct 2014 08:23:08 +0000 (08:23 +0000)]
InstCombine: Optimize icmp eq/ne (shl Const2, A), Const1

The following implements the optimization for sequences of the form:
icmp eq/ne (shl Const2, A), Const1

Such sequences can be transformed to:
icmp eq/ne A, (TrailingZeros(Const1) - TrailingZeros(Const2))

This handles only the equality operators for now. Other operators need
to be handled.

Patch by Ankur Garg!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220162 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix a long-standing miscompile in the load analysis that was uncovered
Chandler Carruth [Sun, 19 Oct 2014 08:17:50 +0000 (08:17 +0000)]
Fix a long-standing miscompile in the load analysis that was uncovered
by my refactoring of this code.

The method isSafeToLoadUnconditionally assumes that the load will
proceed with the preferred type alignment. Given that, it has to ensure
that the alloca or global is at least that aligned. It has always done
this historically when a datalayout is present, but has never checked it
when the datalayout is absent. When I refactored the code in r220156,
I exposed this path when datalayout was present and that turned the
latent bug into a patent bug.

This fixes the issue by just removing the special case which allows
folding things without datalayout. This isn't worth the complexity of
trying to tease apart when it is or isn't safe without actually knowing
the preferred alignment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220161 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSwitch how the datalayout availability test is handled in this code to
Chandler Carruth [Sun, 19 Oct 2014 00:42:16 +0000 (00:42 +0000)]
Switch how the datalayout availability test is handled in this code to
make much more sense and in theory be more correct.

If you trace the code alllll the way back to when it was first
introduced, the comments make it slightly more clear what was going on
here. At that time, the only way Base != V was if DL (then TD) was
non-null. As a consequence, if DL *was* null, that meant we were loading
directly from the alloca or global found above the test. After
refactoring, this has become at least terribly subtle and potentially
incorrect. There are many forms of pointer manipulation that can be
traversed without DataLayout, and some of them would in fact change the
size of object being loaded vs. allocated.

Rather than this subtlety, I've hoisted the actual 'return true' bits
into the code which actually found an alloca or global and based them on
the loaded pointer being that alloca or global. This is both more clear
and safer. I've also added comments about exactly why this set of
predicates is used.

I've also corrected a misleading comment about globals -- if overridden
they may not just have a different size, they may be null and completely
unsafe to load from!

Hopefully this confuses the next reader a bit less. I don't have any
test cases or anything, the patch is motivated strictly to improve the
readability of the code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220156 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse triple predicate functions instead of checking values directly. NFC.
Bob Wilson [Sun, 19 Oct 2014 00:39:30 +0000 (00:39 +0000)]
Use triple predicate functions instead of checking values directly. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220155 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRename 'TD' to 'DL' in this function as the argument is now a DataLayout
Chandler Carruth [Sat, 18 Oct 2014 23:47:22 +0000 (23:47 +0000)]
Rename 'TD' to 'DL' in this function as the argument is now a DataLayout
argument.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220151 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix the other comment to use modern doxygen style and be a bit more
Chandler Carruth [Sat, 18 Oct 2014 23:46:17 +0000 (23:46 +0000)]
Fix the other comment to use modern doxygen style and be a bit more
direct. Notably, comment on the fact that the loaded type is significant
in that it determines how wide of an access must be safe.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220150 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMore formatting cleanup brought to you by clang-format.
Chandler Carruth [Sat, 18 Oct 2014 23:41:25 +0000 (23:41 +0000)]
More formatting cleanup brought to you by clang-format.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220149 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoClean up doxygen syntax and reword comments to flow better, have a brief
Chandler Carruth [Sat, 18 Oct 2014 23:31:55 +0000 (23:31 +0000)]
Clean up doxygen syntax and reword comments to flow better, have a brief
section, and not have unfinished sentence fragments.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220147 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoClean up the formatting and trailing whitespace of a routine before
Chandler Carruth [Sat, 18 Oct 2014 23:19:03 +0000 (23:19 +0000)]
Clean up the formatting and trailing whitespace of a routine before
editting it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220146 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PBQP] Move register-allocation specific PBQP code into RegAllocPBQP.h.
Lang Hames [Sat, 18 Oct 2014 22:23:55 +0000 (22:23 +0000)]
[PBQP] Move register-allocation specific PBQP code into RegAllocPBQP.h.

Just clean-up - no functional change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220145 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[PBQP] Replace the interference-constraints algorithm with a faster version
Lang Hames [Sat, 18 Oct 2014 17:26:07 +0000 (17:26 +0000)]
[PBQP] Replace the interference-constraints algorithm with a faster version
loosely based on linear scan.

On x86-64 this is good for a ~2% drop in compile time on the nightly test suite.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220143 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoPreserve AA metadata when combining (cast (load (...))) -> (load (cast
Chandler Carruth [Sat, 18 Oct 2014 11:00:12 +0000 (11:00 +0000)]
Preserve AA metadata when combining (cast (load (...))) -> (load (cast
(...))).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220141 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[InstCombine] Do an about-face on how LLVM canonicalizes (cast (load
Chandler Carruth [Sat, 18 Oct 2014 06:36:22 +0000 (06:36 +0000)]
[InstCombine] Do an about-face on how LLVM canonicalizes (cast (load
...)) and (load (cast ...)): canonicalize toward the former.

Historically, we've tried to load using the type of the *pointer*, and
tried to match that type as closely as possible removing as many pointer
casts as we could and trading them for bitcasts of the loaded value.
This is deeply and fundamentally wrong.

Repeat after me: memory does not have a type! This was a hard lesson for
me to learn working on SROA.

There is only one thing that should actually drive the type used for
a pointer, and that is the type which we need to use to load from that
pointer. Matching up pointer types to the loaded value types is very
useful because it minimizes the physical size of the IR required for
no-op casts. Similarly, the only thing that should drive the type used
for a loaded value is *how that value is used*! Again, this minimizes
casts. And in fact, the *only* thing motivating types in any part of
LLVM's IR are the types used by the operations in the IR. We should
match them as closely as possible.

I've ended up removing some tests here as they were testing bugs or
behavior that is no longer present. Mostly though, this is just cleanup
to let the tests continue to function as intended.

The only fallout I've found so far from this change was SROA and I have
fixed it to not be impeded by the different type of load. If you find
more places where this change causes optimizations not to fire, those
too are likely bugs where we are assuming that the type of pointers is
"significant" for optimization purposes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220138 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove a test that was ported from the old llvm-gcc frontend test suite.
Chandler Carruth [Sat, 18 Oct 2014 06:36:18 +0000 (06:36 +0000)]
Remove a test that was ported from the old llvm-gcc frontend test suite.

This test is pretty awesome. It is claiming to test devirtualization.
However, the code in question is not in fact devirtualized by LLVM. If
you take the original C++ test case and run it through Clang at -O3 we
fail to devirtualize it completely. It also isn't a sufficiently focused
test case.

The *reason* we fail to devirtualize it isn't because of any missing
instcombine though. Instead, it is because we fail to emit an available
externally vtable and thus the vtable is just an external and completely
opaque. If I cause the vtable to be emitted, we successfully
devirtualize things.

Anyways, I'm just removing it because it is providing negative value at
this point: it isn't representative of the output of Clang really, LLVM
isn't doing the transform it claims to be testing, LLVM's failure to do
the transform isn't actually an LLVM bug at all and we shouldn't be
testing for it here, and finally the test is written in such a way that
it will trivially pass even when the point of the test is failing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220137 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[llvm-objdump] don't test timestamp dump as that is time zone dependent
Nick Kledzik [Sat, 18 Oct 2014 02:28:01 +0000 (02:28 +0000)]
[llvm-objdump] don't test timestamp dump as that is time zone dependent

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220123 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[llvm-objdump] enhance test case for mach-o -private-headers
Nick Kledzik [Sat, 18 Oct 2014 01:50:55 +0000 (01:50 +0000)]
[llvm-objdump] enhance test case for mach-o -private-headers

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220120 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[llvm-objdump] Fix mach-o binding decompression error
Nick Kledzik [Sat, 18 Oct 2014 01:21:02 +0000 (01:21 +0000)]
[llvm-objdump] Fix mach-o binding decompression error

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220119 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[SROA] Change how SROA does vector-based promotion of allocas to handle
Chandler Carruth [Sat, 18 Oct 2014 00:44:02 +0000 (00:44 +0000)]
[SROA] Change how SROA does vector-based promotion of allocas to handle
cases where the alloca type, the load types, and the store types used
all disagree.

Previously, the only way that vector-based promotion occured was if the
alloca type was a vector type. This was one of the *very* few remaining
uses of the alloca's type to guide SROA/mem2reg left in LLVM. It turns
out it was a bad idea.

The alloca type can change very easily based on the mixture of types
loaded and stored to that alloca. We shouldn't be relying on it as
a signal for very much. Instead, the source of truth should be loads and
stores. We should canonicalize the loads and stores as much as possible
and then rely on them exclusively in SROA.

When looking and loads and stores, we may find many different candidate
vector types. This change will let SROA try all of them to find a vector
type which is a viable way to promote the entire alloca to a vector
register.

With this change, it becomes possible to do better canonicalization and
optimization of loads and stores without breaking SROA in random ways,
and that should allow fixing a core source of performance loss in hot
numerical loops such as those in Eigen.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220116 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Add global atomicrmw xchg
Aaron Watry [Fri, 17 Oct 2014 23:33:03 +0000 (23:33 +0000)]
R600/SI: Add global atomicrmw xchg

v2: Add separate offset/no-offset tests

Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220110 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Add global atomicrmw xor
Aaron Watry [Fri, 17 Oct 2014 23:33:01 +0000 (23:33 +0000)]
R600/SI: Add global atomicrmw xor

v2: Add separate offset/no-offset tests

Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220109 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Add global atomicrmw or
Aaron Watry [Fri, 17 Oct 2014 23:32:59 +0000 (23:32 +0000)]
R600/SI: Add global atomicrmw or

v2: Add separate offset/no-offset tests

Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220108 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Add global atomicrmw min/umin
Aaron Watry [Fri, 17 Oct 2014 23:32:57 +0000 (23:32 +0000)]
R600/SI: Add global atomicrmw min/umin

v2: Add separate offset/no-offset tests

Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220107 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Add global atomicrmw max/umax
Aaron Watry [Fri, 17 Oct 2014 23:32:56 +0000 (23:32 +0000)]
R600/SI: Add global atomicrmw max/umax

v2: Add separate offset/no-offset tests

Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220106 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Add global atomicrmw and
Aaron Watry [Fri, 17 Oct 2014 23:32:54 +0000 (23:32 +0000)]
R600/SI: Add global atomicrmw and

v2: Add separate offset/no-offset tests

Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220105 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Add global atomicrmw sub
Aaron Watry [Fri, 17 Oct 2014 23:32:52 +0000 (23:32 +0000)]
R600/SI: Add global atomicrmw sub

v2: Add separate offset/no-offset tests

Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220104 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI: Fix/add tests for atomicrmw add
Aaron Watry [Fri, 17 Oct 2014 23:32:50 +0000 (23:32 +0000)]
R600/SI: Fix/add tests for atomicrmw add

The previous tests claimed to test constant offsets in the function name,
but the tests weren't actually testing them.

Clone the tests, and do testing of all combinations of the following:
1) with/without constant pointer offset
2) 32/64-bit addressing modes
3) Usage and non-usage of the return value from the atomicrmw

Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220103 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Rename atomic_load global tests to atomic_add
Aaron Watry [Fri, 17 Oct 2014 23:32:49 +0000 (23:32 +0000)]
R600: Rename atomic_load global tests to atomic_add

The function name now matches what it's actually testing.

Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220102 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[msan] Fix handling of byval arguments with large alignment.
Evgeniy Stepanov [Fri, 17 Oct 2014 23:29:44 +0000 (23:29 +0000)]
[msan] Fix handling of byval arguments with large alignment.

MSan param-tls slots are 8-byte aligned. This change clips
alignment of memcpy into param-tls to 8.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@220101 91177308-0d34-0410-b5e6-96231b3b80d8