OSDN Git Service

android-x86/external-llvm.git
7 years agoLiveRegUnits: Add accumulateBackward() function
Matthias Braun [Sat, 21 Jan 2017 02:21:04 +0000 (02:21 +0000)]
LiveRegUnits: Add accumulateBackward() function

Re-Commit r292543 with a fix for the situation when the chain end is
MBB.end().

This function can be used to accumulate the set of all read and modified
register in a sequence of instructions.

Use this code in AArch64A57FPLoadBalancing::scavengeRegister() to prove
the concept.

- The AArch64A57LoadBalancing code is using a backwards analysis now
  which is irrespective of kill flags. This is the main motivation for
  this change.

Differential Revision: http://reviews.llvm.org/D22082

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292705 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoFix Grammar. NFCI.
Xin Tong [Sat, 21 Jan 2017 02:11:40 +0000 (02:11 +0000)]
Fix Grammar. NFCI.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292704 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[SLP] Make ReductionOpcode have the right (enum) type. NFC.
Michael Kuperstein [Sat, 21 Jan 2017 02:03:03 +0000 (02:03 +0000)]
[SLP] Make ReductionOpcode have the right (enum) type. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292703 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoMergeFunctions: Preserve debug info in thunks, under option -mergefunc-preserve-debug...
Anmol P. Paralkar [Sat, 21 Jan 2017 02:02:56 +0000 (02:02 +0000)]
MergeFunctions: Preserve debug info in thunks, under option -mergefunc-preserve-debug-info

Summary:
Under option -mergefunc-preserve-debug-info we:
- Do not create a new function for a thunk.
- Retain the debug info for a thunk's parameters (and associated
  instructions for the debug info) from the entry block.
  Note: -debug will display the algorithm at work.
- Create debug-info for the call (to the shared implementation) made by
  a thunk and its return value.
- Erase the rest of the function, retaining the (minimally sized) entry
  block to create a thunk.
- Preserve a thunk's call site to point to the thunk even when both occur
  within the same translation unit, to aid debugability. Note that this
  behaviour differs from the underlying -mergefunc implementation which
  modifies the thunk's call site to point to the shared implementation
  when both occur within the same translation unit.

Reviewers: echristo, eeckstein, dblaikie, aprantl, friss

Reviewed By: aprantl

Subscribers: davide, fhahn, jfb, mehdi_amini, llvm-commits

Differential Revision: https://reviews.llvm.org/D28075

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292702 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ConstantFold] Remove test checking that we don't constant-fold sqrt(-2).
Justin Lebar [Sat, 21 Jan 2017 02:02:27 +0000 (02:02 +0000)]
[ConstantFold] Remove test checking that we don't constant-fold sqrt(-2).

This depended on libm's errno behavior (we constant fold iff libm's
sqrt(-2) does not set errno) and was breaking on mac.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292701 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoLowerTypeTests: Fix use-after-free. Found by asan/msan.
Peter Collingbourne [Sat, 21 Jan 2017 01:57:44 +0000 (01:57 +0000)]
LowerTypeTests: Fix use-after-free. Found by asan/msan.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292700 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AMDGPU] Fix build broken in r292688.
Eugene Zelenko [Sat, 21 Jan 2017 01:34:25 +0000 (01:34 +0000)]
[AMDGPU] Fix build broken in r292688.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292699 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[SLP] Delete useless helper. NFC.
Michael Kuperstein [Sat, 21 Jan 2017 01:33:25 +0000 (01:33 +0000)]
[SLP] Delete useless helper. NFC.

The helper contained a branch for a special case that is unnecessary,
and a cast.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292698 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[libFuzzer] Use CXX to set compiler to use
Vitaly Buka [Sat, 21 Jan 2017 01:21:56 +0000 (01:21 +0000)]
[libFuzzer] Use CXX to set compiler to use

Reviewers: kcc

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D28976

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292697 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[libFuzzer] fix gcc build
Kostya Serebryany [Sat, 21 Jan 2017 01:08:22 +0000 (01:08 +0000)]
[libFuzzer] fix gcc build

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292695 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[NVPTX] Auto-upgrade some NVPTX intrinsics to LLVM target-generic code.
Justin Lebar [Sat, 21 Jan 2017 01:00:32 +0000 (01:00 +0000)]
[NVPTX] Auto-upgrade some NVPTX intrinsics to LLVM target-generic code.

Summary:
Specifically, we upgrade llvm.nvvm.:

 * brev{32,64}
 * clz.{i,ll}
 * popc.{i,ll}
 * abs.{i,ll}
 * {min,max}.{i,ll,u,ull}
 * h2f

These either map directly to an existing LLVM target-generic
intrinsic or map to a simple LLVM target-generic idiom.

In all cases, we check that the code we generate is lowered to PTX as we
expect.

These builtins don't need to be backfilled in clang: They're not
accessible to user code from nvcc.

Reviewers: tra

Subscribers: majnemer, cfe-commits, llvm-commits, jholewinski

Differential Revision: https://reviews.llvm.org/D28793

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292694 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[NVPTX] Move getDivF32Level, usePrecSqrtF32, and useF32FTZ into out of DAGToDAG and...
Justin Lebar [Sat, 21 Jan 2017 01:00:14 +0000 (01:00 +0000)]
[NVPTX] Move getDivF32Level, usePrecSqrtF32, and useF32FTZ into out of DAGToDAG and into TargetLowering.

Summary:
DADToDAG has access to TargetLowering, but not vice versa, so this is
the more general location for these functions.

NFC

Reviewers: tra

Subscribers: jholewinski, llvm-commits

Differential Revision: https://reviews.llvm.org/D28795

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292693 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ConstantFolding] Constant-fold llvm.sqrt(x) like other intrinsics.
Justin Lebar [Sat, 21 Jan 2017 00:59:57 +0000 (00:59 +0000)]
[ConstantFolding] Constant-fold llvm.sqrt(x) like other intrinsics.

Summary:
Currently we return undef, but we're in the process of changing the
LangRef so that llvm.sqrt behaves like the other math intrinsics,
matching the return value of the standard libcall but not setting errno.

This change is legal even without the LangRef change because currently
calling llvm.sqrt(x) where x is negative is spec'ed to be UB.  But in
practice it's also safe because we're simply constant-folding fewer
inputs: Inputs >= -0 get constant-folded as before, but inputs < -0 now
aren't constant-folded, because ConstantFoldFP aborts if the host math
function raises an fp exception.

Reviewers: hfinkel, efriedma, sanjoy

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D28929

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292692 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ValueTracking] Clarify comments on CannotBeOrderedLessThanZero and SignBitMustBeZero.
Justin Lebar [Sat, 21 Jan 2017 00:59:40 +0000 (00:59 +0000)]
[ValueTracking] Clarify comments on CannotBeOrderedLessThanZero and SignBitMustBeZero.

Reviewers: hfinkel, efriedma, sanjoy

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D28926

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292691 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[NVPTX] Add explicit check for llvm.sqrt.f32 to intrinsics.ll.
Justin Lebar [Sat, 21 Jan 2017 00:59:23 +0000 (00:59 +0000)]
[NVPTX] Add explicit check for llvm.sqrt.f32 to intrinsics.ll.

Test-only change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292690 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AMDGPU] Fix some Clang-tidy modernize and Include What You Use warnings; other minor...
Eugene Zelenko [Sat, 21 Jan 2017 00:53:49 +0000 (00:53 +0000)]
[AMDGPU] Fix some Clang-tidy modernize and Include What You Use warnings; other minor fixes (NFC).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292688 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[APFloat] Add asserts and descriptions. NFC.
Tim Shen [Sat, 21 Jan 2017 00:37:53 +0000 (00:37 +0000)]
[APFloat] Add asserts and descriptions. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292687 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[libFuzzer] use print+exit(1) instead of assert to report an error
Kostya Serebryany [Sat, 21 Jan 2017 00:13:50 +0000 (00:13 +0000)]
[libFuzzer] use print+exit(1) instead of assert to report an error

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292685 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[libFuzzer] re-enable LLVMFuzzer-RepeatedMemcmp test, cleanup the test runner nearby
Kostya Serebryany [Sat, 21 Jan 2017 00:01:27 +0000 (00:01 +0000)]
[libFuzzer] re-enable LLVMFuzzer-RepeatedMemcmp test, cleanup the test runner nearby

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292683 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[InstCombine] auto-generate checks; NFC
Sanjay Patel [Fri, 20 Jan 2017 23:39:01 +0000 (23:39 +0000)]
[InstCombine] auto-generate checks; NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292682 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[libFuzzer] call __sanitizer_dump_coverage via EF
Kostya Serebryany [Fri, 20 Jan 2017 23:35:29 +0000 (23:35 +0000)]
[libFuzzer] call __sanitizer_dump_coverage via  EF

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292681 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[PPC] Give unaligned memory access lower cost on processor that supports it
Guozhi Wei [Fri, 20 Jan 2017 23:35:27 +0000 (23:35 +0000)]
[PPC] Give unaligned memory access lower cost on processor that supports it

Newer ppc supports unaligned memory access, it reduces the cost of unaligned memory access significantly. This patch handles this case in PPCTTIImpl::getMemoryOpCost.

This patch fixes pr31492.

Differential Revision: https://reviews.llvm.org/D28630

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292680 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ThinLTO] Don't perform computeDeadSymbols during O0 link, as the result is never...
Mehdi Amini [Fri, 20 Jan 2017 23:34:12 +0000 (23:34 +0000)]
[ThinLTO] Don't perform computeDeadSymbols during O0 link, as the result is never used (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292679 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[NewGVN] Optimize processing for instructions found trivially dead.
Davide Italiano [Fri, 20 Jan 2017 23:29:28 +0000 (23:29 +0000)]
[NewGVN] Optimize processing for instructions found trivially dead.

Don't call `isTriviallyDeadInstructions()` once we discover that
an instruction is dead. Instead, set DFS number zero (as suggested
by Danny) and forget about it (this also speeds up things as we
won't try to reprocess that block).

Differential Revision:  https://reviews.llvm.org/D28930

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292676 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoGlobalISel: prevent heap use-after-free when looking up VReg.
Tim Northover [Fri, 20 Jan 2017 23:25:17 +0000 (23:25 +0000)]
GlobalISel: prevent heap use-after-free when looking up VReg.

Translating the constant can create more VRegs, which can invalidate the
reference into the DenseMap. So we have to look up the value again after all
that's happened.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292675 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoLowerTypeTests: Simplify; always create SizeM1 with type IntPtrTy, move initializatio...
Peter Collingbourne [Fri, 20 Jan 2017 23:22:28 +0000 (23:22 +0000)]
LowerTypeTests: Simplify; always create SizeM1 with type IntPtrTy, move initialization out of if statement.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292674 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAdd indirect call promotion to SamplePGO
Dehao Chen [Fri, 20 Jan 2017 22:56:07 +0000 (22:56 +0000)]
Add indirect call promotion to SamplePGO

Summary: This patch adds metadata for indirect call promotion in the sample profile loader.

Reviewers: xur, davidxl, dnovillo

Reviewed By: davidxl

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D28923

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292672 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[libFuzzer] Don't use `#ifdef` for defined macros, instead use `#if`.
Marcos Pividori [Fri, 20 Jan 2017 22:49:13 +0000 (22:49 +0000)]
[libFuzzer] Don't use `#ifdef` for defined macros, instead use `#if`.

Differential Revision: https://reviews.llvm.org/D28972

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292670 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[libFuzzer] Use clang as linker on Windows, to properly include sanitizer libraries.
Marcos Pividori [Fri, 20 Jan 2017 22:49:08 +0000 (22:49 +0000)]
[libFuzzer] Use clang as linker on Windows, to properly include sanitizer libraries.

In order to use sanitizers on Windows, we need to link against many runtime
libraries which will depend on the target being created (executable or dll) and
the c runtime library used (MT/MD).
By default, cmake uses link.exe for linking, which fails because we don't
specify the appropiate dependencies. As we don't want to consider all of that
possible situations which depends on the implementation of the compiler-rt, the
simplest option is to change the rules for linking executables and shared
libraries, using the compiler instead of link.exe.
Clang driver will consider the sanitizer flags, and automatically provide the
required libraries to the linker.

Differential Revision: https://reviews.llvm.org/D27869

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292669 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[libFuzzer] Properly use compiler options supported on Windows.
Marcos Pividori [Fri, 20 Jan 2017 22:48:47 +0000 (22:48 +0000)]
[libFuzzer] Properly use compiler options supported on Windows.

Replace "-g" by "-gline-tables-only". "-g" is not supported by clang-cl.

Differential Revision: https://reviews.llvm.org/D27868

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292668 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ThinLTO] The "codegen only" path didn't honor the recently added file-based API
Mehdi Amini [Fri, 20 Jan 2017 22:45:34 +0000 (22:45 +0000)]
[ThinLTO] The "codegen only" path didn't honor the recently added file-based API

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292667 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoImprove PGO support for the new inliner
Easwaran Raman [Fri, 20 Jan 2017 22:44:04 +0000 (22:44 +0000)]
Improve PGO support for the new inliner

This adds the following to the new PM based inliner in PGO mode:

* Use block frequency analysis to derive callsite's profile count and use
that to adjust thresholds of hot and cold callsites.

* Incrementally update the BFI of the caller after a callee gets inlined
into it. This incremental update is only within an invocation of the run
method - BFI is not preserved across calls to run.
Update the function entry count of the callee after inlining it into a
caller.

* I've tuned the thresholds for the hot and cold callsites using a hacked
up version of the old inliner that explicitly computes BFI on a set of
internal benchmarks and spec. Once the new PM based pipeline stabilizes
(IIRC Chandler mentioned there are known issues) I'll benchmark this
again and adjust the thresholds if required.
Inliner PGO support.

Differential revision: https://reviews.llvm.org/D28331

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292666 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[pdb] Write the Named Stream mapping to Yaml and binary.
Zachary Turner [Fri, 20 Jan 2017 22:42:09 +0000 (22:42 +0000)]
[pdb] Write the Named Stream mapping to Yaml and binary.

Differential Revision: https://reviews.llvm.org/D28919

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292665 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[pdb] Merge NamedStreamMapBuilder and NamedStreamMap.
Zachary Turner [Fri, 20 Jan 2017 22:41:40 +0000 (22:41 +0000)]
[pdb] Merge NamedStreamMapBuilder and NamedStreamMap.

While the builder pattern has proven useful for certain other
larger types, in this case it was hampering the ability to use
the data structure, as for runtime access we need a map that
we can efficiently read from and write to.  So the two are merged
into a single data structure that can efficiently be read to,
written from, deserialized from bytes, and serialized to bytes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292664 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[PDB] Rename some files to be more intuitive.
Zachary Turner [Fri, 20 Jan 2017 22:41:15 +0000 (22:41 +0000)]
[PDB] Rename some files to be more intuitive.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292663 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoIPO, LTO: Plumb the summary from the LTO API into the pass manager.
Peter Collingbourne [Fri, 20 Jan 2017 22:18:52 +0000 (22:18 +0000)]
IPO, LTO: Plumb the summary from the LTO API into the pass manager.

Differential Revision: https://reviews.llvm.org/D28840

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292661 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ValueTracking] recognize variations of 'clamp' to improve codegen (PR31693)
Sanjay Patel [Fri, 20 Jan 2017 22:18:47 +0000 (22:18 +0000)]
[ValueTracking] recognize variations of 'clamp' to improve codegen (PR31693)

By enhancing value tracking, we allow an existing min/max canonicalization to
kick in and improve codegen for several targets that have min/max instructions.

Unfortunately, recognizing min/max in value tracking may cause us to hit
a hack in InstCombiner::visitICmpInst() more often:
http://lists.llvm.org/pipermail/llvm-dev/2017-January/109340.html
...but I'm hoping we can remove that soon.

Correctness proofs based on Alive:

Name: smaxmin
Pre: C1 < C2
%cmp2 = icmp slt i8 %x, C2
%min = select i1 %cmp2, i8 %x, i8 C2
%cmp3 = icmp slt i8 %x, C1
%r = select i1 %cmp3, i8 C1, i8 %min
=>
%cmp2 = icmp slt i8 %x, C2
%min = select i1 %cmp2, i8 %x, i8 C2
%cmp1 = icmp sgt i8 %min, C1
%r = select i1 %cmp1, i8 %min, i8 C1

Name: sminmax
Pre: C1 > C2
%cmp2 = icmp sgt i8 %x, C2
%max = select i1 %cmp2, i8 %x, i8 C2
%cmp3 = icmp sgt i8 %x, C1
%r = select i1 %cmp3, i8 C1, i8 %max
=>
%cmp2 = icmp sgt i8 %x, C2
%max = select i1 %cmp2, i8 %x, i8 C2
%cmp1 = icmp slt i8 %max, C1
%r = select i1 %cmp1, i8 %max, i8 C1

----------------------------------------
Optimization: smaxmin
Done: 1
Optimization is correct!
----------------------------------------
Optimization: sminmax
Done: 1
Optimization is correct!

Name: umaxmin
Pre: C1 u< C2
%cmp2 = icmp ult i8 %x, C2
%min = select i1 %cmp2, i8 %x, i8 C2
%cmp3 = icmp ult i8 %x, C1
%r = select i1 %cmp3, i8 C1, i8 %min
=>
%cmp2 = icmp ult i8 %x, C2
%min = select i1 %cmp2, i8 %x, i8 C2
%cmp1 = icmp ugt i8 %min, C1
%r = select i1 %cmp1, i8 %min, i8 C1

Name: uminmax
Pre: C1 u> C2
%cmp2 = icmp ugt i8 %x, C2
%max = select i1 %cmp2, i8 %x, i8 C2
%cmp3 = icmp ugt i8 %x, C1
%r = select i1 %cmp3, i8 C1, i8 %max
=>
%cmp2 = icmp ugt i8 %x, C2
%max = select i1 %cmp2, i8 %x, i8 C2
%cmp1 = icmp ult i8 %max, C1
%r = select i1 %cmp1, i8 %max, i8 C1

----------------------------------------
Optimization: umaxmin
Done: 1
Optimization is correct!
----------------------------------------
Optimization: uminmax
Done: 1
Optimization is correct!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292660 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agodocs: Document that !absolute_symbol { all-ones, all-ones } means the full set.
Peter Collingbourne [Fri, 20 Jan 2017 21:56:37 +0000 (21:56 +0000)]
docs: Document that !absolute_symbol { all-ones, all-ones } means the full set.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292657 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ThinLTO] Drop non-prevailing non-ODR weak to declarations
Teresa Johnson [Fri, 20 Jan 2017 21:54:58 +0000 (21:54 +0000)]
[ThinLTO] Drop non-prevailing non-ODR weak to declarations

Summary:
Allow non-ODR weak/linkonce non-prevailing copies to be marked
as available_externally in the index. Add support for dropping these to
declarations in the backend.

Reviewers: mehdi_amini, pcc

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D28806

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292656 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[InstCombine] add tests to show missed canonicalization of min/max; NFC
Sanjay Patel [Fri, 20 Jan 2017 21:49:41 +0000 (21:49 +0000)]
[InstCombine] add tests to show missed canonicalization of min/max; NFC

Unfortunately, recognizing these in value tracking may cause us to hit
a hack in InstCombiner::visitICmpInst() more often:
http://lists.llvm.org/pipermail/llvm-dev/2017-January/109340.html

...but besides being the obviously Right Thing To Do, there's a clear
codegen win from identifying these patterns for several targets.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292655 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoLowerTypeTests: Implement importing of type identifiers.
Peter Collingbourne [Fri, 20 Jan 2017 21:49:34 +0000 (21:49 +0000)]
LowerTypeTests: Implement importing of type identifiers.

To import a type identifier we read the summary and create external
references to the symbols defined when exporting.

Differential Revision: https://reviews.llvm.org/D28546

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292654 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[globalisel] Fix an unused variable warning when NDEBUG is defined.
Daniel Sanders [Fri, 20 Jan 2017 21:40:05 +0000 (21:40 +0000)]
[globalisel] Fix an unused variable warning when NDEBUG is defined.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292653 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[libFuzzer] add an assert to protect against LLVMFuzzerInitialize changing argv[0]
Kostya Serebryany [Fri, 20 Jan 2017 21:34:24 +0000 (21:34 +0000)]
[libFuzzer] add an assert to protect against  LLVMFuzzerInitialize changing argv[0]

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292652 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU/R600: Serialize vector trunc stores to private AS
Jan Vesely [Fri, 20 Jan 2017 21:24:26 +0000 (21:24 +0000)]
AMDGPU/R600: Serialize vector trunc stores to private AS

Add DUMMY_CHAIN SDNode to denote stores of interest

Bugzilla: https://llvm.org/bugs/show_bug.cgi?id=28915
Bugzilla: https://llvm.org/bugs/show_bug.cgi?id=30411

Differential Revision: https://reviews.llvm.org/D27964

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292651 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoNewGVN: Remove pr31686.ll, it is tested by pr31594.ll, which is much smaller and...
Daniel Berlin [Fri, 20 Jan 2017 21:04:58 +0000 (21:04 +0000)]
NewGVN: Remove pr31686.ll, it is tested by pr31594.ll, which is much smaller and simpler

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292649 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoNewGVN: Fix PR 31686 and PR 31698 by rewriting store leader handling.
Daniel Berlin [Fri, 20 Jan 2017 21:04:30 +0000 (21:04 +0000)]
NewGVN: Fix PR 31686 and PR 31698 by rewriting store leader handling.

Summary:

This rewrites store expression/leader handling.  We no longer use the
value operand as the leader, instead, we store it separately.  We also
now store the stored value as part of the expression, and compare it
when comparing stores for equality.  This enables us to get rid of a
bunch of our previous hacks and machinations, as the existing
machinery takes care of everything *except* updating the stored value
on classes.  The only time we have to update it is if the storecount
goes to 0, and when we do, we destroy it.

Since we no longer use the value operand as the leader, during elimination, we have to use the value operand.  Doing this also fixes a bunch of store forwarding cases we were missing.

Any value operand we use is guaranteed to either be updated by previous eliminations, or minimized by future ones.

(IE the fact that we don't use the most dominating value operand when it's not a constant does not affect anything).

Sadly, this change also exposes that we didn't pay attention to the
output of the pr31594.ll test, as it also very clearly exposes the
same store leader bug we are fixing here.

(I added pr31682.ll anyway, but maybe we think that's too large to be useful)

On the plus side, propagate-ir-flags.ll now passes due to the
corrected store forwarding.

This change was 3 stage'd on darwin and linux, with the full test-suite.

Reviewers:
davide
Subscribers:
llvm-commits

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292648 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoLowerTypeTests: Compute SizeM1BitWidth in exportTypeId. NFCI.
Peter Collingbourne [Fri, 20 Jan 2017 20:57:40 +0000 (20:57 +0000)]
LowerTypeTests: Compute SizeM1BitWidth in exportTypeId. NFCI.

This avoids needing to store it in a separate field in TypeIdLowering.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292647 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[libFuzzer] experimental support for 'equivalance fuzzing'
Kostya Serebryany [Fri, 20 Jan 2017 20:57:07 +0000 (20:57 +0000)]
[libFuzzer] experimental support for 'equivalance fuzzing'

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292646 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[WebAssembly] Don't create bitcast-wrappers for varargs.
Dan Gohman [Fri, 20 Jan 2017 20:50:29 +0000 (20:50 +0000)]
[WebAssembly] Don't create bitcast-wrappers for varargs.

WebAssembly varargs functions use a significantly different ABI than
non-varargs functions, and the current code in
WebAssemblyFixFunctionBitcasts doesn't handle that difference. For now,
just avoid creating wrapper functions in the presence of varargs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292645 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ThinLTO] Fix lazy-loading of MDString instruction attachments
Mehdi Amini [Fri, 20 Jan 2017 20:29:16 +0000 (20:29 +0000)]
[ThinLTO] Fix lazy-loading of MDString instruction attachments

CFI is using intrinsics that takes MDString as arguments, and this
was broken during lazy-loading of metadata.

Differential Revision: https://reviews.llvm.org/D28916

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292641 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[x86] add tests to show missed min/max vector codegen (PR31693)
Sanjay Patel [Fri, 20 Jan 2017 20:14:11 +0000 (20:14 +0000)]
[x86] add tests to show missed min/max vector codegen (PR31693)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292640 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[DWARF] [ObjectYAML] Adding APIs for unittesting
Chris Bieneman [Fri, 20 Jan 2017 19:03:14 +0000 (19:03 +0000)]
[DWARF] [ObjectYAML] Adding APIs for unittesting

Summary: This patch adds some new APIs to enable using the YAML DWARF representation in unit tests. The most basic new API is DWARFYAML::EmitDebugSections which converts a YAML string into a series of owned MemoryBuffer objects stored in a StringMap. The string map can then be used to construct a DWARFContext for parsing in place of an ObjectFile.

Reviewers: dblaikie, clayborg

Subscribers: mgorny, fhahn, jgosnell, aprantl, llvm-commits

Differential Revision: https://reviews.llvm.org/D28828

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292634 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRecommit "[InlineCost] Use TTI to check if GEP is free." #3
Haicheng Wu [Fri, 20 Jan 2017 18:51:22 +0000 (18:51 +0000)]
Recommit "[InlineCost] Use TTI to check if GEP is free." #3

This is the third attemp to recommit r292526.

The original summary:

Currently, a GEP is considered free only if its indices are all constant.
TTI::getGEPCost() can give target-specific more accurate analysis. TTI is
already used for the cost of many other instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292633 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[SLP] Initial test for fix of PR31690.
Alexey Bataev [Fri, 20 Jan 2017 18:40:21 +0000 (18:40 +0000)]
[SLP] Initial test for fix of PR31690.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292631 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoARM: tLDR_postidx should be marked mayLoad
Matthias Braun [Fri, 20 Jan 2017 18:30:28 +0000 (18:30 +0000)]
ARM: tLDR_postidx should be marked mayLoad

This fixes -verify-machineinstrs complaints.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292629 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[InstCombine][X86] Add MULDQ/MULUDQ undef handling
Simon Pilgrim [Fri, 20 Jan 2017 18:20:30 +0000 (18:20 +0000)]
[InstCombine][X86] Add MULDQ/MULUDQ undef handling

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292627 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[SLP] A new test for horizontal vectorization for non-power-of-2
Alexey Bataev [Fri, 20 Jan 2017 18:04:29 +0000 (18:04 +0000)]
[SLP] A new test for horizontal vectorization for non-power-of-2
instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292626 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAArch64LoadStoreOptimizer: Update kill flags when merging stores
Matthias Braun [Fri, 20 Jan 2017 18:04:27 +0000 (18:04 +0000)]
AArch64LoadStoreOptimizer: Update kill flags when merging stores

Kill flags need to be updated correctly when moving stores up/down to
form store pair instructions.
Those invalid flags have been ignored before but as of r290014 they are
recognized when using -mllvm -verify-machineinstrs.

Also simplifies test/CodeGen/AArch64/ldst-opt-dbg-limit.mir, renames it
to ldst-opt.mir test and adds a new tests for this change.

Differential Revision: https://reviews.llvm.org/D28875

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292625 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[mips] Fix debug information for __thread variable
Petar Jovanovic [Fri, 20 Jan 2017 17:53:30 +0000 (17:53 +0000)]
[mips] Fix debug information for __thread variable

This patch fixes debug information for __thread variable on Mips
using .dtprelword and .dtpreldword directives.

Patch by Aleksandar Beserminji.

Differential Revision: http://reviews.llvm.org/D28770

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292624 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AMDGPU] Fix some Clang-tidy modernize and Include What You Use warnings; other minor...
Eugene Zelenko [Fri, 20 Jan 2017 17:52:16 +0000 (17:52 +0000)]
[AMDGPU] Fix some Clang-tidy modernize and Include What You Use warnings; other minor fixes (NFC).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292623 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[RegisterCoalescing] Recommit the patch "Remove partial redundent copy".
Wei Mi [Fri, 20 Jan 2017 17:38:54 +0000 (17:38 +0000)]
[RegisterCoalescing] Recommit the patch "Remove partial redundent copy".

The recommit fixes a bug related with live interval update after the partial
redundent copy is moved.

The original patch is to solve the performance problem described in PR27827.
Register coalescing sometimes cannot remove a copy because of interference.
But if we can find a reverse copy in one of the predecessor block of the copy,
the copy is partially redundent and we may remove the copy partially by moving
it to the predecessor block without the reverse copy.

Differential Revision: https://reviews.llvm.org/D28585

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292621 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[InstCombine][SSE] Tests showing missed opportunities to handle muldq/muludq with...
Simon Pilgrim [Fri, 20 Jan 2017 17:06:38 +0000 (17:06 +0000)]
[InstCombine][SSE] Tests showing missed opportunities to handle muldq/muludq with undef arguments

Fixed a typo in existing test names at the same time

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292619 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "Recommit "[InlineCost] Use TTI to check if GEP is free." #2"
Haicheng Wu [Fri, 20 Jan 2017 16:52:22 +0000 (16:52 +0000)]
Revert "Recommit "[InlineCost] Use TTI to check if GEP is free." #2"

This reverts commit r292616 because the test case still has problem.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292618 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRecommit "[InlineCost] Use TTI to check if GEP is free." #2
Haicheng Wu [Fri, 20 Jan 2017 16:36:34 +0000 (16:36 +0000)]
Recommit "[InlineCost] Use TTI to check if GEP is free." #2

This is the second attemp to recommit r292526.

The original summary:

Currently, a GEP is considered free only if its indices are all constant.
TTI::getGEPCost() can give target-specific more accurate analysis. TTI is
already used for the cost of many other instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292616 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRemove trailing whitespace. NFCI.
Simon Pilgrim [Fri, 20 Jan 2017 15:15:59 +0000 (15:15 +0000)]
Remove trailing whitespace. NFCI.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292613 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[CostModel][X86] Removed unused cost. NFCI.
Simon Pilgrim [Fri, 20 Jan 2017 15:14:38 +0000 (15:14 +0000)]
[CostModel][X86] Removed unused cost. NFCI.

SHL v8i32 is already handled in the SSE41 cost table

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292612 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[InstCombine][SSE] Tests showing missed opportunities to constant fold packss/packus
Simon Pilgrim [Fri, 20 Jan 2017 13:21:30 +0000 (13:21 +0000)]
[InstCombine][SSE] Tests showing missed opportunities to constant fold packss/packus

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292609 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[Thumb] Add support for tMUL in the compare instruction peephole optimizer.
Sjoerd Meijer [Fri, 20 Jan 2017 13:10:12 +0000 (13:10 +0000)]
[Thumb] Add support for tMUL in the compare instruction peephole optimizer.

We also want to optimise tests like this: return a*b == 0.  The MULS
instruction is flag setting, so we don't need the CMP instruction but can
instead branch on the result of the MULS. The generated instructions sequence
for this example was: MULS, MOVS, MOVS, CMP. The MOVS instruction load the
boolean values resulting from the select instruction, but these MOVS
instructions are flag setting and were thus preventing this optimisation. Now
we first reorder and move the MULS to before the CMP and generate sequence
MOVS, MOVS, MULS, CMP so that the optimisation could trigger. Reordering of the
MULS and MOVS is safe to do because the subsequent MOVS instructions just set
the CPSR register and don't use it, i.e. the CPSR is dead.

Differential Revision: https://reviews.llvm.org/D27990

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292608 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[InstCombine][SSE] Tests showing missed opportunities to handle packss/packus with...
Simon Pilgrim [Fri, 20 Jan 2017 11:28:07 +0000 (11:28 +0000)]
[InstCombine][SSE] Tests showing missed opportunities to handle packss/packus with undef arguments

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292601 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoPacify -Wreorder.
Benjamin Kramer [Fri, 20 Jan 2017 10:37:53 +0000 (10:37 +0000)]
Pacify -Wreorder.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292599 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAdd an assertion to PlaceholderQueue destructor, ensuring it has been flushed
Mehdi Amini [Fri, 20 Jan 2017 10:18:32 +0000 (10:18 +0000)]
Add an assertion to PlaceholderQueue destructor, ensuring it has been flushed

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292597 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AMDGPU] Add subtarget features for SDWA/DPP
Sam Kolton [Fri, 20 Jan 2017 10:01:25 +0000 (10:01 +0000)]
[AMDGPU] Add subtarget features for SDWA/DPP

Reviewers: vpykhtin, artem.tamazov, tstellarAMD

Subscribers: arsenm, kzhuravl, wdng, nhaehnle, yaxunl, tony-tye

Differential Revision: https://reviews.llvm.org/D28900

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292596 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[PM] Tidy up the spacing of this new, much nicer test file.
Chandler Carruth [Fri, 20 Jan 2017 09:30:03 +0000 (09:30 +0000)]
[PM] Tidy up the spacing of this new, much nicer test file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292592 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[InstCombine][SSE] Add DemandedElts support for PACKSS/PACKUS instructions
Simon Pilgrim [Fri, 20 Jan 2017 09:28:21 +0000 (09:28 +0000)]
[InstCombine][SSE] Add DemandedElts support for PACKSS/PACKUS instructions

Simplify a packss/packus truncation based on the elements of the mask that are actually demanded.

Differential Revision: https://reviews.llvm.org/D28777

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292591 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[PM] Port LoopSink to the new pass manager.
Chandler Carruth [Fri, 20 Jan 2017 08:42:19 +0000 (08:42 +0000)]
[PM] Port LoopSink to the new pass manager.

Like several other loop passes (the vectorizer, etc) this pass doesn't
really fit the model of a loop pass. The critical distinction is that it
isn't intended to be pipelined together with other loop passes. I plan
to add some documentation to the loop pass manager to make this more
clear on that side.

LoopSink is also different because it doesn't really need a lot of the
infrastructure of our loop passes. For example, if there aren't loop
invariant instructions causing a preheader to exist, there is no need to
form a preheader. It also doesn't need LCSSA because this pass is
only involved in sinking invariant instructions from a preheader into
the loop, not reasoning about live-outs.

This allows some nice simplifications to the pass in the new PM where we
can directly walk the loops once without restructuring them.

Differential Revision: https://reviews.llvm.org/D28921

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292589 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[LoopSink] Trivial comment cleanup.
Chandler Carruth [Fri, 20 Jan 2017 08:42:14 +0000 (08:42 +0000)]
[LoopSink] Trivial comment cleanup.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292588 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ARM] Use helpers for adding pred / CC operands. NFC
Diana Picus [Fri, 20 Jan 2017 08:15:24 +0000 (08:15 +0000)]
[ARM] Use helpers for adding pred / CC operands. NFC

Hunt down some of the places where we use bare addReg(0) or addImm(AL).addReg(0)
and replace with add(condCodeOp()) and add(predOps()). This should make it
easier to understand what those operands represent (without having to look at
the definition of the instruction that we're adding to).

Differential Revision: https://reviews.llvm.org/D27984

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292587 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AVX-512] Fix a couple test cases to not pass an undef mask to gather intrinsic....
Craig Topper [Fri, 20 Jan 2017 07:12:30 +0000 (07:12 +0000)]
[AVX-512] Fix a couple test cases to not pass an undef mask to gather intrinsic. This could break if any future optimizations taken advantage of the undef.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292585 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[TargetLowering] Improve comment for setOperationAction().
Jonas Paulsson [Fri, 20 Jan 2017 06:48:47 +0000 (06:48 +0000)]
[TargetLowering] Improve comment for setOperationAction().

Add a sentence that says that the type argument can refer to
either the type of a result, or that of an operand.

Review: Eli Friedman.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292584 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoNewGVN: Fix PR 31682, an overactive assert.
Daniel Berlin [Fri, 20 Jan 2017 06:38:41 +0000 (06:38 +0000)]
NewGVN: Fix PR 31682, an overactive assert.
Part of the assert has been left active for further debugging.
The other part has been turned into a stat for tracking for the
moment.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292583 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[SLP] Add a base test for jumbled store
Mohammad Shahid [Fri, 20 Jan 2017 06:05:33 +0000 (06:05 +0000)]
[SLP] Add a base test for jumbled store

Change-Id: I905ce08a02c76a6896dcfd9629547417c99adc4a

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292581 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agollvm-cxxfilt: add missing includes from previous change
Saleem Abdulrasool [Fri, 20 Jan 2017 05:33:22 +0000 (05:33 +0000)]
llvm-cxxfilt: add missing includes from previous change

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292580 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agollvm-cxxfilt: fix program description
Saleem Abdulrasool [Fri, 20 Jan 2017 05:27:09 +0000 (05:27 +0000)]
llvm-cxxfilt: fix program description

Fix a silly copy-paste error in the tool description.  Take the
opportunity to add crash stack printing which will hopefully never be
needed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292579 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agollvm-cxxfilt: support `-t` to demangle types
Saleem Abdulrasool [Fri, 20 Jan 2017 04:25:26 +0000 (04:25 +0000)]
llvm-cxxfilt: support `-t` to demangle types

By default c++filt demangles functions, though you can optionally pass
`-t` to have it decode types as well, behaving nearly identical to
`__cxa_demangle`.  Add support for this mode.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292576 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoBitVector: Fix undefined behaviour
Matthias Braun [Fri, 20 Jan 2017 04:23:08 +0000 (04:23 +0000)]
BitVector: Fix undefined behaviour

Calling reset() on an empty BitVector would call memset with a nullptr
argument which is undefined behaviour.

This should fix the sanitizer bot.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292575 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "LiveRegUnits: Add accumulateBackward() function"
Matthias Braun [Fri, 20 Jan 2017 03:58:42 +0000 (03:58 +0000)]
Revert "LiveRegUnits: Add accumulateBackward() function"

This seems to be breaking some bots.

This reverts commit r292543.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292574 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "Demangle: only demangle mangled symbols"
Saleem Abdulrasool [Fri, 20 Jan 2017 03:54:04 +0000 (03:54 +0000)]
Revert "Demangle: only demangle mangled symbols"

This reverts SVN r286795.  This was incorrect the demangler is expected
to be able to demangle types as well as functions.  This makes the
behaviour of itaniumDemangle similar to __cxa_demangle once more.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292573 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "Recommit "[InlineCost] Use TTI to check if GEP is free.""
Haicheng Wu [Fri, 20 Jan 2017 03:40:41 +0000 (03:40 +0000)]
Revert "Recommit "[InlineCost] Use TTI to check if GEP is free.""

This reverts commit r292570.  The test still has problem.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292572 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRecommit "[InlineCost] Use TTI to check if GEP is free."
Haicheng Wu [Fri, 20 Jan 2017 03:09:11 +0000 (03:09 +0000)]
Recommit "[InlineCost] Use TTI to check if GEP is free."

This recommits r292526 which is reverted in r292529 after fixing the test case.

The original summary:

Currently, a GEP is considered free only if its indices are all constant.
TTI::getGEPCost() can give target-specific more accurate analysis. TTI is
already used for the cost of many other instructions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292570 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[LoopInfo] Add helper methods to compute two useful orderings of the
Chandler Carruth [Fri, 20 Jan 2017 02:41:20 +0000 (02:41 +0000)]
[LoopInfo] Add helper methods to compute two useful orderings of the
loops in a function.

These are relatively confusing to talk about and compute correctly so it
seems really good to write down their implementation in one place. I've
replaced one place we needed this in the loop PM infrastructure and
I have another place in a pending patch that wants it.

We can't quite use this for the core loop PM walk because there we're
sometimes working on a sub-forest.

I'll add the expected unittests before committing this but wanted to
make sure folks were happy with these names / comments.

Credit goes to Richard Smith for the idea for naming the order where siblings
are in reverse program order but the tree traversal remains preorder.

Differential Revision: https://reviews.llvm.org/D28932

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292569 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[test] Remove a unwanted match for `XFAIL:`.
Greg Parker [Fri, 20 Jan 2017 02:01:04 +0000 (02:01 +0000)]
[test] Remove a unwanted match for `XFAIL:`.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292567 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AArch64][GlobalISel] Widen scalar int->fp conversions.
Ahmed Bougacha [Fri, 20 Jan 2017 01:37:24 +0000 (01:37 +0000)]
[AArch64][GlobalISel] Widen scalar int->fp conversions.

It's incorrect to ignore the higher bits of the integer source.
Teach the legalizer how to widen it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292563 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[PM] Attempt to pacify windows bots.
Michael Kuperstein [Fri, 20 Jan 2017 00:47:32 +0000 (00:47 +0000)]
[PM] Attempt to pacify windows bots.

Another difference in type pretty-printing, this one windows-specific.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292556 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AMDGPU] Prevent spills before exec mask is restored
Stanislav Mekhanoshin [Fri, 20 Jan 2017 00:44:31 +0000 (00:44 +0000)]
[AMDGPU] Prevent spills before exec mask is restored

Inline spiller can decide to move a spill as early as possible in the basic block.
It will skip phis and label, but we also need to make sure it skips instructions
in the basic block prologue which restore exec mask.

Added isPositionLike callback in TargetInstrInfo to detect instructions which
shall be skipped in addition to common phis, labels etc.

Differential Revision: https://reviews.llvm.org/D27997

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292554 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoGlobalISel: Add a note about how we're being a bit loose with memory operands
Justin Bogner [Fri, 20 Jan 2017 00:30:17 +0000 (00:30 +0000)]
GlobalISel: Add a note about how we're being a bit loose with memory operands

The logic in r292461 is conservatively correct, but we should revisit
this later. Add a TODO so we don't forget.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292553 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AArch64][GlobalISel] Split FP conversion legalizer tests. NFC.
Ahmed Bougacha [Fri, 20 Jan 2017 00:30:09 +0000 (00:30 +0000)]
[AArch64][GlobalISel] Split FP conversion legalizer tests. NFC.

Big functions with large vreg # are quite unwieldy to update.

Change it to have one function per test (it does increase boilerplate,
but makes the core hopefully more readable and maintanable).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292552 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AArch64][GlobalISel] Split legalizer combine tests. NFC.
Ahmed Bougacha [Fri, 20 Jan 2017 00:30:06 +0000 (00:30 +0000)]
[AArch64][GlobalISel] Split legalizer combine tests. NFC.

Big functions with large vreg # are quite unwieldy to update.  This test
also relied on legal s8 operations which we're considering removing.

Change it to have one function per test (it does increase boilerplate,
but makes the core hopefully more readable and maintanable), and use
100% legal operations throughout.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292551 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[MIRParser] Allow generic register specification on operand.
Ahmed Bougacha [Fri, 20 Jan 2017 00:29:59 +0000 (00:29 +0000)]
[MIRParser] Allow generic register specification on operand.

This completes r292321 by adding support for generic registers, e.g.:

  %2:_(s32) = G_ADD %0, %1

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292550 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[lit] Limit parallelism of sanitizer tests on Darwin [llvm part, take 2]
Kuba Mracek [Fri, 20 Jan 2017 00:24:32 +0000 (00:24 +0000)]
[lit] Limit parallelism of sanitizer tests on Darwin [llvm part, take 2]

Running lit tests and unit tests of ASan and TSan on macOS has very bad performance when running with a high number of threads. This is caused by xnu (the macOS kernel), which currently doesn't handle mapping and unmapping of sanitizer shadow regions (reserved VM which are several terabytes large) very well. The situation is so bad that increasing the number of threads actually makes the total testing time larger. The macOS buildbots are affected by this. Note that we can't easily limit the number of sanitizer testing threads without affecting the rest of the tests.

This patch adds a special "group" into lit, and limits the number of concurrently running tests in this group. This helps solve the contention problem, while still allowing other tests to run in full, that means running lit with -j8 will still with 8 threads, and parallelism is only limited in sanitizer tests.

Differential Revision: https://reviews.llvm.org/D28420

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292548 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoGlobalISel: Only set FailedISel on dropped dbg intrinsics when using fallback
Justin Bogner [Fri, 20 Jan 2017 00:24:30 +0000 (00:24 +0000)]
GlobalISel: Only set FailedISel on dropped dbg intrinsics when using fallback

It's easier to test the non-fallback path if we just drop these
intrinsics for now, like we did before we added the fallback path.
We'll obviously need to fix this properly, but the fixme for that is
already here.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292547 91177308-0d34-0410-b5e6-96231b3b80d8