7 years agoAMDGPU/SI: Test commit
Changpeng Fang [Fri, 18 Dec 2015 20:04:28 +0000 (20:04 +0000)]
AMDGPU/SI: Test commit
    Summary: This is just my first commit. Test!

    Reviewers: none

    Subscribers: none

    Differential Revision: none

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256022 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "AMDGPU/SI: Test commit"
Changpeng Fang [Fri, 18 Dec 2015 20:04:26 +0000 (20:04 +0000)]
Revert "AMDGPU/SI: Test commit"

This reverts commit a493cb636e0152ad28210934a47c6c44b1437193.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256021 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU/SI: Test commit
Changpeng Fang [Fri, 18 Dec 2015 19:57:41 +0000 (19:57 +0000)]
AMDGPU/SI: Test commit
Summary: This is just my first commit. Test!

Reviewers: none

Subscribers: none

Differential Revision: none

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256020 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDrop support for dematerializing.
Rafael Espindola [Fri, 18 Dec 2015 19:57:26 +0000 (19:57 +0000)]
Drop support for dematerializing.

It was only used on lib/Linker and the use was "dead" since it was used on a
function the IRMover had just moved.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256019 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "Improve DWARFDebugFrame::parse to also handle __eh_frame."
Pete Cooper [Fri, 18 Dec 2015 19:45:38 +0000 (19:45 +0000)]
Revert "Improve DWARFDebugFrame::parse to also handle __eh_frame."

This reverts commit r256008.

Its breaking multiple buildbots, although works for me locally.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256013 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRename variables to reflect linker split (NFC)
Teresa Johnson [Fri, 18 Dec 2015 19:28:59 +0000 (19:28 +0000)]
Rename variables to reflect linker split (NFC)

Renamed variables to be more reflective of whether they are
an instance of Linker, IRLinker or ModuleLinker. Also fix a stale

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256011 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoConvert Arg, ArgList, and Option to dump() to dbgs() rather than errs().
Eric Christopher [Fri, 18 Dec 2015 18:55:26 +0000 (18:55 +0000)]
Convert Arg, ArgList, and Option to dump() to dbgs() rather than errs().

Also add print() functions.

Patch by Justin Lebar!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256010 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAdd a dump method for ArgList.
Eric Christopher [Fri, 18 Dec 2015 18:55:22 +0000 (18:55 +0000)]
Add a dump method for ArgList.

Patch by Justin Lebar!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256009 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoImprove DWARFDebugFrame::parse to also handle __eh_frame.
Pete Cooper [Fri, 18 Dec 2015 18:51:08 +0000 (18:51 +0000)]
Improve DWARFDebugFrame::parse to also handle __eh_frame.

LLVM MC has single methods which can handle the output of EH frame and DWARF CIE's and FDE's.

This code improves DWARFDebugFrame::parse to do the same for parsing.

This also allows llvm-objdump to support the --dwarf=frames option which objdump supports.  This
option dumps the .eh_frame section using the new code in DWARFDebugFrame::parse.


Reviewed by Rafael Espindola.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256008 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRecognize strings for Hexagon-specific variant kinds
Krzysztof Parzyszek [Fri, 18 Dec 2015 18:47:27 +0000 (18:47 +0000)]
Recognize strings for Hexagon-specific variant kinds

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256007 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[WinEH] Update LCSSA to handle catchswitch with handlers inside and outside a loop
Andrew Kaylor [Fri, 18 Dec 2015 18:12:35 +0000 (18:12 +0000)]
[WinEH] Update LCSSA to handle catchswitch with handlers inside and outside a loop

Differential Revision: http://reviews.llvm.org/D15630

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256005 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AArch64] Promote loads from stores
Jun Bum Lim [Fri, 18 Dec 2015 18:08:30 +0000 (18:08 +0000)]
[AArch64] Promote loads from stores

This change promotes load instructions which directly read from stores by
replacing them with mov instructions. If the store is wider than the load,
the load will be replaced with a bitfield extract.
For example :
  STRWui %W1, %X0, 1
  %W0 = LDRHHui %X0, 3
  STRWui %W1, %X0, 1
  %W0 = UBFMWri %W1, 16, 31

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256004 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ThinLTO/LTO] Don't link in unneeded metadata
Teresa Johnson [Fri, 18 Dec 2015 17:51:37 +0000 (17:51 +0000)]
[ThinLTO/LTO] Don't link in unneeded metadata

Third patch split out from http://reviews.llvm.org/D14752.

Only map in needed DISubroutine metadata (imported or otherwise linked
in functions and other DISubroutine referenced by inlined instructions).
This is supported for ThinLTO, LTO and llvm-link --only-needed, with
associated tests for each one.

Depends on D14838.

Reviewers: dexonsmith, joker.eph

Subscribers: davidxl, llvm-commits, joker.eph

Differential Revision: http://reviews.llvm.org/D14843

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256003 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoHandle archives with paths in the names.
Rafael Espindola [Fri, 18 Dec 2015 16:07:17 +0000 (16:07 +0000)]
Handle archives with paths in the names.

We always create archives with just he filename as the member name, but
other archives can put a more complicated path in there.

This patches handles it by computing just the filename as we do when
adding a new member.

If storing the path is important for some reason, we should probably
have an orthogonal option for doing that and do it for both old and new

Fixes pr25877.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256001 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoclang-format to reduce diff in another patch.
Rafael Espindola [Fri, 18 Dec 2015 14:06:34 +0000 (14:06 +0000)]
clang-format to reduce diff in another patch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255999 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoFix error handling in LLVMGetBitcodeModuleInContext.
Rafael Espindola [Fri, 18 Dec 2015 13:58:05 +0000 (13:58 +0000)]
Fix error handling in LLVMGetBitcodeModuleInContext.

It was not setting OutMessage.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255998 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoGlobalsAA: Take advantage of ArgMemOnly, InaccessibleMemOnly and InaccessibleMemOrArg...
Vaivaswatha Nagaraj [Fri, 18 Dec 2015 11:02:52 +0000 (11:02 +0000)]
GlobalsAA: Take advantage of ArgMemOnly, InaccessibleMemOnly and InaccessibleMemOrArgMemOnly attributes

1. Modify AnalyzeCallGraph() to retain function info for external functions
if the function has [InaccessibleMemOr]ArgMemOnly flags.
2. When analyzing the use of a global is function parameter at a call site,
mark the callee also as modifying the global appropriately.
3. Add additional test cases.

Depends on D15499

Reviewers: hfinkel, jmolloy

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D15605

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255994 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[mips][microMIPS][DSP] Implement PACKRL.PH, PICK.PH, PICK.QB, SHILO, SHILOV and WRDSP...
Zlatko Buljan [Fri, 18 Dec 2015 08:59:37 +0000 (08:59 +0000)]
[mips][microMIPS][DSP] Implement PACKRL.PH, PICK.PH, PICK.QB, SHILO, SHILOV and WRDSP instructions
Differential Revision: http://reviews.llvm.org/D14429

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255991 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAdd a test for LLVMGetBitcodeModule.
Rafael Espindola [Fri, 18 Dec 2015 03:57:26 +0000 (03:57 +0000)]
Add a test for LLVMGetBitcodeModule.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255985 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[RS4GC] Use an value handle to help isolate errors quickly
Philip Reames [Fri, 18 Dec 2015 03:53:28 +0000 (03:53 +0000)]
[RS4GC] Use an value handle to help isolate errors quickly

Inspired by the bug reported in 25846.  Whatever we end up doing about that one, the value handle change is a generally good one since it will help catch this type of mistake more quickly.

Patch by: Manuel Jacob

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255984 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoclang-format to reduce diff in another patch.
Rafael Espindola [Fri, 18 Dec 2015 03:04:52 +0000 (03:04 +0000)]
clang-format to reduce diff in another patch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255981 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "[Option] Introduce Arg::print(raw_ostream&) and use llvm::dbgs"
Vedant Kumar [Fri, 18 Dec 2015 02:30:45 +0000 (02:30 +0000)]
Revert "[Option] Introduce Arg::print(raw_ostream&) and use llvm::dbgs"

This reverts commit r255977. This is part of

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255978 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[Option] Introduce Arg::print(raw_ostream&) and use llvm::dbgs
Vedant Kumar [Fri, 18 Dec 2015 02:27:52 +0000 (02:27 +0000)]
[Option] Introduce Arg::print(raw_ostream&) and use llvm::dbgs

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255977 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoReorganize the C API headers to improve build times.
Eric Christopher [Fri, 18 Dec 2015 01:46:52 +0000 (01:46 +0000)]
Reorganize the C API headers to improve build times.

Type specific declarations have been moved to Type.h and error handling
routines have been moved to ErrorHandling.h. Both are included in Core.h
so nothing should change for projects directly including the headers,
but transitive dependencies may be affected.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255965 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoBranchProbabilityTest.cpp: Suppress warnings. [-Wsign-compare]
NAKAMURA Takumi [Fri, 18 Dec 2015 00:18:18 +0000 (00:18 +0000)]
BranchProbabilityTest.cpp: Suppress warnings. [-Wsign-compare]

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255940 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRemove unused class variables.
Eric Christopher [Thu, 17 Dec 2015 23:43:40 +0000 (23:43 +0000)]
Remove unused class variables.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255939 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[X86] Use push-pop for materializing small constants under 'minsize'
Hans Wennborg [Thu, 17 Dec 2015 23:18:39 +0000 (23:18 +0000)]
[X86] Use push-pop for materializing small constants under 'minsize'

Use the 3-byte (4 with REX prefix) push-pop sequence for materializing
small constants. This is smaller than using a mov (5, 6 or 7 bytes
depending on size and REX prefix), but it's likely to be slower, so
only used for 'minsize'.

This is a follow-up to r255656.

Differential Revision: http://reviews.llvm.org/D15549

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255936 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[BranchProbability] Remove the restriction that known and unknown probabilities canno...
Cong Hou [Thu, 17 Dec 2015 22:27:07 +0000 (22:27 +0000)]
[BranchProbability] Remove the restriction that known and unknown probabilities cannot coexist when being normalized.

The current BranchProbability::normalizeProbabilities() forbids known and
unknown probabilities to coexist in the list. This was once used to help
capture probability exceptions but has caused some reported build
failures (https://llvm.org/bugs/show_bug.cgi?id=25838).

This patch removes this restriction by evenly distributing the complement
of the sum of all known probabilities to unknown ones. We could still
treat this as an abnormal behavior, but it is better to emit warnings in
our future profile validator.

Differential revision: http://reviews.llvm.org/D15548

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255934 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[InstCombine] Extend peephole DSE to handle unordered atomics
Philip Reames [Thu, 17 Dec 2015 22:19:27 +0000 (22:19 +0000)]
[InstCombine] Extend peephole DSE to handle unordered atomics

This extends the same line of reasoning used in EarlyCSE w/http://reviews.llvm.org/D15352 to the DSE implementation in InstCombine.

Key points:
 * We only remove unordered or simple stores.
 * The loads producing values consumed by dead stores don't influence whether the store is dead.

Differential Revision: http://reviews.llvm.org/D15354

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255932 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoPolish atomic pointers
JF Bastien [Thu, 17 Dec 2015 22:09:19 +0000 (22:09 +0000)]
Polish atomic pointers

I didn't realize that we already allowed atomic load/store of pointers,
it was added in 2012 by r162146. This patch updates the documentation
and tightens the verifier by using DataLayout to make sure that the
stored size is byte-sized and power-of-two. DataLayout is also used for
integers, and while I'm here I updated the corresponding code for
cmpxchg and rmw.

See the following discussion for context and upcoming changes to
add floating-point and vector atomics:

Reviewers: reames

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D15512

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255931 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoPass -m elf_x84_64 to gold invocations.
Rafael Espindola [Thu, 17 Dec 2015 21:56:27 +0000 (21:56 +0000)]
Pass -m elf_x84_64 to gold invocations.

Fixes pr25868.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255930 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "[AArch64] Add DAG combine for extract extend pattern"
Matthew Simpson [Thu, 17 Dec 2015 21:29:47 +0000 (21:29 +0000)]
Revert "[AArch64] Add DAG combine for extract extend pattern"

This reverts commit r255895. The patch breaks internal tests. Reverting until a
fix is ready.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255928 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDrop function that are deprecated since 2010.
Rafael Espindola [Thu, 17 Dec 2015 21:16:12 +0000 (21:16 +0000)]
Drop function that are deprecated since 2010.

These functions were deprecated in r97608.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255927 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoTest commit
Dave Bartolomeo [Thu, 17 Dec 2015 20:54:16 +0000 (20:54 +0000)]
Test commit

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255926 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[WebAssembly] Switch WebAssemblyMCAsmInfo.h from MCAsmInfo to MCAsmInfoELF.
Dan Gohman [Thu, 17 Dec 2015 20:50:45 +0000 (20:50 +0000)]
[WebAssembly] Switch WebAssemblyMCAsmInfo.h from MCAsmInfo to MCAsmInfoELF.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255925 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[docs] Motivate ninja in GettingStarted.rst
Vedant Kumar [Thu, 17 Dec 2015 20:49:48 +0000 (20:49 +0000)]
[docs] Motivate ninja in GettingStarted.rst

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255924 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[SCEV] Add and use SCEVConstant::getAPInt; NFCI
Sanjoy Das [Thu, 17 Dec 2015 20:28:46 +0000 (20:28 +0000)]
[SCEV] Add and use SCEVConstant::getAPInt; NFCI

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255921 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[InstCombine] Adding "\n" to debug output. NFC.
Weiming Zhao [Thu, 17 Dec 2015 19:53:41 +0000 (19:53 +0000)]
[InstCombine] Adding "\n" to debug output. NFC.

[InstCombine] Adding '\n' to debug output. NFC.

Patch by Zhaoshi Zheng <zhaoshiz@codeaurora.org>

Reviewers: apazos, majnemer, weimingz

Subscribers: arsenm, llvm-commits

Differential Revision: http://reviews.llvm.org/D15403

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255920 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoHardcode the target in this testcase — it depends on the ABI.
Adrian Prantl [Thu, 17 Dec 2015 19:33:56 +0000 (19:33 +0000)]
Hardcode the target in this testcase — it depends on the ABI.
This fixes a failure on Windows buildbots.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255919 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[EarlyCSE] DSE of atomic unordered stores
Philip Reames [Thu, 17 Dec 2015 18:50:50 +0000 (18:50 +0000)]
[EarlyCSE] DSE of atomic unordered stores

The rules for removing trivially dead stores are a lot less complicated than loads. Since we know the later store post dominates the former and the former dominates the later, unless the former has side effects other than the actual store, we can remove it. One slightly surprising thing is that we can freely remove atomic stores, even if the later one isn't atomic. There's no guarantee the atomic one was every visible.

For the moment, we don't handle DSE of ordered atomic stores. We could extend the same chain of reasoning to them, but the catch is we'd then have to model the ordering effect without a store instruction. Since our fences are a stronger than our operation orderings, simple using a fence isn't an obvious win. This arguable calls for a refinement in our fence specification, but that's (much) later work.

Differential Revision: http://reviews.llvm.org/D15352

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255914 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agomake this test less whitespace-sensitive.
Adrian Prantl [Thu, 17 Dec 2015 18:34:37 +0000 (18:34 +0000)]
make this test less whitespace-sensitive.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255913 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRewrite test to use llvm-dwarfdump instead of checking for asm comments.
Adrian Prantl [Thu, 17 Dec 2015 18:25:51 +0000 (18:25 +0000)]
Rewrite test to use llvm-dwarfdump instead of checking for asm comments.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255912 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[ThinLTO] Metadata linking for imported functions
Teresa Johnson [Thu, 17 Dec 2015 17:14:09 +0000 (17:14 +0000)]
[ThinLTO] Metadata linking for imported functions

Second patch split out from http://reviews.llvm.org/D14752.

Maps metadata as a post-pass from each module when importing complete,
suturing up final metadata to the temporary metadata left on the
imported instructions.

This entails saving the mapping from bitcode value id to temporary
metadata in the importing pass, and from bitcode value id to final
metadata during the metadata linking postpass.

Depends on D14825.

Reviewers: dexonsmith, joker.eph

Subscribers: davidxl, llvm-commits, joker.eph

Differential Revision: http://reviews.llvm.org/D14838

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255909 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU/SI: Reserve appropriate number of sgprs for flat scratch init.
Tom Stellard [Thu, 17 Dec 2015 17:05:09 +0000 (17:05 +0000)]
AMDGPU/SI: Reserve appropriate number of  sgprs for flat scratch init.

Reviewers: tstellarAMD

Subscribers: arsenm, llvm-commits

Differential Revision: http://reviews.llvm.org/D15583

Patch by: Changpeng Fang

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255908 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU: Fix off-by-one in SIRegisterInfo::eliminateFrameIndex
Nicolai Haehnle [Thu, 17 Dec 2015 16:46:42 +0000 (16:46 +0000)]
AMDGPU: Fix off-by-one in SIRegisterInfo::eliminateFrameIndex

The method insertNOPs expected the number of wait states to be passed as
parameter, while eliminateFrameIndex passed the immediate argument for the
S_NOP, leading to an off-by-one error. Rename the method to make the
meaning of its parameter clearer. The number of 4 / 5 wait states (which
is what the method has always _tried_ to do according to the comment) is
correct according to the hardware docs.

I stumbled upon this while trying to track down the cause of
https://bugs.freedesktop.org/show_bug.cgi?id=93264. While clearly needed,
this patch unfortunately does not fix that bug...

Reviewers: arsenm, tstellarAMD

Subscribers: arsenm, llvm-commits

Differential Revision: http://reviews.llvm.org/D15542

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255906 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert r254592 (virtual dtor in SCEVPredicate).
Andy Gibbs [Thu, 17 Dec 2015 16:43:53 +0000 (16:43 +0000)]
Revert r254592 (virtual dtor in SCEVPredicate).

Clang has better diagnostics in this case.  It is not necessary therefore
to change the destructor to avoid what is effectively an invalid warning
in gcc.  Instead, better handle the warning flags given to the compiler.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255905 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoMark a couple ModuleLinker member functions as const (NFC)
Teresa Johnson [Thu, 17 Dec 2015 16:34:53 +0000 (16:34 +0000)]
Mark a couple ModuleLinker member functions as const (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255903 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAvoid explicit relocation sorting most of the time.
Rafael Espindola [Thu, 17 Dec 2015 16:22:06 +0000 (16:22 +0000)]
Avoid explicit relocation sorting most of the time.

These days relocations are created and stored in a deterministic way.
The order they are created is also suitable for the .o file, so we don't
need an explicit sort.

The last remaining exception is MIPS.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255902 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "[AArch64] Enable PostRAScheduler for AArch64 generic build"
Rafael Espindola [Thu, 17 Dec 2015 15:12:26 +0000 (15:12 +0000)]
Revert "[AArch64] Enable PostRAScheduler for AArch64 generic build"

This reverts commit r255896. It broke the tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255899 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAlways sort by offset first. NFC.
Rafael Espindola [Thu, 17 Dec 2015 15:08:24 +0000 (15:08 +0000)]
Always sort by offset first. NFC.

Every target changing sortRelocs was first calling the parent
implementation. Just run that first.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255898 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoFix unused variable warning in release builds. NFC.
Diego Novillo [Thu, 17 Dec 2015 14:58:34 +0000 (14:58 +0000)]
Fix unused variable warning in release builds. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255897 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AArch64] Enable PostRAScheduler for AArch64 generic build
MinSeong Kim [Thu, 17 Dec 2015 14:51:22 +0000 (14:51 +0000)]
[AArch64] Enable PostRAScheduler for AArch64 generic build

This patch enables PostRAScheduler specifically for AArch64 generic build,
which is beneficial from the performance perspective.
Speedups up to 2 to 7% for some benchmarks on A57 and A53 are observed.
Also benchmarks from LLVM test-suite did not regress.

Differential Revision: http://reviews.llvm.org/D15557

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255896 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AArch64] Add DAG combine for extract extend pattern
Matthew Simpson [Thu, 17 Dec 2015 14:30:55 +0000 (14:30 +0000)]
[AArch64] Add DAG combine for extract extend pattern

This patch adds a DAG combine for (any_extend (extract_vector_elt v, i)) ->
(extract_vector_elt v, i). The combine enables us to better match some SMOV

Differential Revision: http://reviews.llvm.org/D15515

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255895 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoSimplify. NFC.
Rafael Espindola [Thu, 17 Dec 2015 14:19:52 +0000 (14:19 +0000)]
Simplify. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255894 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDOTGraphTraits: Allow the decision to show a graph to consider the analysis
Tobias Grosser [Thu, 17 Dec 2015 12:48:25 +0000 (12:48 +0000)]
DOTGraphTraits: Allow the decision to show a graph to consider the analysis

The method processFunction() is called to decide if a graph should be shown for
a certain function. To allow DOTGraphTraitViewers to take this decision based
on the analysis results for the given function, we forward a reference to the
analysis result. This will be used by Polly to only visualize functions where
interesting loop regions have been detected.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255889 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRemoved empty directory.
Alexander Kornienko [Thu, 17 Dec 2015 12:19:05 +0000 (12:19 +0000)]
Removed empty directory.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255888 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[X86] Add option for enabling LEA optimization pass, by Andrey Turetsky
Alexey Bataev [Thu, 17 Dec 2015 07:34:39 +0000 (07:34 +0000)]
[X86] Add option for enabling LEA optimization pass, by Andrey Turetsky
Add option to enable/disable LEA optimization pass. By default the pass is disabled.
Differential Revision: http://reviews.llvm.org/D15573

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255881 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[WebAssembly] Convert WebAssemblyTargetObjectFile to TargetLoweringObjectFileELF
Dan Gohman [Thu, 17 Dec 2015 04:55:44 +0000 (04:55 +0000)]
[WebAssembly] Convert WebAssemblyTargetObjectFile to TargetLoweringObjectFileELF

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255877 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAArch64: Simplify emitEpilogue() and related code; NFC
Matthias Braun [Thu, 17 Dec 2015 03:18:47 +0000 (03:18 +0000)]
AArch64: Simplify emitEpilogue() and related code; NFC

This is in preparation to an upcoming patch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255872 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[llvm-objdump] Use report_fatal_error() for a more uniform error handling.
Davide Italiano [Thu, 17 Dec 2015 01:59:50 +0000 (01:59 +0000)]
[llvm-objdump] Use report_fatal_error() for a more uniform error handling.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255871 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[WebAssembly] Experimental ELF writer support
Dan Gohman [Thu, 17 Dec 2015 01:39:00 +0000 (01:39 +0000)]
[WebAssembly] Experimental ELF writer support

This creates the initial infrastructure for writing ELF output files. It
doesn't yet have any implementation for encoding instructions.

Differential Revision: http://reviews.llvm.org/D15555

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255869 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoFix PR25838.
Cong Hou [Thu, 17 Dec 2015 01:29:08 +0000 (01:29 +0000)]
Fix PR25838.

This is a quick fix to PR25838. The issue comes from the restriction that we
cannot normalize probabilities containing both known and unknown ones. A patch
that removes this restriction is under the review now:


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255867 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[PGO] InstrPGO and coverage code refactoring (NFC)
Xinliang David Li [Thu, 17 Dec 2015 00:53:37 +0000 (00:53 +0000)]
[PGO] InstrPGO and coverage code refactoring (NFC)

Introduce a new class InstrProfSymtab to abstract
the PGO symbol table for prof and coverage reader.
The symtab is is to lookup function's PGO name
using function keys. The first user of the class
is CoverageMapping Reader. More will follow.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255862 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoWebAssembly: update expected torture test failures
JF Bastien [Thu, 17 Dec 2015 00:12:06 +0000 (00:12 +0000)]
WebAssembly: update expected torture test failures

We now have 240 expected failures.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255858 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoUse std::unique_ptr. NFC.
Rafael Espindola [Wed, 16 Dec 2015 23:49:14 +0000 (23:49 +0000)]
Use std::unique_ptr. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255852 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[WebAssembly] Fix legalization of shift operators on large integer types.
Dan Gohman [Wed, 16 Dec 2015 23:25:51 +0000 (23:25 +0000)]
[WebAssembly] Fix legalization of shift operators on large integer types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255847 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[WebAssembly] Implement eliminateCallFramePseudo
Derek Schuff [Wed, 16 Dec 2015 23:21:30 +0000 (23:21 +0000)]
[WebAssembly] Implement eliminateCallFramePseudo

Implement eliminateCallFramePsuedo to handle ADJCALLSTACKUP/DOWN
pseudo-instructions. Add a test calling a vararg function which causes non-0
adjustments. This revealed an issue with RegisterCoalescer wherein it
eliminates a COPY from SP32 to a vreg but failes to update the live ranges
of EXPR_STACK, causing a machineinstr verifier failure (so this test
is commented out).

Also add a dynamic alloca test, which causes a callseq_end dag node with
a 0 (instead of undef) second argument to be generated. We currently fail to
select that, so adjust the ADJCALLSTACKUP tablegen code to handle it.

Differential Revision: http://reviews.llvm.org/D15587

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255844 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoChange linkInModule to take a std::unique_ptr.
Rafael Espindola [Wed, 16 Dec 2015 23:16:33 +0000 (23:16 +0000)]
Change linkInModule to take a std::unique_ptr.

Passing in a std::unique_ptr should help find errors when the module
is used after being linked into another module.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255842 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoFix funciton->function typo.
Eric Christopher [Wed, 16 Dec 2015 23:10:53 +0000 (23:10 +0000)]
Fix funciton->function typo.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255841 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoDrop an unnecessary use of writev.
Rafael Espindola [Wed, 16 Dec 2015 22:59:06 +0000 (22:59 +0000)]
Drop an unnecessary use of writev.

It looks like the code this patch deletes is based on a misunderstanding of
what guarantees writev provides. In particular, writev with 1 iovec is
not "more atomic" than a write.

Testing on OS X shows that both write and writev from multiple processes
can be intermixed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255837 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[AArch64] Simplify some TRI/TII getters. NFC.
Ahmed Bougacha [Wed, 16 Dec 2015 22:54:06 +0000 (22:54 +0000)]
[AArch64] Simplify some TRI/TII getters. NFC.

We don't need static_casts when we use the right Subtarget.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255836 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoMove llvm/test/DebugInfo/live-debug-values.ll into X86, due to target triple.
NAKAMURA Takumi [Wed, 16 Dec 2015 22:44:10 +0000 (22:44 +0000)]
Move llvm/test/DebugInfo/live-debug-values.ll into X86, due to target triple.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255834 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoSimplify memory management with std::unique_ptr.
Rafael Espindola [Wed, 16 Dec 2015 22:28:34 +0000 (22:28 +0000)]
Simplify memory management with std::unique_ptr.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255831 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[CodeGen] Make MachineInstrBuilder::copyImplicitOps const. NFC.
Ahmed Bougacha [Wed, 16 Dec 2015 22:15:30 +0000 (22:15 +0000)]
[CodeGen] Make MachineInstrBuilder::copyImplicitOps const. NFC.

This matches the other MIB methods, none of which modify the builder.
Without this, we can't chain copyImplicitOps.
Also reformat the few users, in PPCEarlyReturn.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255828 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[PGO] Handle and report overflow during profile merge for all types of data
Nathan Slingerland [Wed, 16 Dec 2015 21:45:43 +0000 (21:45 +0000)]
[PGO] Handle and report overflow during profile merge for all types of data

Summary: Surface counter overflow when merging profile data. Merging still occurs on overflow but counts saturate to the maximum representable value. Overflow is reported to the user.

Reviewers: davidxl, dnovillo, silvas

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D15547

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255825 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoUse LLVM style variable name (NFC)
Teresa Johnson [Wed, 16 Dec 2015 21:37:48 +0000 (21:37 +0000)]
Use LLVM style variable name (NFC)

Fixes variable name from r255779.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255824 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoCXX_FAST_TLS calling convention: performance improvement for AArch64.
Manman Ren [Wed, 16 Dec 2015 21:04:19 +0000 (21:04 +0000)]
CXX_FAST_TLS calling convention: performance improvement for AArch64.

The access function has a short entry and a short exit, the initialization
block is only run the first time. To improve the performance, we want to
have a short frame at the entry and exit.

We explicitly handle most of the CSRs via copies. Only the CSRs that are not
handled via copies will be in CSR_SaveList.

Frame lowering and prologue/epilogue insertion will generate a short frame
in the entry and exit according to CSR_SaveList. The majority of the CSRs will
be handled by register allcoator. Register allocator will try to spill and
reload them in the initialization block.

We add CSRsViaCopy, it will be explicitly handled during lowering.

1> we first set FunctionLoweringInfo->SplitCSR if conditions are met (the target
   supports it for the given machine function and the function has only return
   exits). We also call TLI->initializeSplitCSR to perform initialization.
2> we call TLI->insertCopiesSplitCSR to insert copies from CSRsViaCopy to
   virtual registers at beginning of the entry block and copies from virtual
   registers to CSRsViaCopy at beginning of the exit blocks.
3> we also need to make sure the explicit copies will not be eliminated.

The target independent portion was committed as r255353.

Differential Revision: http://reviews.llvm.org/D15341

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255821 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[Hexagon] Update e_flags in the ELF definitions
Krzysztof Parzyszek [Wed, 16 Dec 2015 20:57:07 +0000 (20:57 +0000)]
[Hexagon] Update e_flags in the ELF definitions

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255820 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoCXX_FAST_TLS calling convention: target independent portion.
Manman Ren [Wed, 16 Dec 2015 20:45:48 +0000 (20:45 +0000)]
CXX_FAST_TLS calling convention: target independent portion.

Update supportSplitCSR's interface to take machine function instead of the
calling convention.

Review comments for http://reviews.llvm.org/D15341

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255818 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRemove now-unused include
Derek Schuff [Wed, 16 Dec 2015 20:43:10 +0000 (20:43 +0000)]
Remove now-unused include

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255817 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoIterate over phys regs instead
Derek Schuff [Wed, 16 Dec 2015 20:43:08 +0000 (20:43 +0000)]
Iterate over phys regs instead

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255816 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[WebAssembly] Print an extra local decl when the user stack pointer is used
Derek Schuff [Wed, 16 Dec 2015 20:43:06 +0000 (20:43 +0000)]
[WebAssembly] Print an extra local decl when the user stack pointer is used

Differential Revision: http://reviews.llvm.org/D15546

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255815 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago [Hexagon] Misc fixes to r255807
Krzysztof Parzyszek [Wed, 16 Dec 2015 20:07:04 +0000 (20:07 +0000)]
 [Hexagon] Misc fixes to r255807

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255811 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoSet debugger tuning from TargetOptions (NFC)
Paul Robinson [Wed, 16 Dec 2015 19:58:30 +0000 (19:58 +0000)]
Set debugger tuning from TargetOptions (NFC)

Differential Revision: http://reviews.llvm.org/D15427

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255810 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[Hexagon] Update the Hexagon packetizer
Krzysztof Parzyszek [Wed, 16 Dec 2015 19:36:12 +0000 (19:36 +0000)]
[Hexagon] Update the Hexagon packetizer

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255807 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "[ARM] Add ARMv8.2-A FP16 scalar instructions"
Reid Kleckner [Wed, 16 Dec 2015 19:21:03 +0000 (19:21 +0000)]
Revert "[ARM] Add ARMv8.2-A FP16 scalar instructions"

This reverts commit r255762.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255806 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[WebAssembly] Fix the CFG Stackifier to handle unoptimized branches
Dan Gohman [Wed, 16 Dec 2015 19:06:41 +0000 (19:06 +0000)]
[WebAssembly] Fix the CFG Stackifier to handle unoptimized branches

If a branch both branches to and falls through to the same block, treat it as
an explicit branch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255803 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoLPM: Make callers of LPM.deleteLoopFromQueue update LoopInfo directly. NFC
Justin Bogner [Wed, 16 Dec 2015 18:40:20 +0000 (18:40 +0000)]
LPM: Make callers of LPM.deleteLoopFromQueue update LoopInfo directly. NFC

As of r255720, the loop pass manager will DTRT when passes update the
loop info for removed loops, so they no longer need to reach into
LPPassManager APIs to do this kind of transformation. This change very
nearly removes the need for the LPPassManager to even be passed into
loop passes - the only remaining pass that uses the LPM argument is

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255797 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoAMDGPU: Override getCFInstrCost
Matt Arsenault [Wed, 16 Dec 2015 18:37:19 +0000 (18:37 +0000)]
AMDGPU: Override getCFInstrCost

The default cost was 0 with the assumption that it is predictable.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255796 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoMachineScheduler: Add a target hook for deciding which RegPressure sets to
Tom Stellard [Wed, 16 Dec 2015 18:31:01 +0000 (18:31 +0000)]
MachineScheduler: Add a target hook for deciding which RegPressure sets to

This patch adds a function called getRegPressureSetScore() to
TargetRegisterInfo.  The MachineScheduler uses this when comparing
instruction that increase the register pressure of different sets
to determine which set is safer to increase.

This hook is useful for GPU targets where the number of registers in the
class is not the best metric for determing which presser set is safer to

Future work may include adding more parameters to this function, like
for example, the current pressure level of the set or the amount that
the pressure will be increased/decreased.

Reviewers: qcolombet, escha, arsenm, atrick, MatzeB

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D14806

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255795 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[CMake] Adding llvm-profdata to the list of tools clang needs.
Chris Bieneman [Wed, 16 Dec 2015 18:30:36 +0000 (18:30 +0000)]
[CMake] Adding llvm-profdata to the list of tools clang needs.

Bootstrapping clang to generate PGO data (patches coming soon), requires llvm-profdata targets to be generated before clang.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255794 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoReland "[llvm-readobj] Simplify usage of -codeview flag"
Reid Kleckner [Wed, 16 Dec 2015 18:28:12 +0000 (18:28 +0000)]
Reland "[llvm-readobj] Simplify usage of -codeview flag"

Relands r255790 with fixed tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255793 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[SLPVectorizer] Ensure dominated reduction values.
Charlie Turner [Wed, 16 Dec 2015 18:23:44 +0000 (18:23 +0000)]
[SLPVectorizer] Ensure dominated reduction values.

When considering incoming values as part of a reduction phi, ensure the
incoming value is dominated by said phi.

Failing to ensure this property causes miscompiles.

Fixes PR25787.

Many thanks to Mattias Eriksson for reporting, reducing and analyzing the
problem for me.

Differential Revision: http://reviews.llvm.org/D15580

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255792 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoRevert "[llvm-readobj] Simplify usage of -codeview flag"
Reid Kleckner [Wed, 16 Dec 2015 18:19:34 +0000 (18:19 +0000)]
Revert "[llvm-readobj] Simplify usage of -codeview flag"

This reverts commit r255790.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255791 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[llvm-readobj] Simplify usage of -codeview flag
Reid Kleckner [Wed, 16 Dec 2015 18:18:39 +0000 (18:18 +0000)]
[llvm-readobj] Simplify usage of -codeview flag

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255790 91177308-0d34-0410-b5e6-96231b3b80d8

7 years agoGenerate a clang CompilationDatabase when running CMake
Benjamin Kramer [Wed, 16 Dec 2015 18:17:45 +0000 (18:17 +0000)]
Generate a clang CompilationDatabase when running CMake

This generates a compile_commands.json file, which tells tools like
YouCompleteMe and clang_complete exactly how to build each source file.

Patch by Justin Lebar!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255789 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[WebAssembly] Use the new offset syntax for memory operands in inline asm.
Dan Gohman [Wed, 16 Dec 2015 18:14:49 +0000 (18:14 +0000)]
[WebAssembly] Use the new offset syntax for memory operands in inline asm.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255788 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[SystemZ] Sort relocs to avoid code corruption by linker optimization
Ulrich Weigand [Wed, 16 Dec 2015 18:12:40 +0000 (18:12 +0000)]
[SystemZ] Sort relocs to avoid code corruption by linker optimization

The SystemZ linkers provide an optimization to transform a general-
or local-dynamic TLS sequence into an initial-exec sequence if possible.
Do do that, the compiler generates a function call to __tls_get_offset,
which is a brasl instruction annotated with *two* relocations:

- a R_390_PLT32DBL to install __tls_get_offset as branch target
- a R_390_TLS_GDCALL / R_390_TLS_LDCALL to inform the linker that
  the TLS optimization should be performed if possible

If the optimization is performed, the brasl is replaced by an ld load

However, *both* relocs are processed independently by the linker.
Therefore it is crucial that the R_390_PLT32DBL is processed *first*
(installing the branch target for the brasl) and the R_390_TLS_GDCALL
is processed *second* (replacing the whole brasl with an ld).

If the relocs are swapped, the linker will first replace the brasl
with an ld, and *then* install the __tls_get_offset branch target
offset.  Since ld has a different layout than brasl, this may even
result in a completely different (or invalid) instruction; in any
case, the resulting code is corrupted.

Unfortunately, the way the MC common code sorts relocations causes
these two to *always* end up the wrong way around, resulting in
wrong code generation by the linker and crashes.

This patch overrides the sortRelocs routine to detect this particular
pair of relocs and enforce the required order.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255787 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[SystemZ] Fix assertion failure in adjustSubwordCmp
Ulrich Weigand [Wed, 16 Dec 2015 18:04:06 +0000 (18:04 +0000)]
[SystemZ] Fix assertion failure in adjustSubwordCmp

When comparing a zero-extended value against a constant small enough to
be in range of the inner type, it doesn't matter whether a signed or
unsigned compare operation (for the outer type) is being used.  This is
why the code in adjustSubwordCmp had this assertion:

    assert(C.ICmpType == SystemZICMP::Any &&
           "Signedness shouldn't matter here.");

assuming the the caller had already detected that fact.  However, it
turns out that there cases, in particular with always-true or always-
false conditions that have not been eliminated when compiling at -O0,
where this is not true.

Instead of failing an assertion if C.ICmpType is not SystemZICMP::Any
here, we can simply *set* it safely to SystemZICMP::Any, however.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255786 91177308-0d34-0410-b5e6-96231b3b80d8

7 years ago[Hexagon] Make memcpy lowering thread-safe
Tobias Edler von Koch [Wed, 16 Dec 2015 17:29:37 +0000 (17:29 +0000)]
[Hexagon] Make memcpy lowering thread-safe

This removes an unpleasant hack involving a global variable for special
lowering of certain memcpy calls. These are now lowered as intended in
EmitTargetCodeForMemcpy in the same way that other targets do it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@255785 91177308-0d34-0410-b5e6-96231b3b80d8