Commit Graph

255475 Commits

Author SHA1 Message Date
Michael Kuperstein c2af82b4b7 [LoopUnroll] Enable PGO-based loop peeling by default.
This enables peeling of loops with low dynamic iteration count by default,
when profile information is available.

Differential Revision: https://reviews.llvm.org/D27734

llvm-svn: 295796
2017-02-22 00:27:34 +00:00
Tim Shen 01fb2c87b9 [XRay] Change the ppc trampoline asm file into a different name, to not collide with the cc file. NFC.
llvm-svn: 295795
2017-02-22 00:19:43 +00:00
Richard Smith a0abc42911 Fix assertion failure when generating debug information for a variable
declaration declared using class template argument deduction.

Patch by Eric Fiselier (who is busy and asked me to commit this on his behalf)!

Differential Revision: https://reviews.llvm.org/D30082

llvm-svn: 295794
2017-02-22 00:13:14 +00:00
Rui Ueyama 98eafd67d5 Attempt to fix buildbot.
I added this log message to test the /msvclto option, but
this output might confuse FileCheck. This patch attempts to fix
it by removing it.

llvm-svn: 295793
2017-02-22 00:06:18 +00:00
Matt Arsenault 3ea06336fc AMDGPU: Remove some uses of llvm.SI.export in tests
Merge some of the old, smaller tests into more complete versions.

llvm-svn: 295792
2017-02-22 00:02:21 +00:00
Richard Smith b80bbca254 [c++1z] Mark constexpr lambdas as done on status page and start advertising
them via feature test macro __cpp_constexpr.

Thanks to Faisal for implementing this feature!

llvm-svn: 295791
2017-02-21 23:58:29 +00:00
Richard Smith 130cc445e4 Fix deduction of type of pack-expanded non-type template parameter.
We need to look through the PackExpansionType in the parameter type when
deducing, and we need to consider the possibility of deducing arguments for
packs that are not lexically mentioned in the pattern (but are nonetheless
deducible) when figuring out which packs are covered by a pack deduction scope.

llvm-svn: 295790
2017-02-21 23:49:18 +00:00
Matt Arsenault 9417505f7d AMDGPU: Remove llvm.AMDGPU.clamp intrinsic
llvm-svn: 295789
2017-02-21 23:46:04 +00:00
Matt Arsenault 2fdf2a1a18 AMDGPU: Redefine clamp node as clamp 0.0-1.0
Change implementation to use max instead of add.
min/max/med3 do not flush denormals regardless of the mode,
so it is OK to use it whether or not they are enabled.

Also allow using clamp with f16, and use knowledge
of dx10_clamp.

llvm-svn: 295788
2017-02-21 23:35:48 +00:00
Rui Ueyama e6e206d4b4 Do not use errs() or outs() directly. Instead use message(), log() or error()
LLD is a multi-threaded program. errs() or outs() are not guaranteed
to be thread-safe (they are actually not).

LLD's message(), log() or error() are thread-safe. We should use them.

llvm-svn: 295787
2017-02-21 23:22:56 +00:00
Brad Smith 9aa2bf209b Hook up OpenBSD AArch64 support
llvm-svn: 295786
2017-02-21 23:13:09 +00:00
Artem Belevich 29bbdc1c32 [NVPTX] Unify vectorization of load/stores of aggregate arguments and return values.
Original code only used vector loads/stores for explicit vector arguments.
It could also do more loads/stores than necessary (e.g v5f32 would
touch 8 f32 values). Aggregate types were loaded one element at a time,
even the vectors contained within.

This change attempts to generalize (and simplify) parameter space
loads/stores so that vector loads/stores can be used more broadly.
Functionality of the patch has been verified by compiling thrust
test suite and manually checking the differences between PTX
generated by llvm with and without the patch.

General algorithm:
* ComputePTXValueVTs() flattens input/output argument into a flat list
  of scalars to load/store and returns their types and offsets.
* VectorizePTXValueVTs() uses that data to create vectorization plan
  which returns an array of flags marking boundaries of vectorized
  load/stores. Scalars are represented as 1-element vectors.
* Code that generates loads/stores implements a simple state machine
  that constructs a vector according to the plan.

Differential Revision: https://reviews.llvm.org/D30011

llvm-svn: 295784
2017-02-21 22:56:05 +00:00
Matt Arsenault 7d6b71db4f AMDGPU: Formatting fixes
llvm-svn: 295783
2017-02-21 22:50:41 +00:00
Matt Arsenault f0a4823b91 DAG: Check if extract_vector_elt is legal or custom
Avoids test regressions in future AMDGPU commits when
more vector types are custom lowered.

llvm-svn: 295782
2017-02-21 22:47:27 +00:00
Jacob Gravelle 40aefb5fe0 Declare lgamma library builtins as never being const
Summary:
POSIX requires lgamma writes to an external global variable, signgam.
This prevents annotating lgamma with readnone, which is incorrect on
targets that write to signgam.

Reviewers: efriedma, rsmith

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D29778

llvm-svn: 295781
2017-02-21 22:37:27 +00:00
Petr Hosek 5e51f7d24e [ELF] Insert linkerscript symbols directly into symbol table
This change exposes the symbol table insert method and uses it to
insert the linkerscript defined symbols directly into the symbol
table to avoid unnecessarily pulling the object out of an archive.

Differential Revision: https://reviews.llvm.org/D30224

llvm-svn: 295780
2017-02-21 22:32:51 +00:00
Taewook Oh cc89bacabe Fix for pr31836 - pp_nonportable_path on absolute paths: broken delimiters
Summary: This is a patch for PR31836. As the bug replaces the path separators in the included file name with the characters following them, the test script makes sure that there's no "Ccase-insensitive-include-pr31836.h" in the warning message.

Reviewers: rsmith, eric_niebler

Reviewed By: eric_niebler

Subscribers: karies, cfe-commits

Differential Revision: https://reviews.llvm.org/D30000

llvm-svn: 295779
2017-02-21 22:30:55 +00:00
Tim Shen d4ba2f2336 [XRay] Merge xray clang flag tests, and add powerpc64le.
Summary: I'm not sure why they were in different files, but it's kind of harder to maintain. I create this patch partially for initiate a discussion.

Reviewers: dberris

Subscribers: nemanjai, cfe-commits

Differential Revision: https://reviews.llvm.org/D30118

llvm-svn: 295778
2017-02-21 22:30:00 +00:00
Evandro Menezes a8d3301ee1 [AArch64, X86] Add statistics for the MacroFusion pass
llvm-svn: 295777
2017-02-21 22:16:13 +00:00
Evandro Menezes b9b7f4b8d3 [AArch64, X86] Guard against both instrs being wild cards
If both instrs are wild cards, the result can be a crash.

llvm-svn: 295776
2017-02-21 22:16:11 +00:00
Evandro Menezes bc9a13db0e [AArch64] Add test case for fusion of literal generation
Add test case from https://reviews.llvm.org/D28698 that was somehow lost in
transit.

llvm-svn: 295775
2017-02-21 22:16:09 +00:00
Evandro Menezes ec330cc283 [AArch64] Add test case for fusion of AES crypto operations
Add test case from https://reviews.llvm.org/D28491 that was somehow lost in
transit.

llvm-svn: 295774
2017-02-21 22:16:06 +00:00
Eugene Zelenko 49e2fc4f5f [CodeGen] Fix some Clang-tidy modernize and Include What You Use warnings; other minor fixes (NFC).
llvm-svn: 295773
2017-02-21 22:07:52 +00:00
Rui Ueyama f9e8034c9c Add `-z nocopyreloc` option.
This option disable creating copy relocations. ld.bfd and ld.gold
have the same option.

llvm-svn: 295772
2017-02-21 21:41:50 +00:00
Vitaly Buka 5d6631d8b4 [compiler-rt] Prevent symbolizer from starting itself.
Summary:
If symbolizer was instrumented with sanitizer and crash, it may
try to call itself again causing infinite recursion of crashing processes.

Reviewers: eugenis

Subscribers: kubamracek, llvm-commits, dberris

Differential Revision: https://reviews.llvm.org/D30222

llvm-svn: 295771
2017-02-21 21:39:24 +00:00
Zachary Turner e1ca5a294c Try to fix the buildbot on OSX.
Since I'm only seeing failures on OSX, and it's saying
permission denied, I'm suspecting this is due to the addition
of the MAP_RESILIENT_CODESIGN and/or MAP_RESILIENT_MEDIA flags.
Speculatively trying to remove those to get the bots working.

llvm-svn: 295770
2017-02-21 21:31:28 +00:00
Zachary Turner 6bc2dac132 Try to fix Android build.
llvm-svn: 295769
2017-02-21 21:13:10 +00:00
Zachary Turner 392ed9d342 [Support] Add a function to check if a file resides locally.
Differential Revision: https://reviews.llvm.org/D30010

llvm-svn: 295768
2017-02-21 20:55:47 +00:00
Xin Tong ccee0e0c05 Make default value for disable-licm-promotion in licm explicit.
llvm-svn: 295767
2017-02-21 20:53:48 +00:00
Anna Zaks aacf7958c5 [asan] Re-enable a test on i386-darwin.
This test has been reverted in r279918 due to flaky atos support in the OS
some machines in the buildbot fleet were running. This should not be a
problem anymore.

llvm-svn: 295766
2017-02-21 20:46:50 +00:00
Rafael Espindola 23a76be5ad Don't modify archive members unless really needed.
For whatever reason ld64 requires that member headers (not the member
themselves) should be aligned. The only way to do that is to edit the
previous member so that it ends at an aligned boundary.

Since modifying data put in an archive is an undesirable property,
llvm-ar should only do it when it is absolutely necessary.

llvm-svn: 295765
2017-02-21 20:40:54 +00:00
Dehao Chen 7810d4fbd0 Only enable AddDiscriminator pass when -fdebug-info-for-profiling is true
Summary: AddDiscriminator pass is only useful for sample pgo. This patch restricts AddDiscriminator to -fdebug-info-for-profiling so that it does not introduce unecessary debug size increases for non-sample-pgo builds.

Reviewers: dblaikie, aprantl

Reviewed By: dblaikie

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D30220

llvm-svn: 295764
2017-02-21 20:36:21 +00:00
Erik Pilkington 9227e108eb Fix copy and paste mistake in header comment, NFC.
llvm-svn: 295763
2017-02-21 20:31:01 +00:00
Evgeniy Stepanov 1fd19c6e5d Fix PR31896.
Address of an alias of a global with offset is incorrectly lowered as an address of the global (i.e. ignoring offset).

llvm-svn: 295762
2017-02-21 20:17:34 +00:00
Etienne Bergeron 0eec53cb41 [compiler-rt][asan] Fix incorrect macro preventing ICF with MSVC
Summary:
The DLL thunks are stubs added to an instrumented DLL to redirect ASAN API calls
to the real ones in the main executable. These thunks must contain dummy
code before __asan_init got called. Unfortunately, MSVC linker is doing ICF and is
merging functions with the same body.


In our case, this two ASAN thunks were incorrectly merged:
```
asan_interface.inc:16
INTERFACE_FUNCTION(__asan_before_dynamic_init)
```
```
sanitizer_common_interface.inc:16
INTERFACE_FUNCTION(__sanitizer_verify_contiguous_container)
```

The same thunk got patched twice. After the second patching, calls to
`__asan_before_dynamic_init` are redirected to `__sanitizer_verify_contiguous_container`
and trigger a DCHECK on incorrect operands/

The problem was caused by the macro that is only using __LINE__ to prevent
collapsing code.
```
 #define INTERCEPT_SANITIZER_FUNCTION(name)
   extern "C" __declspec(noinline) void name() {
   volatile int prevent_icf = (__LINE__ << 8); (void)prevent_icf;
```

The current patch is adding __COUNTER__ which is safer than __LINE__.

Also, to precent ICF (guarantee that code is different), we are using a unique attribute:
  - the name of the function

Reviewers: rnk

Reviewed By: rnk

Subscribers: llvm-commits, kubamracek, chrisha, dberris

Differential Revision: https://reviews.llvm.org/D30219

llvm-svn: 295761
2017-02-21 20:04:47 +00:00
Zachary Turner 43313b3e89 Try to fix line endings.
llvm-svn: 295759
2017-02-21 19:52:57 +00:00
Sanjay Patel cb731f1538 [InstCombine] canonicalize non-obivous forms of integer min/max
This is part of trying to clean up our handling of min/max patterns in IR.
By converting these to canonical form, we're more likely to recognize them
because there are various places in InstCombine that don't use 
matchSelectPattern or m_SMax and friends.

The backend fixups referenced in the now deleted TODO comment were added with:
https://reviews.llvm.org/rL291392
https://reviews.llvm.org/rL289738

If there's any codegen fallout from this change, we should be able to address
it in DAGCombiner or target-specific lowering. 

llvm-svn: 295758
2017-02-21 19:33:53 +00:00
Matt Arsenault f3ffe75a1b AMDGPU: Remove dead declarations in tests
llvm-svn: 295757
2017-02-21 19:31:33 +00:00
Zachary Turner 3788818730 Remove svn:eol-style property from 2 files.
There are still over 3400 files remaining with this property set, but there are tens of thousands more with the property not set.  Until we decide what to do on a global scale, this at least unblocks me temporarily.

llvm-svn: 295756
2017-02-21 19:29:56 +00:00
Matt Arsenault b2e6811ec1 AMDGPU: Remove dead declarations from MIR tests
llvm-svn: 295755
2017-02-21 19:27:36 +00:00
Matt Arsenault c2a44e4c3c AMDGPU: Remove llvm.AMDGPU.flbit intrinsic
llvm-svn: 295754
2017-02-21 19:27:33 +00:00
Matt Arsenault e0bf7d02f0 AMDGPU: Don't use stack space for SGPR->VGPR spills
Before frame offsets are calculated, try to eliminate the
frame indexes used by SGPR spills. Then we can delete them
after.

I think for now we can be sure that no other instruction
will be re-using the same frame indexes. It should be easy
to notice if this assumption ever breaks since everything
asserts if it tries to use a dead frame index later.

The unused emergency stack slot seems to still be left behind,
so an additional 4 bytes is still wasted.

llvm-svn: 295753
2017-02-21 19:12:08 +00:00
Xin Tong ebfe01c121 [LoopSimplify] Simplify how we compute UniqueExit
Summary: Simplify how we compute UniqueExit. Reuse ExitBlockSet.

Reviewers: sanjoy, efriedma, hfinkel

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D30182

llvm-svn: 295751
2017-02-21 19:10:58 +00:00
Xin Tong a05a6c101d More comments for getUniqueExitBlocks. NFCI
llvm-svn: 295750
2017-02-21 19:08:03 +00:00
Adrian Prantl 11b2d7dad8 Teach the IR verifier to reject conflicting debug info for function arguments.
Conflicting debug info for function arguments causes hard-to-debug
assertions in the DWARF backend, so the Verifier should reject it.
For performance reasons this only checks function arguments from
non-inlined debug intrinsics for now.

rdar://problem/30520286

llvm-svn: 295749
2017-02-21 19:03:15 +00:00
Geoff Berry 5d534b6a11 [CodeGenPrepare] Sink and duplicate more 'and' instructions.
Summary:
Rework the code that was sinking/duplicating (icmp and, 0) sequences
into blocks where they were being used by conditional branches to form
more tbz instructions on AArch64.  The new code is more general in that
it just looks for 'and's that have all icmp 0's as users, with a target
hook used to select which subset of 'and' instructions to consider.
This change also enables 'and' sinking for X86, where it is more widely
beneficial than on AArch64.

The 'and' sinking/duplicating code is moved into the optimizeInst phase
of CodeGenPrepare, where it can take advantage of the fact the
OptimizeCmpExpression has already sunk/duplicated any icmps into the
blocks where they are used.  One minor complication from this change is
that optimizeLoadExt needed to be updated to always mark 'and's it has
determined should be in the same block as their feeding load in the
InsertedInsts set to avoid an infinite loop of hoisting and sinking the
same 'and'.

This change fixes a regression on X86 in the tsan runtime caused by
moving GVNHoist to a later place in the optimization pipeline (see
PR31382).

Reviewers: t.p.northover, qcolombet, MatzeB

Subscribers: aemerson, mcrosier, sebpop, llvm-commits

Differential Revision: https://reviews.llvm.org/D28813

llvm-svn: 295746
2017-02-21 18:53:14 +00:00
Wei Ding 16289cfcfc AMDGPU : AMDGPU : Update AMDGPU Trap Handler ABI.
Differential Revision: http://reviews.llvm.org/D29913

llvm-svn: 295745
2017-02-21 18:48:01 +00:00
Dmitry Preobrazhensky e6e205344e Test commit
llvm-svn: 295740
2017-02-21 18:07:07 +00:00
Simon Pilgrim 8eb515d8c4 [X86] EltsFromConsecutiveLoads SDLoc argument should be const&.
There appears never to have been a time that the reference was updated.

llvm-svn: 295739
2017-02-21 17:42:28 +00:00
Renato Golin fc1ccec9bb [RT ARM] Avoid Linux include with a redefinition
To avoid depending on kernel headers, we just repeat the single define
we need, which is likely never going to change.

Patch by Joakim Sindholt <opensource@zhasha.com>

llvm-svn: 295738
2017-02-21 17:40:26 +00:00