Commit Graph

334482 Commits

Author SHA1 Message Date
Victor Campos
364b8f5fbe [AArch64] Improve codegen of volatile load/store of i128
Summary:
Instead of generating two i64 instructions for each load or store of a
volatile i128 value (two LDRs or STRs), now emit a single LDP or STP.

Reviewers: labrinea, t.p.northover, efriedma

Reviewed By: efriedma

Subscribers: kristof.beyls, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D69559
2019-12-18 10:03:12 +00:00
Jay Foad
97ca7c2cc9 [AArch64] Enable clustering memory accesses to fixed stack objects
Summary:
r347747 added support for clustering mem ops with FI base operands
including support for fixed stack objects in shouldClusterFI, but
apparently this was never tested.

This patch fixes shouldClusterFI to work with scaled as well as
unscaled load/store instructions, and fixes the ordering of memory ops
in MemOpInfo::operator< to ensure that memory addresses always
increase, regardless of which direction the stack grows.

Subscribers: MatzeB, kristof.beyls, hiraditya, javed.absar, arphaman, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D71334
2019-12-18 09:46:11 +00:00
Miloš Stojanović
862a602416 [llvm-exegesis][mips] Add lit test
Adding a basic lit test for MIPS.

Differential Revision: https://reviews.llvm.org/D71605
2019-12-18 10:21:06 +01:00
Anna Welker
7cd1cfdd6b [NFC][TTI] Add Alignment for isLegalMasked[Gather/Scatter]
Add an extra parameter so alignment can be taken under
consideration in gather/scatter legalization.

Differential Revision: https://reviews.llvm.org/D71610
2019-12-18 09:14:39 +00:00
David Stenberg
a6be12dbb1 [cmake] Add llvm-locstats to LLVM_TEST_DEPENDS
Summary:
Make sure that llvm-locstats is updated before running lit tests even
when it's not an explicit target.

Reviewers: djtodoro, krisb, spatel

Reviewed By: djtodoro

Subscribers: mgorny, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D71611
2019-12-18 09:54:21 +01:00
Georgii Rymar
36c3b3a61c [llvm-readobj][test] - Move a comment. NFC.
I've forgot to address this review comment.
2019-12-18 11:50:08 +03:00
Georgii Rymar
8810284af2 [llvm-readob] - Refactor printing of sections flags. NFCI.
This is a natural clean-up after D71462/D71464.
It allows to define known section letters used for GNU style
in one place.

Differential revision: https://reviews.llvm.org/D71591
2019-12-18 11:43:52 +03:00
Georgii Rymar
f31fa4f898 [llvm-readobj][test] - Cleanup hash-histogram.test
In this test case we use 3 precompiled objects to
test how we print a histogram for an GNU hash section.
It does not make sense to use precompiled objects
for that. Also we could have 2 tests: one for 32 and
another for 64 bits target.

This patch does this change. It is not possible to remove
these precompiled objects because they are used elsewhere.

Differential revision: https://reviews.llvm.org/D71606
2019-12-18 11:36:49 +03:00
Georgii Rymar
f8dbb2c62f [llvm-readelf] - Change letters used for SHF_ARM_PURECODE and SHF_X86_64_LARGE flags.
GNU uses `l` for SHF_X86_64_LARGE and `y` for SHF_ARM_PURECODE.
Lets follow.

To do this I had to refactor and refine how we print the help flags description.
It was too generic and inconsistent with GNU readelf.

Differential revision: https://reviews.llvm.org/D71464
2019-12-18 11:31:58 +03:00
Fangrui Song
37b2808059 [ELF] writePlt, writeIplt: replace parameters gotPltEntryAddr and index with const Symbol &. NFC
PPC::writeIplt (IPLT code sequence, D71621) needs to access `Symbol`.

Reviewed By: grimar, ruiu

Differential Revision: https://reviews.llvm.org/D71631
2019-12-18 00:14:03 -08:00
Georgii Rymar
541daa5e6b [llvm-readelf][llvm-readobj] - Reimplement the logic of section flags dumping.
Our logic that dumped the flags was buggy.

For LLVM style it dumped SHF_MASKPROC/SHF_MASKOS named constants, though
they are not flags, but masks.

For GNU style it was just very inconsistent with GNU which has logic
that is not straightforward. Imagine we have sh_flags == 0x90000000.
SHF_EXCLUDE ("E") has a value of 0x80000000 and SHF_MASKPROC is 0xf0000000.
GNU readelf will not print "E" or "Ep" in this case, but will print just
"p". It only will print "E" when no other processor flag is set.
I had to investigate the GNU source to find the algorithm and now our logic should
match it.

Differential revision: https://reviews.llvm.org/D71462
2019-12-18 10:44:40 +03:00
Wang, Pengfei
8cc0b58673 [X86] Add calculation for elements in structures in getting uniform base for the Gather/Scatter intrinsic.
Summary: Add calculation for elements in structures in getting uniform
base for the Gather/Scatter intrinsic.

Reviewers: craig.topper, c-rhodes, RKSimon

Subscribers: hiraditya, llvm-commits, annita.zhang, LuoYuanke

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D71442
2019-12-18 12:24:58 +08:00
Wang, Pengfei
1949235d13 [X86] Add strict fma support
Summary: Add strict fma support

Reviewers: craig.topper, RKSimon, LiuChen3

Subscribers: hiraditya, llvm-commits, LuoYuanke

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D71604
2019-12-18 11:44:00 +08:00
Gabor Horvath
ea93d7d642 [CFG] Add an option to expand CXXDefaultInitExpr into aggregate initialization
This is useful for clients that are relying on linearized CFGs for evaluating
subexpressions and want the default initializer to be evaluated properly.

The upcoming lifetime analysis is using this but it might also be useful
for the static analyzer at some point.

Differential Revision: https://reviews.llvm.org/D71642
2019-12-17 17:56:06 -08:00
Petr Hosek
547659ae56 [unwind] Don't link libpthread and libdl on Fuchsia
This is a follow up to D71135.
2019-12-17 17:21:43 -08:00
Fangrui Song
07522e4e23 [ELF] Fix a comment. NFC 2019-12-17 17:17:33 -08:00
Justin Bogner
b6f5caa48f [docs] Remove git llvm push and git llvm revert from GettingStarted
These sections aren't accurate since the github move.

Differential Revision: https://reviews.llvm.org/D71640
2019-12-17 17:16:20 -08:00
Nemanja Ivanovic
a5da8d90da [PowerPC] Add missing legalization for vector BSWAP
We somehow missed doing this when we were working on Power9 exploitation.
This just adds the missing legalization and cost for producing the vector
intrinsics.

Differential revision: https://reviews.llvm.org/D70436
2019-12-17 19:07:34 -06:00
LLVM GN Syncbot
758825c5b8 gn build: Merge e62e760f29 2019-12-18 00:29:06 +00:00
Eric Christopher
e62e760f29 Temporarily Revert "[gicombiner] Add the MatchDag structure and parse instruction DAG's from the input"
and follow-on patches.

This is breaking a few build bots and local builds with follow-up already
on the patch thread.

This reverts commits 390c8baa54 and
520e3d66e7.
2019-12-17 16:23:29 -08:00
Craig Topper
004fdbe041 [X86] Manually format some setOperationAction calls to line up arguments to improve readability. NFC 2019-12-17 16:11:31 -08:00
Xiaoqing Wu
a17619e0b0 [AArch64][GlobalISel]: Fix a crash in GlobalIsel in dealing with 16bit uadd.with.overflow.
Summary: AArch64 doesn't support uadd.with.overflow.i16 natively. This change adds a legalization rule to convert the 32bit add result to 16bit. This should fix PR43981.

Reviewers: arsenm, qcolombet, paquette, aemerson

Reviewed By: paquette

Subscribers: wdng, rovka, kristof.beyls, hiraditya, Petar.Avramovic, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D71587
2019-12-17 16:05:00 -08:00
Craig Topper
c36773c78e [FPEnv][LegalizeTypes] Make ScalarizeVecOp_STRICT_FP_ROUND do its own replacements and return SDValue()
The caller will assert for nodes with more than 2 results unless
we return a null SDValue.

I tried to test this by copying an AArch64 test for ScalarizeVecOp_FP_ROUND.
While it did hit the assert and this commited fixed that. It also
hit a later problem that couldn't be fixed without adding strict
FP support to AArch64.
2019-12-17 15:17:43 -08:00
Stanislav Mekhanoshin
b8ac5894a1 [AMDGPU] Fixed cost model for packed 16 bit ops
Differential Revision: https://reviews.llvm.org/D71622
2019-12-17 15:14:17 -08:00
Thomas Lively
f1b351e14a [WebAssembly] Implement SIMD {i8x16,i16x8}.avgr_u instructions
Summary:
These instructions were added to the spec proposal in
https://github.com/WebAssembly/simd/pull/126. Their semantics are
equivalent to `(a + b + 1) / 2`. The opcode for the experimental
i32x4.dot_i16x8_s is also bumped due to a collision with the
i8x16.avgr_u opcode.

Reviewers: aheejin

Subscribers: dschuff, sbc100, jgravelle-google, hiraditya, sunfish, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D71628
2019-12-17 15:05:50 -08:00
Mitch Phillips
f827aff859 Revert "[ MC ] Match labels to existing fragments even when switching sections."
This reverts commit 4272372c57.

Caused an MSan buildbot failure. More information available in the patch
that introduced the bug: https://reviews.llvm.org/D71368
2019-12-17 15:04:26 -08:00
Artem Dergachev
6d3f43ec61 [analysis] Discard type qualifiers when casting values retrieved from the Store.
This canonicalizes the representation of unknown pointer symbols,
which reduces the overall confusion in pointer cast representation.

Patch by Vince Bridgers!

Differential Revision: https://reviews.llvm.org/D70836
2019-12-17 15:00:41 -08:00
Craig Topper
84d8fa30f9 [FPEnv][LegalizeTypes][LegalizeDAG][AArch64] Few fixes/improvements for legalizing fp<->int conversion nodes.
This started with adding a test to support get code coverage on
ScalarizeVecOp_UnaryOp_StrictFP by copying an existing AArch64 test
and using constrained sitofp/uitofp intrinsics.

This found 3 separate issues:
-ScalarizeVecOp_UnaryOp_StrictFP needs to do its own replacement
 because the caller can't handle replacing multiple results.
-Missing integer promotion support for sitofp/uitofp
-Chain result not always assigned in ExpandLegalINT_TO_FP.

Committing them together so I can add the test case.
2019-12-17 14:37:00 -08:00
Alex Lorenz
be88a20c90 [driver][darwin] Use explicit -mlinker-version in the -platform_version tests
The driver actually adds a default -mlinker-version, based on HOST_LINK_VERSION
cmake variable. The tests should be explicit about which version they're using to
trigger the right behavior.
2019-12-17 14:25:22 -08:00
Whitney Tsang
36bdc3dc35 [LoopFusion] Move instructions from FC0.Latch to FC1.Latch.
Summary:This PR move instructions from FC0.Latch bottom up to the
beginning of FC1.Latch as long as they are proven safe.

To illustrate why this is beneficial, let's consider the following
example:
Before Fusion:
header1:
  br header2
header2:
  br header2, latch1
latch1:
  br header1, preheader3
preheader3:
  br header3
header3:
  br header4
header4:
  br header4, latch3
latch3:
  br header3, exit3

After Fusion (before this PR):
header1:
  br header2
header2:
  br header2, latch1
latch1:
  br header3
header3:
  br header4
header4:
  br header4, latch3
latch3:
  br header1, exit3

Note that preheader3 is removed during fusion before this PR.
Notice that we cannot fuse loop2 with loop4 as there exists block latch1
in between.
This PR move instructions from latch1 to beginning of latch3, and remove
block latch1. LoopFusion is now able to fuse loop nest recursively.

After Fusion (after this PR):
header1:
  br header2
header2:
  br header3
header3:
  br header4
header4:
  br header2, latch3
latch3:
  br header1, exit3

Reviewer: kbarton, jdoerfert, Meinersbur, dmgreen, fhahn, hfinkel,
bmahjour, etiotto
Reviewed By: kbarton, Meinersbur
Subscribers: hiraditya, llvm-commits
Tag: LLVM
Differential Revision: https://reviews.llvm.org/D71165
2019-12-17 22:10:23 +00:00
David Tenty
84161f18cc [AIX] Avoid unset csect assert for functions defined after their use in TOC
Summary:
If a function is defined after it appears in a TOC expression, we may
try to access an unset containing csect when returning a symbol for the
expression.

Reviewers: Xiangling_L, DiggerLin, jasonliu, hubert.reinterpretcast

Reviewed By: hubert.reinterpretcast

Subscribers: hubert.reinterpretcast, wuzish, nemanjai, hiraditya, kbarton, jsji, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D71125
2019-12-17 16:59:22 -05:00
Tom Stellard
c3bc805f4f AMDGPU/SILoadStoreOptimillzer: Refactor CombineInfo struct
Summary:
Modify CombineInfo to only store information about a single instruction.
This is a little easier to work with and removes a lot of duplicate
initialization code.

Reviewers: arsenm, nhaehnle

Reviewed By: arsenm, nhaehnle

Subscribers: merge_guards_bot, kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D71045
2019-12-17 13:43:10 -08:00
Nemanja Ivanovic
ba5a00167b Fix buildbot failures after removing REQUIRES-ANY
It would appear that the removal of this lit feature was incomplete
and there is a test case that still tests for this. This patch removes
the remaining tests to bring the bots back to green. I would encourage the
author to do a post-commit review on this in case there is a more desirable fix.
2019-12-17 15:27:45 -06:00
Mark de Wever
1a8ff89653 [IR] Use a reference in a range-based for
This avoids unneeded copies when using a range-based for loops.

This avoids new warnings due to D68912 adds -Wrange-loop-analysis to -Wall.

Differential Revision: https://reviews.llvm.org/D70870
2019-12-17 21:57:58 +01:00
Mark de Wever
b750486c5d [Driver] Avoid copies in range-based for loops
This avoids new warnings due to D68912 adds -Wrange-loop-analysis to -Wall.

Differential Revision: https://reviews.llvm.org/D71527
2019-12-17 21:56:04 +01:00
Mark de Wever
3ec6128dae [Sema] Fixes -Wrange-loop-analysis warnings
This avoids new warnings due to D68912 adds -Wrange-loop-analysis to -Wall.

Differential Revision: https://reviews.llvm.org/D71529
2019-12-17 21:54:32 +01:00
Mark de Wever
385339034f [Frontend] Fixes -Wrange-loop-analysis warnings
This avoids new warnings due to D68912 adds -Wrange-loop-analysis to -Wall.

Differential Revision: https://reviews.llvm.org/D71530
2019-12-17 21:52:47 +01:00
Sourabh Singh Tomar
399273e5eb Recommit "[DebugInfo] Refactored macro related generation,
added a test case for macinfo.dwo emission."

This was reverted in caa4120906,
since it was causing an assertion failure on Windows bots.
This revision is revised to fix that.

Original commit message -

[DebugInfo] Refactored macro related generation, added a test case for macinfo.dwo emission.

Reviewers: dblaikie, aprantl, jini.susan.george

Tags: #debug-info #llvm

Differential Revision: https://reviews.llvm.org/D71008
2019-12-18 02:12:59 +05:30
Xin-Xin Wang
b3f789e037 [perf-training] Change profile file pattern string to use %4m instead of %p
Summary: With %p, each test file that we're using to generate profile data will make its own profraw file which is around 60 MB in size. If we have a lot of test files, that quickly uses a lot of space. Use %4m instead to share the profraw files used to store the profile data. We use 4 here based on the default value in https://reviews.llvm.org/source/llvm-github/browse/master/llvm/CMakeLists.txt$604

Reviewers: beanz, phosek, xiaobai, smeenai, vsk

Reviewed By: vsk

Subscribers: vsk, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D71585
2019-12-17 12:12:21 -08:00
Ulrich Weigand
1e89188d35 [FPEnv] Remove unnecessary rounding mode argument for constrained intrinsics
The following intrinsics currently carry a rounding mode metadata argument:

    llvm.experimental.constrained.minnum
    llvm.experimental.constrained.maxnum
    llvm.experimental.constrained.ceil
    llvm.experimental.constrained.floor
    llvm.experimental.constrained.round
    llvm.experimental.constrained.trunc

This is not useful since the semantics of those intrinsics do not in any way
depend on the rounding mode. In similar cases, other constrained intrinsics
do not have the rounding mode argument. Remove it here as well.

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D71218
2019-12-17 21:10:36 +01:00
Alexey Bataev
b6e7084e25 [OPENMP50]Add parsing/sema analysis for nontemporal clause.
Add basic support for parsing/sema analysis of the nontemporal clause in
simd-based directives.
2019-12-17 14:46:32 -05:00
Alexey Bataev
15d47deedd [LIBOPENMP][NVPTX]Fix the build error in the runtime. 2019-12-17 14:46:04 -05:00
Stefan Stipanovic
fff8ec9813 [Attributor] H2S fix.
Summary: Fixing issues that were noticed in D71521

Reviewers: jdoerfert, lebedev.ri, uenoku

Subscribers:

Differential Revision: https://reviews.llvm.org/D71564
2019-12-17 20:41:09 +01:00
Stefan Stipanovic
d020e67575 [Attributor][NFC] Add test for sle comparison in h2s. 2019-12-17 20:36:10 +01:00
Sam Clegg
8aee6b414a [WebAssembly] Convert MC tests to from bitcode to asm
Now that our `.s` format is stable(ish) and useable we should really
convert all our MC and lld tests over to .s format to match other
targets.

This is a test PR that just converts 2 of our MC tests to see what
it might look like.

Differential Revision: https://reviews.llvm.org/D71506
2019-12-17 11:30:57 -08:00
Jonas Paulsson
599d1cc07a [Clang FE, SystemZ] Recognize -mpacked-stack CL option
Recognize -mpacked-stack from the command line and add a function attribute
"mpacked-stack" when passed. This is needed for building the Linux kernel.

If this option is passed for any other target than SystemZ, an error is
generated.

Review: Ulrich Weigand
https://reviews.llvm.org/D71441
2019-12-17 11:26:17 -08:00
Dominic Chen
79b4c897b8
llvm-diff: Perform structural comparison on GlobalVariables, if possible
Summary: Names of GlobalVariables may not be preserved depending on compilation options, so prefer a structural diff

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D71582
2019-12-17 14:21:48 -05:00
Pierre Habouzit
a4e1819c16 [objc_direct] fix uniquing when re-declaring a readwrite-direct property
ObjCMethodDecl::getCanonicalDecl() for re-declared readwrite properties,
only looks in the ObjCInterface for the declaration of the setter
method, which it won't find.

When the method is a property accessor, we must look in extensions for a
possible redeclaration.

Radar-Id: rdar://problem/57991337

Differential Revision: https://reviews.llvm.org/D71588
2019-12-17 11:07:48 -08:00
Jay Foad
0412f518dc [AMDGPU] Fix typo in SIInstrInfo::memOpsHaveSameBasePtr
Summary:
The typo has been present since memOpsHaveSameBasePtr was introduced in
r313208.

It caused SIInstrInfo::shouldClusterMemOps to cluster more mem ops than
it was supposed to.

Subscribers: arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D71616
2019-12-17 18:54:27 +00:00
Sanjay Patel
6a77e36975 [SDAG] adjust isNegatibleForFree calculation to avoid crashing
This is an alternate fix for the bug discussed in D70595.
This also includes minimal tests for other in-tree targets to show the problem more
generally.

We check the number of uses as a predicate for whether some value is free to negate,
but that use count can change as we rewrite the expression in getNegatedExpression().
So something that was marked free to negate during the cost evaluation phase becomes
not free to negate during the rewrite phase (or the inverse - something that was not
free becomes free). This can lead to a crash/assert because we expect that everything
in an expression that is negatible to be handled in the corresponding code within
getNegatedExpression().

This patch adds a hack to work-around the case where we probably no longer detect
that either multiply operand of an FMA isNegatibleForFree which is assumed to be
true when we started rewriting the expression.

Differential Revision: https://reviews.llvm.org/D70975
2019-12-17 13:49:15 -05:00