Simon Pilgrim [Sun, 6 Jun 2021 21:36:26 +0000 (22:36 +0100)]
[CostModel][X86] Add 512-bit bswap costs
Simon Pilgrim [Sun, 6 Jun 2021 21:24:34 +0000 (22:24 +0100)]
[CostModel][X86] Add 512-bit bswap cost tests
David Green [Sun, 6 Jun 2021 21:30:02 +0000 (22:30 +0100)]
[ARM] MVE tests for vmull from a splat. NFC
David Green [Sun, 6 Jun 2021 21:29:44 +0000 (22:29 +0100)]
[AArch64] Extra tests for vector shift. NFC
Simon Pilgrim [Sun, 6 Jun 2021 20:40:59 +0000 (21:40 +0100)]
[CostModel][X86] Improve AVX512 FDIV costs
Add missing v16f32/v8f64 costs and adjust other costs as well based off the SkylakeServer model
Craig Topper [Sun, 6 Jun 2021 19:48:25 +0000 (12:48 -0700)]
[RISCV] Replace && with ||. Spotted by coverity.
We should be exiting when the shift amount is greater than
the bit width regardless of whether it is a power of 2.
Reported by Simon Pilgrim here https://reviews.llvm.org/D96661
This requires getting a shift amount that is out of bounds that
wasn't already optimized by SelectionDAG. This would be pretty
trick to construct a test for.
Or it would require a non-power of 2 shift amount and a mask
that has runs of ones and zeros of the next lowest power of 2 from
that shift amount. I tried a little to produce a test for this,
but didn't get it to work.
Simon Pilgrim [Sun, 6 Jun 2021 19:00:34 +0000 (20:00 +0100)]
[X86][SSE] LowerFP_TO_INT - remove dead code. NFCI.
Non-Strict v2f32->v2i64 cases have already early-returned to be handled by legalization.
Simon Pilgrim [Sun, 6 Jun 2021 18:53:49 +0000 (19:53 +0100)]
[X86][SSE] combineVectorTruncation - simplify PSHUFB-is-better logic. NFCI.
OutSVT is guaranteed to be i8/i16 and we accept any InSVT that isn't i64
maekawatoshiki [Sun, 6 Jun 2021 16:26:47 +0000 (01:26 +0900)]
Revert "[LoopUnrollAndJam] Change LoopUnrollAndJamPass to LoopNest pass"
This reverts commit
21653600034084e8335374ddc1eb8d362158d9a8.
To fix the crash problem in legacy pass manager
Michael Kruse [Sun, 6 Jun 2021 14:18:59 +0000 (09:18 -0500)]
[Clang][OpenMP] Refactor checking for mutually exclusive clauses. NFC.
Multiple clauses are mutually exclusive. This patch refactors the functions that check for pairs of mutually exclusive clauses into a generalized function which also also accepts a list of clause types if which at most one can appear.
NFC patch extracted out of D99459 by request.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D103666
Simon Pilgrim [Sun, 6 Jun 2021 14:33:28 +0000 (15:33 +0100)]
X86MachObjectWriter.cpp - silence null deference warnings. NFCI.
The MCSymbol data should always be present for non-absolute sections so assert that it is to silence static analysis warnings.
Nikita Popov [Sat, 5 Jun 2021 20:56:19 +0000 (22:56 +0200)]
[TargetLowering] Use IRBuilderBase instead of IRBuilder<> (NFC)
Don't require a specific kind of IRBuilder for TargetLowering hooks.
This allows us to drop the IRBuilder.h include from TargetLowering.h.
Differential Revision: https://reviews.llvm.org/D103759
Nikita Popov [Sun, 6 Jun 2021 14:29:00 +0000 (16:29 +0200)]
[LexicalScopesTest] Add missing IRBuilder.h include (NFC)
This currently depends on a transitive include via TargetLowering.h.
Simon Pilgrim [Sun, 6 Jun 2021 14:24:46 +0000 (15:24 +0100)]
X86Operand.h - fix uninitialized variable warnings in constructor. NFCI.
Simon Pilgrim [Sun, 6 Jun 2021 14:05:32 +0000 (15:05 +0100)]
AssumeBundleQueries.cpp - don't dereference a dyn_cast<> result. NFCI.
Use cast<> instead which will assert that the cast is correct and not just return null - the match() should have already failed if the cast isn't valid anyhow.
Fixes static analysis warning.
Michael Kruse [Sun, 6 Jun 2021 12:28:08 +0000 (07:28 -0500)]
[Clang][OpenMP] Add static version of getSingleClause<ClauseT>. NFC.
The current method getSingleClause requires an instance of OMPExecutableDirective to be called. Introduce a static version taking a list of clauses as argument instead that can be used during parsing/Sema before any OMPExecutableDirective has been created.
This is the same approach as taken for getClausesOfKind for getting more more than a single clause of a type which also has a method and static version. NFC patch extracted out of D99459 by request.
Reviewed By: ABataev
Differential Revision: https://reviews.llvm.org/D103665
Nikita Popov [Sun, 6 Jun 2021 13:39:15 +0000 (15:39 +0200)]
[TargetLowering] Move methods out of line (NFC)
Move methods using IRBuilder out of line, so we can drop the
dependency on the header.
Nikita Popov [Sun, 6 Jun 2021 13:47:30 +0000 (15:47 +0200)]
[CodeGen] Add missing includes (NFC)
These currently rely on the IRBuilder.h include in TargetLowering.h.
Make them explicit.
Simon Pilgrim [Sun, 6 Jun 2021 13:15:11 +0000 (14:15 +0100)]
SimplifyCFG.cpp - remove dead early-return code added at rGcc63203908da. NFCI.
We've already checked that ScanIdx == 0 a few lines above.
Simon Pilgrim [Sun, 6 Jun 2021 13:13:08 +0000 (14:13 +0100)]
BreadthFirstIterator.h - fix uninitialized variable warning in default constructor. NFCI.
Simon Pilgrim [Sun, 6 Jun 2021 12:45:11 +0000 (13:45 +0100)]
Fix implicit fall through compiler warning. NFCI.
Simon Pilgrim [Sun, 6 Jun 2021 12:38:39 +0000 (13:38 +0100)]
PatternMatch.h - wrap WrapFlags tests inside brackets to stop static analysis warning about & vs && usage. NFCI.
Simon Pilgrim [Sun, 6 Jun 2021 12:24:02 +0000 (13:24 +0100)]
BPFISelDAGToDAG.cpp - don't dereference a dyn_cast<> result. NFCI.
Use cast<> instead which will assert that the cast is correct and not just return null.
Fixes static analysis warnings.
Simon Pilgrim [Sun, 6 Jun 2021 11:05:39 +0000 (12:05 +0100)]
SmallVector.h - remove unused MathExtras.h header. NFCI.
Simon Pilgrim [Sun, 6 Jun 2021 10:09:55 +0000 (11:09 +0100)]
Fix uninitialized variable warnings. NFCI.
Simon Pilgrim [Sun, 6 Jun 2021 09:39:31 +0000 (10:39 +0100)]
LibcBenchmark.h - add missing implicit cmath header dependency. NFCI.
Noticed while investigating if we can remove an unnecessary MathExtras.h include from SmallVector.h
Simon Pilgrim [Sun, 6 Jun 2021 08:40:20 +0000 (09:40 +0100)]
DirectoryWatcher-linux.cpp - add missing implicit MathExtras.h header dependency. NFCI.
Simon Pilgrim [Sun, 6 Jun 2021 08:39:19 +0000 (09:39 +0100)]
Revert rG0b18c4c0ec03f0321ee83b9976da5777d0e4f53f "SmallVector.h - remove unused MathExtras.h header (REAPPLIED). NFCI."
Buildbots still seem to find implicit header dependencies that I can't locally....
Simon Pilgrim [Sun, 6 Jun 2021 08:30:15 +0000 (09:30 +0100)]
SmallVector.h - remove unused MathExtras.h header (REAPPLIED). NFCI.
Try again to remove this header - I think I've found the implicit dependencies (mainly for <cmath>) on linux builds now.
Liqiang Tao [Sun, 6 Jun 2021 06:45:03 +0000 (14:45 +0800)]
Revert "[llvm] Add interface to order inlining"
Liqiang Tao [Sun, 6 Jun 2021 03:49:37 +0000 (11:49 +0800)]
[llvm] Add interface to order inlining
This patch abstract Calls in Inliner:run() to InlineOrder.
With this patch, it's possible to customize the inlining order, i.e. use queue or priority queue.
Reviewed By: kazu
Differential Revision: https://reviews.llvm.org/D103315
Simon Pilgrim [Sat, 5 Jun 2021 20:32:41 +0000 (21:32 +0100)]
xray-color-helper.cpp - add missing implicit cmath header dependency. NFCI.
Noticed while investigating if we can remove an unnecessary MathExtras.h include from SmallVector.h (necessary for gcc builds but not MSVC)
Simon Pilgrim [Sat, 5 Jun 2021 20:31:55 +0000 (21:31 +0100)]
xray-color-helper.h - sort includes. NFCI.
Simon Pilgrim [Sat, 5 Jun 2021 20:14:18 +0000 (21:14 +0100)]
BranchProbability.cpp - add missing implicit cmath header dependency. NFCI.
Noticed while investigating if we can remove an unnecessary MathExtras.h include from SmallVector.h (necessary for gcc builds but not MSVC)
Simon Pilgrim [Sat, 5 Jun 2021 19:59:10 +0000 (20:59 +0100)]
Revert rG7b839b3542983a313a9bf9f8d8039ceeea35c4d7 - "SmallVector.h - remove unused MathExtras.h header. NFCI."
Breaks on linux buildbots as I seem to have missed some implicit header dependencies....
Alexander Shaposhnikov [Sat, 5 Jun 2021 19:51:36 +0000 (12:51 -0700)]
[lld][MachO] Add support for $ld$install_name symbols
This diff adds support for $ld$install_name symbols.
Test plan: make check-lld-macho
Differential revision: https://reviews.llvm.org/D103746
Simon Pilgrim [Sat, 5 Jun 2021 19:19:42 +0000 (20:19 +0100)]
SmallVector.h - remove unused MathExtras.h header. NFCI.
David Green [Sat, 5 Jun 2021 18:54:42 +0000 (19:54 +0100)]
[AArch64] Remove AArch64ISD::NEG
This NEG node is just a vector negation, easily represented as a SUB
zero. Removing it from the one place it is generated is essentially an
NFC, but can allow some extra folding. The updated tests are now loading
different constant literals, which have already been negated.
Differential Revision: https://reviews.llvm.org/D103703
Simon Pilgrim [Sat, 5 Jun 2021 18:42:00 +0000 (19:42 +0100)]
Fix "not all control paths return a value" MSVC warning. NFCI.
Simon Pilgrim [Sat, 5 Jun 2021 18:20:14 +0000 (19:20 +0100)]
BitstreamWriter.h - add missing implicit MathExtras.h header dependency. NFCI.
Noticed while investigating if we can remove an unnecessary MathExtras.h include from SmallVector.h
Simon Pilgrim [Sat, 5 Jun 2021 18:11:40 +0000 (19:11 +0100)]
ELFTypes.h - add missing implicit MathExtras.h header dependency. NFCI.
Noticed while investigating if we can remove an unnecessary MathExtras.h include from SmallVector.h
Simon Pilgrim [Sat, 5 Jun 2021 18:10:49 +0000 (19:10 +0100)]
[MCA] Support.h - add missing implicit MathExtras.h header dependency. NFCI.
Noticed while investigating if we can remove an unnecessary MathExtras.h include from SmallVector.h
Simon Pilgrim [Sat, 5 Jun 2021 17:04:38 +0000 (18:04 +0100)]
EndianStream.h - add missing implicit MathExtras.h header dependency. NFCI.
Noticed while investigating if we can remove an unnecessary MathExtras.h include from SmallVector.h
Simon Pilgrim [Sat, 5 Jun 2021 17:03:37 +0000 (18:03 +0100)]
MsgPackReader.cpp - add missing implicit MathExtras.h header dependency. NFCI.
Noticed while investigating if we can remove an unnecessary MathExtras.h include from SmallVector.h
Simon Pilgrim [Sat, 5 Jun 2021 17:02:52 +0000 (18:02 +0100)]
NativeFormatting.cpp - add missing implicit MathExtras.h header dependency. NFCI.
Noticed while investigating if we can remove an unnecessary MathExtras.h include from SmallVector.h
Arthur O'Dwyer [Sat, 5 Jun 2021 16:47:30 +0000 (12:47 -0400)]
[libc++] graph_header_deps.py: Update the computing of "root" (public) headers.
The "root nodes" of the graph are displayed in bold. My intent here
was to bold just the public-API headers, e.g. <vector> and
<experimental/coroutine> and <stdlib.h>, but not helper headers
such as <__functional_base> and <__iterator/next.h>. However,
the recent mass helper-header-ification has exposed defects in
this logic: all the new helpers were ending up bolded! Fix this.
Also, add <__undef_macros> to the list of headers we don't display
by default (like <__config>); it's not interesting to see those edges.
Also, add a sample `dot` command line to the `--help` text.
Arthur O'Dwyer [Mon, 31 May 2021 15:33:51 +0000 (11:33 -0400)]
[libc++] Fix result-type and value_type computation in <valarray>.
The `operator[]` of `_UnaryOp` and `_BinaryOp` returns the result of
calling `__op_`, so its return type should be `__result_type`, not
e.g. `_A0::value_type`. However, `_UnaryOp::value_type` also should
never have been `_A0::value_type`; it needs to be the correct type
for the result of the unary op, e.g. `bool` when the op is `logical_not`.
This turns out to matter when multiple operators are nested, e.g.
`+(v == v)` needs to have a `value_type` of `bool`, not `int`,
even when `v` is of type `valarray<int>`.
Differential Revision: https://reviews.llvm.org/D103416
Aaron Puchert [Sat, 5 Jun 2021 13:04:18 +0000 (15:04 +0200)]
Make TableGenGlobalISel an object library
That's how it was originally intended but that wasn't possible because
we still needed to support older CMake versions.
The problem here is that the sources in TableGenGlobalISel are meant to
be linked into both llvm-tblgen and TableGenTests (a unit test), but not
be part of LLVM proper. So they shouldn't be an ordinary LLVM component.
Because they are used in llvm-tblgen, they can't draw in the LLVM dylib
dependency, but then we'd have to do the same thing in TableGenTests to
make sure we don't link both a static Support library and another copy
through the LLVM dylib.
With an object library we're just reusing the object files and don't
have to care about dependencies at all.
Differential Revision: https://reviews.llvm.org/D74588
Mark de Wever [Sat, 5 Jun 2021 11:58:38 +0000 (13:58 +0200)]
[libc++][doc] Update Format status.
The first part of the <format> header patch series is now complete for
review. This updates the status document.
Roman Lebedev [Tue, 1 Jun 2021 08:33:52 +0000 (11:33 +0300)]
[NFC] Promote willNotOverflow() / getStrengthenedNoWrapFlagsFromBinOp() from IndVars into SCEV proper
We might want to use it when creating SCEV proper in createSCEV(),
now that we don't `forgetValue()` in `SimplifyIndvar::strengthenOverflowingOperation()`,
which might have caused us to loose some optimization potential.
Nikita Popov [Sat, 5 Jun 2021 08:51:14 +0000 (10:51 +0200)]
[LoopUnroll] Regenerate test checks (NFC)
Bruce Mitchener [Sat, 5 Jun 2021 08:50:49 +0000 (15:50 +0700)]
Revert "[LLDB/API] Expose args and env from SBProcessInfo."
This reverts commit
8d33437d030af27fff21dd3fd0e66893b0148217.
This broke one of the buildbots.
Nikita Popov [Sat, 29 May 2021 16:33:31 +0000 (18:33 +0200)]
[LoopUnroll] Separate peeling from unrolling
Loop peeling is currently performed as part of UnrollLoop().
Outside test scenarios, it is always performed with an unroll
count of 1. This means that unrolling doesn't actually do anything
apart from performing post-unroll simplification.
When testing, it's currently possible to specify both an explicit
peel count and an explicit unroll count. This doesn't perform any
sensible operation and may result in miscompiles, see
https://bugs.llvm.org/show_bug.cgi?id=45939.
This patch moves peeling from UnrollLoop() into tryToUnrollLoop(),
so that peeling does not also perform a susequent unroll. We only
run the post-unroll simplifications. Specifying both an explicit
peel count and unroll count is forbidden.
In the future, we may want to support both (non-PGO) peeling a
loop and unrolling it, but this needs to be done by first performing
the peel and then recalculating unrolling heuristics on a now
possibly analyzable loop.
Differential Revision: https://reviews.llvm.org/D103362
Alexander Shaposhnikov [Sat, 5 Jun 2021 08:27:42 +0000 (01:27 -0700)]
[lld][MachO] Fix typo in special-symbol-ld-previous.s
Fix typo in the test special-symbol-ld-previous.s. NFC.
Vitaly Buka [Sat, 5 Jun 2021 07:39:21 +0000 (00:39 -0700)]
Revert "Update and improve compiler-rt tests for -mllvm -asan_use_after_return=(never|[runtime]|always)."
Windows is still broken.
This reverts commit
927688a4cd9dcde92e8bf430e6ceefac6ab8c8a0.
Amir Ayupov [Sat, 5 Jun 2021 07:27:47 +0000 (00:27 -0700)]
[MC] Add getLSDASection interface
This diff adds getLSDASection method to MCObjectFileInfo.
Test plan: make check-all
Differential revision: https://reviews.llvm.org/D102298
Kevin Athey [Sat, 5 Jun 2021 03:21:01 +0000 (20:21 -0700)]
Update and improve compiler-rt tests for -mllvm -asan_use_after_return=(never|[runtime]|always).
In addition:
- optionally add global flag to capture compile intent for UAR:
__asan_detect_use_after_return_always.
The global is a SANITIZER_WEAK_ATTRIBUTE.
for issue: https://github.com/google/sanitizers/issues/1394
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D103304
Vitaly Buka [Sat, 5 Jun 2021 07:01:37 +0000 (00:01 -0700)]
Revert "[gn build] manually port
60e5243e593e (weak asan symbol)"
Not needed after
d8a4a2cb93212c493b7b47663c1cda1103de0bfc.
This reverts commit
00542932fe13ea3b989f40b434399c72f5eca4ec.
Bruce Mitchener [Fri, 4 Jun 2021 08:29:34 +0000 (15:29 +0700)]
[LLDB/API] Expose args and env from SBProcessInfo.
This is another step towards implementing the equivalent of
`platform process list` and related functionality.
`uint32_t` is used for the argument count and index despite the
underlying value being `size_t` to be consistent with other
index-based access to arguments.
Differential Revision: https://reviews.llvm.org/D103675
Fangrui Song [Sat, 5 Jun 2021 06:34:43 +0000 (23:34 -0700)]
Fix some -Wunused-but-set-variable in -DLLVM_ENABLE_ASSERTIONS=off build
Alexander Shaposhnikov [Sat, 5 Jun 2021 06:31:40 +0000 (23:31 -0700)]
[lld][MachO] Add first bits to support special symbols
This diff adds first bits to support special symbols $ld$previous* in LLD.
$ld$* symbols modify properties/behavior of the library
(e.g. its install name, compatibility version or hide/add symbols)
for specific target versions.
Test plan: make check-lld-macho
Differential revision: https://reviews.llvm.org/D103505
Vitaly Buka [Sat, 5 Jun 2021 05:27:45 +0000 (22:27 -0700)]
[NFC][asan] Split fake_stack() into two funcions
Vitaly Buka [Sat, 5 Jun 2021 06:23:12 +0000 (23:23 -0700)]
[NFC][asan] Fix include order
Vitaly Buka [Sat, 5 Jun 2021 06:06:43 +0000 (23:06 -0700)]
[NFC][memprof] FIx delete[] usage in test
Vitaly Buka [Sat, 5 Jun 2021 05:35:06 +0000 (22:35 -0700)]
[NFC][asan] format ThreadStackContainsAddress
Vitaly Buka [Sat, 5 Jun 2021 06:05:53 +0000 (23:05 -0700)]
[NFC][memprof] Compile *.c tests as C
Jim Lin [Sat, 5 Jun 2021 04:49:14 +0000 (12:49 +0800)]
[RISCV] Replace (XLenVT (VLOp GPR:$vl)) with VLOpFrag
This is for D100288 to reduce the changes.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D103682
Michael Kruse [Sat, 5 Jun 2021 04:17:41 +0000 (23:17 -0500)]
[Polly] Move MatMul optimization into its own file. NFC.
Functions shared between generalized matrix-multiplication optimization
and other post-reschedule optimizations (tiling, prevect) are moved into
the schedule tree transformation utility ScheduleTreeTransform.
Vitaly Buka [Sat, 5 Jun 2021 03:17:45 +0000 (20:17 -0700)]
Revert "Update and improve compiler-rt tests for -mllvm -asan_use_after_return=(never|[runtime]|always)."
Reverts commits of D103304, it breaks Darwin.
This reverts commit
60e5243e593ec7e0feca42783edb735e9ac2c4a7.
This reverts commit
26b3ea224e7744058c0053e8eee25a8a3acc9f5d.
This reverts commit
17600ec32a6f3e5720e5232ddb23d67daa650f4d.
Fangrui Song [Sat, 5 Jun 2021 03:15:56 +0000 (20:15 -0700)]
[lsan] Bump the thread limit from 1<<13 to 1<<22
This matches asan. Some applications need more than 1<<13 threads.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D103731
Vitaly Buka [Sat, 5 Jun 2021 02:32:24 +0000 (19:32 -0700)]
[NFC][scudo] Fix sign-compare warning in test
Vitaly Buka [Sat, 5 Jun 2021 02:26:03 +0000 (19:26 -0700)]
[NFC][scudo] Convert to TYPED more ScudoCombinedTest
Vitaly Buka [Fri, 4 Jun 2021 22:56:13 +0000 (15:56 -0700)]
[scudo] Remove ScopedString::Length
Differential Revision: https://reviews.llvm.org/D103725
Vitaly Buka [Fri, 4 Jun 2021 20:48:54 +0000 (13:48 -0700)]
[Scudo] Improve ScopedString constructor
Avoid referencing elements beyond internal vector size.
Reviewed By: cryptoad
Differential Revision: https://reviews.llvm.org/D103718
Vitaly Buka [Sat, 5 Jun 2021 01:22:32 +0000 (18:22 -0700)]
[NFC][scudo] Fix sign-compare warning in test
Kevin Athey [Fri, 4 Jun 2021 23:58:39 +0000 (16:58 -0700)]
remove windows tests for -asan_use-after-return=always (as this is currently disabled for Windows)
Aart Bik [Fri, 4 Jun 2021 23:17:16 +0000 (16:17 -0700)]
[mlir][sparse] add option for 32-bit indices in scatter/gather
Controlled by a compiler option, if 32-bit indices can be handled
with zero/sign-extention alike (viz. no worries on non-negative
indices), scatter/gather operations can use the more efficient
32-bit SIMD version.
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D103632
Nico Weber [Fri, 4 Jun 2021 23:49:29 +0000 (19:49 -0400)]
[gn build] manually port
60e5243e593e (weak asan symbol)
Kevin Athey [Fri, 4 Jun 2021 23:44:34 +0000 (16:44 -0700)]
fix SANITIZE_WINDOWS pragma for https://reviews.llvm.org/rG60e5243e593ec7e0feca42783edb735e9ac2c4a7
Rob Suderman [Fri, 4 Jun 2021 23:30:46 +0000 (16:30 -0700)]
[mlir][tosa] Update tosa.rescale for i48 input type
i48 integers require slightly tweaked behavior, specifically supporting zero
point offsetting with slightly higher bitdepth. Updated results lowering
appropriately.
Reviewed By: NatashaKnk
Differential Revision: https://reviews.llvm.org/D102659
Kevin Athey [Fri, 4 Jun 2021 21:30:04 +0000 (14:30 -0700)]
Update and improve compiler-rt tests for -mllvm -asan_use_after_return=(never|[runtime]|always).
In addition:
- optionally add global flag to capture compile intent for UAR:
__asan_detect_use_after_return_always.
The global is a SANITIZER_WEAK_ATTRIBUTE.
for issue: https://github.com/google/sanitizers/issues/1394
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D103304
Ahmed Taei [Fri, 4 Jun 2021 22:41:19 +0000 (15:41 -0700)]
Revert "Add memref.dim canonicalization patterns to TilingCanonicalizationPatterns"
This reverts commit
a52959401d9544ac7037405aaf714bd08f071ea9.
Differential Revision: https://reviews.llvm.org/D103724
Scott Linder [Wed, 26 May 2021 20:38:41 +0000 (20:38 +0000)]
[ADT] Refactor enumerate unit tests
Preparation for landing the tests for llvm::makeVisitor, including
breaking out the a "Counted" base class and explicitly testing
the prvalue case as distinct from the rvalue case.
Differential Revision: https://reviews.llvm.org/D103206
Roman Lebedev [Fri, 4 Jun 2021 22:23:25 +0000 (01:23 +0300)]
[X86] AMD Zen 3: double the LoopMicroOpBufferSize
While the IndVars issue (PR50384) has been resolved,
and the compile performance improved, a new blocker emerged,
the codegen machine instruction scheduling is also quadratic.
So we still can't really specify the right value here.
Filed PR50584.
Peter Steinfeld [Fri, 4 Jun 2021 20:24:19 +0000 (13:24 -0700)]
[flang] Check for undefined derived types
It's possible to specify refer to an undefined derived type as the type of a
component of another derived type and then never define the type of the
component. We were not detecting this situation. To fix this, I
changed the value of isForwardReferenced_ in the symbol's
DerivedTypeDetails and checked for it when performing other derived type
checks.
I also had to record the fact that error messages were previously
emitted for the same problem in some cases so that I could avoid
duplicate messages.
I also added a test.
Differential Revision: https://reviews.llvm.org/D103714
Arnamoy Bhattacharyya [Fri, 4 Jun 2021 21:17:49 +0000 (17:17 -0400)]
[flang][OpenMP] Add semantic check to allow only loop iteration variables in a `linear` clause within `distribute` construct.
Implement the following semantic check:
"A list item may not appear in a linear clause, unless it is the loop iteration variable."
Reviewed By: kiranchandramohan
Differential Revision: https://reviews.llvm.org/D100224
Ten Tzen [Fri, 4 Jun 2021 21:04:46 +0000 (14:04 -0700)]
[Windows SEH]: Fix -O2 crash for Windows -EHa
This patch fixes a Windows -EHa crash induced by previous commit
797ad701522988e212495285dade8efac41a24d4.
The crash was caused by "LifetimeMarker" scope (with option -O2) that should not be considered as SEH Scope.
This change also turns off -fasync-exceptions by default under -EHa option for now.
Differential Revision: https://reviews.llvm.org/D103664#2799944
Terry Wilmarth [Mon, 10 May 2021 17:58:39 +0000 (12:58 -0500)]
[OpenMP] Add experimental nesting mode feature
Nesting mode is a new experimental feature in the OpenMP
runtime. It allows a user to set up nesting for an application in a
way that corresponds to the hardware topology levels on the machine an
application is being run on. For example, if a machine has 2 sockets,
each with 12 cores, then use of nesting mode could set up an outer
level of nesting that uses 2 threads per parallel region, and an inner
level of nesting that uses 12 threads per parallel region.
Nesting mode is controlled with the KMP_NESTING_MODE environment
variable as follows:
1) KMP_NESTING_MODE = 0: Nesting mode is off (default); max-active-levels-var
is set to 1 (the default -- nesting is off, nested parallel regions
are serialized).
2) KMP_NESTING_MODE = 1: Nesting mode is on, and a number of threads
will be assigned for each level discovered in the machine topology;
max-active-levels-var is set to the number of levels discovered.
3) KMP_NESTING_MODE = n, n>1: [Note: this option is experimental and may change
or be removed in the future.] Nesting mode is on, and a number of
threads will be assigned for each topology level discovered on the
machine, up to k<=n levels (since there may be fewer than n levels
discovered in the topology), and beyond the kth level, nested parallel
regions will be serialized; NOTE: max-active-levels-var is 1 (the default --
nesting is off, and nested parallel regions are serialized until the
user changes max-active-levels-var.
If the user sets OMP_NUM_THREADS or OMP_MAX_ACTIVE_LEVELS, they will
override KMP_NESTING_MODE settings for the associated environment
variables. The detected topology may be limited by an affinity mask
setting on the initial thread, or if the user sets KMP_HW_SUBSET. See
also: KMP_HOT_TEAMS_MAX_LEVEL for controlling use of hot teams for
nested parallel regions. Note that this feature only sets numbers of
threads used at nesting levels. The user should make use of
OMP_PLACES and OMP_PROC_BIND or KMP_AFFINITY for affinitizing those
threads, if desired.
Differential Revision: https://reviews.llvm.org/D102188
Kostya Kortchinsky [Fri, 4 Jun 2021 20:38:26 +0000 (13:38 -0700)]
[scudo] Fix String DCHECK
This resolves an issue tripping a `DCHECK`, as I was checking for the
capacity and not the size. We don't need to 0-init the Vector as it's
done already, and make sure we only 0-out the string on clear if it's
not empty.
Differential Revision: https://reviews.llvm.org/D103716
Nico Weber [Fri, 4 Jun 2021 20:41:35 +0000 (16:41 -0400)]
[gn build] (semi-manually) port
07c92b2e9581
Ahmed Taei [Fri, 4 Jun 2021 20:31:05 +0000 (13:31 -0700)]
Add memref.dim canonicalization patterns to TilingCanonicalizationPatterns
Otherwise tiled and padded linalg op will be alive (after distribution).
Differential Revision: https://reviews.llvm.org/D103715
Eli Friedman [Fri, 4 Jun 2021 20:33:29 +0000 (13:33 -0700)]
Regenerate a few tests related to SCEV.
In preparation for https://reviews.llvm.org/D103656
Fangrui Song [Fri, 4 Jun 2021 20:27:56 +0000 (13:27 -0700)]
[InstrProfiling] If no value profiling, make data variable private and (for Windows) use one comdat
`__profd_*` variables are referenced by code only when value profiling is
enabled. If disabled (e.g. default -fprofile-instr-generate), the symbols just
waste space on ELF/Mach-O. We change the comdat symbol from `__profd_*` to
`__profc_*` because an internal symbol does not provide deduplication features
on COFF. The choice doesn't matter on ELF.
(In -DLLVM_BUILD_INSTRUMENTED_COVERAGE=on build, there is now no `__profd_*` symbols.)
On Windows this enables further optimization. We are no longer affected by the
link.exe limitation: an external symbol in IMAGE_COMDAT_SELECT_ASSOCIATIVE can
cause duplicate definition error.
https://lists.llvm.org/pipermail/llvm-dev/2021-May/150758.html
We can thus use llvm.compiler.used instead of llvm.used like ELF (D97585).
This avoids many `/INCLUDE:` directives in `.drectve`.
Here is rnk's measurement for Chrome:
```
This reduced object file size of base_unittests.exe, compiled with coverage, optimizations, and gmlt debug info by 10%:
#BEFORE
$ find . -iname '*.obj' | xargs du -b | awk '{ sum += $1 } END { print sum}'
1047758867
$ du -cksh base_unittests.exe
82M base_unittests.exe
82M total
# AFTER
$ find . -iname '*.obj' | xargs du -b | awk '{ sum += $1 } END { print sum}'
937886499
$ du -cksh base_unittests.exe
78M base_unittests.exe
78M total
```
The change is NFC for Mach-O.
Reviewed By: davidxl, rnk
Differential Revision: https://reviews.llvm.org/D103372
Aaron Ballman [Fri, 4 Jun 2021 19:50:44 +0000 (15:50 -0400)]
Fix a diagnoses-valid bug with using declarations
The following was found by a customer and is accepted by the other primary
C++ compilers, but fails to compile in Clang:
namespace sss {
double foo(int, double);
template <class T>
T foo(T); // note: target of using declaration
} // namespace sss
namespace oad {
void foo();
}
namespace oad {
using ::sss::foo;
}
namespace sss {
using oad::foo; // note: using declaration
}
namespace sss {
double foo(int, double) { return 0; }
template <class T>
T foo(T t) { // error: declaration conflicts with target of using
return t;
}
} // namespace sss
I believe the issue is that MergeFunctionDecl() was calling
checkUsingShadowRedecl() but only considering a FunctionDecl as a
possible shadow and not FunctionTemplateDecl. The changes in this patch
largely mirror how variable declarations were being handled by also
catching FunctionTemplateDecl.
Vitaly Buka [Sun, 30 May 2021 00:18:17 +0000 (17:18 -0700)]
[scudo] Untag pointer in iterateOverChunks
Pointer comparison in Lambda will not work on tagged pointers.
Reviewed By: pcc
Differential Revision: https://reviews.llvm.org/D103496
Vitaly Buka [Tue, 25 May 2021 04:40:44 +0000 (21:40 -0700)]
[scudo] Add memtag_test
Differential Revision: https://reviews.llvm.org/D103074
Vitaly Buka [Fri, 4 Jun 2021 18:25:47 +0000 (11:25 -0700)]
[scudo] Remove disableMemoryTagChecksTestOnly
And replace with ScopedDisableMemoryTagChecks.
Differential Revision: https://reviews.llvm.org/D103708
Martin Storsjö [Fri, 4 Jun 2021 19:27:18 +0000 (22:27 +0300)]
[mailmap] Add the canonical spelling of my name
Nikita Popov [Mon, 31 May 2021 19:22:21 +0000 (21:22 +0200)]
[IndVars] Don't forget value when inferring nowrap flags
When SimplifyIndVars infers IR nowrap flags from SCEV, this may
happen in two ways: Either nowrap flags were already present in
SCEV and just get transferred to IR. Or zero/sign extension of
addrecs infers additional nowrap flags, and those get transferred
to IR. In the latter case, calling forgetValue() ensures that the
newly inferred nowrap flags get propagated to any other SCEV
expressions based on the addrec. However, the invalidation can
also have a major compile-time effect in some cases. For
https://bugs.llvm.org/show_bug.cgi?id=50384 with n=512 compile-
time drops from 7.1s to 0.8s without this invalidation. At the
same time, removing the invalidation doesn't affect any codegen
in test-suite.
Differential Revision: https://reviews.llvm.org/D103424
Rong Xu [Fri, 4 Jun 2021 18:02:11 +0000 (11:02 -0700)]
[SampleFDO] New hierarchical discriminator for FS SampleFDO (llvm-profdata part)
This patch was split from https://reviews.llvm.org/D102246
[SampleFDO] New hierarchical discriminator for Flow Sensitive SampleFDO
This is for llvm-profdata part of change. It sets the bit masks for the
profile reader in llvm-profdata. Also add an internal option
"-fs-discriminator-pass" for show and merge command to process the profile
offline.
This patch also moved setDiscriminatorMaskedBitFrom() to
SampleProfileReader::create() to simplify the interface.
Differential Revision: https://reviews.llvm.org/D103550