Arthur Eubanks [Fri, 20 Aug 2021 17:23:27 +0000 (10:23 -0700)]
[NFC] Remove unused CallBase::addDereferenceableOrNullAttr()
Patrick Holland [Fri, 20 Aug 2021 03:04:03 +0000 (20:04 -0700)]
[MCA] Fixing bug that was causing LSUnit not to realize an instruction finished executing when the instruction has 0 latency.
Differential Revision: https://reviews.llvm.org/D108443
Ivan Zhechev [Fri, 20 Aug 2021 17:14:59 +0000 (18:14 +0100)]
Make test_symbols.py compare files line-by-line
We currently feed full files to Python's unified_diff.
It's not quite what we want though -
line-by-line comparison makes more sense
(we want to be able to identify missing/unnecessary lines)
and is also easier to parse for humans.
This patch makes sure that we compare one line at a time.
This change pretties up the output formatting in the script.
Output before:
```
!DEF:/m/s/xINTENT(IN)(Implicit)ObjectEntityREAL(4)
!DEF:/m/s/yINTENT(INOUT)(Implicit)ObjectEntityREAL(4)
-!-D-E-F-:-f-o-o-b-a-r-
puresubroutines(x,y)bind(c)
!REF:/m/s/x
intent(in)::x
```
Proposed output after:
```
!DEF:/m/s/xINTENT(IN)(Implicit)ObjectEntityREAL(4)
!DEF:/m/s/yINTENT(INOUT)(Implicit)ObjectEntityREAL(4)
-!DEF:foobar
puresubroutines(x,y)bind(c)
!REF:/m/s/x
intent(in)::x
```
Reviewed By: Meinersbur, awarzynski
Differential Revision: https://reviews.llvm.org/D107954
Louis Dionne [Fri, 20 Aug 2021 15:39:46 +0000 (11:39 -0400)]
[libc++] Remove more test-suite workarounds for unsupported GCC versions
Differential Revision: https://reviews.llvm.org/D108466
Albion Fung [Fri, 20 Aug 2021 17:23:32 +0000 (13:23 -0400)]
[libc++][PowerPC] Fix a test case failure when compiled with libcxx
The test case is not ran unless libcxx is used, and a macro
may be undefined. This patch checks for the definition of the
macro before using it.
Fixes http://llvm.org/PR51430
Differential Revision: https://reviews.llvm.org/D108352
Jon Chesterfield [Fri, 20 Aug 2021 17:17:27 +0000 (18:17 +0100)]
Revert "[openmp][nfc] Refactor GridValues"
Failed a nvptx codegen test
This reverts commit
2a47a84b40115b01e03e4d89c1d47ba74beb7bf3.
Shoaib Meenai [Fri, 20 Aug 2021 16:20:28 +0000 (09:20 -0700)]
[cmake] Fix native tooling when cross-compiling on Linux
At least as of CMake 3.20.3, the CMake platform file for Linux doesn't
define the file type prefix and suffix variables, relying on them being
implicitly empty when they're unset. If we're cross-compiling targeting
Windows on a Linux machine, the values of these prefixes and suffixes
populated by the Windows platform file will still be set after including
the Linux platform file, so we'll incorrectly assume the ".exe" suffix
for the host machine. Explicitly unset the variables before including
the platform file, to prevent any previous values from leaking. Thanks
@beanz for suggesting the fix.
Reviewed By: beanz
Differential Revision: https://reviews.llvm.org/D108473
Arthur Eubanks [Thu, 19 Aug 2021 21:41:25 +0000 (14:41 -0700)]
[NFC] Simplify some CallBase attribute methods
Arthur Eubanks [Thu, 19 Aug 2021 21:38:06 +0000 (14:38 -0700)]
[NFC] Remove some unused functions
Andrea Di Biagio [Wed, 18 Aug 2021 19:35:57 +0000 (20:35 +0100)]
[X86][SchedModels] Fix missing ReadAdvance for MULX and ADCX/ADOX (PR51494)
Before this patch, instructions MULX32rm and MULX64rm were missing a ReadAdvance
for the implicit read of register EDX/RDX. This patch fixes the issue, and it
also introduces a new SchedWrite for the two variants of MULX. The general idea
behind this last change is to eventually decrease the number of InstRW in the
scheduling models.
This patch also adds a ReadAdvance for the implicit read of EFLAGS in ADCX/ADOX.
Differential Revision: https://reviews.llvm.org/D108372
Craig Topper [Fri, 20 Aug 2021 16:34:03 +0000 (09:34 -0700)]
[X86] Add missing __inline__ to functions in amxintrin.h
Sanjay Patel [Fri, 20 Aug 2021 16:20:04 +0000 (12:20 -0400)]
[AggressiveInstCombine] guard against applying instruction flags with constant folding
This is a minimized version of a crash reported in:
D108201
Thomas Lively [Fri, 20 Aug 2021 16:21:31 +0000 (09:21 -0700)]
[WebAssembly] Restore builtins and intrinsics for pmin/pmax
Partially reverts
85157c007903, which had removed these builtins and intrinsics
in favor of normal codegen patterns. It turns out that it is possible for the
patterns to be split over multiple basic blocks, however, which means that DAG
ISel is not able to select them to the pmin/pmax instructions. To make sure the
SIMD intrinsics generate the correct instructions in these cases, reintroduce
the clang builtins and corresponding LLVM intrinsics, but also keep the normal
pattern matching as well.
Differential Revision: https://reviews.llvm.org/D108387
Aart Bik [Thu, 19 Aug 2021 22:50:24 +0000 (15:50 -0700)]
[mlir][sparse][python] migrate more code from boilerplate into proper numpy land
The boilerplate was setting up some arrays for testing. To fully illustrate
python - MLIR potential, however, this data should also come from numpy land.
Reviewed By: bixia
Differential Revision: https://reviews.llvm.org/D108336
Louis Dionne [Fri, 20 Aug 2021 16:09:41 +0000 (12:09 -0400)]
[libc++][NFC] Fix minor errors and inconsistencies in the test suite
Thomas Lively [Fri, 20 Aug 2021 16:10:36 +0000 (09:10 -0700)]
[WebAssembly] Make shift values unsigned in wasm_simd128.h
On some platforms, negative shift values mean to shift in the opposite
direction, but this is not true with WebAssembly. To avoid confusion, make the
shift values in the shift intrinsics unsigned.
Differential Revision: https://reviews.llvm.org/D108415
Aaron Ballman [Fri, 20 Aug 2021 16:04:09 +0000 (12:04 -0400)]
Replace an unnecessary null check with an assert; NFC
Thomas Lively [Fri, 20 Aug 2021 15:56:51 +0000 (08:56 -0700)]
[WebAssembly] Add SIMD intrinsics using unsigned integers
For each SIMD intrinsic function that takes or returns a scalar signed integer
value, ensure there is a corresponding intrinsic that returns or an
unsigned value. This is a convenience for users who use -Wsign-conversion so
they don't have to insert explicit casts, especially when the intrinsic
arguments are integer literals that fit into the unsigned integer type but not
the signed type.
Differential Revision: https://reviews.llvm.org/D108412
Jon Chesterfield [Fri, 20 Aug 2021 15:41:25 +0000 (16:41 +0100)]
[openmp][nfc] Refactor GridValues
Remove redundant fields and replace pointer with virtual function
Of fourteen fields, three are dead and four can be computed from the
remainder. This leaves a couple of currently dead fields in place as
they are expected to be used from the deviceRTL shortly. Two of the
fields that can be computed are only used from codegen and require a
log2() implementation so are inlined into codegen instead.
This change leaves the new methods in the same location in the struct
as the previous fields for convenience at review.
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D108380
Corentin Jabot [Fri, 20 Aug 2021 15:10:53 +0000 (11:10 -0400)]
Make wide multi-character character literals ill-formed
This implements P2362, which has not yet been approved by the
C++ committee, but because wide-multi character literals are
implementation defined, clang might not have to wait for WG21.
This change is also being applied in C mode as the behavior is
implementation-defined in C as well and there's no benefit to
having different rules between the languages.
The other part of P2362, making non-representable character
literals ill-formed, is already implemented by clang
Aaron Ballman [Fri, 20 Aug 2021 15:08:58 +0000 (11:08 -0400)]
Use DeclContext::getNonTransparentContext(); NFC
Ben Shi [Tue, 17 Aug 2021 08:40:16 +0000 (16:40 +0800)]
[RISCV] Optimize add in the zba extension with SH*ADD
Optimize (add x, c) to (SH*ADD (c>>b), x) if c is not simm12
while (c>>b) is simm12 and c has b trailing zeros.
Reviewed By: luismarques
Differential Revision: https://reviews.llvm.org/D108193
Louis Dionne [Fri, 20 Aug 2021 14:17:21 +0000 (10:17 -0400)]
[libc++] Fix XFAIL annotation
The triple can sometimes be arm64-apple-macos, where the previous XFAIL
annotation wouldn't match (and hence the test would fail unexpectedly).
Kirill Stoimenov [Thu, 19 Aug 2021 18:11:10 +0000 (18:11 +0000)]
[asan] Implemented getAddressSanitizerParams used by the ASan callback optimization code.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D108397
Jacques Pienaar [Fri, 20 Aug 2021 14:01:54 +0000 (07:01 -0700)]
[mlir][ods] Skip adding TOC in doc gen when present
Enables adding a TOC in the description to be able to interleave
documentation before and after the TOC.
Alexander Potapenko [Fri, 20 Aug 2021 13:50:47 +0000 (15:50 +0200)]
[msan] Hotfix clang/test/CodeGen/sanitize-memory-disable.c
Because KMSAN is not supported on many architectures, explicitly build
the test with -target x86_64-linux-gnu.
Fixes the 'unsupported architecture' and 'unsupported operating system'
errors reported by the clang-armv7-quick (https://lab.llvm.org/buildbot#builders/171/builds/2595)
and llvm-clang-x86_64-sie-ubuntu-fast (https://lab.llvm.org/buildbot#builders/139/builds/9079)
builders.
Differential Revision: https://reviews.llvm.org/D108465
Sanjay Patel [Fri, 20 Aug 2021 13:27:26 +0000 (09:27 -0400)]
[CVP] add tests for unreachable switch default; NFC
Goes with the proposal at D106056.
Jeremy Morse [Fri, 20 Aug 2021 13:48:45 +0000 (14:48 +0100)]
[DebugInfo][InstrRef] Correctly ignore DBG_VALUE_LIST in InstrRef mode
This patch makes InstrRefBasedLDV "safe" to work with DBG_VALUE_LISTs. It
doesn't actually interpret them, but it recognises that they specify
variable locations and avoids propagating false locations, which is better
than the current state. Observe the attached tes
* We avoid propagating DBG_VALUE_LISTs into successor blocks, as they're
not "currently" supported,
* We don't propagate other variable locations across DBG_VALUE_LISTs,
because we know that the variable location is terminated by the
DBG_VALUE_LIST.
Differential Revision: https://reviews.llvm.org/D108143
Aaron Ballman [Fri, 20 Aug 2021 13:49:07 +0000 (09:49 -0400)]
Fix assertion when generating diagnostic for inline namespaces
When calculating the name to display for inline namespaces, we have
custom logic to try to hide redundant inline namespaces from the
diagnostic. Calculating these redundancies requires performing a lookup
in the parent declaration context, but that lookup should not try to
look through transparent declaration contexts, like linkage
specifications. Instead, loop up the declaration context chain until we
find a non-transparent context and use that instead.
This fixes PR49954.
Simon Pilgrim [Fri, 20 Aug 2021 13:30:48 +0000 (14:30 +0100)]
[CostModel][X86] Add costs for f32/f64 scalar and vector types.
The f16 half types are still pretty useless as we don't have it as a legal type (we treat them as i16 most of the time)
Simon Pilgrim [Fri, 20 Aug 2021 12:59:55 +0000 (13:59 +0100)]
MainSwitch::isValidSelectInst - don't dereference dyn_cast<> results.
We've already checked that the pointer isa<PHINode>, so we can use cast<Instruction> safely.
Fixes static analyser warning.
Simon Pilgrim [Fri, 20 Aug 2021 12:49:58 +0000 (13:49 +0100)]
ClangOffloadBundler - getCompatibleOffloadTargets - Fix unknown parameter Wdocumentation warnings. NFC.
Jeremy Morse [Fri, 20 Aug 2021 13:13:37 +0000 (14:13 +0100)]
[DebugInfo][InstrRef] Remove a faulty assertion
This patch removes an assertion, and adds a regression test showing why the
assertion is broken.
For context, LocIdx is a key/index number for machine locations, so that we
can describe locations as a single integer and ignore whether they're on
the stack, in registers or otherwise. Back when InstrRefBasedLDV was added,
I happened to bake in a "special" zero number for various reasons, which
Vedant identified as undesirable in this review comment:
https://reviews.llvm.org/D83047#inline-765495 . I subsequently removed that
special zero number, but it looks like I didn't delete this assertion at
the time, which assumes that a zero LocIdx is invalid.
The attached test shows that this assertion is reachable on valid code --
on x86 $rsp always gets the LocIdx number zero, and if you transfer a
variable value into it, InstrRefBasedLDV crashes on that assertion. The
code might be a bit wild to be storing variables to $rsp like that, however
we shouldn't crash on it.
Differential Revision: https://reviews.llvm.org/D108134
Alexander Potapenko [Tue, 17 Aug 2021 09:34:22 +0000 (11:34 +0200)]
[msan] Add support for disable_sanitizer_instrumentation attribute
Unlike __attribute__((no_sanitize("memory"))), this one will cause MSan
to skip the entire function during instrumentation.
Depends on https://reviews.llvm.org/D108029
Differential Revision: https://reviews.llvm.org/D108199
Guillaume Chatelet [Fri, 20 Aug 2021 13:09:35 +0000 (13:09 +0000)]
[libc] Align to 32B instead of 16B for optimized memcmp
Bjorn Pettersson [Wed, 18 Aug 2021 21:22:35 +0000 (23:22 +0200)]
[NewPM] Use parameterized syntax for a couple of more passes
A couple of passes that are parameterized in new-PM used different
pass names (in cmd line interface) while using the same pass class
name. This patch updates the PassRegistry to model pass parameters
more properly using PASS_WITH_PARAMS.
Reason for the change is to ensure that we have a 1-1 mapping
between class name and pass name (when disregarding the params).
With a 1-1 mapping it is more obvious which pass name to use in
options such as -debug-only, -print-after etc.
The opt -passes syntax is changed for the following passes:
early-cse-memssa => early-cse<memssa>
post-inline-ee-instrument => ee-instrument<post-inline>
loop-extract-single => loop-extract<single>
lower-matrix-intrinsics-minimal => lower-matrix-intrinsics<minimal>
This patch is not updating pass names in docs/Passes.rst. Not quite
sure what the status is for that document (e.g. when it comes to
listing pass paramters). It is only loop-extract-single that is
mentioned in Passes.rst today, out of the passes mentioned above.
Differential Revision: https://reviews.llvm.org/D108362
Louis Dionne [Fri, 20 Aug 2021 12:42:36 +0000 (08:42 -0400)]
[libc++] Update credits.txt per coment on D108263
Louis Dionne [Thu, 19 Aug 2021 16:39:16 +0000 (12:39 -0400)]
[libc++] Bypass calling exception-throwing functions in the dylib with -fno-exceptions
basic_string and vector currently have a hard dependency on the compiled
library because they need to call __vector_base_common::__throw_xxx(),
which are externally instantiated in the compiled library. That makes
sense when exceptions are enabled (because we're trying to localize the
exception-throwing code to the compiled library), but it doesn't really
make sense when exceptions are disabled, and the __throw_xxx functions
are just calling abort() anyways.
This patch simply overrides the __throw_xxx() functions so that they
don't rely on the compiled library when exceptions are disabled.
Differential Revision: https://reviews.llvm.org/D108389
Alexander Potapenko [Fri, 13 Aug 2021 12:17:41 +0000 (14:17 +0200)]
[clang][Codegen] Introduce the disable_sanitizer_instrumentation attribute
The purpose of __attribute__((disable_sanitizer_instrumentation)) is to
prevent all kinds of sanitizer instrumentation applied to a certain
function, Objective-C method, or global variable.
The no_sanitize(...) attribute drops instrumentation checks, but may
still insert code preventing false positive reports. In some cases
though (e.g. when building Linux kernel with -fsanitize=kernel-memory
or -fsanitize=thread) the users may want to avoid any kind of
instrumentation.
Differential Revision: https://reviews.llvm.org/D108029
Simon Pilgrim [Fri, 20 Aug 2021 11:41:38 +0000 (12:41 +0100)]
[clangd] detectClangPath() - remove (dead) return. NFC.
Simon Pilgrim [Fri, 20 Aug 2021 11:36:28 +0000 (12:36 +0100)]
MemProfilerPass::run - remove (dead) duplicate return. NFC.
Simon Pilgrim [Fri, 20 Aug 2021 11:32:12 +0000 (12:32 +0100)]
[AST] getDeclLocForCommentSearch - remove dead return. NFC.
Don't use an else-block as the previous if-block always returns, and remove the (now more obvious) dead return {}.
Florian Mayer [Fri, 20 Aug 2021 11:20:40 +0000 (12:20 +0100)]
Revert "[hwasan] do not check if freed pointer belonged to allocator."
This reverts commit
119146f8ae25c31ea630a15761a6fba6b7eb909c.
Simon Pilgrim [Fri, 20 Aug 2021 11:10:47 +0000 (12:10 +0100)]
[Sema] Remove dead return immediately after another return. NFC.
Denys Shabalin [Fri, 20 Aug 2021 10:35:09 +0000 (12:35 +0200)]
[mlir][linalg] Fix __repr__ implementation in const from opdsl
Reviewed By: gysit
Differential Revision: https://reviews.llvm.org/D108369
Tim Northover [Fri, 20 Aug 2021 10:16:07 +0000 (11:16 +0100)]
AArch64: don't form indexed paired ops if base reg overlaps operands.
The registers involved might not be identical, but can still overlap (e.g.
"str w0, [x0, #4]!").
Roman Lebedev [Fri, 20 Aug 2021 10:17:56 +0000 (13:17 +0300)]
[NFCI][SimplifyCFG] Rewrite `createUnreachableSwitchDefault()`
The only thing that function should do as per it's semantic,
is to ensure that the switch's default is a block consisting only of
an `unreachable` terminator.
So let's just create such a block and update switch's default
to point to it. There should be no need for all this weird dance
around predecessors/successors.
Jingu Kang [Wed, 14 Jul 2021 10:43:29 +0000 (11:43 +0100)]
[AArch64] Enable Upper bound unrolling universally
Differential Revision: https://reviews.llvm.org/D105996
Fraser Cormack [Tue, 17 Aug 2021 14:01:19 +0000 (15:01 +0100)]
[RISCV] Fix reporting of incorrect commutable operand indices
This patch fixes an issue where RISCV's `findCommutedOpIndices` would
incorrectly return the pseudo `CommuteAnyOperandIndex` as a commutable
operand index, rather than fixing a specific index.
Reviewed By: rogfer01
Differential Revision: https://reviews.llvm.org/D108206
Andrzej Warzynski [Fri, 20 Aug 2021 09:11:19 +0000 (09:11 +0000)]
Revert "[flang] Refine output file generation"
This reverts commit
fd21d1e198e381a2b9e7af1701044462b2d386cd.
The test added in this patch [1] is failing on Windows and causing the
Windows BuildBot [2] to fail. I don't see any obvious way to fix this,
so reverting in order to investigate.
[1] llvm-project/flang/test/Driver/output-paths.f90
[2] https://lab.llvm.org/buildbot/#/builders/172/builds/2077
Vladislav Vinogradov [Thu, 19 Aug 2021 13:28:16 +0000 (16:28 +0300)]
[mlir] Fix ControlFlowInterfaces implementation for Async dialect
* Add `RegionBranchTerminatorOpInterface` to `YieldOp`.
* Implement `getSuccessorEntryOperands` in `ExecuteOp`.
* Fix `getSuccessorRegions` implementation in `ExecuteOp`.
Reviewed By: ezhulenev
Differential Revision: https://reviews.llvm.org/D108373
Florian Mayer [Thu, 19 Aug 2021 15:52:45 +0000 (16:52 +0100)]
[hwasan] do not check if freed pointer belonged to allocator.
In that case it is very likely that there will be a tag mismatch anyway.
We handle the case that the pointer belongs to neither of the allocators
by getting a nullptr from allocator.GetBlockBegin.
Reviewed By: hctim, eugenis
Differential Revision: https://reviews.llvm.org/D108383
Vignesh Balasubramanian [Thu, 19 Aug 2021 07:16:07 +0000 (12:46 +0530)]
[OpenMP][OMPD]Code movement required for OMPD
These changes don't come under OMPD guard as it is a movement of existing code to capture parallel behavior correctly.
"Runtime Entry Points for OMPD" like "ompd_bp_parallel_begin" and "ompd_bp_parallel_begin" should be placed at the correct execution point for the debugging tool to access proper handles/data.
Without the below changes, in certain cases, debugging tool will pick the wrong parallel and task handle.
Reviewed By: @hbae
Differential Revision: https://reviews.llvm.org/D100366
Vladislav Vinogradov [Thu, 19 Aug 2021 14:33:58 +0000 (17:33 +0300)]
[mlir][NFC] Use explicit ::mlir namespace in mlir-tblgen generated code
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D108376
Jingu Kang [Fri, 20 Aug 2021 07:46:32 +0000 (08:46 +0100)]
Precommit test for D108204
Justas Janickas [Thu, 19 Aug 2021 14:43:32 +0000 (15:43 +0100)]
[OpenCL] Fix version reporting of C++ for OpenCL 2021
C++ for OpenCL version 2021 and later are expected to consist of a
major version number only. Therefore, a different constructor for
`VersionTuple` needs to be called when reporting language version.
Differential Revision: https://reviews.llvm.org/D108379
Yaron Keren [Fri, 20 Aug 2021 05:28:45 +0000 (08:28 +0300)]
[docs] Clarify how to run cmake and llvm-lit with Visual Studio addressing PR45978
Differential Revision: https://reviews.llvm.org/D108444
Sam McCall [Fri, 20 Aug 2021 06:49:25 +0000 (08:49 +0200)]
[AST] Avoid single-trip loop in ClangAttrEmitter
This triggers coverity warnings, see https://reviews.llvm.org/D107703
Andrzej Warzynski [Wed, 18 Aug 2021 16:14:09 +0000 (16:14 +0000)]
[flang] Refine output file generation
This patch refactors the file generation API in Flang's frontend driver.
It improves the layering between `CreateDefaultOutputFile`,
`CreateOutputFile` (`CompilerInstance` methods) and their various
clients.
List of changes:
* Rename `CreateOutputFile` as `CreateOutputFileImpl` and make it
private. This method is an implementation detail.
* Instead of passing an `std::error_code` out parameter into
`CreateOutputFileImpl`, have it return Expected<>. This is a bit shorter
and more idiomatic LLVM.
* Make `CreateDefaultOutputFile` (which calls `CreateOutputFileImpl`)
issue an error when file creation fails. The error code from
`CreateOutputFileImpl` is used to generate a meaningful diagnostic
message.
* Remove error reporting from `PrintPreprocessedAction::ExecuteAction`.
This is only for cases when output file generation fails. This is
handled in `CreateDefaultOutputFile` instead (see the previous point).
* Inline `AddOutputFile` into its only caller,
`CreateDefaultOutputFile`.
* Switch from `lvm::buffer_ostream` to `llvm::buffer_unique_ostream>`
for non-seekable output streams. This simplifies the logic in the driver
and was introduced for this very reason in [1]
* Moke sure that the diagnostics from the prescanner when running `-E`
(`PrintPreprocessedAction::ExecuteAction`) are printed before the actual
output is generated.
* Update comments, add test.
[1] https://reviews.llvm.org/D93260
Differential Revision: https://reviews.llvm.org/D108390
Guillaume Chatelet [Thu, 19 Aug 2021 20:58:15 +0000 (20:58 +0000)]
[libc] Add an optimized version for memcmp
Differential Revision: https://reviews.llvm.org/D108406
Kazu Hirata [Fri, 20 Aug 2021 06:34:22 +0000 (23:34 -0700)]
[DWARF] Remove parseListTableHeader (NFC)
The last use was removed on Oct 4, 2020 in commit
6d0be74af5555f7bc56ac72cbd98ff270fd1291b.
Sebastian Neubauer [Thu, 19 Aug 2021 11:55:03 +0000 (13:55 +0200)]
[AMDGPU] Fix too many constants with flat scratch
Prevent SIFoldOperands from creating SALU instructions with a constant
and a frame index. Previously, only one operand was checked to be a
frame index, leading to too many constants when flat scratch is enabled
and stack offsets are large.
Differential Revision: https://reviews.llvm.org/D108368
LLVM GN Syncbot [Fri, 20 Aug 2021 06:14:22 +0000 (06:14 +0000)]
[gn build] Port
e6407356ba00
Martin Storsjö [Tue, 10 Aug 2021 10:27:08 +0000 (13:27 +0300)]
[builtins] Move Windows/ARM frontends for fix/float functions into the individual source files
This avoids pulling in all of them if only one of them is needed
(if builtins are built without -ffunction-sections), and matches how
the similar aliases for AEABI are set up.
Differential Revision: https://reviews.llvm.org/D107815
Christian Sigg [Thu, 19 Aug 2021 07:43:54 +0000 (09:43 +0200)]
Update bazel examples.
Use the simplified repo rules from https://github.com/llvm/llvm-project/commit/934f084.
Reviewed By: GMNGeoffrey
Differential Revision: https://reviews.llvm.org/D108237
Anshil Gandhi [Fri, 20 Aug 2021 05:25:34 +0000 (23:25 -0600)]
[HIP] Allow target addr space in target builtins
This patch allows target specific addr space in target builtins for HIP. It inserts implicit addr
space cast for non-generic pointer to generic pointer in general, and inserts implicit addr
space cast for generic to non-generic for target builtin arguments only.
It is NFC for non-HIP languages.
Differential Revision: https://reviews.llvm.org/D102405
Daniel Stone [Fri, 20 Aug 2021 05:24:19 +0000 (22:24 -0700)]
libclc: Fix rounding during type conversion
The rounding during type conversion uses multiple conversions, selecting
between them to try to discover if rounding occurred. This appears to
not have been tested, since it would generate code of the form:
float convert_float_rtp(char x)
{
float r = convert_float(x);
char y = convert_char(y);
[...]
}
which will access uninitialised data. The idea appears to have been to
have done a char -> float -> char roundtrip in order to discover the
rounding, so do this.
Discovered by inspection.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed By: jvesely
Differential Revision: https://reviews.llvm.org/D81999
LLVM GN Syncbot [Fri, 20 Aug 2021 05:16:08 +0000 (05:16 +0000)]
[gn build] Port
642885710e20
Lang Hames [Fri, 20 Aug 2021 05:07:51 +0000 (15:07 +1000)]
[ORC] Add 'Async' suffix to ExecutorProcessControl::MemoryAccess methods.
This prevents the async methods (which shoud be overridden by subclasses) from
hiding the blocking helper methods, avoiding a lot of 'using MemoryAccess::...'
boilerplate.
Lang Hames [Thu, 19 Aug 2021 11:30:29 +0000 (21:30 +1000)]
[ORC] Introduce lookupAndRecordAddrs utility.
Accepts a vector of (SymbolStringPtr, ExecutorAddress*) pairs, looks up all the
symbols, then writes their address to each of the corresponding
ExecutorAddresses.
This idiom (looking up and recording addresses into a specific set of variables)
is used in MachOPlatform and the (temporarily reverted) ELFNixPlatform, and is
likely to be used in other places in the near future, so wrapping it in a
utility function should save us some boilerplate.
Siva Chandra Reddy [Fri, 20 Aug 2021 04:50:50 +0000 (04:50 +0000)]
[libc] Add bcmp to the windows config.
Anton Afanasyev [Thu, 19 Aug 2021 16:33:54 +0000 (19:33 +0300)]
[NFC][AggressiveInstCombine] Simplify code for shift truncation
Anton Afanasyev [Fri, 20 Aug 2021 03:33:19 +0000 (06:33 +0300)]
[Test][AggressiveInstCombine] Split shift tests to `shl`, `lshr` and `ashr`
Anton Afanasyev [Thu, 19 Aug 2021 06:53:13 +0000 (09:53 +0300)]
[Test][AggressiveInstCombine] Add test for arithmetic shift
Anshil Gandhi [Fri, 20 Aug 2021 02:50:36 +0000 (20:50 -0600)]
[Remarks] [AMDGPU] Emit optimization remarks for atomics generating hardware instructions
Produce remarks when atomic instructions are expanded into hardware instructions
in SIISelLowering.cpp. Currently, these remarks are only emitted for atomic fadd
instructions.
Differential Revision: https://reviews.llvm.org/D108150
Albion Fung [Fri, 20 Aug 2021 02:36:43 +0000 (21:36 -0500)]
[PowerPC] Implement cmplxl builtins
This patch implements the builtins for cmplxl by utilising
__builtin_complex. This builtin is implemented to match XL
functionality.
Differential revision: https://reviews.llvm.org/D107138
Rob Suderman [Fri, 20 Aug 2021 00:02:43 +0000 (17:02 -0700)]
[mlir][tosa] Support UInt8 inputs and outputs for tosa.rescale
Tosa rescale can contain uint8 types. Added support for these types
using an unrealized conversion cast. Optimistically it would be better to
use bitcast however it does not support unsigned integers.
Differential Revision: https://reviews.llvm.org/D108427
Maryam Benimmar [Tue, 17 Aug 2021 18:22:29 +0000 (14:22 -0400)]
[AIX][XCOFF] 64-bit relocation reading support
Support XCOFFDumper relocation reading support
This patch is part of D103696 partition
Reviewed By: daltenty, Helflym
Differential Revision: https://reviews.llvm.org/D104646
Jinsong Ji [Thu, 19 Aug 2021 17:24:13 +0000 (13:24 -0400)]
[CRT][LIT] split target_cflags using shlex
Update the commit to use shlex.split, as the win32 is not supported.
See https://reviews.llvm.org/D108329 for discussions.
Yonghong Song [Thu, 19 Aug 2021 22:23:09 +0000 (15:23 -0700)]
[DebugInfo] convert btf_tag attrs to annotations for DIComposite types
Clang patch D106614 added attribute btf_tag support. This patch
generates btf_tag annotations for DIComposite types.
Each btf_tag annotation is represented as a 2D array of
meta strings. Each record may have more than one
btf_tag annotations.
Differential Revision: https://reviews.llvm.org/D106615
Shilei Tian [Fri, 20 Aug 2021 00:49:16 +0000 (20:49 -0400)]
[OpenMP] Use `__kmpc_give_task` in `__kmp_push_task` when encountering a hidden helper task
This patch replaces the current implementation, overwrites `gtid` and `thread`,
with `__kmpc_give_task`.
Reviewed By: AndreyChurbanov
Differential Revision: https://reviews.llvm.org/D106977
Yonghong Song [Mon, 19 Jul 2021 06:43:48 +0000 (23:43 -0700)]
Reland [DebugInfo] generate btf_tag annotations for DIComposite types
Clang patch D106614 added attribute btf_tag support. This patch
generates btf_tag annotations for DIComposite types.
A field "annotations" is introduced to DIComposite, and the
annotations are represented as an DINodeArray, similar to
DIComposite elements. The following example illustrates
how annotations are encoded in IR:
distinct !DICompositeType(..., annotations: !10)
!10 = !{!11, !12}
!11 = !{!"btf_tag", !"a"}
!12 = !{!"btf_tag", !"b"}
Each btf_tag annotation is represented as a 2D array of
meta strings. Each record may have more than one
btf_tag annotations, as in the above example.
Reland with additional fixes for llvm/unittests/IR/DebugTypeODRUniquingTest.cpp.
Differential Revision: https://reviews.llvm.org/D106615
Jessica Paquette [Thu, 19 Aug 2021 22:49:38 +0000 (15:49 -0700)]
[GlobalISel] Add IRTranslator support for @llvm.lround.* -> G_LROUND
Translate the `@llvm.lround.*` family to G_LROUND via
`IRTranslator::translateSimpleIntrinsic`.
Differential Revision: https://reviews.llvm.org/D108418
Jessica Paquette [Thu, 19 Aug 2021 22:41:36 +0000 (15:41 -0700)]
[GlobalISel] Add a G_LROUND instruction
Meant to represent the `@llvm.lround.*` family.
Add the opcode, docs, and verification.
Differential Revision: https://reviews.llvm.org/D108417
Joachim Protze [Thu, 19 Aug 2021 21:28:51 +0000 (23:28 +0200)]
[libomptarget][amdcgn] Add build dependency for llvm-link and opt
D107156 and D107320 are not sufficient when OpenMP is built as llvm runtime
(LLVM_ENABLE_RUNTIMES=openmp) because dependencies only work within the same
cmake instance.
We could limit the dependency to cases where libomptarget/plugins are really
built. But compared to the whole llvm project, building openmp runtime is
negligible and postponing the build of OpenMP runtime after the dependencies
are ready seems reasonable.
The direct dependency introduced in D107156 and D107320 is necessary for the
case where OpenMP is built as llvm project (LLVM_ENABLE_PROJECTS=openmp).
Differential Revision: https://reviews.llvm.org/D108404
Fangrui Song [Thu, 19 Aug 2021 23:42:57 +0000 (16:42 -0700)]
Revert "[InstrProfiling] Make COFF use the ELF comdat scheme (drop link.exe compatibility)"
This reverts commit
fbb8e772ec501a1b71643db90e9c6445e17d7cac.
Accidentally pushed.
Amara Emerson [Wed, 18 Aug 2021 07:19:58 +0000 (00:19 -0700)]
[AArch64][GlobalISel] Add G_VECREDUCE fewerElements support for full scalarization.
For some reductions like G_VECREDUCE_OR on AArch64, we need to scalarize
completely if the source is <= 64b. This change adds support for that in
the legalizer. If the source has a pow-2 num elements, then we can do
a tree reduction using the scalar operation in the individual elements.
Otherwise, we just create a sequential chain of operations.
For AArch64, we only need to scalarize if the input is <64b. If it's great than
64b then we can first do a fewElements step to 64b, taking advantage of vector
instructions until we reach the point of scalarization.
I also had to relax the verifier checks for reductions because the intrinsics
support <1 x EltTy> types, which we lower to scalars for GlobalISel.
Differential Revision: https://reviews.llvm.org/D108276
Fangrui Song [Thu, 19 Aug 2021 23:38:32 +0000 (16:38 -0700)]
[InstrProfiling] Make COFF use the ELF comdat scheme (drop link.exe compatibility)
The COFF specific `DataReferencedByCode` complexity (D103372 D103717) is due to
a link.exe limitation: an external symbol in IMAGE_COMDAT_SELECT_ASSOCIATIVE is
not really dropped, so it can cause duplicate definition error.
Fangrui Song [Thu, 19 Aug 2021 23:36:58 +0000 (16:36 -0700)]
[test] Split icall.ll into comdat/nocomdat variants
darwin/aix don't support comdat. Using IR comdat is incorrect.
Sam Clegg [Wed, 18 Aug 2021 16:57:34 +0000 (12:57 -0400)]
[lld][WebAssembly] Handle weakly defined symbols in shared libraries.
In the case of weakly defined symbols in shared libraries we now
generate both an import and an export. The dynamic linker can then
choose how a winner from among all the shared libraries that define a
given symbol.
Previously any direct usage of a weakly defined symbol would use the
DSO-local definition (For example, even through there would be single
address for a weakly defined function, each DSO could end up directly
calling its local version).
Fixes: https://github.com/emscripten-core/emscripten/issues/13773
Differential Revision: https://reviews.llvm.org/D108413
Thomas Lively [Thu, 19 Aug 2021 23:23:47 +0000 (16:23 -0700)]
[WebAssembly] Make bitmask instructions return unsigned ints
Since they are bitmasks, it will be more common for them to be used and
potentially extended to 64-bit integers as unsigned values rather than signed
values.
Differential Revision: https://reviews.llvm.org/D108401
Amara Emerson [Thu, 19 Aug 2021 22:45:50 +0000 (15:45 -0700)]
[AArch64][GlobalISel] Fix miscompile of <16 x s8> G_EXTRACT_VECTOR_ELT.
When support for copying vector s8 lanes was added recently, this also
had the side effect of fixing a fallback for <16 x s8> extracts since
both used the same helper. However, there was a bug in another helper
to get the regclass for a specific FPR-native type, which was assigning
FPR16 to s8 instead of FPR8.
Kent Ross [Thu, 19 Aug 2021 23:12:37 +0000 (23:12 +0000)]
[libc++][NFC] Update and alphabetize CREDITS.TXT
Reviewed By: ldionne, #libc
Differential Revision: https://reviews.llvm.org/D108263
Kent Ross [Thu, 19 Aug 2021 23:10:47 +0000 (23:10 +0000)]
[libc++] [doc] Add issue tracking for spaceship operator<=> implementation
Add issue tracking and assignment for the implementation of P1614R2: The Mothership has Landed.
Reviewed By: cjdb, #libc, Mordante, Quuxplusone
Differential Revision: https://reviews.llvm.org/D107877
Kent Ross [Thu, 19 Aug 2021 23:06:52 +0000 (23:06 +0000)]
[libc++][NFC] Remove unused include in <compare>.
`<type_traits>` was included in the first iteration of `<compare>` when
it was created as a monolithic header, then never removed. Removing it
now is a beneficial no-op since it is not guaranteed by the standard
and is already included by all of its subheaders.
Reviewed By: cjdb, #libc, Quuxplusone
Differential Revision: https://reviews.llvm.org/D107801
Thomas Lively [Thu, 19 Aug 2021 23:00:07 +0000 (16:00 -0700)]
[WebAssembly] Add explicit casts to silence -Wc++11-narrowing
Yonghong Song [Thu, 19 Aug 2021 22:54:38 +0000 (15:54 -0700)]
Revert "[DebugInfo] generate btf_tag annotations for DIComposite types"
This reverts commit
2fded193e7a8fb5bd8fb339f00fd9de686390530.
Builtbot reports some test failures. Revert now so I can take time
to fix the issues.
Yonghong Song [Mon, 19 Jul 2021 06:43:48 +0000 (23:43 -0700)]
[DebugInfo] generate btf_tag annotations for DIComposite types
Clang patch D106614 added attribute btf_tag support. This patch
generates btf_tag annotations for DIComposite types.
A field "annotations" is introduced to DIComposite, and the
annotations are represented as an DINodeArray, similar to
DIComposite elements. The following example illustrates
how annotations are encoded in IR:
distinct !DICompositeType(..., annotations: !10)
!10 = !{!11, !12}
!11 = !{!"btf_tag", !"a"}
!12 = !{!"btf_tag", !"b"}
Each btf_tag annotation is represented as a 2D array of
meta strings. Each record may have more than one
btf_tag annotations, as in the above example.
Differential Revision: https://reviews.llvm.org/D106615
Thomas Lively [Thu, 19 Aug 2021 22:37:12 +0000 (15:37 -0700)]
[WebAssembly] More convert_low and promote_low codegen
The convert_low and promote_low instructions can widen the lower two lanes of a
four-lane vector, but we were previously scalarizing patterns that widened lanes
besides the low two lanes. The commit adds a shuffle to move the widened lanes
into the low lane positions so the convert_low and promote_low instructions can
be used instead of scalarizing.
Depends on D108266.
Differential Revision: https://reviews.llvm.org/D108341
Thomas Lively [Thu, 19 Aug 2021 22:24:28 +0000 (15:24 -0700)]
[WebAssembly] Pattern match SIMD convert_low and promote_low during ISel
Since the simplest DAG patterns for convert_low and promote_low instructions
involved v2i32, v2f32, v4i64, and v4f64 types, which are not legal in the
WebAssembly backend and would be eliminated by type legalization, we were
previously matching those patterns in a DAG combine before the type legalization
stage. However in cases where the vectors were wider than 128 bits, the patterns
we matched were not created until the type legalization stage when the wide
vectors were split up. Type legalization would continue to eliminate the illegal
types we were matching as well, so the code ended up scalarized.
To make the ISel for these instructions more robust, match the scalarized
patterns rather than the patterns containing illegal types. Add tests with
double-wide vectors to show that this works as intended.
Fixes PR51098.
Depends on D107502.
Differential Revision: https://reviews.llvm.org/D108266