Jacob Hegna [Wed, 12 Oct 2022 02:05:19 +0000 (02:05 +0000)]
[MLGO] Force persistency in tflite buffers.
When training large models, we encounter use-after-free bugs when
writing to the input tensors for various MLGO models. This patch fixes the
issue by marking the tensors as "persistent".
Differential Revision: https://reviews.llvm.org/D135739
Krzysztof Parzyszek [Wed, 12 Oct 2022 18:35:59 +0000 (11:35 -0700)]
[Hexagon] Implement TLI::isExtractSubvectorCheap hook
Florian Hahn [Wed, 12 Oct 2022 19:38:30 +0000 (20:38 +0100)]
[ConstraintElim] Move GEP decomposition code to separate fn (NFC).
Breaks up a large function and allows for the use to early exits.
Dan Albert [Thu, 6 Oct 2022 22:22:55 +0000 (15:22 -0700)]
[LLD] Enable --no-undefined-version by default.
Allowing incorrect version scripts is not a helpful default. Flip that
to help users find their bugs at build time rather than at run time.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D135402
Martin Storsjö [Sat, 1 Oct 2022 15:49:32 +0000 (18:49 +0300)]
[AArch64] Generate SEH info for PAC instructions
Without this, unwinding through functions that does use PAC
would fail, if PAC actually was active.
Differential Revision: https://reviews.llvm.org/D135103
Martin Storsjö [Wed, 12 Oct 2022 11:27:22 +0000 (14:27 +0300)]
[AArch64] [SEH] Rename pac_sign_return_address to pac_sign_lr
This new opcode was initially documented as "pac_sign_return_address"
in https://github.com/MicrosoftDocs/cpp-docs/pull/4202, but was
soon afterwards renamed into "pac_sign_lr" in
https://github.com/MicrosoftDocs/cpp-docs/pull/4209, as the other
name was unwieldy, and there were no other external references to
that name anywhere.
Rename our external .seh assembler directive - it hasn't been merged
for very long yet, so there's probably no external use to account for.
Rename all other internal references to the opcode similarly.
Differential Revision: https://reviews.llvm.org/D135762
Valentin Clement [Wed, 12 Oct 2022 19:14:46 +0000 (12:14 -0700)]
[flang] Fix after introduction of ConditionallySpeculatable
86771d0b65ee introduced some fialure for flang
https://buildkite.com/llvm-project/premerge-checks/builds/116406#
0183ca8c-bc14-4314-9b1d-
c764cf47e0aa
Update FIROps.td to fix the issues.
Joseph Huber [Wed, 12 Oct 2022 16:31:46 +0000 (11:31 -0500)]
[Clang] Do not crash when an invalid offload architecture is set
If an invalid architecture is set we currently return an empty string.
This will cause the offloading toolchain to continue to be built and
hit an assertion elsewhere due to the invalid architecture. This patch
fixes that so we now correctly exit.
Reviewed By: tra
Differential Revision: https://reviews.llvm.org/D135791
Louis Dionne [Wed, 5 Oct 2022 18:03:31 +0000 (14:03 -0400)]
[libc++] Add test for bug that had been introduced in D98573 and fixed in D119633
Differential Revision: https://reviews.llvm.org/D135297
Arthur Eubanks [Wed, 12 Oct 2022 18:58:01 +0000 (11:58 -0700)]
[gn build] Set _LIBCPP_ABI_NAMESPACE to __$libcxx_abi_version
Matches CMake build.
Otherwise currently everything is literally std::_LIBCPP_ABI_NAMESPACE::foo.
Fangrui Song [Wed, 12 Oct 2022 18:55:26 +0000 (11:55 -0700)]
[Frontend] Recognize environment variable SOURCE_DATE_EPOCH
See https://reproducible-builds.org/docs/source-date-epoch/ . The environment
variable ``SOURCE_DATE_EPOCH`` been recognized by many compilers.
In GCC, if `SOURCE_DATE_EPOCH` is set, it specifies a UNIX timestamp to be used
in replacement of the current date and time in the `__DATE__` and `__TIME__`
macros. Note: GCC as of today does not update `__TIMESTAMP__` (the modification
time of the current source file) but
https://wiki.debian.org/ReproducibleBuilds/TimestampsFromCPPMacros expresses the
intention to update it.
This patches parses SOURCE_DATE_EPOCH and changes all the three macros.
In addition, in case gmtime/localtime returns null (e.g. on 64-bit Windows
gmtime returns null when the timestamp is larger than
32536850399
(3001-01-19T21:59:59Z)), use `??? ?? ????` as used by GCC.
Reviewed By: ychen
Differential Revision: https://reviews.llvm.org/D135045
Fangrui Song [Wed, 12 Oct 2022 18:52:05 +0000 (11:52 -0700)]
[llvm-objcopy] Support --decompress-debug-sections when zlib is disabled
When zlib is disabled at build time, the diagnostic `LLVM was not compiled with
LLVM_ENABLE_ZLIB: cannot decompress` for --decompress-debug-sections may be
inaccurate: if zstd is enabled, we should still support zstd decompression.
It's not useful to test zlib and zstd. Just remove the diagnostic and add a new
one before `compression::decompress`.
This fixes compress-debug-sections-zstd.test
Reviewed By: mariusz-sikora-at-amd, jhenderson, phosek
Differential Revision: https://reviews.llvm.org/D135744
Florian Hahn [Wed, 12 Oct 2022 18:49:29 +0000 (19:49 +0100)]
[ConstraintElimination] Add tests with 128 bit pointers.
Nikolas Klauser [Wed, 12 Oct 2022 13:45:09 +0000 (15:45 +0200)]
[libc++][NFC] Remove __lcpp prefix in math.h
Reviewed By: ldionne, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D135774
Mark de Wever [Sun, 2 Oct 2022 17:42:46 +0000 (19:42 +0200)]
[NFC][libc++] Moves transitive includes location.
This moves some includes missed in D133212.
Reviewed By: ldionne, #libc, philnik
Differential Revision: https://reviews.llvm.org/D135180
Nikolas Klauser [Wed, 12 Oct 2022 13:41:22 +0000 (15:41 +0200)]
[libc++][NFC] Replace _LIBCPP_INLINE_VISIBILITY in math.h
Reviewed By: ldionne, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D135773
Peter Rong [Wed, 12 Oct 2022 18:26:57 +0000 (11:26 -0700)]
[PowerPC] Pre-commit unit test change for D132978
Volodymyr Sapsai [Wed, 5 Oct 2022 00:25:03 +0000 (17:25 -0700)]
[modules] Allow to validate system headers less often with `-fmodules-validate-once-per-build-session`.
Make flags `-fmodules-validate-system-headers` and
`-fmodules-validate-once-per-build-session` orthogonal, so they have
their own independent responsibilities - if system headers should be
validated and how often.
rdar://
87988889
Differential Revision: https://reviews.llvm.org/D135232
Sanjoy Das [Thu, 6 Oct 2022 22:57:25 +0000 (15:57 -0700)]
Introduce a ConditionallySpeculatable op interface
This patch takes the first step towards a more principled modeling of undefined behavior in MLIR as discussed in the following discourse threads:
1. https://discourse.llvm.org/t/semantics-modeling-undefined-behavior-and-side-effects/4812
2. https://discourse.llvm.org/t/rfc-mark-tensor-dim-and-memref-dim-as-side-effecting/65729
This patch in particular does the following:
1. Introduces a ConditionallySpeculatable OpInterface that dynamically determines whether an Operation can be speculated.
2. Re-defines `NoSideEffect` to allow undefined behavior, making it necessary but not sufficient for speculation. Also renames it to `NoMemoryEffect`.
3. Makes LICM respect the above semantics.
4. Changes all ops tagged with `NoSideEffect` today to additionally implement ConditionallySpeculatable and mark themselves as always speculatable. This combined trait is named `Pure`. This makes this change NFC.
For out of tree dialects:
1. Replace `NoSideEffect` with `Pure` if the operation does not have any memory effects, undefined behavior or infinite loops.
2. Replace `NoSideEffect` with `NoSideEffect` otherwise.
The next steps in this process are (I'm proposing to do these in upcoming patches):
1. Update operations like `tensor.dim`, `memref.dim`, `scf.for`, `affine.for` to implement a correct hook for `ConditionallySpeculatable`. I'm also happy to update ops in other dialects if the respective dialect owners would like to and can give me some pointers.
2. Update other passes that speculate operations to consult `ConditionallySpeculatable` in addition to `NoMemoryEffect`. I could not find any other than LICM on a quick skim, but I could have missed some.
3. Add some documentation / FAQs detailing the differences between side effects, undefined behavior, speculatabilty.
Reviewed By: rriddle, mehdi_amini
Differential Revision: https://reviews.llvm.org/D135505
Sam McCall [Tue, 20 Sep 2022 19:39:47 +0000 (21:39 +0200)]
[AST] Preserve more structure in UsingEnumDecl node.
- store NestedNameSpecifier & Loc for the qualifiers
This information was entirely missing from the AST.
- expose the location information for qualifier/identifier/typedefs as typeloc
This allows many traversals/astmatchers etc to handle these generically along
with other references. The decl vs type split can help preserve typedef
sugar when https://github.com/llvm/llvm-project/issues/57659 is resolved.
- fix the SourceRange of UsingEnumDecl to include 'using'.
Fixes https://github.com/clangd/clangd/issues/1283
Differential Revision: https://reviews.llvm.org/D134303
Han Zhu [Wed, 12 Oct 2022 06:58:27 +0000 (23:58 -0700)]
[X86] Pre-commit more unit tests for D134477
Benjamin Maxwell [Wed, 12 Oct 2022 11:25:51 +0000 (11:25 +0000)]
Add test to show missed optimization for masked load/stores
This test shows instcombine failing to remove a alloca and memcpy for
for a constant array that is read with a masked load.
This will be addressed in a subsequent commit.
Mingming Liu [Mon, 5 Sep 2022 02:30:28 +0000 (19:30 -0700)]
[AArch64] Pre-commit test case to show sub-optimal codegen for Github issue #57502
Pre-commit test cases to show cases when UZP1 (TRUNC, TRUNC) could be
combined into TRUNC (UZP1) (with some proper bit conversions in the middle) to generate more efficient code.
Differential Revision: https://reviews.llvm.org/D133280
Nathan Sidwell [Fri, 7 Oct 2022 14:03:32 +0000 (07:03 -0700)]
[libcxxabi]: Resync llvm & libcxxabi's demangler
Sadly the demangler copies have diverged. This resyncs them by
a) pulling the meaningful llvm changes into libcxxabi's gold copy.
b) rerunning the sync script.
I notice uses of placement new, which assume the allocator succeeds --
that's incorrect in general, but an orthogonal problem.
Reviewed By: bader
Differential Revision: https://reviews.llvm.org/D135446
Arthur Eubanks [Tue, 11 Oct 2022 18:14:53 +0000 (11:14 -0700)]
[PrintPipeline] Handle CoroConditionalWrapper and add more verification
Add a check (can be disabled via a flag) that the pipeline we generate is actually parsable.
Can be disabled because we don't expect to handle every pass in -print-pipeline-passes.
Fixes #58280.
Reviewed By: ChuanqiXu
Differential Revision: https://reviews.llvm.org/D135703
Mehdi Amini [Thu, 6 Oct 2022 19:23:10 +0000 (19:23 +0000)]
Apply clang-tidy fixes for modernize-use-using in SparseBufferRewriting.cpp (NFC)
Mehdi Amini [Thu, 6 Oct 2022 19:22:18 +0000 (19:22 +0000)]
Apply clang-tidy fixes for performance-unnecessary-copy-initialization in CodegenUtils.cpp (NFC)
Sanjay Patel [Wed, 12 Oct 2022 15:36:27 +0000 (11:36 -0400)]
[InstCombine] add test for udiv with shl divisor; NFC
This would solve an example from issue #58137 more
generally, but it may require adding a canonicalization
for shift + shift to shift + add.
Mark de Wever [Sat, 6 Aug 2022 08:47:53 +0000 (10:47 +0200)]
[libc++] Implements constexpr <charconv>.
Implements:
- P2291R3 Add Constexpr Modifiers to Functions to_chars and from_chars for
Integral Types in <charconv> Header
Reviewed By: #libc, ldionne
Differential Revision: https://reviews.llvm.org/D131317
Mark de Wever [Wed, 5 Oct 2022 17:54:44 +0000 (19:54 +0200)]
[libc++][format] Improves CTAD.
This partly reverts D133535 and enables CTAD for more parts in format.
Reviewed By: ldionne, #libc
Differential Revision: https://reviews.llvm.org/D135292
Nicolas Vasilache [Wed, 12 Oct 2022 12:52:30 +0000 (05:52 -0700)]
[mlir][Linalg] Quarantine usage of LinalgTransformationFilter to TestTilingInterface.
This revision also retires code that has now become dead.
Context: https://discourse.llvm.org/t/psa-retire-linalg-filter-based-patterns/63785
Differential Revision: https://reviews.llvm.org/D135771
Guillaume Chatelet [Wed, 12 Oct 2022 15:35:18 +0000 (15:35 +0000)]
Revert "[libc] New version of the mem* framework"
This reverts commit
4c19439d249256db720e323a446e39d05496732f.
Mirko Brkusanin [Wed, 12 Oct 2022 15:20:46 +0000 (17:20 +0200)]
[SelectionDAG] Use consistent type sizes for opcode
Sanjay Patel [Wed, 12 Oct 2022 15:18:55 +0000 (11:18 -0400)]
[InstCombine] fold sdiv with common shl amount in operands
(X << Z) / (Y << Z) --> X / Y
https://alive2.llvm.org/ce/z/CLKzqT
This requires a surprising "nuw" constraint because we have
to guard against immediate UB via signed-div overflow with
-1 divisor.
This extends
008a89037a49ca0d9 and is another transform
derived from issue #58137.
Guillaume Chatelet [Tue, 4 Oct 2022 08:49:02 +0000 (08:49 +0000)]
[libc] New version of the mem* framework
This version is more composable and also simpler at the expense of being more explicit and more verbose.
This patch is not meant to be submitted but gives an idea of the change.
Codegen can be checked in https://godbolt.org/z/6z1dEoWbs by removing the "static inline" before individual functions.
Unittests are coming.
Suggested review order:
- utils
- op_base
- op_builtin
- op_generic
- op_x86 / op_aarch64
- *_implementations.h
Differential Revision: https://reviews.llvm.org/D135134
Alexey Bataev [Mon, 10 Oct 2022 14:40:35 +0000 (07:40 -0700)]
[SLP]Fix insertpoint of the extractellements instructions to avoid reshuffle crash.
Need to set the insertpoint for extractelement to point to the first
instruction in the node to avoid possible crash during external uses
combine process. Without it we may endup with the incorrect
transformation.
Differential Revision: https://reviews.llvm.org/D135591
Andre Vieira [Wed, 12 Oct 2022 15:12:23 +0000 (16:12 +0100)]
[libc] Update benchmarks/README.md
This patch updates the instructions on how to run the mem* benchmarks in
README.md.
Differential Revision: https://reviews.llvm.org/D135582
Sanjay Patel [Wed, 12 Oct 2022 14:58:03 +0000 (10:58 -0400)]
[InstCombine] fold udiv with common shl amount in operands
(X << Z) / (Y << Z) --> X / Y
https://alive2.llvm.org/ce/z/E5eaxU
This fixes the motivating example from issue #58137,
but it is not the most general transform. We should
probably also convert left-shift in the divisor to
right-shift in the dividend for that, but that exposes
another missed canonicalization for shifts and adds.
Nikita Popov [Wed, 12 Oct 2022 14:59:42 +0000 (16:59 +0200)]
[FunctionAttrs] Regenerate test checks (NFC)
Jordan Rupprecht [Wed, 12 Oct 2022 15:11:10 +0000 (08:11 -0700)]
[NFC] Ignore unused var in no-asserts builds
Simon Pilgrim [Wed, 12 Oct 2022 14:57:45 +0000 (15:57 +0100)]
[UpdateTestChecks] Add basic BPF triple handling
Working on Issue #57872 - its really useful to be able to autogenerate checks
Louis Dionne [Wed, 5 Oct 2022 15:35:53 +0000 (11:35 -0400)]
[libc++] Remove _LIBCPP_HAS_NO_CXX20_COROUTINES
As far as I can tell, all the compilers we support have support for
C++20 coroutines now.
Differential Revision: https://reviews.llvm.org/D135274
Louis Dionne [Wed, 12 Oct 2022 14:40:06 +0000 (10:40 -0400)]
[libc++] Disable the GDB pretty printer tests again, since they break the CI
Nikita Popov [Wed, 12 Oct 2022 12:34:03 +0000 (14:34 +0200)]
[CGStmt] Use helper functions to set memory attributes (NFC)
Alexey Bataev [Thu, 6 Oct 2022 18:01:54 +0000 (11:01 -0700)]
[SLP]Fix PR58177: Improve isUndefVector function to avoid extra freeze.
Freeze instruction in some cases makes codegen worse, so need to be very
careful when emitting it. Instead improve analysis in isUndefVector
function to generate mask of unused elements and use it in the analysis.
Differential Revision: https://reviews.llvm.org/D135382
Nicolas Vasilache [Wed, 12 Oct 2022 12:21:52 +0000 (05:21 -0700)]
[mlir][Linalg] Drop filter-based splitReduction
This transformation is available and tested via the transform dialect.
Differential Revision: https://reviews.llvm.org/D135767
Nicolas Vasilache [Wed, 12 Oct 2022 12:06:18 +0000 (05:06 -0700)]
[mlir][Linalg] NFC - Drop LinalgLoweringPattern
Linalg ops are generally lowered to loops using a pass and the individual application
of patterns has no concrete uses.
If/when necessary, this can later be reintroduced via a proper transform dialect op.
Differential Revision: https://reviews.llvm.org/D135766
Kai Nacke [Wed, 12 Oct 2022 14:10:05 +0000 (14:10 +0000)]
[SystemZ] Revert migration of SystemZ/Large/branch-01.ll
Root cause for the regression is that value %6598 is
defined as a label, and later as an i32 value.
Reviewed By: uweigand
Differential Revision: https://reviews.llvm.org/D135778
Florian Hahn [Wed, 12 Oct 2022 14:21:38 +0000 (15:21 +0100)]
[LV] Add extra tests for epilogue vectorization with widened inductions.
Extend test coverage to also include inductions with step > 1 and also
with runtime trip counts.
Valentin Clement [Wed, 12 Oct 2022 14:18:18 +0000 (16:18 +0200)]
[flang] Use fir.rebox for tbp fir.dispatch call with allocatable or pointer
Polymorphic entities with allocatable or pointer attribute cannot be passed
directly as passed-object when the type-bound procedure is expecting a
simply dummy polymorphic entity. Use fir.rebox to adapt the fir.class box
to the tbp type.
Depends on D135649
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D135654
Tobias Gysi [Wed, 12 Oct 2022 14:12:21 +0000 (17:12 +0300)]
[mlir] Fix bitwidth of memref-to-llvm constant.
One constant generated in MemRefToLLVM had a hardcoded bitwidth of
64 bits. The fix uses the typeConverter to create a constant that
matches the bitwidth of the provided by the data layout. The issue was
detected in an attempt to add a verifier to the LLVM ICmp operation that
checks that the types of the compared arguments match.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D135775
Louis Dionne [Mon, 19 Sep 2022 20:17:02 +0000 (16:17 -0400)]
[runtimes] Always define cxx_shared, cxx_static & other targets
However, mark them as EXCLUDE_FROM_ALL when we don't want to build them.
Simply declaring the targets should be of no harm, and it allows other
projects to mention these targets regardless of whether they end up
being built or not.
While the diff may not make that obvious, this patch basically
moves the definition of e.g. `cxx_shared` out of the `if (LIBCXX_ENABLE_SHARED)`
and instead marks it as `EXCLUDE_FROM_ALL` conditionally on whether
LIBCXX_ENABLE_SHARED is passed. It then does the same for libunwind
and libc++abi targets.
Differential Revision: https://reviews.llvm.org/D134221
gonglingqin [Wed, 12 Oct 2022 13:11:29 +0000 (21:11 +0800)]
[LoongArch] Handle missing CondCodes
Support SETLE/SETEQ and expand SETGE/SETNE/SETGT
Differential Revision: https://reviews.llvm.org/D135511
Sanjay Patel [Wed, 12 Oct 2022 13:05:34 +0000 (09:05 -0400)]
[InstCombine] propagate "exact" through folds of div
These folds were added recently with:
6b869be8100d
8da2fa856f1b
...but they didn't account for the "exact" attribute,
and that can be safely propagated:
https://alive2.llvm.org/ce/z/F_WhnR
https://alive2.llvm.org/ce/z/ft9Cgr
Sanjay Patel [Wed, 12 Oct 2022 12:53:39 +0000 (08:53 -0400)]
[InstCombine] add "exact" to div tests to show missing propagation; NFC
Sanjay Patel [Wed, 12 Oct 2022 12:50:13 +0000 (08:50 -0400)]
[InstCombine] add helper function for div+shl folds; NFC
There are at least 2 similar patterns that could be added here,
and the existing fold can be improved because it fails to
propagate "exact".
Sanjay Patel [Tue, 11 Oct 2022 21:06:01 +0000 (17:06 -0400)]
[InstCombine] add tests for div with shl operands; NFC
Valentin Clement [Wed, 12 Oct 2022 13:24:21 +0000 (15:24 +0200)]
[flang] Lower type-bound procedure call needing dynamic dispatch to fir.dispatch
Lower call with polymorphic entities to fir.dispatch operation. This patch only
focus one lowering with simple scalar polymorphic entities. A follow-up patch
will deal with allocatble, pointer and array of polymorphic entities as they
require box manipulation for the passed-object.
Reviewed By: jeanPerier
Differential Revision: https://reviews.llvm.org/D135649
gonglingqin [Wed, 12 Oct 2022 12:43:22 +0000 (20:43 +0800)]
[LoongArch] Add earlyclobber of destination register to atomic instructions
If the AM* atomic memory access instruction has the same register number as
rd and rj, the execution will trigger an Instruction Non-defined Exception.
If the AM* atomic memory access instruction has the same register number as
rd and rk, the execution result is uncertain.
Reference: https://github.com/loongson/LoongArch-Documentation
Differential Revision: https://reviews.llvm.org/D135641
Arseniy Zaostrovnykh [Wed, 12 Oct 2022 12:46:32 +0000 (14:46 +0200)]
Fix false positive related to handling of [[noreturn]] function pointers
Before this change, the `NoReturnFunctionChecker` was missing function pointers
with a `[[noreturn]]` attribute, while `CFG` was constructed taking that into
account, which leads CSA to take impossible paths. The reason was that the
`NoReturnFunctionChecker` was looking for the attribute in the type of the
entire call expression rather than the type of the function being called.
This change makes the `[[noreturn]]` attribute of a function pointer visible
to `NoReturnFunctionChecker`. This leads to a more coherent behavior of the
CSA on the AST involving.
Reviewed By: xazax.hun
Differential Revision: https://reviews.llvm.org/D135682
Peixin Qiao [Wed, 12 Oct 2022 12:25:34 +0000 (20:25 +0800)]
[flang][NFC] Fix FIRTypes.td
Fix two typos and remove the incorrectly added `extraClassDeclaration` in
`FIRTypes.td`.
Reviewed By: clementval
Differential Revision: https://reviews.llvm.org/D135665
Chen Zheng [Wed, 12 Oct 2022 12:15:15 +0000 (12:15 +0000)]
[PowerPC][NFC] refactor some test cases.
Tonko Sabolčec [Wed, 12 Oct 2022 11:46:23 +0000 (11:46 +0000)]
[lldb] Fix member access in GetExpressionPath
This change fixes two issues in ValueObject::GetExpressionPath method:
1. Accessing members of struct references used to produce expression
paths such as "str.&str.member" (instead of the expected
"str.member"). This is fixed by assigning the flag tha the child
value is a dereference when calling Dereference() on references
and adjusting logic in expression path creation.
2. If the parent of member access is dereference, the produced
expression path was "*(ptr).member". This is incorrect, since it
dereferences the member instead of the pointer. This is fixed by
wrapping dereference expression into parenthesis, resulting with
"(*ptr).member".
Reviewed By: werat, clayborg
Differential Revision: https://reviews.llvm.org/D132734
Adrian Kuegel [Wed, 12 Oct 2022 12:07:31 +0000 (14:07 +0200)]
Adrian Kuegel [Wed, 12 Oct 2022 08:18:38 +0000 (10:18 +0200)]
[mlir][Linalg] Add MapOp to Linalg structured ops.
This will allow to model elementwise ops with this special op instead of using
GenericOp.
Also allow MapOp and ReduceOp to have no result if the output type is not a tensor.
This is needed for buffer semantics.
Differential Revision: https://reviews.llvm.org/D135754
Florian Hahn [Wed, 12 Oct 2022 11:47:18 +0000 (12:47 +0100)]
[VPlan] Remove dead recipes before sinking.
optimizeInductions may leave dead recipes which can prevent sinking.
Sinking on the other hand should not introduce new dead recipes, so
clean up dead recipes before sinking.
Reviewed By: Ayal
Differential Revision: https://reviews.llvm.org/D133762
Nicolas Vasilache [Wed, 12 Oct 2022 11:14:33 +0000 (04:14 -0700)]
[mlir][Linalg] NFC - Drop filter from LinalgGeneralizationPattern
Differential Revision: https://reviews.llvm.org/D135761
Nicolas Vasilache [Wed, 12 Oct 2022 10:26:55 +0000 (03:26 -0700)]
[mlir][Linalg] NFC - Drop CodegenStrategy and remaining strategy passes
Differential Revision: https://reviews.llvm.org/D135759
Tobias Gysi [Wed, 12 Oct 2022 11:23:10 +0000 (14:23 +0300)]
[mlir][llvm] Tablegen based operation import from LLVM IR.
The revision uses tablegen generated builders to convert the most common
LLVM IR instructions to MLIR LLVM dialect operations. All instructions
with special handlers, except for alloca and fence, still use manual
handlers. The revision also introduces an additional "instructions.ll"
test file to test the import of instructions that have tablegen builders
(except for the resume instruction whose test remains untouched). A part
of the test cases are new, for example the integer instruction test,
while others are migrated from the "basic.ll" test file.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D135709
Max Kazantsev [Wed, 12 Oct 2022 11:14:06 +0000 (18:14 +0700)]
[NFC] Perform all legality checks for non-trivial unswitch in one function
They have been scattered over the code. For better structuring, perform
them in one place. Potential CT drop is possible because we collect exit
blocks twice, but it's small price to pay for much better code structure.
Luo, Yuanke [Wed, 12 Oct 2022 10:42:32 +0000 (18:42 +0800)]
Don't widen shuffle element with AVX512
Fix crash issue of D129537 and reopen it.
Currently the X86 shuffle lowering would widen the element type for
shuffle if the mask element value is adjacent. For below example
%t2 = add nsw <16 x i32> %t0, %t1
%t3 = sub nsw <16 x i32> %t0, %t1
%t4 = shufflevector <16 x i32> %t2, <16 x i32> %t3,
<16 x i32> <i32 16, i32 17, i32 2, i32 3, i32 4,
i32 5, i32 6, i32 7, i32 8, i32 9, i32 10,
i32 11, i32 12, i32 13, i32 14, i32 15>
ret <16 x i32> %t4
Compiler would transform the shuffle to
%t4 = shufflevector <8 x i64> %t2, <8 x i64> %t3,
<8 x i64> <i32 8, i32 1, i32 2, i32 3, i32 4,
i32 5, i32 6, i32 7>
This may lose the oppotunity to let ISel select mask instruction when
avx512 is enabled.
This patch is to prevent the tranform when avx512 feature is enabled.
Thank Simon for the idea.
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D130830
OCHyams [Wed, 12 Oct 2022 11:11:50 +0000 (12:11 +0100)]
[Dexter] Ignore step information in __libc_start_call_main
The test dexter/feature_tests/commands/perfect/expect_step_kind/direction.cpp
was failing on a machine because __libc_start_call_main was not identified as a
"frame below main" (a frame we don't want to gather information from), causing
dexter to count one more step than expected in the test.
Add __libc_start_call_main to the list of "frames below main". There may
be a more robust way of doing this but this is a pragmatic solution we can
use for now.
Florian Hahn [Wed, 12 Oct 2022 11:06:40 +0000 (12:06 +0100)]
[VPlan] Add test for sinking pointer induction increments.
Extra test for D133762.
Alex Zinenko [Wed, 12 Oct 2022 11:02:01 +0000 (11:02 +0000)]
[mlir][doc] move transform dialect docs to .md, NFC
The description of the Transform dialect has become quite lengthy to be
kept as a Tablegen string literal. Move it to a proper Markdown file.
Alex Zinenko [Wed, 12 Oct 2022 10:55:54 +0000 (10:55 +0000)]
[mlir] document transform dialect types
Bjorn Pettersson [Wed, 12 Oct 2022 10:00:25 +0000 (12:00 +0200)]
[test][DebugInfo] Convert some test cases to opaque pointers. NFC
Conversion performed using the script at:
https://gist.github.com/nikic/
98357b71fd67756b0f064c9517b62a34
Adrian Kuegel [Wed, 12 Oct 2022 10:45:35 +0000 (12:45 +0200)]
[mlir] Apply ClangTidyLegacy findings.
Use '= default' to define a trivial default constructor.
Pierre van Houtryve [Fri, 7 Oct 2022 08:12:36 +0000 (08:12 +0000)]
[llvm-objdump] Support nonzero section addresses in addSymbolizer
The previous calculations seem to have assumed that the section address would be zero.
This is true for relocatable object files, but certainly not for linked files like shared libraries.
Fixed the calculations to make them identical to the "real" `getInstruction` call below & added a regression test.
Reviewed By: scott.linder, simon_tatham
Differential Revision: https://reviews.llvm.org/D135430
David Green [Wed, 12 Oct 2022 10:11:32 +0000 (11:11 +0100)]
Revert "[AArch64] Add support for 128-bit non temporal loads."
This reverts commit
661403b85c219a83baa37335a870d4d93dc4b1c3 as the
custom lowering of loads prevents expanding unaligned loads with
strict-align.
Martin Storsjö [Tue, 11 Oct 2022 12:11:39 +0000 (15:11 +0300)]
[AArch64] Exclude instructions after setting the FP from SEH prologues
After setting up the FP, the rest of the prologue doesn't need to
be replayed for unwinding the stack frame.
This allows reverting the functional parts of
2f7fbf837625267193351cc334e506a3a9161958 (but fixing inconsistent
duplicate setting of HasWinCFI).
Differential Revision: https://reviews.llvm.org/D135686
David Spickett [Wed, 12 Oct 2022 09:19:29 +0000 (09:19 +0000)]
[LLDB] Fix x86_64 build
28e65a6a63ab39be97d1a88fe7b4d0fa2f532643 changed the parameter
type of SetType but I forgot to build on x86 as well as arm64.
Cullen Rhodes [Wed, 12 Oct 2022 08:36:03 +0000 (08:36 +0000)]
[AArch64][SVE] Add instcombine for PTEST_ANY(X=OP(PG,...), X) -> PTEST_ANY(PG, X))
Given this is an OR reduction the two are equivalent and later
optimizations (AArch64InstrInfo::optimizePTestInstr) may rewrite the
sequence to use the flag-setting variant of instruction X, to remove the
PTEST altogether.
Reviewed By: paulwalker-arm, bsmith
Differential Revision: https://reviews.llvm.org/D134946
Cullen Rhodes [Fri, 7 Oct 2022 14:29:17 +0000 (14:29 +0000)]
[AArch64][SVE] NFC: auto-generate llvm/test/Transforms/InstCombine/AArch64/sve-intrinsics-ptest.ll
Tobias Hieta [Wed, 12 Oct 2022 08:04:21 +0000 (10:04 +0200)]
[CMake] Add option LLVM_FORCE_CREATE_SYMLINKS
On Windows we don't create symlinks for the binaries (clang++, clang-cl)
since the support requires special setup (group policy settings and
you need to know exactly our distribution story). But if you know
about these things and have a controlled environment there is a lot
of storage to be saved, so let's add a manual opt-in for using symlinks
on Windows with LLVM_FORCE_CREATE_SYMLINKS=ON.
Reviewed By: phosek
Differential Revision: https://reviews.llvm.org/D135578
Dmitry Polukhin [Thu, 6 Oct 2022 14:56:19 +0000 (07:56 -0700)]
[clang-tidy] Dump effective diagnostics level in YAML output
Before this patch YAML output had default diagnostic level instead of effective level reported to the user on stdout. Wrapper scripts for clang-tidy usually use YAML output and they pick wrong diagnostics level without this patch.
Test Plan: check-clang-tools
Differential Revision: https://reviews.llvm.org/D135367
Nikita Popov [Tue, 11 Oct 2022 11:56:06 +0000 (13:56 +0200)]
[Attributes] Make intrinsic attribute generation more flexible (NFC)
Currently attributes for intrinsics are emitted using the
ArrayRef<AttrKind> based constructor for AttributeLists. This works
out fine for simple enum attributes, but doesn't really generalize
to attributes that accept values. We're already doing something
awkward for alignment attributes, and I'd like to have a cleaner
solution to this with
https://discourse.llvm.org/t/rfc-unify-memory-effect-attributes/65579 in mind.
The new generation approach is to instead directly construct
Attributes, giving us access to the full generality of that
interface. To avoid significantly increasing the size of the
generated code, we now also deduplicate the attribute sets. The
code generated per unique AttributeList looks like this:
case 204: {
AS[0] = {1, getIntrinsicArgAttributeSet(C, 5)};
AS[1] = {AttributeList::FunctionIndex, getIntrinsicFnAttributeSet(C, 10)};
NumAttrs = 2;
break;
}
and then the helper functions contain something like
case 5:
return AttributeSet::get(C, {
Attribute::get(C, Attribute::NoCapture),
});
and
case 10:
return AttributeSet::get(C, {
Attribute::get(C, Attribute::NoUnwind),
Attribute::get(C, Attribute::ArgMemOnly),
});
A casualty of this change is the intrin-properties.td test, as I
don't think that FileCheck allows matching this kind of output.
Differential Revision: https://reviews.llvm.org/D135679
David Spickett [Wed, 28 Sep 2022 15:32:33 +0000 (15:32 +0000)]
[LLDB] Change EmulateInstriction::ReadRegister to return Optional
Making it easier to understand and harder to misuse.
This only applies to the ReadRegister(const RegisterInfo ®_info) variant.
Depends on D135671
Reviewed By: clayborg
Differential Revision: https://reviews.llvm.org/D135672
Cullen Rhodes [Wed, 12 Oct 2022 08:17:45 +0000 (08:17 +0000)]
[AArch64][SVE] Fix BRKNS bug in optimizePTestInstr
The BRKNS instruction is unlike the other instructions that set flags
since it has an all active implicit predicate, so the existing
PTEST(PG, BRKN(PG, A, B)) -> BRKNS(PG, A, B)
in AArch64InstrInfo::optimizePTestInstr is incorrect, however
PTEST(PTRUE_B(31), BRKN(PG, A, B)) -> BRKNS(PG, A, B)
is correct.
Spotted by @paulwalker-arm in D134946.
Reviewed By: paulwalker-arm
Differential Revision: https://reviews.llvm.org/D135655
Cullen Rhodes [Wed, 28 Sep 2022 09:25:21 +0000 (09:25 +0000)]
[AArch64] NFC: Auto-generate llvm/test/CodeGen/AArch64/sve-ptest-removal-brk.ll
David Spickett [Wed, 28 Sep 2022 14:52:12 +0000 (14:52 +0000)]
[LLDB] Change RegisterValue::SetType param to const RegisterInfo&
No one was pasing nullptr here.
Depends on D135670
Reviewed By: clayborg
Differential Revision: https://reviews.llvm.org/D135671
David Spickett [Wed, 28 Sep 2022 14:44:17 +0000 (14:44 +0000)]
[LLDB] Pass const RegisterInfo& to RegisterValue::SetValueFromData
Familiar story, callers are either checking upfront that the pointer
wasn't null or not checking at all. SetValueFromData itself didn't
check either.
So make the parameter a ref and fixup the few places where a nullptr
check seems needed.
Depends on D135668
Reviewed By: clayborg
Differential Revision: https://reviews.llvm.org/D135670
Alex Zinenko [Tue, 11 Oct 2022 15:23:48 +0000 (15:23 +0000)]
[mlir] add transform dialect entry point
Introduce `transform::applyTransforms` as a top-level entry point to the
Transform dialect-driven transformation infrastructure, by analogy with
`applyFull/PartialConversion`. Clients are expected to use this function
and no longer need to maintain the transformation state. Make the
constructor of the TransformState private for that purpose.
Reviewed By: springerm
Differential Revision: https://reviews.llvm.org/D135681
David Spickett [Wed, 28 Sep 2022 14:24:57 +0000 (14:24 +0000)]
[LLDB] Change RegisterValue::SetFromMemoryData to const RegisterInfo&
All callers were either assuming their pointer was not null before calling
this, or checking beforehand.
Reviewed By: clayborg
Differential Revision: https://reviews.llvm.org/D135668
Martin Storsjö [Tue, 11 Oct 2022 08:11:31 +0000 (11:11 +0300)]
[LLD] [MinGW] Look for libs named "<libname>.lib" even with -static
This matches how ld.bfd works in practice; this fixes
https://github.com/mstorsjo/llvm-mingw/issues/305.
Adding a test for the new lib name combination that this allows, but
also adding a few negative tests for combinations that aren't
matched when -static is specified (because this change in itself
didn't break any of the existing tests either).
The logic in how ld.bfd looks for various libraries based on
an -l<libname> argument is rather complex; the
ldemul_open_dynamic_archive function looks for various combinations:
https://sourceware.org/git/?p=binutils-gdb.git;a=blob;f=ld/emultempl/pep.em;h=
e68d1e69f17ad73af065b6bed19ae89ded913172;hb=
b51c2fec1da205ea3e7354cbb3e253018d64873c#l2066
This function is only called if looking for dynamic libraries
(i.e. if -static wasn't specified):
https://sourceware.org/git/?p=binutils-gdb.git;a=blob;f=ld/ldfile.c;h=
731ae5f7aedcf921bd36a1b32a3e0f5bfa189071;hb=
b51c2fec1da205ea3e7354cbb3e253018d64873c#l365
However even this function is skipped, it still looks for libraries
in the form of "lib<libname>.a" (this is what lld did before):
https://sourceware.org/git/?p=binutils-gdb.git;a=blob;f=ld/ldfile.c;h=
731ae5f7aedcf921bd36a1b32a3e0f5bfa189071;hb=
b51c2fec1da205ea3e7354cbb3e253018d64873c#l440
But it also calls a format specific function called
ldemul_find_potential_libraries, which for PE targets looks for
files named "<libname>.lib":
https://sourceware.org/git/?p=binutils-gdb.git;a=blob;f=ld/emultempl/pep.em;h=
e68d1e69f17ad73af065b6bed19ae89ded913172;hb=
b51c2fec1da205ea3e7354cbb3e253018d64873c#l2175
Differential Revision: https://reviews.llvm.org/D135651
Martin Storsjö [Tue, 11 Oct 2022 11:29:36 +0000 (14:29 +0300)]
[MC] [Win64EH] Generate ARM64 packed unwind info with signed return addresses
Differential Revision: https://reviews.llvm.org/D135660
Martin Storsjö [Wed, 5 Oct 2022 10:42:46 +0000 (13:42 +0300)]
[AArch64] Add support for the SEH opcode for return address signing
This was documented upstream in
https://github.com/MicrosoftDocs/cpp-docs/pull/4202.
Differential Revision: https://reviews.llvm.org/D135276
Petr Hosek [Thu, 18 Aug 2022 08:25:13 +0000 (08:25 +0000)]
[runtimes] Use a response file for runtimes test suites
We don't know which test suites are going to be included by runtimes
builds so we cannot include those before running the sub-build, but
that's not possible during the LLVM build configuration. We instead use
a response file that's populated by the runtimes build as a level of
indirection.
This addresses the issue described in:
https://discourse.llvm.org/t/cmake-regeneration-is-broken/62788
Differential Revision: https://reviews.llvm.org/D132438
jacquesguan [Thu, 29 Sep 2022 03:19:25 +0000 (11:19 +0800)]
[mlir][Math] Change regex to match fp value on different target.
Link: https://github.com/llvm/llvm-project/issues/58048
Reviewed By: ftynse, Mogball
Differential Revision: https://reviews.llvm.org/D134850
Vitaly Buka [Wed, 12 Oct 2022 06:24:13 +0000 (23:24 -0700)]
Revert "Fix locating of COMPILER_RT_LIBCXX_PATH"
clang-aarch64-sve-vla crashes in msan
clang-s390x-linux-lnt fails to compile libcxx
This reverts commit
14b2b67cd7d5e439c4bc5c8051f68a5fbb683480.
Max Kazantsev [Wed, 12 Oct 2022 06:17:28 +0000 (13:17 +0700)]
[SimpleLoopUnswitch][NFC] Separate legality checks from cost computation
These are semantically two different stages, but were entwined in the
old implementation. Now cost computation does not do legality checks,
and they all are done beforehead.