Commit Graph

425540 Commits

Author SHA1 Message Date
Martin Storsjö 40c937cba2 [ARM] Fix restoring stack for varargs with SEH split frame pointer push
Previously, the "add sp, #12" ended up inserted after "bx lr".

Differential Revision: https://reviews.llvm.org/D126872
2022-06-03 09:32:00 +03:00
Timm Bäder 9f97720268 [clang][driver] Dynamically select gcc-toolset/devtoolset
Instead of adding all devtoolset and gcc-toolset prefixes to the list of
prefixes, just scan the /opt/rh/ directory for the one with the highest
version number and only add that one.

Differential Revision: https://reviews.llvm.org/D125862
2022-06-03 08:12:27 +02:00
Alexander Batashev b34fb277df [mlir][cf] Implement missing SwitchOp::build function
Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D126594
2022-06-03 09:08:04 +03:00
bzcheeseman 47231248f5 [LLVM][Docs] Update for HowToSetUpLLVMStyleRTTI.rst, NFC.
This patch updates the document with some advanced use cases and examples on how to set up and use LLVM-style RTTI. It includes a few motivating examples to get readers comfortable with the concepts.

Reviewed By: lattner

Differential Revision: https://reviews.llvm.org/D126943
2022-06-02 22:34:38 -07:00
Max Kazantsev 8555e59a71 [NFC][MemDep] Remove unnecessary Worklist.clear
This execution path leads to return 'false' where the Worklist
will be deallocated anyways. No need to clear it separately.
2022-06-03 12:31:44 +07:00
Serguei Katkov 24e16e4af2 [SSAUpdaterImpl] Do not generate phi node with all the same incoming values
If all available vals to basic block are the same - do not build new phi node and
just use this value.

Reviewed By: sameerds
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D126525
2022-06-03 12:24:33 +07:00
Tue Ly 614567a7bf [libc] Automatically add -mfma flag for architectures supporting FMA.
Detect if the architecture supports FMA instructions and if
the targets depend on fma.

Reviewed By: gchatelet

Differential Revision: https://reviews.llvm.org/D123615
2022-06-03 01:21:20 -04:00
Douglas Chen 78b16ccf2b [M68k] Instruction selection to choose neg x when mul x -1 (Fix issue 48588)
This patch is trying to fix issue 48588(https://github.com/llvm/llvm-project/issues/48588)

I found the results of Instruction Selection between SelectionDAG and FastISEL for the `%mul = mul i32 %A, 4294967295`:
(seldag-isel) mul --> sub --> SUB32dp
(fast-isel)   mul --> sub --> NEG32d

My patch to fix this issue is by overriding a virtual function M68kDAGToDAGISel::IsProfitableToFold(). Return `false` when it was trying to match with SUB, then it will match with NEG.

Reviewed By: myhsu

Differential Revision: https://reviews.llvm.org/D116886
2022-06-03 13:20:30 +08:00
Thomas Raoux 271a48e029 [mlir][VectorToGPU] Fix bug generating incorrect ldmatrix ops
ldmatrix transpose can only be used with types that are 16bits wide.

Differential Revision: https://reviews.llvm.org/D126846
2022-06-03 04:30:22 +00:00
Serguei Katkov c4d955dd7f [MachineSSAUpdate] Add a test for redundant phi generation. 2022-06-03 11:27:14 +07:00
Thomas Raoux 205c08b54d [mlir][scf] Add option to loop pipelining to not peel the epilogue
Add an option to predicate the epilogue within the kernel instead of
peeling the epilogue. This is a useful option to prevent generating
large amount of code for deep pipeline. This currently require a user
lamdba to implement operation predication.

Differential Revision: https://reviews.llvm.org/D126753
2022-06-03 04:20:20 +00:00
Craig Topper 1d67adbfbf [RISCV] Give CSImm12MulBy4 PatLeaf priority over CSImm12MulBy8. NFC
The immediate range check for CSImm12MulBy8 included some values
covered by CSImm12MulBy4. I assume CSImm12MulBy4 had priority due
to pattern order in the td file, but this makes the priority
explicit in the predicate.
2022-06-02 20:51:14 -07:00
Fangrui Song e33af271ea [llvm-c-test] Default to opaque pointers 2022-06-02 20:34:52 -07:00
Fangrui Song 8d846849f8 [llvm-c][test] Convert tests to opaque pointers
echo.ll is unchanged to test typed pointers.
2022-06-02 20:27:10 -07:00
River Riddle ee1cf1f645 [mlir][NFC] Simplify the various `parseSourceFile<T>` overloads
These effectively all share the same implementation, i.e. forward
to the non-templated overload and then construct the container op.
2022-06-02 19:18:55 -07:00
Amir Ayupov e2142ff47c [BOLT][NFC] Make ICP::verifyProfile static
Follow LLVM style guide suggestion to avoid function definitions in anonymous
namespaces: https://llvm.org/docs/CodingStandards.html#anonymous-namespaces

Reviewed By: rafauler

Differential Revision: https://reviews.llvm.org/D124896
2022-06-02 19:09:29 -07:00
Shilei Tian b917433835 [NFC][Doc] Finish atomic compare 2022-06-02 21:50:07 -04:00
Shilei Tian c4a90db720 [Clang][OpenMP] Add the codegen support for `atomic compare capture`
This patch adds the codegen support for `atomic compare capture` in clang.

Reviewed By: ABataev

Differential Revision: https://reviews.llvm.org/D120290
2022-06-02 21:38:21 -04:00
Amir Ayupov 65a84195ca [BOLT][DOCS] Add PACKAGE_VERSION to doxygen config
Clang's doxygen documentation specifies LLVM revision. Do the same for BOLT.

Reviewed By: rafauler

Differential Revision: https://reviews.llvm.org/D126912
2022-06-02 18:05:44 -07:00
Daniil Suchkov f1940a5895 Revert "[LoopInterchange] New cost model for loop interchange"
Reverting the commit due to numerous buildbot failures.

This reverts commit 006334470d.
2022-06-03 00:52:08 +00:00
Mike Rice 48d6a6c9ad [OpenMP][NFC] update status for 'omp_all_memory' directive to 'done' 2022-06-02 17:31:33 -07:00
Akira Hatanaka 66e08995b0 [Sema] Reject list-initialization of enumeration types from a
brace-init-list containing a single element of a different scoped
enumeration type

It is rejected because it doesn't satisfy the condition that the element
has to be implicitly convertible to the underlying type of the
enumeration.

http://eel.is/c++draft/dcl.init.list#3.8

Differential Revision: https://reviews.llvm.org/D126084
2022-06-02 17:25:11 -07:00
Aart Bik f8b692dd31 [mlir][python][f16] add ctype python binding support for f16
Similar to complex128/complex64, float16 has no direct support
in the ctypes implementation. This fixes the issue by using a
custom F16 type to change the view in and out of MLIR code

Reviewed By: wrengr

Differential Revision: https://reviews.llvm.org/D126928
2022-06-02 17:21:24 -07:00
Akira Hatanaka b64f6e5722 Add a release note for the scope enum initialization bug fix in
https://reviews.llvm.org/D126084
2022-06-02 17:13:05 -07:00
River Riddle bb81b3b274 [vscode-mlir] Bump to version 0.8
Since version 0.7 we've added:

* Initial language support for TableGen
* Tweaked syntax highlighting for PDLL
* Added a new command to view intermediate PDLL output
2022-06-02 16:35:09 -07:00
River Riddle bf352e0b2e [mlir:PDLL] Add better support for providing Constraint/Pattern/Rewrite documentation
This commit enables providing long-form documentation more seamlessly to the LSP
by revamping decl documentation. For ODS imported constructs, we now also import
descriptions and attach them to decls when possible. For PDLL constructs, the LSP will
now try to provide documentation by parsing the comments directly above the decls
location within the source file. This commit also adds a new parser flag
`enableDocumentation` that gates the import and attachment of ODS documentation,
which is unnecessary in the normal build process (i.e. it should only be used/consumed
by tools).

Differential Revision: https://reviews.llvm.org/D124881
2022-06-02 16:31:07 -07:00
Arjun P 8bc2cff95a [MLIR][Presburger] Simplex: remove redundant member vars nRow, nCol
Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D126790
2022-06-03 00:30:48 +01:00
Chia-hung Duan 633ad1d864 [mlir:MultiOpDriver] Quick fix the assertion position
The assertion should come after null check
2022-06-02 23:25:35 +00:00
Nicolas van Kempen 987f9cb6b9 [clang-tidy] Add proper emplace checks to modernize-use-emplace
modernize-use-emplace only recommends going from a push_back to an
emplace_back, but does not provide a recommendation when emplace_back is
improperly used. This adds the functionality of warning the user when
an unecessary temporary is created while calling emplace_back or other "emplacy"
functions from the STL containers.

Reviewed By: kuhar, ivanmurashko

Differential Revision: https://reviews.llvm.org/D101471
2022-06-03 00:14:57 +01:00
Congzhe Cao 006334470d [LoopInterchange] New cost model for loop interchange
This patch proposed to use a new cost model for loop interchange, which
is obtained from loop cache analysis.

Given a loopnest, what loop cache analysis returns is a vector of loops
[loop0, loop1, loop2, ...] where loop0 should be replaced as the outermost
loop, loop1 should be placed one more level inside, and loop2 one more level
inside, etc. What loop cache analysis does is not only more comprehensive than
the current cost model, it is also a "one-shot" query which means that we only
need to query it once during the entire loop interchange pass, which is better
than the current cost model where we query it every time we check whether it is
profitable to interchange two loops. Thus complexity is reduced, especially after
D120386 where we do more interchanges to get the globally optimal loop access pattern.

Updates made to test cases are mostly minor changes and some corrections.
Test coverage for loop interchange is not reduced.

Currently we did not completely remove the legacy cost model, but keep it as
fall-back in case the new cost model did not run successfully. This is because
currently we have some limitations in delinearization, which sometimes makes
loop cache analysis bail out. The longer term goal is to enhance delinearization
and eventually remove the legacy cost model compeletely.

Reviewed By: bmahjour, #loopoptwg

Differential Revision: https://reviews.llvm.org/D124926
2022-06-02 19:07:14 -04:00
Paul Pluzhnikov 4ad17d2e96 Clean "./" from __FILE__ expansion.
This is alternative to https://reviews.llvm.org/D121733
and helps with Clang header modules in which FILE
may expand to "./foo.h" or "foo.h" depending on whether the file was
included directly or not.

Only do this when UseTargetPathSeparator is true, as we are already
changing the path in that case.

Reviewed By: ayzhao

Differential Revision: https://reviews.llvm.org/D126396
2022-06-02 18:00:19 -04:00
Shilei Tian 3a96256b7e [Clang][OpenMP] Avoid using `IgnoreImpCasts` if possible
This patch removes all `IgnoreImpCasts` in Sema, and only uses it if necessary. If the expression is not of the same type as the pointer value, a cast is inserted.

Reviewed By: ABataev

Differential Revision: https://reviews.llvm.org/D126602
2022-06-02 17:45:02 -04:00
Paul Robinson aa1cdf87b5 [PS5] Ignore 'packed' on one-byte bitfields, matching PS4 2022-06-02 14:41:18 -07:00
Mehdi Amini 4e5ce2056e Revert "[mlir] Add integer range inference analysis"
This reverts commit 1350c9887d.

Shared library build is broken with undefined references.
2022-06-02 21:24:06 +00:00
Matt Arsenault dd7e407d81 AMDGPU: Move SpilledReg from MFI to SIRegisterInfo
This isn't the most natural place for it, but it avoids a circular
include dependency in an out of tree patch.
2022-06-02 17:11:24 -04:00
Julien Pages 2dfe419446 [AMDGPU] Improve codegen of extractelement/insertelement in some cases
This patch improves the codegen of extractelement and insertelement for vector
containing 8 elements. Before, a dag combine transformation was generating a
sequence of 8 select/cmp.
This patch changes the upper limit for this transformation and the movrel
instruction will eventually be used instead. Extractlement/insertelement for
vectors containing less than 8 elements are unchanged.

Differential Revision: https://reviews.llvm.org/D126389
2022-06-02 17:05:55 -04:00
Alexander Smarus 4e1b89064f cmake fill `cmake_args` when cross-compiling external project with non-clang compiler
This makes it possible to crosscompile runtimes with cl.exe on Windows.

An external project is completely misconfigured otherwise because
cmake_args is set only for native builds or builds crosscompiled with
clang.

Differential Revision: https://reviews.llvm.org/D122578
Reviewed By: beanz, compnerd
2022-06-02 21:02:58 +00:00
David Blaikie cb08f4aa44 Support warn_unused_result on typedefs
While it's not as robust as using the attribute on enums/classes (the
type information may be lost through a function pointer, a declaration
or use of the underlying type without using the typedef, etc) but I
think there's still value in being able to attribute a typedef and have
all return types written with that typedef pick up the
warn_unused_result behavior.

Specifically I'd like to be able to annotate LLVMErrorRef (a wrapper for
llvm::Error used in the C API - the underlying type is a raw pointer, so
it can't be attributed itself) to reduce the chance of unhandled errors.

Differential Revision: https://reviews.llvm.org/D102122
2022-06-02 20:57:31 +00:00
Craig Topper dbead2388b [RISCV] Add custom isel for (add X, imm) used by load/stores.
If the imm is out of range for an ADDI, we will materialize it in
a register using multiple instructions. If the ADD is used by a
load/store, doPeepholeLoadStoreADDI can try to pull an ADDI from
the constant materialization into the load/store offset. This only
works if the ADD has a single use, otherwise the peephole would have
to rebuild multiple nodes.

This patch instead tries to solve the problem when the add is selected.
We check that the add is only used by loads/stores and if it is
we will select it to (ADDI (ADD X, Imm-Lo12), Lo12). This will enable
the simple case in doPeepholeLoadStoreADDI that can bypass an ADDI
used as a pointer. As a result we can remove the more complicated
peephole from doPeepholeLoadStoreADDI.

Reviewed By: reames

Differential Revision: https://reviews.llvm.org/D126576
2022-06-02 13:45:32 -07:00
Florian Hahn 78c6b1488f
[CaptureTracking] Increase limit and use it for all visited uses.
Currently the MaxUsesToExplore limit only applies to the number of users
per value, not the total number of users to explore.

The current limit of 20 pessimizes IR with opaque pointers in some
cases. Without opaque pointers, we have deeper pointer def-use chains in
general due to extra bitcasts and geps for structs with index 0.

With opaque pointers the def-use chain is not as deep but wider, due to
bitcasts & 0-geps missing.

To improve the situation for opaque pointers, this patch does 2 things:

 1. Apply the limit to the total number of uses visited. From the
    wording in the description of the option it seems like this may be
    the original intention. With the current implementation we could
    still end up walking a lot of uses.
 2. Increase the limit to 100. This is quite arbitrary, but enables
    a good number of additional optimizations.

Those adjustments have a noticeable compile-time impact though. In part
that is likely due to additional transformations (and conversely
the current baseline misses optimizations after switching to opaque
pointers).

This recovers some regressions that showed up after enabling opaque
pointers.

Limit=100:

* NewPM-O3: +0.21%
* NewPM-ReleaseThinLTO: +0.87%
* NewPM-ReleaseLTO-g: +0.46%

https://llvm-compile-time-tracker.com/compare.php?from=2e50ecb2ef4e1da1aeab05bcf66380068e680991&to=7e6fbe519d958d09f32f01d5d44a622f551e2031&stat=instructions

Limit=60:

* NewPM-O3: +0.14%
* NewPM-ReleaseThinLTO: +0.41%
* NewPM-ReleaseLTO-g: +0.21%

https://llvm-compile-time-tracker.com/compare.php?from=aeb19817d66f1a15754163c7f48e01e9ebdd6d45&to=520563fdc146319aae90d06f88d87f2e9e1247b7&stat=instructions

Limit=40:
* NewPM-O3: +0.11%
* NewPM-ReleaseThinLTO: +0.12%
* NewPM-ReleaseLTO-g: +0.09%

https://llvm-compile-time-tracker.com/compare.php?from=aeb19817d66f1a15754163c7f48e01e9ebdd6d45&to=c9182576e9fe3f1c84a71479665aef91a416318c&stat=instructions

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D126236
2022-06-02 21:43:58 +01:00
Fangrui Song e09f77d394 [ELF] Remove support for legacy .zdebug sections
.zdebug is unlikely used any longer: gcc -gz switched from legacy
.zdebug to SHF_COMPRESSED with binutils 2.26 (2016), which has been
several years. clang 14 dropped -gz=zlib-gnu support. According to
Debian Code Search (`gz=zlib-gnu`), no project uses -gz=zlib-gnu.

Remove .zdebug support to (a) simplify code and (b) allow removal of llvm-mc's
--compress-debug-sections=zlib-gnu.

In case the old object file `a.o` uses .zdebug, run `objcopy --decompress-debug-sections a.o`

Reviewed By: peter.smith

Differential Revision: https://reviews.llvm.org/D126793
2022-06-02 13:37:19 -07:00
Fangrui Song dfa9221aa7 [docs] Mention LLVMContext::setOpaquePointers for C++ API 2022-06-02 13:28:42 -07:00
Krzysztof Drewniak 1350c9887d [mlir] Add integer range inference analysis
This commit defines a dataflow analysis for integer ranges, which
uses a newly-added InferIntRangeInterface to compute the lower and
upper bounds on the results of an operation from the bounds on the
arguments. The range inference is a flow-insensitive dataflow analysis
that can be used to simplify code, such as by statically identifying
bounds checks that cannot fail in order to eliminate them.

The InferIntRangeInterface has one method, inferResultRanges(), which
takes a vector of inferred ranges for each argument to an op
implementing the interface and a callback allowing the implementation
to define the ranges for each result. These ranges are stored as
ConstantIntRanges, which hold the lower and upper bounds for a
value. Bounds are tracked separately for the signed and unsigned
interpretations of a value, which ensures that the impact of
arithmetic overflows is correctly tracked during the analysis.

The commit also adds a -test-int-range-inference pass to test the
analysis until it is integrated into SCCP or otherwise exposed.

Finally, this commit fixes some bugs relating to the handling of
region iteration arguments and terminators in the data flow analysis
framework.

Depends on D124020

Depends on D124021

Reviewed By: rriddle, Mogball

Differential Revision: https://reviews.llvm.org/D124023
2022-06-02 20:24:11 +00:00
Mingming Liu 8601f269f1 [Inline][Remark][NFC] Optionally provide inline context to inline
advisor.

This patch has no functional change, and merely a preparation patch for
main functional change. The motivating use case is to annotate inline
remark pass name with context information (e.g. prelink or postlink,
CGSCC or always-inliner), see D125495 for more details.

Differential Revision: https://reviews.llvm.org/D126824
2022-06-02 13:14:30 -07:00
Adrian Prantl e7b929d756 Adapt IRForTarget::RewriteObjCConstStrings() for D126689.
With opaque pointers, the LLVM IR expected by this function changed.
2022-06-02 13:06:40 -07:00
Philip Reames 76ac916d63 [RISCV] Inline one copy of needVSETVLI into the other [NFC]
Calling the non-MI version directly was unsound (as fixed in dcdb0bf2), so remove that version to decrease likelyhood of future mistakes.
2022-06-02 13:06:18 -07:00
Xiang Li 6bea9ff913 [HLSL] Add WaveActiveCountBits as Langugage builtin function for HLSL
One clang builtins are introduced
 uint WaveActiveCountBits( bool bBit ) as Langugage builtin function for HLSL.

The detail for WaveActiveCountBits is at
https://github.com/microsoft/DirectXShaderCompiler/wiki/Wave-Intrinsics#uint-waveactivecountbits-bool-bbit-

This is only clang part change to make WaveActiveCountBits into AST.
llvm intrinsic for WaveActiveCountBits will be add in separate PR.

Reviewed By: Anastasia

Differential Revision: https://reviews.llvm.org/D126857
2022-06-02 13:06:01 -07:00
Sanjay Patel 1882c25f12 [InstCombine] add tests for mul with low-bit mask operand; NFC 2022-06-02 16:01:23 -04:00
Sanjay Patel 8689463bfb [InstCombine] make pattern matching more consistent; NFC
We could go either way on this and several similar matches.
Just matching as a binop is possibly slightly more efficient;
we don't need to re-confirm the opcode of the instruction.
2022-06-02 16:01:23 -04:00
Maksim Panchenko 986e5dedf2 [BOLT][NFC] Fix braces in BinaryEmitter
Reviewed By: rafauler

Differential Revision: https://reviews.llvm.org/D126844
2022-06-02 12:45:25 -07:00