platform/upstream/llvm.git
12 months ago[FSAFDO] Switch to xxh3_64bits
Fangrui Song [Wed, 19 Jul 2023 21:23:28 +0000 (14:23 -0700)]
[FSAFDO] Switch to xxh3_64bits

Following recent changes switching from xxh64 to xxh32 for better
hashing performance. This particular instance may or may not have
noticeable performance difference, but this change makes us toward
removing xxHash64.

12 months ago[-Wunsafe-buffer-usage] Do not assert that function parameters have names
Ziqing Luo [Wed, 19 Jul 2023 21:02:06 +0000 (14:02 -0700)]
[-Wunsafe-buffer-usage] Do not assert that function parameters have names

It is possible that a function parameter does not have a name even in
a function definition.  This patch deals with such cases in generating
function overload fix-its for safe buffers.

Reviewed by: NoQ (Artem Dergachev)

Differential revision: https://reviews.llvm.org/D155641

12 months ago[lldb] Make frame var --regex always search globals
Felipe de Azevedo Piovezan [Fri, 14 Jul 2023 19:59:24 +0000 (15:59 -0400)]
[lldb] Make frame var --regex always search globals

Currently frame var --regex sometimes searches globals, sometimes it doesn't.
This happens because `StackFrame::GetVariableList` always returns the biggest
list it has, regardless of whether only globals were requested or not. In other
words, if a previous call to `GetVariableList` requested globals, all subsequent
calls will see them.

The implication here is that users of `StackFrame::GetVariableList` are expected
to filter the results of this function. This is what we do for a vanilla
`frame var` command. But it is not what we do when `--regex` is used. This
commit solves the issue by:

1. Making `--regex` imply `--globals`. This matches the behavior of `frame var
<some_name>`, which will also search the global scope.
2. Making the `--regex` search respect the command object options.

See the added test for an example of the oddities this patch addresses. Without
the patch, the test fails. However it could be made to pass by calling a plain
`frame var` before calling `frame var --regex A::`.

Differential Revision: https://reviews.llvm.org/D155334

12 months ago[COFF] Switch to xxh3_64bits
Fangrui Song [Wed, 19 Jul 2023 21:08:13 +0000 (14:08 -0700)]
[COFF] Switch to xxh3_64bits

Similar to recent changes to ELF (e.g., commit
f4b4bc2f18dc0e44afde05735fb673d3de4d5c39) and Mach-O to improve hashing
performance.

12 months ago[lldb-vscode] Creating a new flag for adjusting the behavior of evaluation repl expre...
John Harrison [Thu, 13 Jul 2023 22:50:06 +0000 (15:50 -0700)]
[lldb-vscode] Creating a new flag for adjusting the behavior of evaluation repl expressions to allow users to more easily invoke lldb commands.

This adds a new flag and lldb runtime command to allow users to manage the behavior of the lldb-vscode evaluate repl request.

When evaluating a repl context this now has runtime managed flag for control how the repl behaviors with the follow values:

* `variable` - the existing behavior, with this mode requests are evaluted in the current frame context as variable expressions. To trigger a lldb command prefix an expression with ` and it will be evaluted as an lldb command.
* `command` - all expressions are evaluated as lldb commands.
* `auto` - An alternative mode that will attempt to determine if the expression is an lldb command or a variable expression. Based off the intepreted results the expression will be evaluted either as a command or an expression.

Additionally, I enabled completions and ensured they work with the new repl expression behavior to provide auto-completes.

This commit includes updates to the tests to verify the new behavior after the previous failures from submitting https://reviews.llvm.org/D154030.

Differential Revision: https://reviews.llvm.org/D155248

12 months ago[mlir][spirv] Add D155747 to `.git-blame-ignore-revs`
Jakub Kuderski [Wed, 19 Jul 2023 20:52:36 +0000 (16:52 -0400)]
[mlir][spirv] Add D155747 to `.git-blame-ignore-revs`

12 months ago[mlir][spirv] Split op implementation file into subfiles. NFC.
Jakub Kuderski [Wed, 19 Jul 2023 20:48:46 +0000 (16:48 -0400)]
[mlir][spirv] Split op implementation file into subfiles. NFC.

The main op implementation file for SPIR-V grew past 5k LOC. This makes it
take a long time to compile and index with LSPs like clangd.

Pull out the first few SPIR-V extension ops into their own `.cpp` files,
just like we do with `.td` op definitions. This includes the
KHR/NV/Intel coop matrix and the integer dot prod extensions.

I plan to further split this in future revisions.

Reviewed By: antiagainst

Differential Revision: https://reviews.llvm.org/D155747

12 months ago[clang-tidy][NFC] Fixes in release notes
Piotr Zegar [Wed, 19 Jul 2023 19:11:24 +0000 (19:11 +0000)]
[clang-tidy][NFC] Fixes in release notes

Correct link in release notes and format of some notes.

12 months ago[flang] Fix complex libm use logic problems
Razvan Lupusoru [Wed, 19 Jul 2023 18:29:46 +0000 (11:29 -0700)]
[flang] Fix complex libm use logic problems

Fix the various complex libm selection logic issues from D155310:
- disableMlirComplex is set to false. This means that using mlir complex
is enabled. Yet, the current code still selects libm.
- If we enable mlir complex, we should not check if use approx is
enabled. Namely, we should use mlir complex either if we enable mlir
complex OR use approx is enabled.

To fix the issues, we flip the logic of `disableMlirComplex` to enable
instead. We set it to false by default since the intention from D155310
is to use libm by default. Then we use a logical `&&` with use approx
so that we select libm when BOTH mlir complex and use approx are
disabled.

Reviewed By: vzakhari

Differential Revision: https://reviews.llvm.org/D155737

12 months ago[libc][NFC] Move the RPC types to the `rpc_client.h` header
Joseph Huber [Wed, 19 Jul 2023 20:23:25 +0000 (15:23 -0500)]
[libc][NFC] Move the RPC types to the `rpc_client.h` header

Summary:
Simple cleanup of the interface so we do not depend on the installed
headers and get everything we need just including rpc_client.h.

12 months ago[llvm][utils] Add DenseMap data formatters
Dave Lee [Sat, 29 Oct 2022 22:07:49 +0000 (15:07 -0700)]
[llvm][utils] Add DenseMap data formatters

Add summary and synthetic data formatters for `llvm::DenseMap`.

This implementation avoids expression evaluation by using a heuristic. However, as
heuristics go, there is a corner case: A single deleted entry (a single "tombstone"),
will result in a child value with an invalid key but a valid value.

Instead of calling `getEmptyKey()` and `getTombstoneKey()` to determine which buckets
are empty, and which contain real key-values, the heuristic scans all buckets to
identify keys that exist only once. These singleton keys are considered valid.

The empty key will always exist multiple times. However the tombstone key may exist
zero, one, or many times. The heuristic has no problems when there are zero or many
tombstones, but when there is exactly one deleted entry (one tombstone), then the
heuristic will incorrectly identify it as valid.

Differential Revision: https://reviews.llvm.org/D137028

12 months ago[RISCV] Add a DAG combine for (czero_eq X, (xor Y, 1)) -> (czero_ne X, Y) if Y is...
Craig Topper [Wed, 19 Jul 2023 19:33:08 +0000 (12:33 -0700)]
[RISCV] Add a DAG combine for (czero_eq X, (xor Y, 1)) -> (czero_ne X, Y) if Y is 0 or 1.

This is an alternative to D155288 that can handle other sources of
xori like FP compares. Unfortunately, it misses the i64 setge case
on RV32 in condops.ll.

Reviewed By: asb

Differential Revision: https://reviews.llvm.org/D155328

12 months ago[mlir][Linalg] Fix warnings/errors on D155518.
Mahesh Ravishankar [Wed, 19 Jul 2023 18:45:10 +0000 (18:45 +0000)]
[mlir][Linalg] Fix warnings/errors on D155518.

This fixes some warnings (that were caught as errors) in https://reviews.llvm.org/D155518/.

Differential Revision: https://reviews.llvm.org/D155738

12 months ago[gn] add include dir for 1f7c7d4bdd7b
Nico Weber [Wed, 19 Jul 2023 19:17:50 +0000 (15:17 -0400)]
[gn] add include dir for 1f7c7d4bdd7b

12 months ago[mlir][sparse] Renaming `CreationPolicy` to `Policy`
wren romano [Tue, 18 Jul 2023 01:02:04 +0000 (18:02 -0700)]
[mlir][sparse] Renaming `CreationPolicy` to `Policy`

This change is mostly for brevity's sake; but it also paves the way for the `Policy` enum to be reuseable for other situations that require the same three-way semantics.

Reviewed By: Peiming

Differential Revision: https://reviews.llvm.org/D155532

12 months ago[libc++][print] Disables tests with msan.
Mark de Wever [Wed, 19 Jul 2023 18:56:51 +0000 (20:56 +0200)]
[libc++][print] Disables tests with msan.

These tests break with msan on the sanitizer-aarch64-linux-bootstrap-msan
builder. Note the x86_64 builder is not affected. To unbreak the CI
temporary disable the tests completely with msan.

The breakage was introduced by D150044.

12 months agoDefine llvm::thread::DefaultStackSize to 4 megabytes on AIX
Wael Yehia [Thu, 1 Jun 2023 19:56:07 +0000 (15:56 -0400)]
Define llvm::thread::DefaultStackSize to 4 megabytes on AIX

Link time thinLTO spawns pthreads to parallelize optimization and
codegen of the input bitcode files. On AIX, the default pthread
stack size limit is ~192k for 64-bit programs; insufficient for a
normal LLVM compilation.

Reviewed By: ZarkoCA, MaskRay

Differential Revision: https://reviews.llvm.org/D155731

12 months ago[SLP]Fix a crash when trying to cast scalable vector type to fixed.
Alexey Bataev [Wed, 19 Jul 2023 18:20:39 +0000 (11:20 -0700)]
[SLP]Fix a crash when trying to cast scalable vector type to fixed.

Need to check for FixedVectorType, not a vector type, since later
compiler performs unconditional cast to FixedVectorType and gets the
number of elements in this type.

12 months ago[ELF] --build-id=fast: switch to xxh3_64bits
Fangrui Song [Wed, 19 Jul 2023 18:37:42 +0000 (11:37 -0700)]
[ELF] --build-id=fast: switch to xxh3_64bits

12 months ago[ELF] splitNonStrings: switch to xxh3_64bits
Fangrui Song [Wed, 19 Jul 2023 18:28:47 +0000 (11:28 -0700)]
[ELF] splitNonStrings: switch to xxh3_64bits

This is primarily used for .rodata.cst* duplicate elimination. The
sections are usually much smaller than .debug_str (D154813), so the
speedup is negligible. We do this switch for consistency as we want to
eliminate xxh64 in lld.

12 months ago[lld-macho]Use install_name as Identifier for code-sign, if available.
Vy Nguyen [Mon, 17 Jul 2023 19:37:51 +0000 (15:37 -0400)]
[lld-macho]Use install_name  as Identifier for code-sign, if available.

Detail:
LD64 uses the name provided via  -[dylib]install_name as "Identifier", when available.
For compatiblity, LLD should do that too.

Differential Revision: https://reviews.llvm.org/D155508

12 months ago[OpenMP][Docs] Add some things to the OpenMP support
Joseph Huber [Wed, 19 Jul 2023 17:32:19 +0000 (12:32 -0500)]
[OpenMP][Docs] Add some things to the OpenMP support

This patch adds some information that we have support for in the OpenMP
clang support page.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D155727

12 months ago[ELF][test] Refactor merge.s
Fangrui Song [Wed, 19 Jul 2023 18:13:26 +0000 (11:13 -0700)]
[ELF][test] Refactor merge.s

12 months ago[sanitizer-common] Run module msan init before early sigaction test
Daniel Thornburgh [Tue, 18 Jul 2023 21:51:10 +0000 (14:51 -0700)]
[sanitizer-common] Run module msan init before early sigaction test

MSAN wrappers can be inserted for e.g. the access to stderr in the
constructor of the test, which can segfault if the constructor function
runs before these data structures have been initialized.

Reviewed By: dvyukov

Differential Revision: https://reviews.llvm.org/D155648

12 months ago[X86] matchBinaryShuffle - relax PACKSS for v2i64 -> v4i32 shuffle truncation pattern...
Simon Pilgrim [Wed, 19 Jul 2023 17:58:12 +0000 (18:58 +0100)]
[X86] matchBinaryShuffle - relax PACKSS for v2i64 -> v4i32 shuffle truncation pattern match.

Similar to combineVectorSignBitsTruncation, we don't require all-signbits source inputs, just enough signbits to reach into the lowest i16 to safely use PACKSSDW.

12 months ago[CodeGenPrepare] Refactor optimizeSelectInst (NFC)
Momchil Velikov [Wed, 19 Jul 2023 17:31:25 +0000 (18:31 +0100)]
[CodeGenPrepare] Refactor optimizeSelectInst (NFC)

Refactor to use BasicBlockUtils functions and make life easier for
a subsequent patch for updating the dominator tree.

Reviewed By: dmgreen

Differential Revision: https://reviews.llvm.org/D154053

12 months agoFix MSVC "'GetVMSetForLMul': not all control paths return a value" warning. NFC.
Simon Pilgrim [Wed, 19 Jul 2023 17:55:37 +0000 (18:55 +0100)]
Fix MSVC "'GetVMSetForLMul': not all control paths return a value" warning. NFC.

12 months ago[InstrProf][NFC] Ignore -Wcast-qual after D153911 to fix build failure on AIX
Jake Egan [Wed, 19 Jul 2023 17:50:33 +0000 (13:50 -0400)]
[InstrProf][NFC] Ignore -Wcast-qual after D153911 to fix build failure on AIX

/scratch/powerllvm/powerllvm_env/aix-ppc64/clang-ppc64-aix/llvm-project/compiler-rt/lib/profile/InstrProfilingPlatformAIX.c:202:40: error: cast from 'const int (*)[0]' to 'void *' drops const qualifier [-Werror,-Wcast-qual]
                               (void *)&dummy_name, (void *)&dummy_vnds};
                                       ^
1 error generated.

12 months ago[mlir][Linalg] Cleanup the drop unit dims pass in Linalg.
Mahesh Ravishankar [Wed, 19 Jul 2023 04:58:26 +0000 (04:58 +0000)]
[mlir][Linalg] Cleanup the drop unit dims pass in Linalg.

TL;DR the following API functions have been merged

```
void populateFoldUnitExtentDimsViaReshapesPatterns(RewritePatternSet &patterns);
void populateFoldUnitExtentDimsViaSlicesPatterns(RewritePatternSet &patterns);
```

into

```
void populateFoldUnitExtentDimsPatterns(RewritePatternSet &patterns,
                                        ControlDropUnitDims &options);
```

To use the previous functionality use

```
ControlDropUnitDims options;
// By default options.rankReductionStrategy is
// ControlDropUnitDims::RankReductionStrategy::ReassociativeReshape.
populateFoldUnitExtentDimsPatterns(patterns, options);
```

and

```
ControlDropUnitDims options;
options.rankReductionStrategy = ControlDropUnitDims::RankReductionStrategy::ExtractInsertSlice
populateFoldUnitExtentDimsPatterns(patterns, options);

```

This pass is quite old and needed to be updated based on the current
approach to transformations in Linalg

- Instead of two patterns, one to just remove loop dimensions that are
  unit extent (and using 0 in the indexing maps), and another to drop
  the unit-extents in the operand shapes, combine into a single
  transformation. This avoid creating an intermediate step with
  indexing maps having 0's in the domains exp ressions.

- Expose the core transformation as a utility function and add a
  pattern that calls this transformation.

This is a mostly NFC change, apart from the API change and dropping
the patterns/test that only dropped the loops that are unit extents.

Differential Revision: https://reviews.llvm.org/D155518

12 months ago[ThinLTO][AIX] Enable thinlto on AIX
Wael Yehia [Wed, 19 Jul 2023 01:48:03 +0000 (21:48 -0400)]
[ThinLTO][AIX] Enable thinlto on AIX

Starting from AIX 7.2 TL5 SP6 and AIX 7.3 TL2 the system linker supports thinLTO.

Reviewed By: ZarkoCA, MaskRay

Differential Revision: https://reviews.llvm.org/D155700

12 months ago[mlir][openacc] Relax verifier for the acc.reduction.recipe
Valentin Clement [Wed, 19 Jul 2023 17:30:42 +0000 (10:30 -0700)]
[mlir][openacc] Relax verifier for the acc.reduction.recipe

Allow the init and combiner regions to have more
arguments to pass information.

Reviewed By: razvanlupusoru

Differential Revision: https://reviews.llvm.org/D155656

12 months ago[RISCV] Run mem2reg on the scalar C builtin tests to remove allocas and simplify...
Craig Topper [Wed, 19 Jul 2023 17:30:42 +0000 (10:30 -0700)]
[RISCV] Run mem2reg on the scalar C builtin tests to remove allocas and simplify checks. NFC

As requested on D155647.

12 months ago[AMDGPUAttributor][FIX] No endless recursion for recursive initializers
Johannes Doerfert [Wed, 19 Jul 2023 17:11:29 +0000 (10:11 -0700)]
[AMDGPUAttributor][FIX] No endless recursion for recursive initializers

Fixes: https://github.com/llvm/llvm-project/issues/63956

12 months ago[Attributor][NFCI] Add a shortcut for constants
Johannes Doerfert [Wed, 19 Jul 2023 17:13:03 +0000 (10:13 -0700)]
[Attributor][NFCI] Add a shortcut for constants

12 months ago[Attributor][NFCI] Avoid updating AAs that depend on missing callees
Johannes Doerfert [Wed, 19 Jul 2023 06:27:16 +0000 (23:27 -0700)]
[Attributor][NFCI] Avoid updating AAs that depend on missing callees

12 months agoRefactor some BasicBlockUtils functions (NFC)
Momchil Velikov [Wed, 19 Jul 2023 16:27:56 +0000 (17:27 +0100)]
Refactor some BasicBlockUtils functions (NFC)

Add a more "flexible" `SplitBlockAndInsertIfThenElse` function
and re-implement some others on top of it.

Reviewed By: dmgreen

Differential Revision: https://reviews.llvm.org/D154052

12 months ago[Clang][CodeGen]`vtable`, `typeinfo` et al. are globals
Alex Voicu [Wed, 19 Jul 2023 17:04:31 +0000 (18:04 +0100)]
[Clang][CodeGen]`vtable`, `typeinfo` et al. are globals

All data structures and values associated with handling virtual functions / inheritance, as well as RTTI, are globals and thus can only reside in the global address space. This was not taken fully taken into account because for most targets, global & generic appear to coincide. However, on targets where global & generic ASes differ (e.g. AMDGPU), this was problematic, since it led to the generation of invalid bitcasts (which would trigger asserts in Debug) and less than optimal code. This patch does two things:

ensures that vtables, vptrs, vtts, typeinfo are generated in the right AS, and populated accordingly;
removes a bunch of bitcasts which look like left-overs from the typed ptr era.

Reviewed By: yxsamliu

Differential Revision: https://reviews.llvm.org/D153092

12 months ago[RISCV] Upgrade Zvfh version to 1.0 and move out of experimental state.
Craig Topper [Wed, 19 Jul 2023 17:03:57 +0000 (10:03 -0700)]
[RISCV] Upgrade Zvfh version to 1.0 and move out of experimental state.

This has been ratified according to https://wiki.riscv.org/display/HOME/Recently+Ratified+Extensions

Differential Revision: https://reviews.llvm.org/D155668

12 months ago[lld-macho] Switch to xxh3_64bits
Fangrui Song [Wed, 19 Jul 2023 16:58:43 +0000 (09:58 -0700)]
[lld-macho] Switch to xxh3_64bits

xxh3 is substantially faster than xxh64.
For lld/ELF, there is substantial speedup in `.debug_str` duplicate
elimination (D154813). Use xxh3 for lld-macho as well.

Reviewed By: #lld-macho, oontvoo

Differential Revision: https://reviews.llvm.org/D155677

12 months ago[DWARFLinkerParallel] Switch to xxh3_64bits
Fangrui Song [Wed, 19 Jul 2023 16:56:53 +0000 (09:56 -0700)]
[DWARFLinkerParallel] Switch to xxh3_64bits

xxh3 is substantially faster than xxh64.
For lld `.debug_str` there is substantial speedup (D154813).

@avl reports that it is around 1-2% improvement for different number of input sets and for different threads configuration.

Reviewed By: avl

Differential Revision: https://reviews.llvm.org/D155675

12 months ago[libc] Fix global constructor being emitted for the RPC client
Joseph Huber [Wed, 19 Jul 2023 15:58:45 +0000 (10:58 -0500)]
[libc] Fix global constructor being emitted for the RPC client

The indirection here is for some reason causing an unnecessary
constructor. If we leave this uninitialized we will get the default
constructor which simply zero initliaizes the global. I've checked the
output and confirmed that it uses the `zeroinitializer` so this should
be safe.

Reviewed By: JonChesterfield

Differential Revision: https://reviews.llvm.org/D155720

12 months ago[RISCV] Fold ops into vmv.v.v as vmerge with all-ones mask
Luke Lau [Wed, 12 Jul 2023 00:13:14 +0000 (01:13 +0100)]
[RISCV] Fold ops into vmv.v.v as vmerge with all-ones mask

A vmv.v.v shares the same encoding as a vmerge that isn't masked, so we can
also fold it into its operands if we treat it as a vmerge with an all-ones
mask.  We take care here not to actually transform the existing vmv into a
vmerge, otherwise things like True.hasOneUse() become inaccurate. Instead this
just returns an equivalent list of operands.
This is an alternative to D153351.

Reviewed By: reames

Differential Revision: https://reviews.llvm.org/D155101

12 months ago[RISCV] Fold vmerge into its ops with smaller VL if known
Luke Lau [Tue, 11 Jul 2023 16:52:23 +0000 (17:52 +0100)]
[RISCV] Fold vmerge into its ops with smaller VL if known

Currently when folding vmerge into its operands, we stop if the VLs aren't
identical.  However since the body of (vmerge (vop)) is the intersection of
vmerge and vop's bodies, we can use the smaller of the two VLs if we know it
ahead of time.  This patch relaxes the constraint on VL if they are both
constants, or if either of them are VLMAX.

Reviewed By: reames

Differential Revision: https://reviews.llvm.org/D155071

12 months ago[RISCV] Add tests for merges with differing VLs that could be folded
Luke Lau [Tue, 11 Jul 2023 12:14:50 +0000 (13:14 +0100)]
[RISCV] Add tests for merges with differing VLs that could be folded

Reviewed By: reames

Differential Revision: https://reviews.llvm.org/D155069

12 months ago[nfc][openacc] Add missing comma in acc dialect operation macros
Razvan Lupusoru [Wed, 19 Jul 2023 16:08:35 +0000 (09:08 -0700)]
[nfc][openacc] Add missing comma in acc dialect operation macros

A comma is missing which is incorrect if macro is used.

Reviewed By: clementval

Differential Revision: https://reviews.llvm.org/D155722

12 months ago[SLP][X86] Regenerate some test checks to reduce diff in D154891
Simon Pilgrim [Wed, 19 Jul 2023 16:02:03 +0000 (17:02 +0100)]
[SLP][X86] Regenerate some test checks to reduce diff in D154891

12 months ago[X86] matchBinaryShuffle - match PACKSS for v2i64 -> v4i32 all-signbits shuffle trunc...
Simon Pilgrim [Wed, 19 Jul 2023 13:55:54 +0000 (14:55 +0100)]
[X86] matchBinaryShuffle - match PACKSS for v2i64 -> v4i32 all-signbits shuffle truncation patterns.

Ideally matchShuffleWithPACK should be able to handle this, but it needs a major rewrite to handle illegal types.

12 months ago[libunwind] Fix build error on 32 bit Arm after -Wcast-qual was added
David Spickett [Wed, 19 Jul 2023 09:19:13 +0000 (09:19 +0000)]
[libunwind] Fix build error on 32 bit Arm after -Wcast-qual was added

New warning was added in https://reviews.llvm.org/D153911 which caused:
https://buildkite.com/llvm-project/libcxx-ci/builds/28407#01896b79-2a5e-4554-ac31-2abec5a8b281

../../libunwind/src/UnwindLevel1-gcc-ext.c:172:47: error: cast from 'const unsigned int *' to 'unsigned int *' drops const qualifier [-Werror,-Wcast-qual]
    ex.pr_cache.ehtp = (_Unwind_EHT_Header *) unwindInfo;

I don't see any reason there should be a const here in the first place,
so just remove it.

Reviewed By: #libunwind, michaelplatings, MaskRay

Differential Revision: https://reviews.llvm.org/D155685

12 months ago[flang] Represent unknown extent correctly in getTypeAsString
Valentin Clement [Wed, 19 Jul 2023 15:53:50 +0000 (08:53 -0700)]
[flang] Represent unknown extent correctly in getTypeAsString

Reviewed By: razvanlupusoru

Differential Revision: https://reviews.llvm.org/D155655

12 months ago[IR] Remove declaration of no longer existing constructor (NFC)
Nikita Popov [Wed, 19 Jul 2023 15:28:47 +0000 (17:28 +0200)]
[IR] Remove declaration of no longer existing constructor (NFC)

The definition was dropped in D155585, so drop the declaration
as well.

Thanks to Roman Divacky for pointing this out!

12 months ago[mlir][transform][gpu][python] Add .td file for bindings.
Ingo Müller [Tue, 18 Jul 2023 15:02:32 +0000 (15:02 +0000)]
[mlir][transform][gpu][python] Add .td file for bindings.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D155602

12 months ago[ELF] Use llvm::xxh3_64bits for MergeInputSection::splitStrings
Fangrui Song [Wed, 19 Jul 2023 15:34:25 +0000 (08:34 -0700)]
[ELF] Use llvm::xxh3_64bits for MergeInputSection::splitStrings

See D154812 for the speedup.

Reviewed By: PiotrZSL

Differential Revision: https://reviews.llvm.org/D154813

12 months ago[libc++] Work around dynamic linking of stringbuf::str() on Windows
Piotr Fusik [Fri, 14 Jul 2023 16:30:41 +0000 (18:30 +0200)]
[libc++] Work around dynamic linking of stringbuf::str() on Windows

https://github.com/llvm/llvm-project/issues/40363 caused the C++20
`str() const &` and `str() &&` to be dllimport'ed despite _LIBCPP_HIDE_FROM_ABI.
This is a temporary solution until #40363 is fixed.

Reviewed By: #libc, hans, ldionne, Mordante

Differential Revision: https://reviews.llvm.org/D155185

12 months ago[AggressiveInstCombine] Fold strcmp for short string literals
Maksim Kita [Wed, 19 Jul 2023 15:07:40 +0000 (17:07 +0200)]
[AggressiveInstCombine] Fold strcmp for short string literals

Fold strcmp() against 1-char string literals.

This designates AggressiveInstCombine as the pass for libcalls
simplifications that may need to change the control flow graph.

Fixes https://github.com/llvm/llvm-project/issues/58003.

Differential Revision: https://reviews.llvm.org/D154725

12 months ago[libc] Add basic support for calling host functions from the GPU
Joseph Huber [Tue, 11 Jul 2023 19:14:08 +0000 (14:14 -0500)]
[libc] Add basic support for calling host functions from the GPU

This patch adds the `rpc_host_call` function as a GPU extension. This is
exported from the `libc` project to use the RPC interface to call a
function pointer via RPC any copying the arguments by-value. The
interface can only support a single void pointer argument much like
pthreads. The function call here is the bare-bones version of what's
required for OpenMP reverse offloading. Full support will require
interfacing with the mapping table, nowait support, etc.

I decided to test this interface in `libomptarget` as that will be the
primary consumer and it would be more difficult to make a test in `libc`
due to the testing infrastructure not really having a concept of the
"host" as it runs directly on the GPU as if it were a CPU target.

Reviewed By: jplehr

Differential Revision: https://reviews.llvm.org/D155003

12 months ago[mlir][spirv] Add cooperative matrix store op
Jakub Kuderski [Wed, 19 Jul 2023 15:01:07 +0000 (11:01 -0400)]
[mlir][spirv] Add cooperative matrix store op

Implement cooperative matrix store for the `SPV_KHR_cooperative_matrix`
extension: https://github.com/KhronosGroup/SPIRV-Registry/blob/master/extensions/KHR/SPV_KHR_cooperative_matrix.html.

Reviewed By: antiagainst

Differential Revision: https://reviews.llvm.org/D155631

12 months ago[mlir][spirv] Add cooperative matrix load op
Jakub Kuderski [Wed, 19 Jul 2023 14:55:25 +0000 (10:55 -0400)]
[mlir][spirv] Add cooperative matrix load op

Implement cooperative matrix load for the `SPV_KHR_cooperative_matrix`
extension: https://github.com/KhronosGroup/SPIRV-Registry/blob/master/extensions/KHR/SPV_KHR_cooperative_matrix.html.

Also some minor fixes in common code for custom parsing.

Reviewed By: antiagainst

Differential Revision: https://reviews.llvm.org/D155616

12 months ago[openacc] Add attribute to hold declare data clause information
Razvan Lupusoru [Tue, 18 Jul 2023 20:39:51 +0000 (13:39 -0700)]
[openacc] Add attribute to hold declare data clause information

For variables in declare clauses, their producing operation should be
marked with the data clause for ease of lookup and consistency
verification. Thus add an attribute that can be used for this purpose
plus verification that declare data operation matches the declare
data clause on variable.

Reviewed By: clementval

Differential Revision: https://reviews.llvm.org/D155640

12 months ago[AggressiveInstCombine] Fold strcmp for short string literals tests (NFC)
Maksim Kita [Wed, 19 Jul 2023 14:52:55 +0000 (16:52 +0200)]
[AggressiveInstCombine] Fold strcmp for short string literals tests (NFC)

Precommit tests for D154725.

Differential Revision: https://reviews.llvm.org/D155053

12 months ago[mlir][flang] Convert TBAA metadata to an attribute representation
Markus Böck [Tue, 18 Jul 2023 12:30:00 +0000 (14:30 +0200)]
[mlir][flang] Convert TBAA metadata to an attribute representation

The current representation of TBAA is the very last in-tree user of the `llvm.metadata` operation.
Using ops to model metadata has a few disadvantages:
* Building a graph has to be done through some weakly typed indirection mechanism such as `SymbolRefAttr`
* Creating the metadata has to be done through a builder within a metadata op.
* It is not multithreading safe as operation insertion into the same block is not thread-safe

This patch therefore converts TBAA metadata into an attribute representation, in a similar manner as it has been done for alias groups and access groups in previous patches.

This additionally has the large benefit of giving us more "correctness by construction" as it makes things like cycles in a TBAA graph, or references to an incorrectly typed metadata node impossible.

Differential Revision: https://reviews.llvm.org/D155444

12 months ago[mlir][transform][linalg][python] Add mix-in for FuseIntoContainingOp.
Ingo Müller [Wed, 19 Jul 2023 11:55:00 +0000 (11:55 +0000)]
[mlir][transform][linalg][python] Add mix-in for FuseIntoContainingOp.

The class did not have any mix-in until now. The new mix-in has two
overloads for the constructor of the class: one with all arguments and
one without the result types, which are defaulted to `AnyOpType`.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D155695

12 months agoReland "[clangd] Always run preamble indexing on a separate thread"
Kadir Cetinkaya [Wed, 19 Jul 2023 10:31:40 +0000 (12:31 +0200)]
Reland "[clangd] Always run preamble indexing on a separate thread"

This reverts commit 92c0546941190973e9201a08fa10edf27cb6992d.

Prevents tsan issues by resetting ref-counted-pointers eagerly, before
passing the copies into a new thread.

12 months ago[mlir][transform][linalg][python] Fix test for TileToForallOp mixin.
Ingo Müller [Wed, 19 Jul 2023 14:19:57 +0000 (14:19 +0000)]
[mlir][transform][linalg][python] Fix test for TileToForallOp mixin.

The test was introduced recently by https://reviews.llvm.org/D155090,
but https://reviews.llvm.org/D155567, which I worked on concurrently
affected the output of that test and the two patches were never tested
together.

Differential Revision: https://reviews.llvm.org/D155709

12 months ago[mlir][bufferization] Add lowering of bufferization.dealloc to memref.dealloc
Martin Erhart [Wed, 19 Jul 2023 14:13:27 +0000 (14:13 +0000)]
[mlir][bufferization] Add lowering of bufferization.dealloc to memref.dealloc

Adds a generic lowering that suppors all cases of bufferization.dealloc
and one specialized, more efficient lowering for the simple case. Using
a helper function with for loops in the general case enables
O(|num_dealloc_memrefs|+|num_retain_memrefs|) size of the lowered code.

Depends on D155467

Reviewed By: springerm

Differential Revision: https://reviews.llvm.org/D155468

12 months agoRevert "[libc] Treat the locks array as a bitfield"
Joseph Huber [Wed, 19 Jul 2023 14:25:34 +0000 (09:25 -0500)]
Revert "[libc] Treat the locks array as a bitfield"

Summary:
This caused test failures on the gfx90a buildbot. This works on my
gfx1030 and the Nvidia buildbots, so we'll need to investigate what is
going wrong here. For now revert it to get the bots green.

This reverts commit 05abcc579244b68162b847a6780d27b22bd58f74.

12 months ago[RISCV] Add DAG combine for CTTZ/CTLZ in the case of input 0
Djordje Todorovic [Wed, 19 Jul 2023 09:25:51 +0000 (11:25 +0200)]
[RISCV] Add DAG combine for CTTZ/CTLZ in the case of input 0

Within the AggressiveInstCombine Pass we have
an analysis/optimization that matches that
pattern of the Table Based CTZ. Some Targets do
not support/define ctz(0), but since the
AggressiveInstCombine is just an extension of
InstCombine, it should be a target-independent
canonicalization Pass, and therefore, we decided
to introduce several instructions, such as select
and compare that produce canonical IR, even if
the input is 0. The task for the Targets that do
support that input is to handle such a case and
to produce an optimal assembly.

This patch optimizes the CTTZ/CTLZ instructions
if the input is 0 by performing the`DAG combine`,
by generating the cttz(x) & 0x1f pattern (the
same goes for ctlz as well).

Differential Revision: https://reviews.llvm.org/D151449

12 months ago[mlir][transform][linalg][python] Add extended TileToForallOp.
Ingo Müller [Wed, 12 Jul 2023 10:38:24 +0000 (10:38 +0000)]
[mlir][transform][linalg][python] Add extended TileToForallOp.

This patch adds a mixin for TileToForallOp to
_structured_transform_ops_ext.py with syntactic sugar for construction
such ops. First, the types of the results are made optional and filled
with common default values if omitted. Second, for num_threads and
tile_sizes, the three possible forms (static, dynamic, or packed), can
now all be given through the same respective argument, which gets
dispatched to the correct form-specific argument automatically.

Reviewed By: nicolasvasilache, ftynse

Differential Revision: https://reviews.llvm.org/D155090

12 months ago[clangd] Fix -Wunused-variable of 'FIDIt' (NFC)
Jie Fu [Wed, 19 Jul 2023 13:59:53 +0000 (21:59 +0800)]
[clangd] Fix -Wunused-variable of 'FIDIt' (NFC)

/data/llvm-project/clang-tools-extra/clangd/index/SymbolCollector.cpp:878:16: error: unused variable 'FIDIt' [-Werror,-Wunused-variable]
    const auto FIDIt = IncludeFiles.find(SID);
               ^
1 error generated.

12 months ago[clangd] Update symbol collector to use include-cleaner.
Viktoriia Bakalova [Wed, 14 Jun 2023 12:21:08 +0000 (12:21 +0000)]
[clangd] Update symbol collector to use include-cleaner.

Differential Revision: https://reviews.llvm.org/D152900

12 months ago[flang] Implement tand intrinsic
David Truby [Thu, 6 Jul 2023 14:59:16 +0000 (15:59 +0100)]
[flang] Implement tand intrinsic

This implements the tand intrinsic by performing a multiplication
by pi/180 to the argument before calling tan inline.

This is a commonly provided extension that is used by OpenRadioss

Differential Revision: https://reviews.llvm.org/D154614

12 months ago[flang] Use libm functions for complex operations by default
David Truby [Fri, 14 Jul 2023 16:02:44 +0000 (17:02 +0100)]
[flang] Use libm functions for complex operations by default

This patch changes the default lowering for complex operations to use
the more accurate libm operations as opposed to the mlir complex
operations. This is necessary due to precision issues in the mlir
complex dialect that cause failures in e.g. the LAPACK tests.

The mlir complex dialect lowering will still be used when
`-fapprox-func` is set (and by extension `-ffast-math` and `-Ofast`)

Differential Revision: https://reviews.llvm.org/D155310

12 months ago[X86] getFauxShuffleMask - add SIGN_EXTEND_VECTOR_INREG handling for all-signbits...
Simon Pilgrim [Wed, 19 Jul 2023 13:32:25 +0000 (14:32 +0100)]
[X86] getFauxShuffleMask - add SIGN_EXTEND_VECTOR_INREG handling for all-signbits sources

Add suport for shuffle combines (via combineEXTEND_VECTOR_INREG) to begin from SIGN_EXTEND_VECTOR_INREG nodes

12 months ago[InstCombine] Preserve metadata when combining select+binop
Paweł Bylica [Wed, 19 Jul 2023 11:55:19 +0000 (13:55 +0200)]
[InstCombine] Preserve metadata when combining select+binop

Fixes https://github.com/llvm/llvm-project/issues/63910

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D155461

12 months ago[InstCombine][NFC] Add tests for preserving metadata
Paweł Bylica [Wed, 19 Jul 2023 11:54:51 +0000 (13:54 +0200)]
[InstCombine][NFC] Add tests for preserving metadata

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D155594

12 months ago[lldb][NFC] Refactor test to enable subsequent reuse
Felipe de Azevedo Piovezan [Thu, 13 Jul 2023 13:12:01 +0000 (09:12 -0400)]
[lldb][NFC] Refactor test to enable subsequent reuse

On a subsequent commit, I intend to update the expression and call the evaluate
function more times. This refactors enables reusing some of the existing code
for that.

Differential Revision: https://reviews.llvm.org/D155197

12 months ago[lldb][NFC] Factor out CommandObject code filtering results based on scope
Felipe de Azevedo Piovezan [Fri, 14 Jul 2023 17:34:27 +0000 (13:34 -0400)]
[lldb][NFC] Factor out CommandObject code filtering results based on scope

We will need this code in a subsequent commit.

Differential Revision: https://reviews.llvm.org/D155332

12 months ago[X86] Add test coverage for Issue #63946
Simon Pilgrim [Wed, 19 Jul 2023 13:04:30 +0000 (14:04 +0100)]
[X86] Add test coverage for Issue #63946

12 months ago[X86] matchUnaryShuffle - match SIGN_EXTEND_VECTOR_INREG patterns for 'all-signbits...
Simon Pilgrim [Wed, 19 Jul 2023 12:55:11 +0000 (13:55 +0100)]
[X86] matchUnaryShuffle - match SIGN_EXTEND_VECTOR_INREG patterns for 'all-signbits' sources

Adapt the existing ANY/ZERO_EXTEND_VECTOR_INREG shuffle matching to also recognise SIGN_EXTEND_VECTOR_INREG patterns to handle cases where we're effectively "splatting" all-signbits sources.

12 months ago[ARM] Correctly handle execute-only in EmitStructByval
John Brawn [Wed, 5 Jul 2023 15:21:51 +0000 (16:21 +0100)]
[ARM] Correctly handle execute-only in EmitStructByval

Currently when compiling for an execute-only target without movt then
EmitStructByval will generate a constant pool load which isn't
compatible with execute-only. Handle this by emitting tMOVi32imm,
and also simplify the existing movt handling by emitting t2MOVi32imm
or MOVi32imm.

Differential Revision: https://reviews.llvm.org/D154944

12 months ago[ARM] Restructure MOVi32imm expansion to not do pointless instructions
John Brawn [Thu, 6 Jul 2023 09:43:40 +0000 (10:43 +0100)]
[ARM] Restructure MOVi32imm expansion to not do pointless instructions

The expansion of the various MOVi32imm pseudo-instructions works by
splitting the operand into components (either halfwords or bytes) and
emitting instructions to combine those components into the final
result. When the operand is an immediate with some components being
zero this can result in pointless instructions that just add zero.

Avoid this by restructuring things so that a separate function handles
splitting the operand into components, then don't emit the component
if it is a zero immediate. This is straightforward for movw/movt,
where we just don't emit the movt if it's zero, but the thumb1
expansion using mov/add/lsl is more complex, as even when we don't
emit a given byte we still need to get the shift correct.

Differential Revision: https://reviews.llvm.org/D154943

12 months ago[ARM] don't emit constant pool for Thumb1 XO/stack guard combo
Ties Stuij [Wed, 19 Jul 2023 10:26:01 +0000 (11:26 +0100)]
[ARM] don't emit constant pool for Thumb1 XO/stack guard combo

Currently for armv6-m and armv8-m.baseline, we emit constant pool code when we
use execute-only (XO) in combination with stack guards.

XO is a new feature for armv6-m, and this patch is part of a series of patches
that substitutes constant pool generation with the tMOVi32imm equivalent.

However XO for armv8-m.baseline has been available for about 6 years, and so
for armv8-m.baseline this is a bugfix.

Reviewed By: simonwallis2, olista01

Differential Revision: https://reviews.llvm.org/D155170

12 months ago[flang] Simplify macros for the vec_sld and vec_sldw intrinsics
Kelvin Li [Tue, 18 Jul 2023 14:40:39 +0000 (10:40 -0400)]
[flang] Simplify macros for the vec_sld and vec_sldw intrinsics

The ignore_tkr directive is applied to the third argument so that the number of
interfaces can be reduced.

Differential Revision: https://reviews.llvm.org/D155624

12 months ago[clang-tidy] Warn only for copyable/movable classes in cppcoreguidelines-avoid-const...
Carlos Galvez [Mon, 17 Jul 2023 19:54:23 +0000 (19:54 +0000)]
[clang-tidy] Warn only for copyable/movable classes in cppcoreguidelines-avoid-const-or-ref-members

Since that's what the guidelines require.

Fixes #63733

Differential Revision: https://reviews.llvm.org/D155625

12 months ago[RISCV] Replace zihintntl with zicond in ISAInfo unittest
Piyou Chen [Wed, 19 Jul 2023 08:58:02 +0000 (01:58 -0700)]
[RISCV] Replace zihintntl with zicond in ISAInfo unittest

Some ISAInfo unittest base on experimental-zihintntl, but zihintntl will become none-experimental. This patch use another experimental extension zicond to replace zihintntl.

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D155673

12 months ago[TableGen] Deprecate old GI Combiner Emitter
pvanhout [Tue, 11 Jul 2023 10:22:57 +0000 (12:22 +0200)]
[TableGen] Deprecate old GI Combiner Emitter

Will be removed in a month or so.

Reviewed By: aemerson

Differential Revision: https://reviews.llvm.org/D154939

12 months ago[CodeGen] Extend ComplexDeinterleaving pass to recognise patterns using integer types
Igor Kirillov [Mon, 26 Jun 2023 17:26:43 +0000 (17:26 +0000)]
[CodeGen] Extend ComplexDeinterleaving pass to recognise patterns using integer types

AArch64 introduced CMLA and CADD instructions as part of SVE2. This
change allows to generate such instructions when this architecture
feature is available.

Differential Revision: https://reviews.llvm.org/D153808

12 months ago[DAG] hoistLogicOpWithSameOpcodeHands - add support for SIGN_EXTEND_INREG nodes.
Simon Pilgrim [Wed, 19 Jul 2023 10:56:14 +0000 (11:56 +0100)]
[DAG] hoistLogicOpWithSameOpcodeHands - add support for SIGN_EXTEND_INREG nodes.

This can reuse the existing *_EXTEND node handling (with special handling for the valuetype arg)

12 months ago[AMDGPU] Combine the SDAG and GISel versions of the fmed3.ll test.
Ivan Kosarev [Wed, 19 Jul 2023 10:34:25 +0000 (11:34 +0100)]
[AMDGPU] Combine the SDAG and GISel versions of the fmed3.ll test.

Reviewed By: foad

Differential Revision: https://reviews.llvm.org/D155590

12 months ago[AMDGPU][AsmParser][NFC] Translate parsed DS instructions to MCInsts automatically.
Ivan Kosarev [Wed, 19 Jul 2023 10:00:13 +0000 (11:00 +0100)]
[AMDGPU][AsmParser][NFC] Translate parsed DS instructions to MCInsts automatically.

Part of <https://github.com/llvm/llvm-project/issues/62629>.

Reviewed By: foad

Differential Revision: https://reviews.llvm.org/D155620

12 months ago[InstrProf] Ignore -Wcast-qual after D153911 to fix build failure (NFC)
Jie Fu [Wed, 19 Jul 2023 09:50:11 +0000 (17:50 +0800)]
[InstrProf] Ignore -Wcast-qual after D153911 to fix build failure (NFC)

/data/llvm-project/compiler-rt/lib/profile/InstrProfilingMerge.c:46:60: error: cast from 'const char *' to 'struct __llvm_profile_header *' drops const qualifier [-Werror,-Wcast-qual]
  __llvm_profile_header *Header = (__llvm_profile_header *)ProfileData;
                                                           ^
/data/llvm-project/compiler-rt/lib/profile/InstrProfilingMerge.c:49:30: error: cast from 'const char *' to 'struct __llvm_profile_data *' drops const qualifier [-Werror,-Wcast-qual]
      (__llvm_profile_data *)(ProfileData + sizeof(__llvm_profile_header) +
                             ^
/data/llvm-project/compiler-rt/lib/profile/InstrProfilingMerge.c:77:41: error: cast from 'const struct __llvm_profile_data *' to 'struct __llvm_profile_data *' drops const qualifier [-Werror,-Wcast-qual]
       DstData = (__llvm_profile_data *)__llvm_profile_begin_data();
                                        ^
/data/llvm-project/compiler-rt/lib/profile/InstrProfilingMerge.c:109:60: error: cast from 'const char *' to 'struct __llvm_profile_header *' drops const qualifier [-Werror,-Wcast-qual]
  __llvm_profile_header *Header = (__llvm_profile_header *)ProfileData;
                                                           ^
/data/llvm-project/compiler-rt/lib/profile/InstrProfilingMerge.c:116:30: error: cast from 'const char *' to 'struct __llvm_profile_data *' drops const qualifier [-Werror,-Wcast-qual]
      (__llvm_profile_data *)(ProfileData + sizeof(__llvm_profile_header) +
                             ^
/data/llvm-project/compiler-rt/lib/profile/InstrProfilingMerge.c:129:40: error: cast from 'const struct __llvm_profile_data *' to 'struct __llvm_profile_data *' drops const qualifier [-Werror,-Wcast-qual]
      DstData = (__llvm_profile_data *)__llvm_profile_begin_data(),
                                       ^
/data/llvm-project/compiler-rt/lib/profile/InstrProfilingMerge.c:179:34: error: cast from 'const char *' to 'struct ValueProfData *' drops const qualifier [-Werror,-Wcast-qual]
    VPMergeHook((ValueProfData *)SrcValueProfData, DstData);
                                 ^
/data/llvm-project/compiler-rt/lib/profile/InstrProfilingMerge.c:181:46: error: cast from 'const char *' to 'struct ValueProfData *' drops const qualifier [-Werror,-Wcast-qual]
        SrcValueProfData + ((ValueProfData *)SrcValueProfData)->TotalSize;
                                             ^
...

12 months ago[DAG] hoistLogicOpWithSameOpcodeHands - add support for *_EXTEND_VECTOR_INREG nodes.
Simon Pilgrim [Wed, 19 Jul 2023 09:50:07 +0000 (10:50 +0100)]
[DAG] hoistLogicOpWithSameOpcodeHands - add support for *_EXTEND_VECTOR_INREG nodes.

This can reuse the existing *_EXTEND node handling.

12 months ago[bazel] remove PythonAttr TD after 67a910bbff772ebf4c47e8b434b59cdc4820bb68
Mikhail Goncharov [Wed, 19 Jul 2023 09:21:42 +0000 (11:21 +0200)]
[bazel] remove PythonAttr TD after 67a910bbff772ebf4c47e8b434b59cdc4820bb68

Differential Revision: https://reviews.llvm.org/D155686

12 months ago[mlir][nvvm] Introduce Syncronization Ops for WGMMA
Guray Ozen [Wed, 19 Jul 2023 08:29:41 +0000 (10:29 +0200)]
[mlir][nvvm] Introduce Syncronization Ops for WGMMA

This work introduces : `wgmma.fence.aligned`, `wgmma.commit.group.sync.aligned` and `wgmma.wait.group.sync.aligned` Ops. They are used to syncronize warpgroup level matrix multiply-accumulate instructions, as known as WGMMA.

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D155676

12 months ago[AMDGPU] Insert s_nop before s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
Jay Foad [Wed, 19 Jul 2023 07:29:14 +0000 (08:29 +0100)]
[AMDGPU] Insert s_nop before s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)

Differential Revision: https://reviews.llvm.org/D155681

12 months ago[ARM] Add a regression test for D154281
Jay Foad [Mon, 17 Jul 2023 15:20:32 +0000 (16:20 +0100)]
[ARM] Add a regression test for D154281

This is a reduced version of one of the tests that was broken by the
original commit of D154281 "[CodeGen] Store SP adjustment in
MachineBasicBlock. NFCI.".

Differential Revision: https://reviews.llvm.org/D155471

12 months agoRevert "[MachineLICM] Handle Subloops"
Jingu Kang [Wed, 19 Jul 2023 09:30:30 +0000 (10:30 +0100)]
Revert "[MachineLICM] Handle Subloops"

This reverts commit 33e60484d750291e99301e29e60fe72c8fa48ccd.

12 months ago[clang][Docs] Added release note for D142609
Shivam Gupta [Wed, 19 Jul 2023 09:19:43 +0000 (14:49 +0530)]
[clang][Docs] Added release note for D142609

12 months ago[VectorCombine] Use poison insteaf of undef as placeholder [NFC]
Nuno Lopes [Wed, 19 Jul 2023 09:24:41 +0000 (10:24 +0100)]
[VectorCombine] Use poison insteaf of undef as placeholder [NFC]
These vector lanes are never accessed. They are used for shifting a value into the right lane
and therefore only 1 value of the whole vector is actually used

12 months agoReapply: [MemCpyOpt] implement single BB stack-move optimization which unify the...
khei4 [Tue, 18 Jul 2023 09:58:20 +0000 (18:58 +0900)]
Reapply: [MemCpyOpt] implement single BB stack-move optimization which unify the static unescaped allocas

This reverts commit 8f3864ba4323a253bcf29825d23cd325b52c4106.
Differential Revision: https://reviews.llvm.org/D153453

12 months ago[MemCpyOpt] add terminator user test for D153453(NFC)
khei4 [Tue, 18 Jul 2023 09:55:26 +0000 (18:55 +0900)]
[MemCpyOpt] add terminator user test for D153453(NFC)
Differential Revision: https://reviews.llvm.org/D155571