platform/upstream/llvm.git
16 months ago[mlir][linalg][transform][python] Add type arg to MatchOp extension.
Ingo Müller [Tue, 18 Jul 2023 09:07:17 +0000 (09:07 +0000)]
[mlir][linalg][transform][python] Add type arg to MatchOp extension.

The extension class to MatchOp has a class method called match_op_names.
The previous version of that function did not allow to specify the
result type. This, however, may be useful/necessary if the op consuming
the resulting handle requires a particular type (such as the
bufferization.EmptyTensorToAllocTensorOp). This patch adds an overload
to match_op_names that allows to specify the result type.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D155567

16 months ago[X86] LowerEXTEND_VECTOR_INREG - add sign_extend_vector_inreg fast path for all-signb...
Simon Pilgrim [Wed, 19 Jul 2023 09:10:15 +0000 (10:10 +0100)]
[X86] LowerEXTEND_VECTOR_INREG - add sign_extend_vector_inreg fast path for all-signbits source values

If the source operand is already all-signbits we don't need to create the sign extended elements - just splat the source element to the destination element width

16 months ago[mlir][bazel] Fix build.
Oleg Shyshkov [Wed, 19 Jul 2023 09:10:53 +0000 (11:10 +0200)]
[mlir][bazel] Fix build.

16 months ago[libc][NFC] Rename files
Guillaume Chatelet [Tue, 18 Jul 2023 15:29:42 +0000 (15:29 +0000)]
[libc][NFC] Rename files

This patch mostly renames files so it better reflects the function they declare.

Reviewed By: michaelrj

Differential Revision: https://reviews.llvm.org/D155607

16 months ago[lldb] XFAIL command-disassemble-mixed.c on Windows
David Spickett [Wed, 19 Jul 2023 09:04:13 +0000 (09:04 +0000)]
[lldb] XFAIL command-disassemble-mixed.c on Windows

Either clang-cl has different assembly output or we're not mapping
source to assembly properly on Windows. We (Linaro) will find out which.

https://lab.llvm.org/buildbot/#/builders/219/builds/4195

16 months ago[NFC] Adjust test for pr63595
Chuanqi Xu [Wed, 19 Jul 2023 08:50:26 +0000 (16:50 +0800)]
[NFC] Adjust test for pr63595

The original test case is not strictly correct but our ODR checking
system doesn't find it.

16 months agoRevert "[clangd] Remove unused private field 'Opts' in UpdateIndexCallbacks (NFC)"
Kadir Cetinkaya [Wed, 19 Jul 2023 08:45:06 +0000 (10:45 +0200)]
Revert "[clangd] Remove unused private field 'Opts' in UpdateIndexCallbacks (NFC)"

This reverts commit c4fa97fca7e92c736fcb09779c84b42c25ffae70.

16 months ago[libc][memfunctions] Explicit error when platform in not supported
Guillaume Chatelet [Tue, 18 Jul 2023 16:32:04 +0000 (16:32 +0000)]
[libc][memfunctions] Explicit error when platform in not supported

Reviewed By: JonChesterfield, jhuber6

Differential Revision: https://reviews.llvm.org/D155597

16 months ago[LoopPeel] Clear dispositions after peeling
Nikita Popov [Mon, 26 Jun 2023 13:33:09 +0000 (15:33 +0200)]
[LoopPeel] Clear dispositions after peeling

Block dispositions of values defined inside the loop may change
during peeling, so clear them. We already do this for other kinds
of unrolling.

Differential Revision: https://reviews.llvm.org/D153762

16 months agoRevert "[clangd] Always run preamble indexing on a separate thread"
Kadir Cetinkaya [Wed, 19 Jul 2023 07:14:23 +0000 (09:14 +0200)]
Revert "[clangd] Always run preamble indexing on a separate thread"

This reverts commit 036a1b2202cb71aacfa72ef15145a88dc50a02cf.

Triggering failures under tsan, https://lab.llvm.org/buildbot/#/builders/131/builds/48349

16 months ago[Clang][Driver] Pass through the --be8 endian flag to linker in BareMetal driver...
Simi Pallipurath [Fri, 30 Jun 2023 10:48:47 +0000 (11:48 +0100)]
[Clang][Driver] Pass through the --be8 endian flag to linker in BareMetal driver For Arm.

When linking a big-endian image for Arm, clang has
to select  between BE8 and BE32 formats. The default
is dependent on the selected target architecture.
For ARMv6 and later architectures the default is
BE8, for older architectures the default is BE32.
For BE8 and BE32, compiler outputs a big endian ELF
relocatable object file with the instructions and
data both big endian. The difference is that at
link time, for BE8 a linker must endian reverse
the instructions to little endian. For BE8, the
clang has to pass --be8 to the linker for Arm.

At the moment clang is not passing the --be8 flag
to linker for the baremetal target architectures
above ArmV6 for Arm. This patch passes through --be8
and -BE or EL to the linker, taking into account the
target and the -mbig-endian and -mlittle-endian flag.
Also there are few more changes in the baremetal
driver so that the code can cope with AArch64 being
big-endian as well.

Reviewed By: michaelplatings, MaskRay

Differential Revision: https://reviews.llvm.org/D154786

16 months ago[clangd] Make an include always refer to itself. Background: clang-review expects...
Viktoriia Bakalova [Tue, 18 Jul 2023 15:48:24 +0000 (15:48 +0000)]
[clangd] Make an include always refer to itself. Background: clang-review expects all referents to have definition, declaration or reference(s).

Differential Revision: https://reviews.llvm.org/D155614

16 months ago[AMDGPU] Regenerate is.fpclass checks
Jay Foad [Wed, 19 Jul 2023 08:20:08 +0000 (09:20 +0100)]
[AMDGPU] Regenerate is.fpclass checks

16 months ago[libcxx] Link to fewer MSVC CRT libraries
Martin Storsjö [Sun, 16 Jul 2023 21:35:04 +0000 (00:35 +0300)]
[libcxx] Link to fewer MSVC CRT libraries

The library msvcrt.lib pulls in ucrt.lib and vcruntime.lib anyway,
there's no need to manually link against the individual dependencies.

This matches how the tests link against libraries - they only link
against msvcrt and msvcprt, not directly against ucrt and vcruntime.

Differential Revision: https://reviews.llvm.org/D155555

16 months ago[libcxx] [test] Remove a stray double space from a test config. NFC.
Andrew Ng [Fri, 14 Jul 2023 08:49:24 +0000 (11:49 +0300)]
[libcxx] [test] Remove a stray double space from a test config. NFC.

Differential Revision: https://reviews.llvm.org/D155553

16 months ago[mlir] Update SVE integration tests to use mlir-cpu-runner
Andrzej Warzynski [Mon, 10 Apr 2023 13:43:39 +0000 (13:43 +0000)]
[mlir] Update SVE integration tests to use mlir-cpu-runner

With the recent addition of "-mattr" and "-march" to the list of options
supported by mlir-cpu-runner [1], the SVE integration
tests can be updated to use mlir-cpu-runner instead of lli. This will
allow better code re-use and more consistency

This patch updates 2 tests to demonstrate the new logic. The remaining
tests will be updated in the follow-up patches.

[1] https://reviews.llvm.org/D146917

Depends on D155403

Differential Revision: https://reviews.llvm.org/D155405

16 months ago[CMake] Clean up old code for handling MSVC runtime setting the old way
Martin Storsjö [Mon, 17 Jul 2023 08:16:05 +0000 (11:16 +0300)]
[CMake] Clean up old code for handling MSVC runtime setting the old way

This was left in place to reduce the risk of breaking anything,
and to keep the diff smaller, in D155233.

Differential Revision: https://reviews.llvm.org/D155431

16 months ago[IR] Deprecate opaque pointer compatibility APIs
Nikita Popov [Tue, 18 Jul 2023 10:13:59 +0000 (12:13 +0200)]
[IR] Deprecate opaque pointer compatibility APIs

This deprecates various compatibility APIs that have been
introduced as part of the opaque pointer migration.

These will be removed at some point after the LLVM 17 release.

Differential Revision: https://reviews.llvm.org/D155585

16 months agoReland [AArch64] Fix an immediate out of range for large realignments on Windows
Martin Storsjö [Mon, 17 Jul 2023 12:19:14 +0000 (15:19 +0300)]
Reland [AArch64] Fix an immediate out of range for large realignments on Windows

Also add a missing FrameSetup flag on the existing add instruction.

This fixes https://github.com/llvm/llvm-project/issues/63701.

Since the previous iteration, change ADDXrr to ADDXrx64, which
works with this use of SP.

Differential Revision: https://reviews.llvm.org/D155447

16 months ago[X86] Add PBNDKB instruction.
Freddy Ye [Wed, 19 Jul 2023 08:14:06 +0000 (16:14 +0800)]
[X86] Add PBNDKB instruction.

For more details about this instruction, please refer to the latest ISE document: https://www.intel.com/content/www/us/en/develop/download/intel-architecture-instruction-set-extensions-programming-reference.html

Reviewed By: pengfei, skan

Differential Revision: https://reviews.llvm.org/D155142

16 months ago[NFC] Add a blank line after directives in StandardCPlusPlusModules.rst
Chuanqi Xu [Wed, 19 Jul 2023 08:10:55 +0000 (16:10 +0800)]
[NFC] Add a blank line after  directives in StandardCPlusPlusModules.rst

16 months ago[clang][analyzer] StdLibraryFunctionsChecker: Allow NULL buffer in `fread` and `fwrit...
Balázs Kéri [Wed, 19 Jul 2023 07:23:32 +0000 (09:23 +0200)]
[clang][analyzer] StdLibraryFunctionsChecker: Allow NULL buffer in `fread` and `fwrite` if size is zero.

Reviewed By: donat.nagy

Differential Revision: https://reviews.llvm.org/D154509

16 months ago[NFC][AMDGPULowerModuleLDSPass] Use shorter APIs in markUsedByKernel
Juan Manuel MARTINEZ CAAMAÑO [Wed, 19 Jul 2023 07:37:42 +0000 (09:37 +0200)]
[NFC][AMDGPULowerModuleLDSPass] Use shorter APIs in markUsedByKernel

* Use shorter versions of the LLVM API

Reviewed By: JonChesterfield

Differential Revision: https://reviews.llvm.org/D155589

16 months ago[NFC][AMDGPULowerModuleLDSPass] Cleanup of getTableLookupKernelIndex
Juan Manuel MARTINEZ CAAMAÑO [Wed, 19 Jul 2023 07:37:29 +0000 (09:37 +0200)]
[NFC][AMDGPULowerModuleLDSPass] Cleanup of getTableLookupKernelIndex

* Do a single lookup when querying the map
* Use shorter versions of the LLVM API

Reviewed By: JonChesterfield

Differential Revision: https://reviews.llvm.org/D155588

16 months ago[mlir][LLVM] add `llvm.ssa.copy` intrinsic
Markus Böck [Wed, 19 Jul 2023 07:32:17 +0000 (09:32 +0200)]
[mlir][LLVM] add `llvm.ssa.copy` intrinsic

This is quite the niche intrinsic, whose whole purpose is to be able to essentially split an SSA value to be able to attach additional information to the new value.
It therefore really acts like a noop that just passes through its argument. It interestingly does not have any support in CodeGen and is therefore required to also be deleted by any pass creating it.

Differential Revision: https://reviews.llvm.org/D155678

16 months ago[IR] Remove Type::getPointerElementType() (NFC)
Nikita Popov [Wed, 19 Jul 2023 07:40:48 +0000 (09:40 +0200)]
[IR] Remove Type::getPointerElementType() (NFC)

This method has been deprecated since the LLVM 15 release and
can no longer meaningfully be used, so remove it.

16 months ago[NFC] [C++20] [Modules] Add a performance tip to the document
Chuanqi Xu [Wed, 19 Jul 2023 07:46:52 +0000 (15:46 +0800)]
[NFC] [C++20] [Modules] Add a performance tip to the document

It is a known problem that the compiler isn't efficiency when there are
a lot of duplications in the modules to developers. But it may not be
verbose to the (potential) new users of C++20 modules.

It may be necessary to mention this in the document.

16 months ago[clangd] Remove unused private field 'Opts' in UpdateIndexCallbacks (NFC)
Jie Fu [Wed, 19 Jul 2023 07:28:38 +0000 (15:28 +0800)]
[clangd] Remove unused private field 'Opts' in UpdateIndexCallbacks (NFC)

/Users/jiefu/llvm-project/clang-tools-extra/clangd/ClangdServer.cpp:167:32: error: private field 'Opts' is not used [-Werror,-Wunused-private-field]
  const ClangdServer::Options &Opts;
                               ^
1 error generated.

16 months ago[mlir][NFC] Avoid `OpBuilder::setListener` when possible
Matthias Springer [Wed, 19 Jul 2023 06:59:02 +0000 (08:59 +0200)]
[mlir][NFC] Avoid `OpBuilder::setListener` when possible

`setListener` is dangerous because an already registered listener may accidentally be overwritten/replaced. (A `ForwardingListener` must be used in such cases.) This change updates a few trivial call sites of `setListener`, where no forwarding listener is needed.

Differential Revision: https://reviews.llvm.org/D155599

16 months ago[RISCV] Make SubtargetFeature description strings consistent with AssemblerPredicate.
Craig Topper [Wed, 19 Jul 2023 07:12:19 +0000 (00:12 -0700)]
[RISCV] Make SubtargetFeature description strings consistent with AssemblerPredicate.

Mostily capitalizing 'Z' for multi-letter extensions.

16 months ago[analyzer] Model lifetime of a variable declared in for condition in CFG correctly
tomasz-kaminski-sonarsource [Tue, 18 Jul 2023 14:55:50 +0000 (16:55 +0200)]
[analyzer] Model lifetime of a variable declared in for condition in CFG correctly

Per [stmt.for] p1 (https://eel.is/c++draft/stmt.for#1) the following
`for` and `while` statements are equivalent
```
for (; A c = b; b.c) {
  A d;
}

while (A c = b) {
  A d;
  b.c;
}
```
As a consequence, the variable declared for the condition expression
should be destroyed after the increment expression.

This fixed the handling of the object lifetime `continue`, and now
destructors, scope and lifetime elements are present for continue
path in following code:
```
for (; A c = b; b.c) {
  if (cond)
     continue;
  A d;
}
```

Reviewed By: xazax.hun

Differential Revision: https://reviews.llvm.org/D155547

16 months ago[mlir][nfc] Clarify the limitation on scalable vectors
Andrzej Warzynski [Fri, 14 Jul 2023 11:55:13 +0000 (11:55 +0000)]
[mlir][nfc] Clarify the limitation on scalable vectors

When converting/lowering to LLVM, MLIR does not support n-D scalable
vectors at the moment. This patch simply clarifies an assert that's
meant to catch such unsupported cases.

Note that we may want to relax this condition in the future.

Differential Revision: https://reviews.llvm.org/D154302

16 months ago[examples] Fix -Wcast-qual in OrcV2Examples after D153911 (NFC)
Jie Fu [Wed, 19 Jul 2023 06:37:12 +0000 (14:37 +0800)]
[examples] Fix -Wcast-qual in OrcV2Examples after D153911 (NFC)

/Users/jiefu/llvm-project/llvm/examples/OrcV2Examples/OrcV2CBindingsBasicUsage/OrcV2CBindingsBasicUsage.c:77:52: error: cast from 'char **' to 'const char **' must have all intermediate pointers const qualified to be safe [-Werror,-Wcast-qual]
  LLVMParseCommandLineOptions(argc, (const char **)argv, "");
                                                   ^
/Users/jiefu/llvm-project/llvm/examples/OrcV2Examples/OrcV2CBindingsAddObjectFile/OrcV2CBindingsAddObjectFile.c:63:52: error: cast from 'char **' to 'const char **' must have all intermediate pointers const qualified to be safe [-Werror,-Wcast-qual]
  LLVMParseCommandLineOptions(argc, (const char **)argv, "");
                                                   ^
/Users/jiefu/llvm-project/llvm/examples/OrcV2Examples/OrcV2CBindingsDumpObjects/OrcV2CBindingsDumpObjects.c:62:52: error: cast from 'char **' to 'const char **' must have all intermediate pointers const qualified to be safe [-Werror,-Wcast-qual]
  LLVMParseCommandLineOptions(argc, (const char **)argv, "");
                                                   ^
...

16 months ago[Clang] Fix the location of default init expressions
Corentin Jabot [Tue, 18 Jul 2023 10:16:16 +0000 (12:16 +0200)]
[Clang] Fix the location of default init expressions

Default member initializations constructed from
a parenthesized aggregate initialization should be constructed
at the location of the left paren, to be consistent with
brace initialization.
Otherwise we get diagmostics and source_location in the wrong places.

Fixes #63903

Reviewed By: aaron.ballman, #clang-language-wg

Differential Revision: https://reviews.llvm.org/D155573

16 months ago[clangd] Always run preamble indexing on a separate thread
Kadir Cetinkaya [Tue, 18 Jul 2023 16:12:32 +0000 (18:12 +0200)]
[clangd] Always run preamble indexing on a separate thread

This has been the default in our production setup for weeks now,
showing great improvements to latency and no problems around stability or
correctness of the results.

Differential Revision: https://reviews.llvm.org/D155619

16 months ago[Clang] Fix -Wconstant-logical-operand when LHS is a constant
Shivam Gupta [Wed, 19 Jul 2023 05:01:35 +0000 (10:31 +0530)]
[Clang] Fix -Wconstant-logical-operand when LHS is a constant

This fix PR37919

The below code produces -Wconstant-logical-operand for the first statement,
but not the second.

void foo(int x) {
if (x && 5) {}
if (5 && x) {}
}

Reviewed By: nickdesaulniers

Differential Revision: https://reviews.llvm.org/D142609

16 months ago[NFC] [C++20] [Modules] Add C++20 modules related changes to Release Note
Chuanqi Xu [Wed, 19 Jul 2023 06:11:32 +0000 (14:11 +0800)]
[NFC] [C++20] [Modules] Add C++20 modules related changes to Release Note

Given LLVM17 is going to be branched, add some uncovered C++20 modules related changes
to the Release Notes.

16 months agoRevert "[NFC] [C++20] [Modules] Add C++20 modules related changes to Release Note"
Chuanqi Xu [Wed, 19 Jul 2023 06:28:07 +0000 (14:28 +0800)]
Revert "[NFC] [C++20] [Modules] Add C++20 modules related changes to Release Note"

This reverts commit 95bdd6eed6f3f4b8a95523516120068ac18eccff. The bot
shows there are some grammar issues.

16 months ago[mlir] Remove dead code in Analysis/FlatLinearValueConstraints.
Alexander Belyaev [Mon, 10 Jul 2023 10:40:50 +0000 (12:40 +0200)]
[mlir] Remove dead code in Analysis/FlatLinearValueConstraints.

Differential Revision: https://reviews.llvm.org/D154830

16 months ago[NFC] [C++20] [Modules] Add C++20 modules related changes to Release Note
Chuanqi Xu [Wed, 19 Jul 2023 06:11:32 +0000 (14:11 +0800)]
[NFC] [C++20] [Modules] Add C++20 modules related changes to Release Note

Given LLVM17 is going to be branched, add some uncovered C++20 modules related changes
to the Release Notes.

16 months ago[test] Make fuzzer/value-profile-div.test x86 specific
Fangrui Song [Wed, 19 Jul 2023 05:11:27 +0000 (22:11 -0700)]
[test] Make fuzzer/value-profile-div.test x86 specific

The test requires that an integer division by zero causes a trap, leading to a
signal like SIGFPE. This is the case on x86. On many other architectures, such
as Arm (unless software division is used and `__aeabi_idiv0` is patched),
Power, RISC-V, LoongArch, there is no trap. Therefore, it's more appropriate to
invert the condition to run the test.

Reviewed By: Ami-zhang

Differential Revision: https://reviews.llvm.org/D155644

16 months ago[Demangle] Include <exception> for IWYU
Fangrui Song [Wed, 19 Jul 2023 05:10:58 +0000 (22:10 -0700)]
[Demangle] Include <exception> for IWYU

The called std::terminate currently gets the declaration transitively
through llvm/Demangle/Utility.h, removing <exception> from Utility.h
would break ItaniumDemangle.cpp.

16 months agoRevert "[LoongArch] Change 'using namespace llvm;' to 'namespace llvm {' in LoongArch...
Weining Lu [Wed, 19 Jul 2023 05:07:06 +0000 (13:07 +0800)]
Revert "[LoongArch] Change 'using namespace llvm;' to 'namespace llvm {' in LoongArchTargetParser.cpp. NFC"

This reverts commit 586321467a0d48573ae643e837a6c4eaf6bc75db.

Reason to revert:
These changes are inconsistent with the [[https://llvm.org/docs/CodingStandards.html#use-namespace-qualifiers-to-implement-previously-declared-functions|LLVM stype guide]].

16 months ago[tools] Use "#pragma GCC" instead of "#pragma clang" to ignore -Wcast-qual in c-index...
Jie Fu [Wed, 19 Jul 2023 04:05:05 +0000 (12:05 +0800)]
[tools] Use "#pragma GCC" instead of "#pragma clang" to ignore -Wcast-qual in c-index-test.c (NFC)

16 months ago[AVR] Expand shifts of all types except int8 and int16
Patryk Wychowaniec [Wed, 19 Jul 2023 03:54:34 +0000 (11:54 +0800)]
[AVR] Expand shifts of all types except int8 and int16

Currently our AVRShiftExpand pass expands only 32-bit shifts, with the
assumption that other kinds of shifts (e.g. 64-bit ones) are
automatically reduced to 8-bit ones by LLVM during ISel.

However this is not always true and causes problems in the rust-lang runtime.

This commit changes the logic a bit, so that instead of expanding only
32-bit shifts, we expand shifts of all types except 8-bit and 16-bit.

This is not the most optimal solution, because 64-bit shifts can be
expanded to 32-bit shifts which has been deeply optimized.

I've checked the generated code using rustc + simavr, and all shifts
seem to behave correctly.

Spotted in the wild in rustc:
https://github.com/rust-lang/compiler-builtins/issues/523
https://github.com/rust-lang/rust/issues/112140

Reviewed By: benshi001

Differential Revision: https://reviews.llvm.org/D154785

16 months ago[AVR] Enable verifyInstructionPredicates for AVR
Jianjian GUAN [Tue, 18 Jul 2023 09:37:01 +0000 (17:37 +0800)]
[AVR] Enable verifyInstructionPredicates for AVR

This patch fixes the failed test of verifyInstructionPredicates which is caused by verifyInstructionPredicates. verifyInstructionPredicates will add JMPk without checking the target predicate.

Reviewed By: benshi001

Differential Revision: https://reviews.llvm.org/D155570

16 months ago[tools] Fix buildbot build failure
Jie Fu [Wed, 19 Jul 2023 03:39:00 +0000 (11:39 +0800)]
[tools] Fix buildbot build failure

/buildbot/worker/arc-folder/llvm-project/clang/tools/c-index-test/c-index-test.c:234: warning: ignoring '#pragma clang diagnostic' [-Wunknown-pragmas]
/buildbot/worker/arc-folder/llvm-project/clang/tools/c-index-test/c-index-test.c:235: warning: ignoring '#pragma clang diagnostic' [-Wunknown-pragmas]
/buildbot/worker/arc-folder/llvm-project/clang/tools/c-index-test/c-index-test.c:236:10: warning: cast discards 'const' qualifier from pointer target type [-Wcast-qual]
/buildbot/worker/arc-folder/llvm-project/clang/tools/c-index-test/c-index-test.c:237:10: warning: cast discards 'const' qualifier from pointer target type [-Wcast-qual]
/buildbot/worker/arc-folder/llvm-project/clang/tools/c-index-test/c-index-test.c:238: warning: ignoring '#pragma clang diagnostic' [-Wunknown-pragmas]
/buildbot/worker/arc-folder/llvm-project/clang/tools/c-index-test/c-index-test.c:3765: warning: ignoring '#pragma clang diagnostic' [-Wunknown-pragmas]
/buildbot/worker/arc-folder/llvm-project/clang/tools/c-index-test/c-index-test.c:3766: warning: ignoring '#pragma clang diagnostic' [-Wunknown-pragmas]
/buildbot/worker/arc-folder/llvm-project/clang/tools/c-index-test/c-index-test.c:3767:10: warning: cast discards 'const' qualifier from pointer target type [-Wcast-qual]
/buildbot/worker/arc-folder/llvm-project/clang/tools/c-index-test/c-index-test.c:3768: warning: ignoring '#pragma clang diagnostic' [-Wunknown-pragmas]

16 months ago[NFC][XCOFF] Remove curly braces from single line if statement
Jake Egan [Wed, 19 Jul 2023 03:33:48 +0000 (23:33 -0400)]
[NFC][XCOFF] Remove curly braces from single line if statement

Missed this detail from D155199.

16 months ago[LoongArch] Align functions and loops better according to uarch
WANG Xuerui [Wed, 19 Jul 2023 01:20:53 +0000 (09:20 +0800)]
[LoongArch] Align functions and loops better according to uarch

The LA464 micro-architecture is very sensitive to alignment of hot code,
with performance variation of up to ~12% in the go1 benchmark suite of
the Go language (as observed by me during my work on the Go loong64
port).
[[ https://go.dev/cl/479816 | Manual alignment of certain loops ]] and [[ https://go.dev/cl/479817 | automatic alignment of loop heads ]]
helps a lot there, by reducing much of the random variation and
generally increasing performance, so we naturally want to do the same
here.

Practically, LA464 is the only LoongArch micro-architecture in wide use,
and we are currently supporting just that. The first "4" in "LA464"
stands for "4-issue", in particular its instruction fetch and decode
stages are 4-wide; so functions and branch targets should be preferably
aligned to at least 16 bytes for best throughput.

The Loongson team has benchmarked various combinations of function,
loop, and branch target alignments with GCC.
[[ https://gcc.gnu.org/pipermail/gcc-patches/2023-May/619980.html | The results ]]
show that "16-byte label alignment together with 32-byte function
alignment gives best results in terms of SPEC score". A "label" in GCC
means a branch target; while we don't currently align branch targets,
we do align loops, so in this patch we default to 32-byte function
alignment and 16-byte loop alignment.

Reviewed By: SixWeining

Differential Revision: https://reviews.llvm.org/D148622

16 months ago[mlir] Fix -Wcast-qual in mlir/test/CAPI/ir.c after D153911 (NFC)
Jie Fu [Wed, 19 Jul 2023 03:16:55 +0000 (11:16 +0800)]
[mlir] Fix -Wcast-qual in mlir/test/CAPI/ir.c after D153911 (NFC)

/Users/jiefu/llvm-project/mlir/test/CAPI/ir.c:1120:18: error: cast from 'const void *' to 'unsigned char *' drops const qualifier [-Werror,-Wcast-qual]
      (uint8_t *)mlirDenseElementsAttrGetRawData(uint8Elements);
                 ^
/Users/jiefu/llvm-project/mlir/test/CAPI/ir.c:1121:35: error: cast from 'const void *' to 'signed char *' drops const qualifier [-Werror,-Wcast-qual]
  int8_t *int8RawData = (int8_t *)mlirDenseElementsAttrGetRawData(int8Elements);
                                  ^
/Users/jiefu/llvm-project/mlir/test/CAPI/ir.c:1123:19: error: cast from 'const void *' to 'unsigned int *' drops const qualifier [-Werror,-Wcast-qual]
      (uint32_t *)mlirDenseElementsAttrGetRawData(uint32Elements);
                  ^
/Users/jiefu/llvm-project/mlir/test/CAPI/ir.c:1125:18: error: cast from 'const void *' to 'int *' drops const qualifier [-Werror,-Wcast-qual]
      (int32_t *)mlirDenseElementsAttrGetRawData(int32Elements);
                 ^
/Users/jiefu/llvm-project/mlir/test/CAPI/ir.c:1127:19: error: cast from 'const void *' to 'unsigned long long *' drops const qualifier [-Werror,-Wcast-qual]
      (uint64_t *)mlirDenseElementsAttrGetRawData(uint64Elements);
                  ^
/Users/jiefu/llvm-project/mlir/test/CAPI/ir.c:1129:18: error: cast from 'const void *' to 'long long *' drops const qualifier [-Werror,-Wcast-qual]
      (int64_t *)mlirDenseElementsAttrGetRawData(int64Elements);
                 ^
/Users/jiefu/llvm-project/mlir/test/CAPI/ir.c:1130:34: error: cast from 'const void *' to 'float *' drops const qualifier [-Werror,-Wcast-qual]
  float *floatRawData = (float *)mlirDenseElementsAttrGetRawData(floatElements);
                                 ^
/Users/jiefu/llvm-project/mlir/test/CAPI/ir.c:1132:17: error: cast from 'const void *' to 'double *' drops const qualifier [-Werror,-Wcast-qual]
      (double *)mlirDenseElementsAttrGetRawData(doubleElements);
                ^
/Users/jiefu/llvm-project/mlir/test/CAPI/ir.c:1134:19: error: cast from 'const void *' to 'unsigned short *' drops const qualifier [-Werror,-Wcast-qual]
      (uint16_t *)mlirDenseElementsAttrGetRawData(bf16Elements);
                  ^
/Users/jiefu/llvm-project/mlir/test/CAPI/ir.c:1136:19: error: cast from 'const void *' to 'unsigned short *' drops const qualifier [-Werror,-Wcast-qual]
      (uint16_t *)mlirDenseElementsAttrGetRawData(f16Elements);
                  ^
10 errors generated.

16 months ago[tools] Ignore -Wcast-qual in c-index-test.c after D153911 (NFC)
Jie Fu [Wed, 19 Jul 2023 02:55:12 +0000 (10:55 +0800)]
[tools] Ignore -Wcast-qual in c-index-test.c after D153911 (NFC)

/Users/jiefu/llvm-project/clang/tools/c-index-test/c-index-test.c:234:18: error: cast from 'const char *' to 'char *' drops const qualifier [-Werror,-Wcast-qual]
    free((char *)unsaved_files[i].Filename);
                 ^
/Users/jiefu/llvm-project/clang/tools/c-index-test/c-index-test.c:235:18: error: cast from 'const char *' to 'char *' drops const qualifier [-Werror,-Wcast-qual]
    free((char *)unsaved_files[i].Contents);
                 ^
/Users/jiefu/llvm-project/clang/tools/c-index-test/c-index-test.c:3762:32: error: cast from 'const char *' to 'void *' drops const qualifier [-Werror,-Wcast-qual]
  return (CXIdxClientContainer)"TU";
                               ^
3 errors generated.

16 months ago[RISCV] Use the stack for MVT::f16 for fastcc when there are no other registers available
eopXD [Mon, 17 Jul 2023 19:30:00 +0000 (12:30 -0700)]
[RISCV] Use the stack for MVT::f16 for fastcc when there are no other registers available

In D155502, we added code for the compiler to check GPR-s for f16
under zhinx. This commit adds code to hit the stack when we run out of
GPR-s.

With this patch and D155502, resolves #63922

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D155507

16 months ago[BOLT][Utils] Pass cmp-rev to nfc-check-setup
Amir Ayupov [Wed, 19 Jul 2023 02:43:08 +0000 (19:43 -0700)]
[BOLT][Utils] Pass cmp-rev to nfc-check-setup

Pass the revision to checkout to (cmp-rev) as nfc-check-setup option.
Simpifies the comparison against arbitrary commit, not just the previous one.

Reviewed By: #bolt, rafauler

Differential Revision: https://reviews.llvm.org/D155657

16 months ago[Clang][RISCV] Improve diagnostic message for full multiply intrinsics
eopXD [Mon, 17 Jul 2023 03:18:11 +0000 (20:18 -0700)]
[Clang][RISCV] Improve diagnostic message for full multiply intrinsics

The full multiply intrinsics are not included for EEW=64 in Zve64*.
They require the V extension to be enabled.

This commit improves diagnostic message from

```
<source>:4:10: error: call to undeclared function '__riscv_vsmul_vv_i64m1';
    4 |   return __riscv_vsmul_vv_i64m1(op1, op2, __RISCV_VXRM_RNU, vl);
```

to

```
test.c:5:10: error: builtin requires: v
    5 |   return __riscv_vsmul_vv_i64m1(op1, op2, __RISCV_VXRM_RNU, vl);
```

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D155416

16 months ago[NFC][XCOFF] Use common function to calculate file offset
Jake Egan [Wed, 19 Jul 2023 02:08:46 +0000 (22:08 -0400)]
[NFC][XCOFF] Use common function to calculate file offset

The file offset code is repeated in nearly identical form for every derivation of SectionEntry, so make it into a method in SectionEntry instead.

Reviewed By: scott.linder

Differential Revision: https://reviews.llvm.org/D155199

16 months ago[Clang][RISCV] Guard RVV intrinsics types that is not available when ELEN < 64
eopXD [Mon, 17 Jul 2023 02:24:09 +0000 (19:24 -0700)]
[Clang][RISCV] Guard RVV intrinsics types that is not available when ELEN < 64

(ELEN, LMUL) pairs of (8, mf8), (16, mf4), (32, mf2), (64, m1)
requires at least `zve64x`.

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D155414

16 months ago[clang-format][doc] Replace single back quotes with double ones
Owen Pan [Wed, 19 Jul 2023 01:23:55 +0000 (18:23 -0700)]
[clang-format][doc] Replace single back quotes with double ones

16 months agoFix types of arm64 MSVC __readx18/__writex18 intrinsics
Akira Hatanaka [Wed, 19 Jul 2023 01:15:56 +0000 (18:15 -0700)]
Fix types of arm64 MSVC __readx18/__writex18 intrinsics

Using `L` for type `long` results in clang passing 64-bit integers to
these intrinsics on LP64 operating systems. This isn't correct as the
intrinsics accept 32-bit integers.

Use `N` instead of `L` so that 32-bit integers are passed to the
intrinsics on LP64 operating systems too. This is the same fix as the
following two commits:

33703fb9f908113f93bd9af83a79eb56f5131735
afa47c91ce5085d446ebb5ac1312dc98b6a68a6c

Differential Revision: https://reviews.llvm.org/D155012

16 months ago[gn build] Port ef70fe4d264d
LLVM GN Syncbot [Wed, 19 Jul 2023 01:09:48 +0000 (01:09 +0000)]
[gn build] Port ef70fe4d264d

16 months ago[gn] port 20341c3ad6f64a
Nico Weber [Wed, 19 Jul 2023 01:09:32 +0000 (21:09 -0400)]
[gn] port 20341c3ad6f64a

16 months ago[gn] port 3f65f718332c
Nico Weber [Wed, 19 Jul 2023 01:00:16 +0000 (21:00 -0400)]
[gn] port 3f65f718332c

16 months ago[ELF] --icf: switch to xxh3_64bits
Fangrui Song [Wed, 19 Jul 2023 00:57:31 +0000 (17:57 -0700)]
[ELF] --icf: switch to xxh3_64bits

for more efficient section content hashing. Also clean up the tests a
bit and properly test the formatting of --print-icf-sections.

16 months ago[OpenMP][NFCI] Avoid storing non-constant values in ICV
Johannes Doerfert [Tue, 18 Jul 2023 23:33:31 +0000 (16:33 -0700)]
[OpenMP][NFCI] Avoid storing non-constant values in ICV

If we store a constant in an ICV it is easier for the optimizer to
propagate it. Since we often use the full block for the thread limit and
the parallel team size, we can instead replace that dynamic value with a
constant that otherwise cannot occur, here 0.

16 months ago[OpenMP][NFCI] Split assertion message from assertion expression
Johannes Doerfert [Tue, 18 Jul 2023 23:05:08 +0000 (16:05 -0700)]
[OpenMP][NFCI] Split assertion message from assertion expression

We ended up with `llvm.assume(icmp ne ptr as(4) null, as(4) @str)`
because the string in address space 4 was not known to be non-null.
There is no need to create these assumes.

16 months ago[cmake] Add -Wcast-qual to C flags if LLVM_ENABLE_WARNINGS is defined.
Alex MacLean [Tue, 18 Jul 2023 23:44:26 +0000 (23:44 +0000)]
[cmake] Add -Wcast-qual to C flags if LLVM_ENABLE_WARNINGS is defined.

Disable this warning for specific cast in llvm_regcomp().

Reviewed By: dblaikie

Differential Revision: https://reviews.llvm.org/D153911

16 months agoPseudoObjectExpr: Prefer ArrayRef over iterator_range when iterating with pointers
David Blaikie [Tue, 18 Jul 2023 23:40:19 +0000 (23:40 +0000)]
PseudoObjectExpr: Prefer ArrayRef over iterator_range when iterating with pointers

Simpler to use ArrayRef directly here rather than a more
generic/customizable range helper like iterator_range

16 months ago[RISCV] Simplify VROR_IV_V_X_I multiclass. NFC
Craig Topper [Tue, 18 Jul 2023 23:39:40 +0000 (16:39 -0700)]
[RISCV] Simplify VROR_IV_V_X_I multiclass. NFC

Remove template parameters that always have their default value.

16 months ago[RISCV] Remove unused classes VPseudoTernary, VPseudoTernaryNoMaskNoPolicy, and VPseu...
Michael Maitland [Tue, 18 Jul 2023 20:08:58 +0000 (13:08 -0700)]
[RISCV] Remove unused classes VPseudoTernary, VPseudoTernaryNoMaskNoPolicy, and VPseudoConversionW_V

These classes are dead and can be removed.

Differential Revision: https://reviews.llvm.org/D155634

16 months ago[libc][NFC] reuse variable in printf string conv
Michael Jones [Tue, 11 Jul 2023 22:06:08 +0000 (15:06 -0700)]
[libc][NFC] reuse variable in printf string conv

The amount of spaces to pad with is stored in the variable
padding_spaces, previously the actual write calls used the same formula
to calculate the value. This simplifies and clarifies the values by just
reusing the variable.

Reviewed By: lntue

Differential Revision: https://reviews.llvm.org/D155113

16 months ago[libc] Set min precision for strtofloat fuzz
Michael Jones [Wed, 12 Jul 2023 22:21:52 +0000 (15:21 -0700)]
[libc] Set min precision for strtofloat fuzz

MPFR has a minimum precision of 2, but the strtofloat fuzz sometimes
would request a precision of 1 for the case of the minimum subnormal.
This patch tells the fuzzer to ignore any case where the precision would
go below 2.

Reviewed By: lntue

Differential Revision: https://reviews.llvm.org/D155130

16 months ago[Clang] Add warnings for CWG2521
Po-yao Chang [Thu, 13 Jul 2023 15:42:51 +0000 (23:42 +0800)]
[Clang] Add warnings for CWG2521

1. Teach -Wuser-defined-literals to warn on using double underscores in
   literal suffix identifiers.
2. Add -Wdeprecated-literal-operator to warn about the use of the first
   grammar production of literal-operator-id, which defaults to off for now.

Differential Revision: https://reviews.llvm.org/D152632

16 months ago[lld] A Unified LTO Bitcode Frontend
Matthew Voss [Tue, 18 Jul 2023 23:13:58 +0000 (16:13 -0700)]
[lld] A Unified LTO Bitcode Frontend

The unified LTO pipeline creates a single LTO bitcode structure that can
be used by Thin or Full LTO. This means that the LTO mode can be chosen
at link time and that all LTO bitcode produced by the pipeline is
compatible, from an optimization perspective. This makes the behavior of
LTO a bit more predictable by normalizing the set of LTO features
supported by each LTO bitcode file.

Example usage:

clang -flto -funified-lto -fuse-ld=lld foo.c

clang -flto=thin -funified-lto -fuse-ld=lld foo.c

clang -c -flto -funified-lto foo.c  # -flto={full,thin} are identical in
terms of compilation actions
clang -flto=thin -fuse-ld=lld foo.o # pass --lto=thin to ld.lld

clang -c -flto -funified-lto foo.c clang -flto -fuse-ld=lld foo.o

The RFC discussing the details and rational for this change is here:
https://discourse.llvm.org/t/rfc-a-unified-lto-bitcode-frontend/61774

Differential Revision: https://reviews.llvm.org/D123805

16 months ago[clang][docs] Defensively turn off exception behavior in dump_ast_matchers.py
Rashmi Mudduluru [Tue, 18 Jul 2023 22:27:33 +0000 (15:27 -0700)]
[clang][docs] Defensively turn off exception behavior in dump_ast_matchers.py

Differential Revision: https://reviews.llvm.org/D155134

16 months ago[clang][docs] Update LibASTMatchersReference.html
Rashmi Mudduluru [Fri, 14 Jul 2023 15:50:15 +0000 (08:50 -0700)]
[clang][docs] Update LibASTMatchersReference.html

Differential Revision: https://reviews.llvm.org/D155304

16 months ago[clang][dataflow] Add function to `WatchedLiteralsSolver` that reports whether the...
Yitzhak Mandelbaum [Tue, 18 Jul 2023 20:28:35 +0000 (20:28 +0000)]
[clang][dataflow] Add function to `WatchedLiteralsSolver` that reports whether the iteration limit has been reached.

This change provides a centralized record of whether the solver is
"exhausted". In general, once the solver is exhausted, further uses do not
produce meaningful conclusions. Clients can use this information, for example,
to choose to discard all results based on the solver, instead of trying to
distinguish those reached before the limit was reached. Providing it at in the
solver avoids the need to propagate it from each callsite to the client of the
solver.

Differential Revision: https://reviews.llvm.org/D155636

16 months ago[BOLT] Switch to using layout order in YAML
Amir Ayupov [Tue, 18 Jul 2023 21:29:34 +0000 (14:29 -0700)]
[BOLT] Switch to using layout order in YAML

Use layout order in YAML profile reading/writing. Preserve old behavior (DFS order)
under `-profile-use-dfs` option.

Reviewed By: spupyrev

Differential Revision: https://reviews.llvm.org/D155514

16 months ago[BOLT][NFC] Rename icf-dfs option variable to ICFUseDFS
Amir Ayupov [Tue, 18 Jul 2023 21:29:29 +0000 (14:29 -0700)]
[BOLT][NFC] Rename icf-dfs option variable to ICFUseDFS

Rename to avoid collision with profile-use-dfs.

Differential Revision: https://reviews.llvm.org/D155513

16 months ago[clang-format] Add TypeNames option to disambiguate types/objects
Owen Pan [Fri, 14 Jul 2023 08:25:34 +0000 (01:25 -0700)]
[clang-format] Add TypeNames option to disambiguate types/objects

If a non-keyword identifier is found in TypeNames, then a *, &, or && that
follows it is annotated as TT_PointerOrReference.

Differential Revision: https://reviews.llvm.org/D155273

16 months ago[flang][openacc] Support static slice in reduction lowering
Valentin Clement [Tue, 18 Jul 2023 21:13:46 +0000 (14:13 -0700)]
[flang][openacc] Support static slice in reduction lowering

Lower static array slices reduction with the correct
type.

Reviewed By: razvanlupusoru

Differential Revision: https://reviews.llvm.org/D155519

16 months ago[clang-tidy] Allow explicit throwing in bugprone-exception-escape for special functions
Piotr Zegar [Tue, 18 Jul 2023 20:26:30 +0000 (20:26 +0000)]
[clang-tidy] Allow explicit throwing in bugprone-exception-escape for special functions

Functions declared explicitly with noexcept(false) or throw(exception)
will be excluded from the analysis, as even though it is not recommended for
functions like swap, main, move constructors and assignment operators,
and destructors, it is a clear indication of the developer's intention and
should be respected.

Fixes: #40583, #55143

Reviewed By: carlosgalvezp

Differential Revision: https://reviews.llvm.org/D153423

16 months ago[clang-format] Fix indent for selective formatting
Sedenion [Tue, 18 Jul 2023 20:39:14 +0000 (13:39 -0700)]
[clang-format] Fix indent for selective formatting

The problem was that the LevelIndentTracker remembered
the indentation level of previous deeper levels when
leaving a scope. Afterwards, when it entered again a
deeper level, it blindly reused the previous
indentation level. In case the --lines option was used
such that the previous deeper level was not formatted,
that previous level was whatever happened to be there
in the source code. The formatter simply believed it.

This is fixed by letting the LevelIndentTracker forget
the previous deeper levels when stepping out of them
(=> change in LevelIndentTracker::nextLine()).
Note that this used to be the case until LLVM 14.0.6,
but was changed in https://reviews.llvm.org/D129064
(#56352) to fix a crash. Our commit here essentially
reverts that crash fix. It was incorrect/incomplete.

Fixes #58464.
Fixes #59178.
Fixes #62799.

Differential Revision: https://reviews.llvm.org/D151047

16 months ago[AArch64] NFC. Add a test exposing a bug in FixupStatepointCallerSaved pass
Daniil Suchkov [Tue, 18 Jul 2023 19:57:52 +0000 (12:57 -0700)]
[AArch64] NFC. Add a test exposing a bug in FixupStatepointCallerSaved pass

This pass doesn't take register classes into account, so it ends up
trying to spill a non-GP register onto stack which is not correct.

16 months ago[scudo] Fix memtag tests.
Christopher Ferris [Tue, 18 Jul 2023 02:33:16 +0000 (19:33 -0700)]
[scudo] Fix memtag tests.

A few memtag tests were using EXPECT_DEBUG_DEATH which executes the
statement unless NDEBUG is set. This meant the statement would
cause an abort, but the EXPECT_DEATH part was not done. Change these
to simply EXPECT_DEATH, these tests already verify that SCUDO_DEBUG
is set before checking for DEATH.

Reviewed By: Chia-hungDuan

Differential Revision: https://reviews.llvm.org/D155538

16 months ago[ODS] Quick fix
Amanda Tang [Tue, 18 Jul 2023 20:34:08 +0000 (20:34 +0000)]
[ODS] Quick fix

Quick fix for a failing test in https://reviews.llvm.org/rG5267ed05bc4612e91409d63b4dbc4e01751acb75

Reviewed By: jpienaar

Differential Revision: https://reviews.llvm.org/D155637

16 months ago[Support] Add llvm::xxh3_64bits
Fangrui Song [Tue, 18 Jul 2023 20:36:11 +0000 (13:36 -0700)]
[Support] Add llvm::xxh3_64bits

ld.lld SHF_MERGE|SHF_STRINGS duplicate elimination is computation heavy
and utilitizes llvm::xxHash64, a simplified version of XXH64.
Externally many sources confirm that a new variant XXH3 is much faster.

I have picked a few hash implementations and computed the
proportion of time spent on hashing in the overall link time (a debug
build of clang 16 on a machine using AMD Zen 2 architecture):

* llvm::xxHash64: 3.63%
* official XXH64 (`#define XXH_VECTOR XXH_SCALAR`): 3.53%
* official XXH3_64bits (`#define XXH_VECTOR XXH_SCALAR`): 1.21%
* official XXH3_64bits (default, essentially `XXH_SSE2`): 1.22%
* this patch llvm::xxh3_64bits: 1.19%

The remaining part of lld remains unchanged. Consequently, a lower ratio
indicates that hashing is faster. Therefore, it is evident that XXH3 from xxhash
is significantly faster than both the official version and our llvm::xxHash64.

(
string length: count
1-3: 393434
4-8: 2084056
9-16: 2846249
17-128: 5598928
129-240: 1317989
241-: 328058
)

This patch adds heavily simplified https://github.com/Cyan4973/xxHash,
taking account of many simplification ideas from Devin Hussey's xxhash-clean.

Important x86-64 optimization ideas:

* Make XXH3_len_129to240_64b and XXH3_hashLong_64b noinline
* Unroll XXH3_len_17to128_64b
* __restrict does not affect Clang code generation

Beside SHF_MERGE|SHF_STRINGS duplicate elimination, llvm/ADT/StringMap.h
StringMapImpl::LookupBucketFor and a few places in lld can potentially be
accelerated by switching to llvm::xxh3_64bits.

Link: https://github.com/llvm/llvm-project/issues/63750
Reviewed By: serge-sans-paille

Differential Revision: https://reviews.llvm.org/D154812

16 months ago[clang-format] Refactoring and asserts in LevelIndentTracker. (NFC)
Sedenion [Tue, 18 Jul 2023 20:23:45 +0000 (13:23 -0700)]
[clang-format] Refactoring and asserts in LevelIndentTracker. (NFC)

adjustToUnmodifiedLine: The code does something only for non-PP-directives.
This is now reflected by putting the if-check to the top. This also ensures
that the assert() there is executed only if IndentForLevel is actually
accessed.

getIndent(): assert valid index into IndentForLevel.

Added explanation regarding the intention of IndentForLevel.

Differential Revision: https://reviews.llvm.org/D155094

16 months ago[SLP][NFC]Improve compile-time by using map {TreeEntry *, Instruction *}
Alexey Bataev [Mon, 17 Jul 2023 16:47:01 +0000 (09:47 -0700)]
[SLP][NFC]Improve compile-time by using map {TreeEntry *, Instruction *}
in getLastInstructionInBundle(), NFC.

Instead of building EntryToLastInstruction before the vectorization,
build it automatically during the calls to getLastInstructionInBundle()
function.

16 months ago[libc++][test] Fixes transitives includes.
Mark de Wever [Tue, 18 Jul 2023 20:21:10 +0000 (22:21 +0200)]
[libc++][test] Fixes transitives includes.

16 months ago[Flang][OpenMP] Disable function filtering for host
Andrew Gozillon [Tue, 18 Jul 2023 20:08:58 +0000 (15:08 -0500)]
[Flang][OpenMP] Disable function filtering for host

This should be a temporary fix while we work
towards enabling function filtering for host
again via a future patch.

16 months ago[bazel] fix layering check to run bazel with 16.x
Mikhail Goncharov [Tue, 18 Jul 2023 20:05:14 +0000 (22:05 +0200)]
[bazel] fix layering check to run bazel with 16.x

Preparation to update bazel builder to use LLVM 16 release
where layering check was enabled https://reviews.llvm.org/D132779

Current setup missed some backsliding in layering check as it has
only on for projects with the check enforced.

Disabled it completely for libc and fixed for DWARFLinkerParallel.
It would be great to re-enable it for libc later.

16 months ago[RISCV] Remove unnecessary _32 and _64 suffixes from some scalar crypto builtins.
Craig Topper [Tue, 18 Jul 2023 19:55:26 +0000 (12:55 -0700)]
[RISCV] Remove unnecessary _32 and _64 suffixes from some scalar crypto builtins.

The names of these builtins aren't ambiguous so they don't need the suffix.

16 months ago[libc++][print] Fixes tests.
Mark de Wever [Tue, 18 Jul 2023 19:49:26 +0000 (21:49 +0200)]
[libc++][print] Fixes tests.

The exception message improvements for `<format>` were not applied to
print. The issue has been reported by Ian Anderson.

16 months ago[mlir-lsp] Add client information to the InitializationParams
River Riddle [Tue, 18 Jul 2023 08:34:57 +0000 (01:34 -0700)]
[mlir-lsp] Add client information to the InitializationParams

This is specified in the spec, but we just never really needed it. This
allows for users of the LSP libraries to inspect information about the
client that is connected to the server.

Differential Revision: https://reviews.llvm.org/D155566

16 months ago[RISCV] Minor style cleanups in post ISEL combines
Philip Reames [Tue, 18 Jul 2023 19:22:08 +0000 (12:22 -0700)]
[RISCV] Minor style cleanups in post ISEL combines

16 months ago[libc++][format] Fixes times before epoch.
Mark de Wever [Mon, 10 Jul 2023 05:59:58 +0000 (07:59 +0200)]
[libc++][format] Fixes times before epoch.

The number of days should be rounded down, for both positive and
negative times since epoch. The original code truncated, which is
correct for positive values, but wrong for negative values.

Depends on D138826

Reviewed By: #libc, ldionne

Differential Revision: https://reviews.llvm.org/D154865

16 months ago[clang] scan-view: Remove unused python import
Tulio Magno Quites Machado Filho [Tue, 18 Jul 2023 19:09:50 +0000 (16:09 -0300)]
[clang] scan-view: Remove unused python import

Python's module imp is not being used and is not available on Python
3.12 anymore.

Reviewed By: tbaeder

Differential Revision: https://reviews.llvm.org/D155192

16 months ago[libc++][format] Improves diagnostics.
Mark de Wever [Sun, 4 Jun 2023 19:32:10 +0000 (21:32 +0200)]
[libc++][format] Improves diagnostics.

Improves both the compile-time and run-time errors.
At compile-time it does a bit more work to get more specific errors.
This could be done at run-time too, but that has a performance penalty.
Since it's expected most use-cases use format* instead of vformat* the
compile-time errors are more common.

For example when using

  std::format_to("{:-c}", 42);

Before compile output would contain

  std::__throw_format_error("The format-spec should consume the input or end with a '}'");

Now it contains

  std::__throw_format_error("The format specifier does not allow the sign option");

Given a better indication the sign option is not allowed. Note the
output is still not user-friendly; C++ doesn't have good facilities to
generate nice messages from the library.

In general all messages have been reviewed and improved, using a more
consistent style and using less terms used in the standard. For example

  format-spec -> format specifier
  arg-id -> argument index

Reviewed By: #libc, ldionne

Differential Revision: https://reviews.llvm.org/D152624

16 months ago[RISCV] Remove 'Z' from Zbb builtin type strings. NFC
Craig Topper [Tue, 18 Jul 2023 18:15:44 +0000 (11:15 -0700)]
[RISCV] Remove 'Z' from Zbb builtin type strings. NFC

'Z' means uint32_t, but 'i' is always 32 bits for RISC-V so
we don't need this.

16 months ago[HWASAN][LSAN] Fix false positive memory leak reports on X86_64
Kirill Stoimenov [Fri, 14 Jul 2023 20:07:29 +0000 (20:07 +0000)]
[HWASAN][LSAN] Fix false positive memory leak reports on X86_64

Before this patch when running HWASAN on x86_64 with with memory tagging support we got a bunch of false memory leak reports. The reason for that is that the heuristic used to detect if an 8 bytes could be a user pointer was not valid when memory tagging is used as the top byte could contain non-zero information.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D155338

16 months ago[libc++][format] Improves run-time diagnostics.
Mark de Wever [Sun, 4 Jun 2023 19:32:10 +0000 (21:32 +0200)]
[libc++][format] Improves run-time diagnostics.

After parsing a std-format-spec it's validated, depending on the type used some
format options are not allowed. This improves the error messages in the
exceptions thrown upon failure.

Depends on D155364

Reviewed By: #libc, ldionne

Differential Revision: https://reviews.llvm.org/D155366