Nikolas Klauser [Mon, 4 Jul 2022 20:43:50 +0000 (22:43 +0200)]
[libc++] default-construct moved-from test-iterators
This way we ensure that we don't use-after-move the iterators.
Reviewed By: Mordante, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D129044
Nikolas Klauser [Mon, 4 Jul 2022 01:27:58 +0000 (03:27 +0200)]
[libc++] Prefer __has_builtin for detecting compiler-provided type_traits
Both clang and GCC support using `__has_builtin` for detecting compiler-provided type_traits. Use it instead of `__has_keyword` or `__has_feature` to remove special-casing for GCC-provided builtins
Reviewed By: ldionne, #libc
Spies: libcxx-commits
Differential Revision: https://reviews.llvm.org/D129056
Sunho Kim [Wed, 6 Jul 2022 11:12:22 +0000 (20:12 +0900)]
[ORC][ORC_RT][AArch64] Implement TLS descriptor in ELFNixPlatform.
Implements TLS descriptor relocations in JITLink ELF/AARCH64 backend and support the relevant runtime functions in ELFNixPlatform.
Unlike traditional TLS model, TLS descriptor model requires linker to return the "offset" from thread pointer via relocaiton not the actual pointer to thread local variable. There is no public libc api for adding new allocations to TLS block dynamically which thread pointer points to. So, we support this by taking delta from thread base pointer to the actual thread local variable in our allocated section.
Reviewed By: lhames
Differential Revision: https://reviews.llvm.org/D128601
Dimitrije Milosevic [Wed, 6 Jul 2022 10:24:58 +0000 (12:24 +0200)]
[MIPS] Resolve issues in building ASAN for N32 ABI
Building the compiler-rt's AddressSanitizer for
the n32 MIPS ABI currently fails, due to a few reasons:
- defined(__mips64), which is set solely based on
the architecture type (32-bit/64-bit), was still used
in some places. Therefore, defined(__mips64) is swapped
with SANITIZER_MIPS64, which takes the ABI into account
as well - defined(__mips64) && _MIPS_SIM == ABI64.
- The n32 ABI still uses 64-bit *Linux* system calls,
even though the word size is 32 bits.
- After the transition to canonical system calls (D124212),
the n32 ABI still didn't use them, even though they
are supported.
Differential Revision: https://reviews.llvm.org/D127098
Dimitrije Milosevic [Wed, 6 Jul 2022 10:22:04 +0000 (12:22 +0200)]
[MIPS] Fix the ASAN shadow offset hook for the N32 ABI
Currently, LLVM doesn't have the correct shadow offset
mapping for the n32 ABI.
This patch introduces the correct shadow offset value
for the n32 ABI - 1ULL << 29.
Differential Revision: https://reviews.llvm.org/D127096
Markus Böck [Wed, 6 Jul 2022 10:27:44 +0000 (12:27 +0200)]
[mlir] Define proper DenseMapInfo for Interfaces
Prior to this patch, using any kind of interface (op interface, attr interface, type interface) as the key of a llvm::DenseSet, llvm::DenseMap or other related containers, leads to invalid pointer dereferences, despite compiling.
The gist of the problem is that a llvm::DenseMapInfo specialization for the base type (aka one of Operation*, Type or Attribute) are selected when using an interface as a key, which uses getFromOpaquePointer with invalid pointer addresses to construct instances for the empty key and tombstone key values. The interface is then constructed with this invalid base type and an attempt is made to lookup the implementation in the interface map, which then dereferences the invalid pointer address. (For more details and the exact call chain involved see the GitHub issue below)
The current workaround is to use the more generic base type (eg. instead of DenseSet<FunctionOpInterface>, DenseSet<Operation*>), but this is strictly worse from a code perspective (doesn't enforce the invariant, code is less self documenting, having to insert casts etc).
This patch fixes that issue by defining a DenseMapInfo specialization of Interface subclasses which uses a new constructor to construct an instance without querying a concept. That allows getEmptyKey and getTombstoneKey to construct an interface with invalid pointer values.
Fixes https://github.com/llvm/llvm-project/issues/54908
Differential Revision: https://reviews.llvm.org/D129038
Jay Foad [Wed, 6 Jul 2022 10:00:07 +0000 (11:00 +0100)]
[AMDGPU] Move all -global-isel RUN lines into main test file
This test previously had some -global-isel RUN lines, but others
were in a separate test file in GlobalISel/.
Weining Lu [Wed, 6 Jul 2022 09:46:10 +0000 (17:46 +0800)]
[LoongArch] Add codegen support for division operations
These operations include sdiv/udiv/srem/urem.
As the ISA [https://loongson.github.io/LoongArch-Documentation/LoongArch-Vol1-EN.html#_div_wudu_mod_wudu]
described, when the divisor is 0, the result can be any value, but no
exception will be triggered. Unlike gcc, which by default emit code
that checks divide-by-zero after the division or modulus instruction,
we only emit this check when the `-loongarch-check-zero-division`
option is passed.
Differential Revision: https://reviews.llvm.org/D128572
Jay Foad [Wed, 6 Jul 2022 09:48:30 +0000 (10:48 +0100)]
[AMDGPU] Add a MIR test for D127781
wanglei [Wed, 6 Jul 2022 07:58:13 +0000 (15:58 +0800)]
[LoongArch] Add LoongArch support to update_llc_test_checks
Add LoongArch assembly scrubbing and triple support to update_llc_test_checks.
Depends on D128432
Reviewed By: MaskRay, xen0n
Differential Revision: https://reviews.llvm.org/D128433
Martin Storsjö [Thu, 23 Jun 2022 11:19:47 +0000 (14:19 +0300)]
[lldb] Stop passing both i386 and i686 in parallel as architectures on Windows
When an object file returns multiple architectures, it is treated
as a fat binary - which really isn't the case of i386 vs i686 where
the object file actually has one architecture.
This allows getting rid of hardcoded architecture triples in
PlatformWindows.
The parallel i386 and i686 architecture strings stem from
5e6f45201f0b62c1e7a24fc396f3ea6e10dc880d / D7120 and
ad587ae4ca143d388c0ec4ef2faa1b5eddedbf67 / D4658.
Differential Revision: https://reviews.llvm.org/D128617
Nikita Popov [Wed, 6 Jul 2022 08:42:09 +0000 (10:42 +0200)]
[ValueTracking] Accept Instruction in isSafeToSpeculativelyExecute() (NFC)
As constant expressions can no longer trap, it only makes sense to
call isSafeToSpeculativelyExecute on Instructions, so limit the
API to accept only them, rather than general Operators or Values.
Lucas Prates [Mon, 4 Jul 2022 15:17:31 +0000 (16:17 +0100)]
[Docs] Add release note for ARM's new -mframe-chain option
This adds a release note entry for the new -mframe-chain option
introduced on D125094.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D129085
Chenbing Zheng [Wed, 6 Jul 2022 09:03:14 +0000 (17:03 +0800)]
[InstCombine] remove useless insertelement
extractelement (bitcast (insertelement (Vec, b)), a) ->
extractelement (bitcast (Vec), a)
Reviewed By: RKSimon
Differential Revision: https://reviews.llvm.org/D128890
Thomas Symalla [Mon, 4 Jul 2022 15:23:50 +0000 (17:23 +0200)]
[NFC][AMDGPU] Cleanup the SIOptimizeExecMasking pass.
This patch removes a bit of code duplication and
moves the v_cmpx optimization out of the
runOnMachineFunction pass.
Reviewed By: foad
Differential Revision: https://reviews.llvm.org/D129086
Nikita Popov [Wed, 6 Jul 2022 08:46:13 +0000 (10:46 +0200)]
[SelectionDAGISel] Drop unused variable (NFC)
Nikita Popov [Wed, 6 Jul 2022 08:42:49 +0000 (10:42 +0200)]
[LoopVectorizationLegality] Drop unused variable (NFC)
David Spickett [Tue, 5 Jul 2022 16:14:24 +0000 (16:14 +0000)]
[llvm][AArch64] Fix "+all" feature for sysreg aliases
For example the predres extension adds one instruction that
is a sys alias. Previously this wasn't disassembled properly
with "+all".
This was because a check for "+all" was added to haveFeatures
in AArch64SysReg but not in SysAlias.
Reviewed By: MaskRay, lenary
Differential Revision: https://reviews.llvm.org/D129147
Nikita Popov [Wed, 6 Jul 2022 08:29:47 +0000 (10:29 +0200)]
[IR] Remove Constant::canTrap() (NFC)
As integer div/rem constant expressions are no longer supported,
constants can no longer trap and are always safe to speculate.
Remove the Constant::canTrap() method and its usages.
Nikita Popov [Mon, 4 Jul 2022 14:26:51 +0000 (16:26 +0200)]
[ConstExpr] Remove div/rem constant expressions
D128820 stopped creating div/rem constant expressions by default;
this patch removes support for them entirely.
The getUDiv(), getExactUDiv(), getSDiv(), getExactSDiv(), getURem()
and getSRem() on ConstantExpr are removed, and ConstantExpr::get()
now only accepts binary operators for which
ConstantExpr::isSupportedBinOp() returns true. Uses of these methods
may be replaced either by corresponding IRBuilder methods, or
ConstantFoldBinaryOpOperands (if a constant result is required).
On the C API side, LLVMConstUDiv, LLVMConstExactUDiv, LLVMConstSDiv,
LLVMConstExactSDiv, LLVMConstURem and LLVMConstSRem are removed and
corresponding LLVMBuild methods should be used.
Importantly, this also means that constant expressions can no longer
trap! This patch still keeps the canTrap() method to minimize diff --
I plan to drop it in a separate NFC patch.
Differential Revision: https://reviews.llvm.org/D129148
Nikita Popov [Wed, 6 Jul 2022 08:01:17 +0000 (10:01 +0200)]
[ConstFold] Salvage some div/rem folding test (NFC)
The div/rem constant expressions are going away in D129148. Convert
some tests to use InstSimplify instead, to show that the constant
folding still happens.
wanglei [Wed, 6 Jul 2022 07:44:10 +0000 (15:44 +0800)]
[LoongArch] Implement prologue/epilogue insertion
Differential Revision: https://reviews.llvm.org/D128432
Nikita Popov [Wed, 6 Jul 2022 07:45:01 +0000 (09:45 +0200)]
[PowerPC] Simplify test for PR33636 (NFC)
There was a lot of unnecessary code here. Add the -O0 flag to
avoid using constant expressions, otherwise it may get folded
away during EarlyCSE.
Verified that this test fails prior to the fixing commit.
wanglei [Wed, 6 Jul 2022 07:27:46 +0000 (15:27 +0800)]
[LoongArch] Support lowering FrameIndex
Ensure callee-saved registers are accessed relative to the stackpointer.
Differential Revision: https://reviews.llvm.org/D128430
Adrian Kuegel [Wed, 6 Jul 2022 06:47:21 +0000 (08:47 +0200)]
Fangrui Song [Wed, 6 Jul 2022 06:31:09 +0000 (23:31 -0700)]
[ELF] Optimize getInputSections. NFC
In the majority of cases (e.g. orphan sections), an OutputSection has at most
one InputSectionDescription (isd). By changing the return type to
ArrayRef<InputSection *> we can just reference the isd->sections. For
OutputSections with more than one InputSectionDescription we use a caller
provided SmallVector to copy the elements as before.
Reviewed By: peter.smith
Differential Revision: https://reviews.llvm.org/D129111
Carl Ritson [Wed, 6 Jul 2022 05:03:49 +0000 (14:03 +0900)]
[AMDGPU] Additional liveness tests for si-optimize-exec-masking-pre-ra
Merge tests and fixes from D128110 and D128315 on top of already
committed D128800.
Original author: arsenm
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D128882
Elvina Yakubova [Wed, 6 Jul 2022 05:19:22 +0000 (08:19 +0300)]
[BOLT] Change mutex implementation
Changed acquire implemetaion to __atomic_test_and_set() and release
to __atomic_clear() so it eliminates inline asm usage and is arch
independent.
Elvina Yakubova,
Advanced Software Technology Lab, Huawei
Reviewers: yota9, maksfb, rafauler
Differential Revision: https://reviews.llvm.org/D129162
Alex Bradbury [Wed, 6 Jul 2022 04:16:25 +0000 (05:16 +0100)]
[WebAssembly][NFC] Refactor table handling in WebAssembly::wasmSymbolSetType
Use the isExternRefType and isFuncRefType helpers rather than
reimplementing that logic in this function (which acts as a blocker to
work to prototype alternative IR-level representations of reference
types).
This relands
8ccc7e0aa461 (which had compilation errors).
Alex Bradbury [Wed, 6 Jul 2022 04:07:22 +0000 (05:07 +0100)]
Revert "[WebAssembly][NFC] Refactor table handling in WebAssembly::wasmSymbolSetType"
This reverts commit
8ccc7e0aa461350bce02d70669c11f4a5e300ee7.
Wrong version of the patch was pushed.
Alex Bradbury [Wed, 6 Jul 2022 03:54:13 +0000 (04:54 +0100)]
[WebAssembly][NFC] Refactor table handling in WebAssembly::wasmSymbolSetType
Use the isExternRefType and isFuncRefType helpers rather than
reimplementing that logic in this function (which acts as a blocker to
work to prototype alternative IR-level representations of reference
types).
jacquesguan [Fri, 1 Jul 2022 08:02:28 +0000 (16:02 +0800)]
[mlir][Vector] Fold InsertOp(SplatOp(X), SplatOp(X)) to SplatOp(X).
This patch folds InsertOp(SplatOp(X), SplatOp(X)) to SplatOp(X).
Differential Revision: https://reviews.llvm.org/D129058
Alexandre Ganea [Wed, 6 Jul 2022 00:27:53 +0000 (20:27 -0400)]
[Clang][unittests] Silence trucation warning with MSVC 2022
Differential Revision: https://reviews.llvm.org/D129152
Alexandre Ganea [Tue, 5 Jul 2022 16:40:22 +0000 (12:40 -0400)]
[Clang] Silence warning when building with MSVC 2022
Previously, the warning seen:
[22/95] Building CXX object tools\clang\lib\StaticAnalyzer\Checker...bj.clangStaticAnalyzerCheckers.dir\NoReturnFunctionChecker.cpp.objC:\git\llvm-project\clang\lib\StaticAnalyzer\Checkers\NoReturnFunctionChecker.cpp(149): warning C4305: 'if': truncation from 'size_t' to 'bool'
C:\git\llvm-project\clang\include\clang/Analysis/SelectorExtras.h(28): note: see reference to function template instantiation 'clang::Selector clang::getKeywordSelector<const char,const char,const char,const char>(clang::ASTContext &,const char *,const char *,const char *,const char *)' being compiled
C:\git\llvm-project\clang\lib\StaticAnalyzer\Checkers\NoReturnFunctionChecker.cpp(125): note: see reference to function template instantiation 'void clang::lazyInitKeywordSelector<const char,const char,const char,const char>(clang::Selector &,clang::ASTContext &,const char *,const char *,const char *,const char *)' being compiled
Alexandre Ganea [Tue, 5 Jul 2022 16:15:36 +0000 (12:15 -0400)]
[mlir] Silence warnings when building with Clang ToT
Previously, these warnings were seen:
[6599/7564] Building native mlir-pdll...
[270/278] Building CXX object tools/mlir/lib/Tools/PDLL/CodeGen/CMakeFiles/obj.MLIRPDLLCodeGen.dir/CPPGen.cpp.obj
In file included from C:/git/llvm-project/mlir/lib/Tools/PDLL/CodeGen/CPPGen.cpp:20:
C:/git/llvm-project/mlir/include\mlir/Tools/PDLL/ODS/Operation.h(202,16): warning: unqualified friend declaration referring to type outside of the nearest enclosing namespace is a Microsoft extension; add a nested name specifier [-Wmicrosoft-unqualified-friend]
friend class Dialect;
^
::mlir::
1 warning generated.
[278/278] Linking CXX executable bin\mlir-pdll.exe
[6857/7564] Building CXX object tools/mlir/unittests/ExecutionEngine/CMakeFiles/MLIRExecutionEngineTests.dir/Invoke.cpp.obj
In file included from C:/git/llvm-project/mlir/unittests/ExecutionEngine/Invoke.cpp:20:
C:/git/llvm-project/mlir/include\mlir/ExecutionEngine/MemRefUtils.h(104,27): warning: shift count >= width of type [-Wshift-count-overflow]
assert(sizeof(T) < (1ul << 32) && "Elemental type overflows");
^ ~~
C:\Program Files (x86)\Windows Kits\10\include\10.0.22000.0\ucrt\assert.h(40,17): note: expanded from macro 'assert'
(!!(expression)) || \
^~~~~~~~~~
1 warning generated.
Félix Cloutier [Tue, 7 Jun 2022 22:45:07 +0000 (15:45 -0700)]
Allow non-variadic functions to be attributed with `__attribute__((format))`
Clang only allows you to use __attribute__((format)) on variadic functions. There are legit use cases for __attribute__((format)) on non-variadic functions, such as:
(1) variadic templates
```c++
template<typename… Args>
void print(const char *fmt, Args… &&args) __attribute__((format(1, 2))); // error: format attribute requires variadic function
```
(2) functions which take fixed arguments and a custom format:
```c++
void print_number_string(const char *fmt, unsigned number, const char *string) __attribute__((format(1, 2)));
// ^error: format attribute requires variadic function
void foo(void) {
print_number_string(“%08x %s\n”, 0xdeadbeef, “hello”);
print_number_string(“%d %s”, 0xcafebabe, “bar”);
}
```
This change allows Clang users to attach __attribute__((format)) to non-variadic functions, including functions with C++ variadic templates. It replaces the error with a GCC compatibility warning and improves the type checker to ensure that received arrays are treated like pointers (this is a possibility in C++ since references to template types can bind to arrays).
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D112579
rdar://
84629099
Julian Lettner [Tue, 5 Jul 2022 23:27:11 +0000 (16:27 -0700)]
[Sanitizer][Darwin] Fix log-path_test.cpp
In my previous change [1], I added log output that made a test that
expected "no output" fail. The easiest solution is to only print the
new hint/warning when we at least ask for `verbosity=1`.
[1] https://reviews.llvm.org/D128936
Radar-Id: rdar://
96437354
Med Ismail Bennani [Tue, 5 Jul 2022 23:25:23 +0000 (16:25 -0700)]
[lldb/Core] Fix finite progress event reporting
This patch should fix event handling for finite progress reports.
Previously, the event handler would get stuck when receiving a finite
progress report, and stop displaying upcoming reports.
This was due to the fact that we were checking if the progress event was
completed by calling `GetCompleted` but returns the completion amount
instead of saying whether it's completed.
That caused the current event id to remain the same, preventing all the
following progress reports to be shown to the user.
This patch also adds some logging to facilitate debugging progress events.
rdar://
91788326
Differential Revision: https://reviews.llvm.org/D128768
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
Craig Topper [Tue, 5 Jul 2022 23:01:41 +0000 (16:01 -0700)]
[RISCV] Add more SHXADD patterns where the input is (and (shl/shr X, C2), C1)
It might be possible to rewrite (and (shl/shr X, C2), C1) in a way
that gives us a left shift that can be folded into a SHXADD.
River Riddle [Tue, 5 Jul 2022 23:20:30 +0000 (16:20 -0700)]
[mlir:Parser] Don't use strings for the "ugly" form of Attribute/Type syntax
This commit refactors the syntax of "ugly" attribute/type formats to not use
strings for wrapping. This means that moving forward attirbutes and type formats
will always need to be in some recognizable form, i.e. if they use incompatible
characters they will need to manually wrap those in a string, the framework will
no longer do it automatically.
This has the benefit of greatly simplifying how parsing attributes/types work, given
that we currently rely on some extremely complicated nested parser logic which is
quite problematic for a myriad of reasons; unecessary complexity(we create a nested
source manager/lexer/etc.), diagnostic locations can be off/wrong given string escaping,
etc.
Differential Revision: https://reviews.llvm.org/D118505
Vadim Paretsky [Tue, 5 Jul 2022 22:20:32 +0000 (17:20 -0500)]
[OpenMP] add 4 custom APIs supporting MSVC OMP codegen
This check-in adds 4 APIs to support MSVC, specifically:
* 3 APIs (__kmpc_sections_init, __kmpc_next_section,
__kmpc_end_sections) to support the dynamic scheduling of OMP sections.
* 1 API (__kmpc_copyprivate_light, a light-weight version of
__kmpc_copyrprivate) to support the OMP single copyprivate clause.
Differential Revision: https://reviews.llvm.org/D128403
Amara Emerson [Mon, 4 Jul 2022 23:43:56 +0000 (16:43 -0700)]
[AArch64][GlobalISel] Add support for sret demotion.
To do this, we need to implement a target hook and make a minor change to the
call lowering to demote arguments to sret if they can't be handled by the
calling conventions.
Fixes issue 56295
Differential Revision: https://reviews.llvm.org/D129098
Lang Hames [Tue, 5 Jul 2022 22:01:15 +0000 (15:01 -0700)]
Fix running orc-rt tests with LLVM_BUILD_EXTERNAL_COMPILER_RT (again).
Add missing dependency on lli when building compiler-rt with
LLVM_BUILD_EXTERNAL_COMPILER_RT. Previously we would non-deterministically fail
the tests due to the missing binary.
This is essentially identical to
0e5ea403e8d, which added an earlier dependence
on llvm-jitlink.
rdar://
96467892
Anton Afanasyev [Fri, 1 Jul 2022 14:17:59 +0000 (17:17 +0300)]
[RISCV][Driver] Add libm linking to `RISCVToolchain` for C++
GCC automatically links math library by adding `-lm` to linker command
line, since C++ runtime `libstdc++` requires libm, so add it to
`RISCVToochain` as well.
Differential Revision: https://reviews.llvm.org/D129065
Muhammad Omair Javaid [Tue, 5 Jul 2022 21:39:49 +0000 (01:39 +0400)]
[LLDB] Remove TestLoadUnload.py Arm/Linux Xfail decorator
This is a follow up on my last commit where one of the decorator was
left unremoved.
This patch removes Xfail decorator from TestLoadUnload.py as it is now
passing on Arm/Linux buildbot.
Chris Bieneman [Tue, 5 Jul 2022 20:38:06 +0000 (15:38 -0500)]
[NFC][HLSL] Add tests for vector alias. Remove dead code.
Based on feedback from @Aaron.Ballman.
Remove the unused static ID char (can re-add it later if needed).
Add test to cover some invalid HLSL vector instantations ensuring
that the appropriate error messages are generated.
Muhammad Omair Javaid [Tue, 5 Jul 2022 21:13:54 +0000 (01:13 +0400)]
[LLDB] Remove TestLoadUnload.py Arm/Linux Xfail decorator
This patch removes Xfail decorator from TestLoadUnload.py as it is now
passing on Arm/Linux buildbot.
Dmitri Gribenko [Tue, 5 Jul 2022 21:08:28 +0000 (23:08 +0200)]
[clang-tools-extra] Fix a link in ReleaseNotes.rst
LLVM GN Syncbot [Tue, 5 Jul 2022 21:04:41 +0000 (21:04 +0000)]
[gn build] Port
05130a6ba7d9
Dmitri Gribenko [Tue, 5 Jul 2022 20:48:53 +0000 (22:48 +0200)]
new clang-tidy checker for assignments within condition clause of if statement
new clang-tidy checker for assignments within the condition clause of an 'if' statement.
Reviewed By: gribozavr2
Differential Revision: https://reviews.llvm.org/D127114
Craig Topper [Tue, 5 Jul 2022 20:04:19 +0000 (13:04 -0700)]
[RISCV] Add more SHXADD tests. NFC
Craig Topper [Tue, 5 Jul 2022 18:58:03 +0000 (11:58 -0700)]
[RISCV] Use a switch statement in PreprocessISelDAG. NFC
This should make it easier to add more peepholes in the future.
Craig Topper [Tue, 5 Jul 2022 16:37:58 +0000 (09:37 -0700)]
[RISCV] Update PreprocessISelDAG to use RemoveDeadNodes.
Instead of deleting nodes as we go, delete all dead nodes if a
change is made. This allows adding peepholes that might make
multiple nodes dead.
Jay Foad [Thu, 23 Jun 2022 10:28:02 +0000 (11:28 +0100)]
[PowerPC] PPCTLSDynamicCall does not preserve LiveIntervals
According to D127731, PPCTLSDynamicCall does not preserve
LiveIntervals, so stop claiming that it does and remove the code
that tried to repair them. NFCI.
Differential Revision: https://reviews.llvm.org/D128421
Joe Nash [Tue, 5 Jul 2022 18:15:07 +0000 (14:15 -0400)]
[AMDGPU] NFC. Add a test of the error message for assembling global_atomic_cmpswap_x2
Joseph Huber [Fri, 1 Jul 2022 15:48:15 +0000 (11:48 -0400)]
[Libomptarget][NFC] Make Libomptarget use the LLVM naming convention
Libomptarget grew out of a project that was originally not in LLVM. As
we develop libomptarget this has led to an increasingly large clash
between the naming conventions used. This patch fixes most of the
variable names that did not confrom to the LLVM standard, that is
`VariableName` for variables and `functionName` for functions.
This patch was primarily done using my editor's linting messages, if
there are any issues I missed arising from the automation let me know.
Reviewed By: saiislam
Differential Revision: https://reviews.llvm.org/D128997
Sam McCall [Wed, 29 Jun 2022 11:48:00 +0000 (13:48 +0200)]
[pseudo] Add error-recovery framework & brace-based recovery
The idea is:
- a parse failure is detected when all heads die when trying to shift the next token
- we can recover by choosing a nonterminal we're partway through parsing, and
determining where it ends through nonlocal means (e.g. matching brackets)
- we can find candidates by walking up the stack from the (ex-)heads
- the token range is defined using heuristics attached to grammar rules
- the unparsed region is represented in the forest by an Opaque node
This patch has the core GLR functionality.
It does not allow recovery heuristics to be attached as extensions to
the grammar, but rather infers a brace-based heuristic.
Expected followups:
- make recovery heuristics grammar extensions (depends on D127448)
- add recovery to our grammar for bracketed constructs and sequence nodes
- change the structure of our augmented `_ := start` rules to eliminate some
special-cases in glrParse.
- (if I can work out how): avoid some spurious recovery cases described in comments
(Previously mistakenly committed as
a0f4c10ae227a62c2a63611e64eba83f0ff0f577)
Differential Revision: https://reviews.llvm.org/D128486
Sanjay Patel [Tue, 5 Jul 2022 17:35:47 +0000 (13:35 -0400)]
[InstCombine] avoid conflict between CHECK prefix and value name in test file; NFC
Tests can fail if a value name (%sqrt becomes SQRT) and that is the same as the RUN prefix.
Sanjay Patel [Tue, 5 Jul 2022 13:44:29 +0000 (09:44 -0400)]
[InstCombine] add tests for sqrt libcalls; NFC
Yuanfang Chen [Tue, 5 Jul 2022 17:56:19 +0000 (10:56 -0700)]
[SimplifyCFG] Skip hoisting common instructions that return token type
By LangRef, hoisting token-returning instructions obsures the origin
so it should be skipped. Found this issue while investigating a
CoroSplit pass crash.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D129025
Markus Böck [Tue, 5 Jul 2022 18:04:50 +0000 (20:04 +0200)]
[mlir][tblgen] Consistently use `$_ctxt` instead of `$_ctx`
With the exceptions of AttrOrTypeParameter and DerivedAttr, all of MLIR consistently uses $_ctxt as the substitute variable for the MLIRContext in TableGen C++ code.
Usually this does not matter unless one where to reuse some code in multiple fields but it is still needlessly inconsistent and prone to error.
This patch fixes that by consistently using _$ctxt everywhere.
Differential Revision: https://reviews.llvm.org/D129153
Joseph Huber [Thu, 30 Jun 2022 15:41:38 +0000 (11:41 -0400)]
[LinkerWrapper] Add AMDGPU specific options to the LLD invocation
We use LLD to perform AMDGPU linking. This linker accepts some arguments
through the `-plugin-opt` facilities. These options match what `Clang`
will output when given the same input.
Reviewed By: yaxunl
Differential Revision: https://reviews.llvm.org/D128923
Pengxuan Zheng [Sat, 2 Jul 2022 02:23:03 +0000 (19:23 -0700)]
[llvm-lib] Ignore /NODEFAULTLIB flag
It doesn't look like there is anything llvm-lib needs to handle based on
Microsoft's description of the flag.
https://docs.microsoft.com/en-us/cpp/build/reference/managing-a-library?view=msvc-170
Reviewed By: thakis
Differential Revision: https://reviews.llvm.org/D129032
Jonas Devlieghere [Tue, 5 Jul 2022 17:12:57 +0000 (10:12 -0700)]
Revert "[lldb/test] Don't use preexec_fn for launching inferiors"
This reverts commit
b15b1421bc9a11b318b65b489e5fd58dd917db1f because it
breaks GreenDragon [1]. The bot has been red for several days, so
reverting to green while I take a look.
[1] https://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake/45012/
Gabor Marton [Wed, 1 Jun 2022 14:29:51 +0000 (16:29 +0200)]
[analyzer] Fix assertion in simplifySymbolCast
Depends on D128068.
Added a new test code that fails an assertion in the baseline.
That is because `getAPSIntType` works only with integral types.
Differential Revision: https://reviews.llvm.org/D126779
Gabor Marton [Fri, 17 Jun 2022 16:19:55 +0000 (18:19 +0200)]
[analyzer] Do not emit redundant SymbolCasts
In `RegionStore::getBinding` we call `evalCast` unconditionally to align
the stored value's type to the one that is being queried. However, the
stored type might be the same, so we may end up having redundant
`SymbolCasts` emitted.
The solution is to check whether the `to` and `from` type are the same
in `makeNonLoc`.
Note, we can't just do type equivalence check at the beginning of `evalCast`
because when `evalCast` is called from `getBinding` then the original type
(`OriginalTy`) is not set, so one operand is missing for the comparison. In
`evalCastSubKind(nonloc::SymbolVal)` when the original type is not set,
we get the `from` type via `SymbolVal::getType()`.
Differential Revision: https://reviews.llvm.org/D128068
LLVM GN Syncbot [Tue, 5 Jul 2022 16:35:01 +0000 (16:35 +0000)]
[gn build] Port
b8dbc6ffea93
Zaara Syeda [Tue, 5 Jul 2022 16:16:08 +0000 (12:16 -0400)]
[LSR] Fix bug for optimizing unused IVs to final values
This is a fix for a crash reported for https://reviews.llvm.org/D118808
The fix is to only consider PHINodes which are induction phis.
Fixes #55529
Differential Revision: https://reviews.llvm.org/D125990
Chris Bieneman [Thu, 16 Jun 2022 20:35:48 +0000 (15:35 -0500)]
[HLSL] Add ExternalSemaSource & vector alias
HLSL vector types are ext_vector types, but they are also exposed via a
template syntax `vector<T, #>`. This is morally equavalent to the code:
```c++
template <typename T, int Size>
using vector = T __attribute__((ext_vector_type(Size)))
```
The problem is that templates aren't supported before HLSL 2021, and
type aliases still aren't supported in HLSL.
To resolve this (and other issues where HLSL can't represent its own
types), we rely on an external AST & Sema source being registered for
HLSL code.
This patch adds the HLSLExternalSemaSource and registers the vector
type alias.
Depends on D127802
Differential Revision: https://reviews.llvm.org/D128012
Min-Yih Hsu [Fri, 1 Jul 2022 22:29:02 +0000 (15:29 -0700)]
[mlir][LLVMIR] Apply CallOp/CallableInterface on suitable operations
- Applying CallOpInterface on CallOp and InvokeOp.
- Applying CallableInterface on LLVMFuncOp.
We're testing the changes using CallGraph, which uses both interfaces.
Differential Revision: https://reviews.llvm.org/D129026
Jay Foad [Tue, 5 Jul 2022 16:17:01 +0000 (17:17 +0100)]
[AMDGPU] GFX11 trivial NFC tweaks
A few miscellaneous comment, whitespace and indentation tweaks.
Nikita Popov [Tue, 5 Jul 2022 16:17:02 +0000 (18:17 +0200)]
[Thumb2] Remove unneeded IR from MIR test (NFC)
Apart from the global, the IR does not appear to be relevant for
the test. Drop it, to remove the dependence on the sdiv constant
expression.
David Green [Tue, 5 Jul 2022 16:16:18 +0000 (17:16 +0100)]
[VectorCombine] Improve shuffle select shuffle-of-shuffles
This in an extension to the code added in D123911 which added vector
combine folding of shuffle-select patterns, attempting to reduce the
total amount of shuffling required in patterns like:
%x = shuffle %i1, %i2
%y = shuffle %i1, %i2
%a = binop %x, %y
%b = binop %x, %y
shuffle %a, %b, selectmask
This patch extends the handing of shuffles that are dependent on one
another, which can arise from the SLP vectorizer, as-in:
%x = shuffle %i1, %i2
%y = shuffle %x
The input shuffles can also be emitted, in which case they are treated
like identity shuffles. This patch also attempts to calculate a better
ordering of input shuffles, which can help getting lower cost input
shuffles, pushing complex shuffles further down the tree.
This is a recommit with some additional checks for supported forms and
out-of-bounds mask elements, with some extra tests.
Differential Revision: https://reviews.llvm.org/D128732
Nikita Popov [Tue, 5 Jul 2022 16:01:25 +0000 (18:01 +0200)]
[Assembler] Name globals in test (NFC)
This makes it easier to modify the test without having to renumber
everything.
Stella Stamenova [Tue, 5 Jul 2022 15:52:54 +0000 (08:52 -0700)]
Revert "[mlir][sparse] add more unittest cases to sparse dialect merger"
This broke the windows mlir bot: https://lab.llvm.org/buildbot/#/builders/13/builds/22743
This reverts commit
daeb2dcea09820d92f81db84623cf1c6df825e14 and
537db49596f65a05c0309cf3333fc44f1657e999.
Simon Pilgrim [Tue, 5 Jul 2022 15:44:21 +0000 (16:44 +0100)]
[DAG] visitREM - use isAllOnesOrAllOnesSplat instead of isConstOrConstSplat
We were only using the N1C scalar/splat value once, so for clarity use isAllOnesOrAllOnesSplat instead if we actually need it.
Simon Pilgrim [Tue, 5 Jul 2022 15:27:30 +0000 (16:27 +0100)]
[DAG] foldAddSubOfSignBit - don't bother creating the new shift node unless constant folding succeeds
Noticed by inspection - the new shift is only ever used if the constant fold occurs
Groverkss [Tue, 5 Jul 2022 15:15:43 +0000 (16:15 +0100)]
[MLIR][Presburger] Rename attachments to identifiers in PresburgerSpace
"attachment" was a temporary name chosen for the information attached to a
variable in a PresburgerSpace. After the disambiguation of "variables" and
"identifiers" in PresburgerSpace, we use the word "identifiers" for this
information, since this information is used to "identify" these variables.
Reviewed By: arjunp
Differential Revision: https://reviews.llvm.org/D128751
Jay Foad [Thu, 23 Jun 2022 14:32:50 +0000 (15:32 +0100)]
[AMDGPU] Add patterns for GFX11 v_minmax and v_maxmin instructions
Differential Revision: https://reviews.llvm.org/D128445
Ben Dunbobbin [Tue, 5 Jul 2022 15:01:27 +0000 (16:01 +0100)]
[BUILD] Add missed CMakeLists.txt change from dfb77f2
See: https://reviews.llvm.org/D128195
Matthias Springer [Tue, 5 Jul 2022 15:01:40 +0000 (17:01 +0200)]
[mlir][memref][NFC] Silence compiler warnings
Matthias Springer [Tue, 5 Jul 2022 14:54:38 +0000 (16:54 +0200)]
[mlir][interfaces][NFC] Remove ViewLikeInterface::expandToRank
This helper function is no longer needed.
Differential Revision: https://reviews.llvm.org/D129145
Nikita Popov [Tue, 5 Jul 2022 08:56:54 +0000 (10:56 +0200)]
Revert "[SimplifyCFG] Thread branches on same condition in more cases (PR54980)"
This reverts commit
4e545bdb355a470d601e9bb7f7b2693c99e61a3e.
The newly added test is the third infinite combine loop caused by
this change. In this case, it's a combination of the branch to
common dest and jump threading folds that keeps peeling off loop
iterations.
The core problem here is that we ideally would not thread over
loop backedges, both because it is potentially non-profitable
(it may break canonical loop structure) and because it may result
in these kinds of loops. Unfortunately, due to the lack of a
dominator tree in SimplifyCFG, there is no good way to prevent
this. While we have LoopHeaders, this is an optional structure and
we don't do a good job of keeping it up to date. It would be fine
for a profitability check, but is not suitable for a correctness
check.
So for now I'm just giving up here, as I don't see a good way to
robustly prevent infinite combine loops.
Fixes https://github.com/llvm/llvm-project/issues/56203.
Matthias Springer [Tue, 5 Jul 2022 14:39:29 +0000 (16:39 +0200)]
[mlir][memref] Improve type inference for rank-reducing subviews
The result shape of a rank-reducing subview cannot be inferred in the general case. Just the result rank is not enough. The only thing that we can infer is the layout map.
This change also improves the bufferization patterns of tensor.extract_slice and tensor.insert_slice to fully support rank-reducing operations.
Differential Revision: https://reviews.llvm.org/D129144
Joe Nash [Mon, 27 Jun 2022 17:20:21 +0000 (13:20 -0400)]
[AMDGPU] gfx11 CodeGen for new DPP instructions
Modifies the GCNDPPCombine pass to enable DPP formation for the new DPP
instruction in gfx11, namely VOP3 encoded instructions with DPP and VOPC
with DPP.
Depends on D128656
Reviewed By: #amdgpu, rampitec
Differential Revision: https://reviews.llvm.org/D128682
Matthias Springer [Tue, 5 Jul 2022 13:36:02 +0000 (15:36 +0200)]
[mlir][tensor][bufferize][NFC] Clean up test case
Insert -split-input-file flag to make the test cases more stable.
Differential Revision: https://reviews.llvm.org/D129143
LLVM GN Syncbot [Tue, 5 Jul 2022 13:57:20 +0000 (13:57 +0000)]
[gn build] Port
d1af09ad9617
Haojian Wu [Fri, 1 Jul 2022 12:50:07 +0000 (14:50 +0200)]
[pseudo] Implement guard extension.
- Extend the GLR parser to allow conditional reduction based on the
guard functions;
- Implement two simple guards (contextual-override/final) for cxx.bnf;
- layering: clangPseudoCXX depends on clangPseudo (as the guard function need
to access the TokenStream);
Differential Revision: https://reviews.llvm.org/D127448
Nikita Popov [Wed, 29 Jun 2022 12:27:04 +0000 (14:27 +0200)]
[ConstExpr] Don't create div/rem expressions
This removes creation of udiv/sdiv/urem/srem constant expressions,
in preparation for their removal. I've added a
ConstantExpr::isDesirableBinOp() predicate to determine whether
an expression should be created for a certain operator.
With this patch, div/rem expressions can still be created through
explicit IR/bitcode, forbidding them entirely will be the next step.
Differential Revision: https://reviews.llvm.org/D128820
Eric Li [Mon, 4 Jul 2022 20:46:07 +0000 (20:46 +0000)]
[clang][dataflow] Handle null pointers of type std::nullptr_t
Treat `std::nullptr_t` as a regular scalar type to avoid tripping
assertions when analyzing code that uses `std::nullptr_t`.
Differential Revision: https://reviews.llvm.org/D129097
Joe Nash [Thu, 23 Jun 2022 19:57:01 +0000 (15:57 -0400)]
[AMDGPU] gfx11 Generate VOPD Instructions
We form VOPD instructions in the GCNCreateVOPD pass by combining
back-to-back component instructions. There are strict register
constraints for creating a legal VOPD, namely that the matching operands
(e.g. src0x and src0y, src1x and src1y) must be in different register
banks. We add a PostRA scheduler
mutation to put possible VOPD components back-to-back.
Depends on D128442, D128270
Reviewed By: #amdgpu, rampitec
Differential Revision: https://reviews.llvm.org/D128656
Haojian Wu [Tue, 5 Jul 2022 13:40:02 +0000 (15:40 +0200)]
[pseudo] Fix the build for the benchmark tool.
Vladislav Khmelevsky [Tue, 28 Jun 2022 16:54:59 +0000 (19:54 +0300)]
[RuntimeDyld] Fix R_AARCH64_TSTBR14 relocation
Wrong mask was used to get branch instruction imm value.
Differential Revision: https://reviews.llvm.org/D128740
Nikita Popov [Mon, 27 Jun 2022 13:09:24 +0000 (15:09 +0200)]
[SCEV] Fix isImpliedViaMerge() with values from previous iteration (PR56242)
When trying to prove an implied condition on a phi by proving it
for all incoming values, we need to be careful about values coming
from a backedge, as these may refer to a previous loop iteration.
A variant of this issue was fixed in D101829, but the dominance
condition used there isn't quite right: It checks that the value
dominates the incoming block, which doesn't exclude backedges
(values defined in a loop will usually dominate the loop latch,
which is the incoming block of the backedge).
Instead, we should be checking for domination of the phi block.
Any values defined inside the loop will not dominate the loop
header phi.
Fixes https://github.com/llvm/llvm-project/issues/56242.
Differential Revision: https://reviews.llvm.org/D128640
Andi-Bogdan Postelnicu [Tue, 5 Jul 2022 09:07:15 +0000 (09:07 +0000)]
[Compiler-RT] Remove FlushViewOfFile call when unmapping gcda files on win32.
This patch was pushed for calixte@mozilla.com
- this function (Windows only) is called when gcda are dumped on disk;
- according to its documentation, it's only useful in case of hard failures, this is highly improbable;
- it drastically decreases the time in the tests and consequently it avoids timeouts when we use slow disks.
Differential Revision: https://reviews.llvm.org/D129128
Haojian Wu [Mon, 4 Jul 2022 12:15:51 +0000 (14:15 +0200)]
[pseudo] Use the prebuilt cxx grammar for the lit tests, NFC.
Differential Revision: https://reviews.llvm.org/D129074
Ivan Kosarev [Tue, 5 Jul 2022 13:06:36 +0000 (14:06 +0100)]
[AMDGPU][NFC] Refine matching SMRD offsets.
Tell the matcher what we are looking for instead of matching everything
and then discarding the result if doesn't fit.
Reviewed By: foad
Differential Revision: https://reviews.llvm.org/D128171
Ivan Kosarev [Tue, 5 Jul 2022 12:39:46 +0000 (13:39 +0100)]
[AMDGPU][GlobalISel] Support register offsets for SMRDs.
Reviewed By: foad
Differential Revision: https://reviews.llvm.org/D128836
Sam McCall [Mon, 4 Jul 2022 18:35:40 +0000 (20:35 +0200)]
[pseudo] Eliminate LRTable::Action. NFC
The last remaining uses are in tests/test builders.
Replace with a builder struct.
Differential Revision: https://reviews.llvm.org/D129093
Nikita Popov [Tue, 5 Jul 2022 10:21:06 +0000 (12:21 +0200)]
[SimplifyCFG] Thread all predecessors with same value at once
If there are multiple predecessors that have the same condition
value (and thus same "real destination"), these were previously
handled by copying the threaded block for each predecessor.
Instead, we can reuse one block for all of them. This makes the
behavior of SimplifyCFG's jump threading match that of the
actual JumpThreading pass.
This also avoids the infinite combine loop reported in:
https://reviews.llvm.org/D124159#3624387