The test may fail when running from a directory that contains the string used in
CHECK-NOT. We observe flakiness rate of around 3/100000. Increasing the length
helps reducing the rate of failures.

Reviewed By: DavidSpickett

Differential Revision: https://reviews.llvm.org/D148099

This is useful to materialize copies explicitly before bufferization and
transform them, avoiding the need to rediscover them after bufferization.

Differential Revision: https://reviews.llvm.org/D148108

Add a new transform op combinator that implements an "if-then-else"
style of mechanism for applying transformations. Its main purpose is to
serve as a higher-level driver when applying multiple transform scripts
to potentially overlapping pieces of the payload IR. This is similar to
how the various rewrite drivers operate in C++, but at a higher level
and with more declarative expressions. This is not intended to replace
existing pattern-based rewrites, but to to drive more complex
transformations that are exposed in the transform dialect and are too
complex to be expressed as simple declarative rewrites.

Reviewed By: springerm

Differential Revision: https://reviews.llvm.org/D148013

We only need the sign bits from these nodes

Another step towards Issue #60007

Add some tests to generalize the clearing of known bits using
computeKnownFPClass instead of isKnownNeverNaN/isKnownNeverInfinity.

Copy the logic from isKnownNeverNaN for fadd/fsub.

Avoids some regressions in the math libraries in a future
patch.

In strict mode the 'roundin_mode' is set to 'dynamic'. Using this pragma to
get out of strict mode doesn't have any effect on the 'rounding_mode'.
See https://godbolt.org/z/zoGTf4j1G
This patch fixes that.

Differential Revision: https://reviews.llvm.org/D147733

Part of Issue #60007

Most of these folds bitcast to the PTEST operand type anyway, and its only relevant for the PTEST->MOVMSK fold, which I'm looking at expanding to attempt to fold to PTEST->TESTP as well.

Noticed while beginning triage of Issue #60007

Reviewed By: vitalybuka, dvyukov

Differential Revision: https://reviews.llvm.org/D147337

This extends the existing tests to include negative zeros and additional
combinations. See D147724 and D147723.

Add post-inc variations of tests.

[Reland][ADT][ConcurrentHashTable] adapt ConcurrentHashTable and its users to LLVM_ENABLE_THREADS=0 mode.

This patch hides thread specific handling under LLVM_ENABLE_THREADS guard.
It also removes usages of thread_local variables, since it has a weak
support on some platforms. Instead, the patch uses single mutex for locking
allocator. That may be replaced with more effective allocator later.
f.e. D142318

Differential Revision: https://reviews.llvm.org/D147649

If we go to line 302, with one of MCE or MAB is not nullptr, then we could
leak mem here.
Use unique_ptr to maintain these 2 pointer can avoid it.

Reviewed By: LuoYuanke

Differential Revision: https://reviews.llvm.org/D148003

/data/llvm-project/mlir/lib/Dialect/Bufferization/IR/BufferizableOpInterface.cpp:342:2: error: extra ';' outside of a function is incompatible with C++
98 [-Werror,-Wc++98-compat-extra-semi]
}; // namespace

Having to choose from only static or dynamic layout for all function is limiting.

Differential Revision: https://reviews.llvm.org/D148074

Add an additional explanation of what is wrong if a constraint is
not satisfied, in some cases.
Additionally the bug report generation is changed to use raw_ostream.

Reviewed By: Szelethus, NoQ

Differential Revision: https://reviews.llvm.org/D144003

Because the comment parser does not support slices,
we emit a warning for comments that do contain
a splice within their delimiter, and do not add them as
documentation comment.

Fixes #62054

Reviewed By: shafik, aaron.ballman

Differential Revision: https://reviews.llvm.org/D148029

This header isn't used by any public header, so there shouldn't
be any need to install it or treat it as a heder.

Once it's part of the src subdirectory, I guess one could consider
giving it a more traditional name too.

Differential Revision: https://reviews.llvm.org/D147855

Use uintptr_t instead. On Windows, long is 32 bit even on 64 bit
architectures.

Differential Revision: https://reviews.llvm.org/D147639

This test succeeds on aarch64 windows now (tested both in mingw
mode and with clang-cl), most probably since commit
ccd96b3e / D146667 ([builtins][test] Fix
divmodti4_test.c on Windows).

Differential Revision: https://reviews.llvm.org/D148000

When the libcxx test framework is executed within libunwind, there
are no standard C++ headers available (libunwind builds with
-nostdinc++, but doesn't add any libcxx headers to the include path).

Check that a test that includes <iostream> can be compiled before trying
to build and execute a test program that includes it.

Previously, the compile error here would block all libunwind tests from
executing altogether.

Differential Revision: https://reviews.llvm.org/D147630

No replacement ops are needed for tracked ops who's handles are all dead.

Differential Revision: https://reviews.llvm.org/D147510

This patch removes forward declarations of llvm::Any as they are not
used.  Removing them gets us a bit closer to the world where we use
std::any instead of llvm::Any because we won't forward-declare std::
classes.

From line 693 in file llc.cpp, it uses new operator to creates a ModulePass
and assigned to MMIWP. If the condition after take the true branch, it has
chance to go in to line 702 or line 709, the function will return without
cleaning the memory.
The second issue existed for the same reason, the ref TPC get the pointer
created from LLVMTM.createPassConfig, and will leak memory if goes into
line 709.
This patch uses delete in the issued branch.

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D147993

This revision adds additional "expensive-checks" checks to the transform dialect that detect the most common cases of:

* Missing `consumesHandle` side effects on transform ops.
* Patterns that remove operations but do not notify the transform dialect.

In essence, these additional checks are looking for dangling pointers to erased payload ops in the transform dialect state and crash the program execution (by dereferencing free'd memory) or triggering an assertion failure. It is recommended to run these extra checks with ASAN. Otherwise, certain failures may not be detected. The ASAN error message can also be used to find the faulty transform op/pattern.

This change also fixes a few faulty transform ops.

Differential Revision: https://reviews.llvm.org/D147447

`Builder.GetInsertBlock()` may return null sometimes. https://godbolt.org/z/n1Ph47jP1

Reviewed By: efriedma

Differential Revision: https://reviews.llvm.org/D147867

Differential Revision: https://reviews.llvm.org/D147997

Optimize (mul x, imm) to (ADD (SLLI x, s0), (SLLI x, s1)) or
(SUB (SLLI x, s0), (SLLI x, s1)) if possible.

Reviewed By: SixWeining

Differential Revision: https://reviews.llvm.org/D147692

Reviewed By: SixWeining

Differential Revision: https://reviews.llvm.org/D147691