This simplifies completeness comparisons against OpenCLBuiltins.td and
also makes the header no longer "claim" the identifiers "image",
"image_array", "coord", "sampler", "sample", "gradientX", "gradientY",
"lod", and "color".
Continues the direction set out in D119560.
The current implementation of ShuffleVectorOp assumes all vectors are
scalable. LLVM IR allows shufflevector operations on scalable vectors,
and the current translation between LLVM Dialect and LLVM IR does the
rigth thing when the shuffle mask is all zeroes. This is required to
do a splat operation on a scalable vector, but it doesn't make sense
for scalable vectors outside of that operation, i.e.: with non-all zero
masks.
Differential Revision: https://reviews.llvm.org/D118371
Using getSplatValue causes poor codegen due to not always being able to remove the EXTRACT_VECTOR_ELT created inside getSplatValue.
The vXi16 shifts/rotates are still showing occasional regressions but vXi8 is a definite improvement.
In D115022, we introduced an optimization where OpResults of a `linalg.generic` may bufferize in-place with an "in" OpOperand if the corresponding "out" OpOperand is not used in the computation.
This optimization can lead to unexpected behavior if the newly chosen OpOperand is in the same alias set as another OpOperand (that is used in the computation). In that case, the newly chosen OpOperand must bufferize out-of-place. This can be confusing to users, as always choosing the "out" OpOperand (regardless of whether it is used) would be expected when having the notion of "destination-passing style" in mind.
With this change, we go back to always bufferizing in-place with "out" OpOperands by default, but letting users override the behavior with a bufferization option.
Differential Revision: https://reviews.llvm.org/D120182
All current callers set the argument to false. monitor_signals=true used
to be used in the Process plugins (which needed to know when the
debugged process gets a signal), but this implementation has several
serious issues, which means that individual process plugins now
orchestrate the monitoring of debugged processes themselves.
This allows us to simplify the implementation (no need to play with
process groups), and the interface (we only catch fatal events, so the
callback is always called just once).
Differential Revision: https://reviews.llvm.org/D120425
Previously only accessing values for `index` and signless int types
would work; signed and unsigned ints would hit an assert in
`IntegerAttr::getInt`. This exposes `IntegerAttr::get{S,U}Int` to the C
API and calls the appropriate function from the python bindings.
Reviewed By: ftynse
Differential Revision: https://reviews.llvm.org/D120194
Fixes https://github.com/llvm/llvm-project/issues/53876.
This is a solution for standard C++ casts: const_cast, dynamic_cast, reinterpret_cast, static_cast.
A general approach handling all possible casts is not possible without semantic information.
Consider the code:
```
static_cast<T>(*function_pointer_variable)(arguments);
```
vs.
```
some_return_type<T> (*function_pointer_variable)(parameters);
// Later used as:
function_pointer_variable = &some_function;
return function_pointer_variable(args);
```
In the latter case, it's not a cast but a variable declaration of a pointer to function.
Without knowing what `some_return_type<T>` is (and clang-format does not know it), it's hard to distinguish between the two cases. Theoretically, one could check whether "parameters" are types (not a cast) and "arguments" are value/expressions (a cast), but that might be inefficient (needs lots of lookahead).
Reviewed By: MyDeveloperDay, HazardyKnusperkeks, owenpan
Differential Revision: https://reviews.llvm.org/D120140
Sanitizer coverage point should be the previous instruction PC of the
caller and the offset to the previous instruction might be different
on each CPU architecture.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D119233
Fixes https://github.com/llvm/llvm-project/issues/53962.
Given the config:
```
BasedOnStyle: LLVM
QualifierAlignment: Custom
QualifierOrder: ['constexpr', 'type']
```
The code:
```
template <typename F>
requires std::invocable<F>
constexpr constructor();
```
was incorrectly formatted to:
```
template <typename F>
requires
constexpr std::invocable<F> constructor();
```
because we considered `std::invocable<F> constexpr` as a type, not recognising the requires clause.
This patch avoids moving the qualifier across the boundary of the requires clause (checking `ClosesRequiresClause`).
Reviewed By: HazardyKnusperkeks, owenpan
Differential Revision: https://reviews.llvm.org/D120309
When assembling for example an SVE instruction with the `.arch +sve2` directive,
+sve should be implied by setting +sve2, similar to what would happen if
one would pass the mattr=+sve2 flag on the command-line.
The AsmParser doesn't set the implied features, meaning that the SVE
instruction does not assemble. This patch fixes that.
Note that the same does not hold when disabling a feature. For example,
+nosve2 does not imply +nosve.
Reviewed By: c-rhodes
Differential Revision: https://reviews.llvm.org/D120259
This implements the parsing and recognition of module partition CMIs
and removes the FIXMEs in the parser.
Module partitions are recognised in the base computation of visibility,
however additional amendments to visibility follow in subsequent patches.
Differential Revision: https://reviews.llvm.org/D118586
This fixes issue #45084 (https://llvm.org/PR45739).
Remove unnecessary trailing spaces after the "RUB" international
currency symbol (and after the plain number in some parts of the
put_long_double test).
Both of these test files are
`XFAIL: netbsd || linux || LIBCXX-WINDOWS-FIXME`, and then have some of
their test cases commented out when `__APPLE__`. This patch comments-in
those test cases and adjusts them all to work on Apple, while leaving the
test `XFAIL`ed on NetBSD, Linux, and Windows.
Differential Revision: https://reviews.llvm.org/D120316
Now that integer min/max intrinsics have good support in both
InstCombine and other passes, start canonicalizing SPF min/max
to intrinsic min/max.
Once this sticks, we can stop matching SPF min/max in various
places, and can remove hacks we have for preventing infinite loops
and breaking of SPF canonicalization.
Differential Revision: https://reviews.llvm.org/D98152
PC-relative memory operand could reference a different object from
the one located at the target address, e.g. when a negative offset
is used. Check relocations for the real referenced object.
Reviewed By: rafauler
Differential Revision: https://reviews.llvm.org/D120379
Further improve error handling in BOLT by reporting `RewriteInstance` errors in
a library and fuzzer-friendly way instead of exiting.
Follow-up to D119658
Reviewed By: rafauler
Differential Revision: https://reviews.llvm.org/D120224
* detect `def_tls.o undef_nontls.o` violation
* place error checking code (checking duplicate symbol) together
* allow `--defsym tls1=tls2 def_tls.o`
As a degraded error checking, `--defsym tls1=42` violation will not be detected.
Introduce -fgpu-default-stream={legacy|per-thread} option to
support per-thread default stream for HIP runtime.
When -fgpu-default-stream=per-thread, HIP kernels are
launched through hipLaunchKernel_spt instead of
hipLaunchKernel. Also HIP_API_PER_THREAD_DEFAULT_STREAM=1
is defined by the preprocessor to enable other per-thread stream
API's.
Reviewed by: Artem Belevich
Differential Revision: https://reviews.llvm.org/D120298
Previously, we only support float64. We now support float32 and float64. When
constructing a tensor without providing a data type, the default is float32.
Fix the tests to data type consistency. All PyTACO application tests now use
float32 to match the default data type of TACO. Other tests may use float32 or
float64.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D120356
Symbol.h depends on InputFiles.h. This change moves us toward dropping the
weird dependency.
The call sites will become slightly uglier (`cast<SharedFile>(s->file)`), but
the compromise is acceptable.
D118090 causes a pretty significant (19%) regression in some Eigen
benchmarks. Investigating is a bit time consuming as the compilation
unit where this occurs is large. Rather than revert, this patch adds a
flag controlling that behavior (enabled by default).
Now that sparse tensor types are first-class citizens and the sparse compiler
is taking shape, it is time to make sure other compiler optimizations compose
well with sparse tensors. Mostly, this should be completely transparent (i.e.,
dense and sparse take the same path). However, in some cases, optimizations
only make sense in the context of sparse tensors. This is a first example of
such an optimization, where fusing a sampled elt-wise multiplication only makes
sense when the resulting kernel has a potential lower asymptotic complexity due
to the sparsity.
As an extreme example, running SDDMM with 1024x1024 matrices and a sparse
sampling matrix with only two elements runs in 463.55ms in the unfused
case but just 0.032ms in the fused case, with a speedup of 14485x that
is only possible in the exciting world of sparse computations!
Reviewed By: mravishankar
Differential Revision: https://reviews.llvm.org/D120429
This patch is a follow-up of D120100 to address some feedbacks from
@labath.
This should mainly fix the race issue with the even listener by moving
the listener setup to the main thread.
This also changes the SBDebugger::GetProgressFromEvent SWIG binding
arguments to be output only, so the user don't have to provide them.
Finally, this updates the test to check it the out arguments are returned
in a tuple and re-enables the test on all platforms.
Differential Revision: https://reviews.llvm.org/D120284
Signed-off-by: Med Ismail Bennani <medismail.bennani@gmail.com>
The `SplitIndirectBrCriticalEdges` function was originally designed for
`CodeGenPrepare` and skipped splitting of edges when the destination
block didn't contain any `PHI` instructions. This only makes sense when
reducing COPYs like `CodeGenPrepare`. In the case of
`PGOInstrumentation` or `GCOVProfiling` it would result in missed
counters and wrong result in functions with computed goto.
Differential Revision: https://reviews.llvm.org/D120096
Jay Foad