Reland "[flang][runtime] Use cuda::std::complex in F18 runtime CUDA build. (#109078)" (#109207)

`std::complex` operators do not work for the CUDA device compilation
of F18 runtime. This change make

Reland "[flang][runtime] Use cuda::std::complex in F18 runtime CUDA build. (#109078)" (#109207)

`std::complex` operators do not work for the CUDA device compilation
of F18 runtime. This change makes use of `cuda::std::complex` from
`libcudacxx`.
`cuda::std::complex` does not have specializations for `long double`,
so the change is accompanied with a clean-up for `long double` usage.

Additional change on top of #109078 is to use `cuda::std::complex`
only for the device compilation, otherwise the host compilation
fails because `libcudacxx` may not support `long double` specialization
at all (depending on the compiler).

show more ...

Revert "[flang][runtime] Use cuda::std::complex in F18 runtime CUDA build." (#109173)

Reverts llvm/llvm-project#109078

[flang][runtime] Use cuda::std::complex in F18 runtime CUDA build. (#109078)

`std::complex` operators do not work for the CUDA device compilation
of F18 runtime. This change makes use of `cuda::std

[flang][runtime] Use cuda::std::complex in F18 runtime CUDA build. (#109078)

`std::complex` operators do not work for the CUDA device compilation
of F18 runtime. This change makes use of `cuda::std::complex` from
`libcudacxx`.
`cuda::std::complex` does not have specializations for `long double`,
so the change is accompanied with a clean-up for `long double` usage.

show more ...

[flang] Lower MATMUL to type specific runtime calls. (#97547)

Lower MATMUL to the new runtime entries added in #97406.

[flang][runtime] Split MATMUL[_TRANSPOSE] into separate entries. (#97406)

Device compilation is much faster for separate MATMUL[_TRANPOSE]
entries than for a single one that covers all data types.

[flang][runtime] Split MATMUL[_TRANSPOSE] into separate entries. (#97406)

Device compilation is much faster for separate MATMUL[_TRANPOSE]
entries than for a single one that covers all data types.
The lowering changes and the removal of the generic entries will follow.

show more ...

[flang][runtime] Added Fortran::common::optional for use on device.

This is a simplified implementation of std::optional that can be used
in the offload builds for the device code. The methods are p

[flang][runtime] Added Fortran::common::optional for use on device.

This is a simplified implementation of std::optional that can be used
in the offload builds for the device code. The methods are properly
marked with RT_API_ATTRS so that the device compilation succedes.

Reviewers: klausler, jeanPerier

Reviewed By: jeanPerier

Pull Request: https://github.com/llvm/llvm-project/pull/85177

show more ...

[flang][runtime] Enable more APIs in the offload build. (#76486)

[flang][runtime] Enable more APIs in the offload build. (#75996)

This patch enables more numeric (mod, sum, matmul, etc.) APIs,
and some others.

I added new macros to disable warnings about usin

[flang][runtime] Enable more APIs in the offload build. (#75996)

This patch enables more numeric (mod, sum, matmul, etc.) APIs,
and some others.

I added new macros to disable warnings about using C++ STD methods
like operators of std::complex, which do not have __device__ attribute.
This may probably result in unresolved references, if the header files
implementation relies on libstdc++. I will need to follow up on this.

show more ...

[flang] Improved performance of runtime Matmul/MatmulTranspose.

This patch mostly affects performance of the code produced by
HLIFR lowering. If MATMUL argument is an array slice, then
HLFIR lowerin

[flang] Improved performance of runtime Matmul/MatmulTranspose.

This patch mostly affects performance of the code produced by
HLIFR lowering. If MATMUL argument is an array slice, then
HLFIR lowering passes the slice to the runtime, whereas
FIR lowering would create a contiguous temporary for the slice.
Performance might be better than the generic implementation
for cases where the leading dimension is contiguous.
This patch improves CPU2000/178.galgel making HLFIR version
faster than FIR version (due to avoiding the temporary copies
for MATMUL arguments).

Reviewed By: klausler

Differential Revision: https://reviews.llvm.org/D159134

show more ...

[NFC][flang] Distinguish MATMUL and MATMUL-TRANSPOSE printouts.

When MatmulTranpose reports incorrect shapes of the arguments
it cannot represent itself as MATMUL, because the reading
of the first a

[NFC][flang] Distinguish MATMUL and MATMUL-TRANSPOSE printouts.

When MatmulTranpose reports incorrect shapes of the arguments
it cannot represent itself as MATMUL, because the reading
of the first argument's shape will be confusing.

Reviewed By: kiranchandramohan

Differential Revision: https://reviews.llvm.org/D155911

show more ...

[flang] add fused matmul-transpose to the runtime

This fused operation should run a lot faster than first transposing the
lhs array and then multiplying the matrices separately.

Based on flang/runt

[flang] add fused matmul-transpose to the runtime

This fused operation should run a lot faster than first transposing the
lhs array and then multiplying the matrices separately.

Based on flang/runtime/matmul.cpp

Depends on D145959

Reviewed By: klausler

Differential Revision: https://reviews.llvm.org/D145960

show more ...