Skip to content

Add wait to MKL calls #518

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
amontoison merged 2 commits into from
Aug 23, 2025
Merged

Add wait to MKL calls #518

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3ebc1df
Add wait to MKL calls
michel2323 Aug 22, 2025
2da6dde
More fixes
michel2323 Aug 23, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions deps/generate_interfaces.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -452,6 +452,7 @@ function generate_cpp(library::String, filename::Vector{String}, output::String;
else
if !(name ∈ void_output)
write(oneapi_cpp, " auto status = oneapi::mkl::$library::$variant$name($parameters, {});\n")
write(oneapi_cpp, " device_queue->val.wait_and_throw();\n")
else
write(oneapi_cpp, " oneapi::mkl::$library::$variant$name($parameters);\n")
end
Expand Down
1 change: 1 addition & 0 deletions deps/onemkl_epilogue.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
extern "C" int onemklXsparse_matmat(syclQueue_t device_queue, matrix_handle_t A, matrix_handle_t B, matrix_handle_t C, onemklMatmatRequest req, matmat_descr_t descr, int64_t *sizeTempBuffer, void *tempBuffer) {
auto status = oneapi::mkl::sparse::matmat(device_queue->val, (oneapi::mkl::sparse::matrix_handle_t) A, (oneapi::mkl::sparse::matrix_handle_t) B, (oneapi::mkl::sparse::matrix_handle_t) C, convert(req), (oneapi::mkl::sparse::matmat_descr_t) descr, sizeTempBuffer, tempBuffer, {});
device_queue->val.wait_and_throw();
return 0;
}

Expand Down
18 changes: 18 additions & 0 deletions deps/onemkl_prologue.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -393,13 +393,15 @@ extern "C" int onemklHgemm_batch(syclQueue_t device_queue, onemklTranspose trans
int64_t *ldb, uint16_t *beta, short **c,
int64_t *ldc, int64_t group_count, int64_t *group_size) {
gemmBatchInfo gemmInfo(device_queue, group_count, transa, transb);
device_queue->val.wait_and_throw();
auto status = oneapi::mkl::blas::column_major::gemm_batch(device_queue->val,
&gemmInfo.m_transa[0], &gemmInfo.m_transb[0],
m, n, k, reinterpret_cast<sycl::half *>(alpha),
reinterpret_cast<const sycl::half **>(&a[0]), lda,
reinterpret_cast<const sycl::half **>(&b[0]), ldb,
reinterpret_cast<sycl::half *>(beta), reinterpret_cast<sycl::half **>(&c[0]),
ldc, group_count, group_size, {});
device_queue->val.wait_and_throw();
return 0;
}

Expand All @@ -410,13 +412,15 @@ extern "C" int onemklSgemm_batch(syclQueue_t device_queue, onemklTranspose trans
int64_t *ldb, float *beta, float **c,
int64_t *ldc, int64_t group_count, int64_t *group_size) {
gemmBatchInfo gemmInfo(device_queue, group_count, transa, transb);
device_queue->val.wait_and_throw();
auto status = oneapi::mkl::blas::column_major::gemm_batch(device_queue->val,
&gemmInfo.m_transa[0], &gemmInfo.m_transb[0],
m, n, k, alpha,
(const float **)&a[0], lda,
(const float **)&b[0], ldb,
beta, &c[0], ldc,
group_count, group_size, {});
device_queue->val.wait_and_throw();
return 0;
}

Expand All @@ -427,13 +431,15 @@ extern "C" int onemklDgemm_batch(syclQueue_t device_queue, onemklTranspose trans
int64_t *ldb, double *beta, double **c,
int64_t *ldc, int64_t group_count, int64_t *group_size) {
gemmBatchInfo gemmInfo(device_queue, group_count, transa, transb);
device_queue->val.wait_and_throw();
auto status = oneapi::mkl::blas::column_major::gemm_batch(device_queue->val,
&gemmInfo.m_transa[0], &gemmInfo.m_transb[0],
m, n, k, alpha,
(const double **)&a[0], lda,
(const double **)&b[0], ldb,
beta, &c[0], ldc,
group_count, group_size, {});
device_queue->val.wait_and_throw();
return 0;
}

Expand All @@ -445,6 +451,7 @@ extern "C" int onemklCgemm_batch(syclQueue_t device_queue, onemklTranspose trans
int64_t *ldb, float _Complex *beta, float _Complex **c,
int64_t *ldc, int64_t group_count, int64_t *group_size) {
gemmBatchInfo gemmInfo(device_queue, group_count, transa, transb);
device_queue->val.wait_and_throw();
auto status = oneapi::mkl::blas::column_major::gemm_batch(device_queue->val,
&gemmInfo.m_transa[0], &gemmInfo.m_transb[0],
m, n, k, reinterpret_cast<std::complex<float> *>(alpha),
Expand All @@ -455,6 +462,7 @@ extern "C" int onemklCgemm_batch(syclQueue_t device_queue, onemklTranspose trans
reinterpret_cast<std::complex<float> *>(beta),
reinterpret_cast<std::complex<float> **>(&c[0]), ldc,
group_count, group_size, {});
device_queue->val.wait_and_throw();
return 0;
}

Expand All @@ -467,6 +475,7 @@ extern "C" int onemklZgemm_batch(syclQueue_t device_queue, onemklTranspose trans
double _Complex **c,
int64_t *ldc, int64_t group_count, int64_t *group_size) {
gemmBatchInfo gemmInfo(device_queue, group_count, transa, transb);
device_queue->val.wait_and_throw();
auto status = oneapi::mkl::blas::column_major::gemm_batch(device_queue->val,
&gemmInfo.m_transa[0], &gemmInfo.m_transb[0],
m, n, k, reinterpret_cast<std::complex<double> *>(alpha),
Expand All @@ -477,6 +486,7 @@ extern "C" int onemklZgemm_batch(syclQueue_t device_queue, onemklTranspose trans
reinterpret_cast<std::complex<double> *>(beta),
reinterpret_cast<std::complex<double> **>(&c[0]), ldc,
group_count, group_size, {});
device_queue->val.wait_and_throw();
return 0;
}

Expand All @@ -487,12 +497,14 @@ extern "C" int onemklStrsm_batch(syclQueue_t device_queue, onemklSide left_right
int64_t group_count, int64_t *group_size) {
trsmBatchInfo trsmInfo(device_queue, left_right, upper_lower, transa,
unit_diag, group_count);
device_queue->val.wait_and_throw();

auto status = oneapi::mkl::blas::column_major::trsm_batch(device_queue->val,
&trsmInfo.m_leftright[0], &trsmInfo.m_upperlower[0],
&trsmInfo.m_transa[0], &trsmInfo.m_unitdiag[0],
m, n, alpha, (const float **)&a[0], lda,
&b[0], ldb, group_count, group_size, {});
device_queue->val.wait_and_throw();
return 0;
}

Expand All @@ -504,12 +516,14 @@ extern "C" int onemklDtrsm_batch(syclQueue_t device_queue, onemklSide left_right
int64_t *group_size) {
trsmBatchInfo trsmInfo(device_queue, left_right, upper_lower, transa,
unit_diag, group_count);
device_queue->val.wait_and_throw();

auto status = oneapi::mkl::blas::column_major::trsm_batch(device_queue->val,
&trsmInfo.m_leftright[0], &trsmInfo.m_upperlower[0],
&trsmInfo.m_transa[0], &trsmInfo.m_unitdiag[0],
m, n, alpha, (const double **)&a[0], lda, &b[0],
ldb, group_count, group_size, {});
device_queue->val.wait_and_throw();
return 0;
}

Expand All @@ -521,6 +535,7 @@ extern "C" int onemklCtrsm_batch(syclQueue_t device_queue, onemklSide left_right
int64_t group_count, int64_t *group_size) {
trsmBatchInfo trsmInfo(device_queue, left_right, upper_lower, transa,
unit_diag, group_count);
device_queue->val.wait_and_throw();

auto status = oneapi::mkl::blas::column_major::trsm_batch(device_queue->val,
&trsmInfo.m_leftright[0], &trsmInfo.m_upperlower[0],
Expand All @@ -529,6 +544,7 @@ extern "C" int onemklCtrsm_batch(syclQueue_t device_queue, onemklSide left_right
reinterpret_cast<const std::complex<float> **>(&a[0]),
lda, reinterpret_cast<std::complex<float> **>(&b[0]),
ldb, group_count, group_size, {});
device_queue->val.wait_and_throw();
return 0;
}

Expand All @@ -540,6 +556,7 @@ extern "C" int onemklZtrsm_batch(syclQueue_t device_queue, onemklSide left_right
int64_t group_count, int64_t *group_size) {
trsmBatchInfo trsmInfo(device_queue, left_right,
upper_lower, transa, unit_diag, group_count);
device_queue->val.wait_and_throw();

auto status = oneapi::mkl::blas::column_major::trsm_batch(device_queue->val,
&trsmInfo.m_leftright[0], &trsmInfo.m_upperlower[0],
Expand All @@ -548,5 +565,6 @@ extern "C" int onemklZtrsm_batch(syclQueue_t device_queue, onemklSide left_right
reinterpret_cast<const std::complex<double> **>(&a[0]),
lda, reinterpret_cast<std::complex<double> **>(&b[0]),
ldb, group_count, group_size, {});
device_queue->val.wait_and_throw();
return 0;
}
Loading