Skip to content

Extend getrs with trans_code argument #2563

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

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

Extend getrs with trans_code argument #2563

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
932a23a
Extend getrs with trans_code argument
vlad-perevezentsev Aug 18, 2025
5b5a2a5
Pass trans_code to getrs in dpnp_solve()
vlad-perevezentsev Aug 18, 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
20 changes: 15 additions & 5 deletions dpnp/backend/extensions/lapack/getrs.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -166,6 +166,7 @@ std::pair<sycl::event, sycl::event>
const dpctl::tensor::usm_ndarray &a_array,
const dpctl::tensor::usm_ndarray &ipiv_array,
const dpctl::tensor::usm_ndarray &b_array,
const int trans_code,
const std::vector<sycl::event> &depends)
{
const int a_array_nd = a_array.get_ndim();
Expand Down Expand Up @@ -264,11 +265,20 @@ std::pair<sycl::event, sycl::event>
const std::int64_t lda = std::max<size_t>(1UL, n);
const std::int64_t ldb = std::max<size_t>(1UL, n);

// Use transpose::T if the LU-factorized array is passed as C-contiguous.
// For F-contiguous we use transpose::N.
oneapi::mkl::transpose trans = is_a_array_c_contig
? oneapi::mkl::transpose::T
: oneapi::mkl::transpose::N;
oneapi::mkl::transpose trans;
switch (trans_code) {
case 0:
trans = oneapi::mkl::transpose::N;
break;
case 1:
trans = oneapi::mkl::transpose::T;
break;
case 2:
trans = oneapi::mkl::transpose::C;
break;
default:
throw py::value_error("`trans_code` must be 0 (N), 1 (T), or 2 (C)");
}

char *a_array_data = a_array.get_data();
char *b_array_data = b_array.get_data();
Expand Down
1 change: 1 addition & 0 deletions dpnp/backend/extensions/lapack/getrs.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,7 @@ extern std::pair<sycl::event, sycl::event>
const dpctl::tensor::usm_ndarray &a_array,
const dpctl::tensor::usm_ndarray &ipiv_array,
const dpctl::tensor::usm_ndarray &b_array,
const int trans_code,
const std::vector<sycl::event> &depends = {});

extern void init_getrs_dispatch_vector(void);
Expand Down
3 changes: 2 additions & 1 deletion dpnp/backend/extensions/lapack/lapack_py.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -160,7 +160,8 @@ PYBIND11_MODULE(_lapack_impl, m)
"the solves of linear equations with an LU-factored "
"square coefficient matrix, with multiple right-hand sides",
py::arg("sycl_queue"), py::arg("a_array"), py::arg("ipiv_array"),
py::arg("b_array"), py::arg("depends") = py::list());
py::arg("b_array"), py::arg("trans_code"),
py::arg("depends") = py::list());

m.def("_orgqr_batch", &lapack_ext::orgqr_batch,
"Call `_orgqr_batch` from OneMKL LAPACK library to return "
Expand Down
7 changes: 7 additions & 0 deletions dpnp/linalg/dpnp_utils_linalg.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2632,6 +2632,12 @@ def dpnp_solve(a, b):
_manager = dpu.SequentialOrderManager[exec_q]
dev_evs = _manager.submitted_events

# TODO: remove after PR #2558 is merged
# Temporarily set trans_code=1 (transpose) because the LU-factorized
# array is C-contiguous.
# For F-contiguous arrays use 0 (non-transpose)
trans_code = 1

# use DPCTL tensor function to fill the сopy of the input array
# from the input array
ht_ev, a_copy_ev = ti._copy_usm_ndarray_into_usm_ndarray(
Expand Down Expand Up @@ -2688,6 +2694,7 @@ def dpnp_solve(a, b):
a_h.get_array(),
ipiv_h.get_array(),
b_h.get_array(),
trans_code,
depends=[b_copy_ev, getrf_ev],
)
_manager.add_event_pair(ht_ev, getrs_ev)
Expand Down
Loading