Skip to content

Release GIL when calling mkl_lapack::orgqr #2850

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
antonwolfy wants to merge 3 commits into
base: master
Choose a base branch
from
+10 −0
Open

Release GIL when calling mkl_lapack::orgqr #2850

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
129f7b9
Release GIL before calling mkl_lapack::orgqr()
antonwolfy Apr 13, 2026
9f17f5c
Populate the changelog
antonwolfy Apr 13, 2026
91bd264
Apply suggestions from code review
antonwolfy Apr 14, 2026
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 CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,6 +84,7 @@ Also, that release drops support for Python 3.9, making Python 3.10 the minimum
* Resolved an issue with strides calculation in `dpnp.diagonal` to return correct values for empty diagonals [#2814](https://github.com/IntelPython/dpnp/pull/2814)
* Fixed test tolerance issues for float16 intermediate precision that became visible when testing against conda-forge's NumPy [#2828](https://github.com/IntelPython/dpnp/pull/2828)
* Ensured device aware dtype handling in `dpnp.identity` and `dpnp.gradient` [#2835](https://github.com/IntelPython/dpnp/pull/2835)
* Resolved a deadlock in `dpnp.linalg.qr` by releasing the GIL before OneMKL `orgqr` call to prevent host tasks contention [#2850](https://github.com/IntelPython/dpnp/pull/2850)

### Security

Expand Down
9 changes: 9 additions & 0 deletions dpnp/backend/extensions/lapack/orgqr.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -87,8 +87,17 @@ static sycl::event orgqr_impl(sycl::queue &exec_q,

sycl::event orgqr_event;
try {
// Release GIL to avoid serialization of host task submissions
Comment thread
vlad-perevezentsev marked this conversation as resolved.
// to the same queue in OneMKL
py::gil_scoped_release lock{};

scratchpad = sycl::malloc_device<T>(scratchpad_size, exec_q);

// mkl_lapack::orgqr() is done through GPU-to-Host reverse offload:
// exec_q.submit([&](sycl::handler& cgh) {
// cgh.depends_on(depends);
// cgh.host_task([=]() { orgqr_host(...); });
// }).wait();
orgqr_event = mkl_lapack::orgqr(
exec_q,
m, // The number of rows in the matrix; (0 ≤ m).
Expand Down
Loading