Merge branch 'master' into add-api-functions-to-usmarray

Author: ndgrigorian

.github/workflows/openssf-scorecard.yml

@@ -72,6 +72,6 @@ jobs:
 
       # Upload the results to GitHub's code scanning dashboard.
       - name: "Upload to code-scanning"
-        uses: github/codeql-action/upload-sarif@c10b8064de6f491fea524254123dbe5e09572f13 # v4.35.1
+        uses: github/codeql-action/upload-sarif@95e58e9a2cdfd71adc6e0353d5c52f41a045d225 # v4.35.2
         with:
           sarif_file: results.sarif

CHANGELOG.md

@@ -60,6 +60,7 @@ Also, that release drops support for Python 3.9, making Python 3.10 the minimum
 * Moved all SYCL kernel functors from `backend/extensions/` to a unified `backend/kernels/` directory hierarchy [#2816](https://github.com/IntelPython/dpnp/pull/2816)
 * `dpnp` uses pybind11 3.0.3 [#2834](https://github.com/IntelPython/dpnp/pull/2834)
 * Disabled `dpnp.tensor` tests by default in `conda build --test` to prevent OOM failures during package testing. Set `SKIP_TENSOR_TESTS=0` to re-enable them on systems with enough memory [#2860](https://github.com/IntelPython/dpnp/pull/2860)
+* `dpnp` uses pybind11 3.0.4 [#2865](https://github.com/IntelPython/dpnp/pull/2865)
 
 ### Deprecated
 
@@ -88,6 +89,7 @@ Also, that release drops support for Python 3.9, making Python 3.10 the minimum
 * 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)
 * Fixed `dpnp.tensor.round` to use device-aware output dtype for boolean input [#2851](https://github.com/IntelPython/dpnp/pull/2851)
+* 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
 

CMakeLists.txt

@@ -106,8 +106,8 @@ find_package(Python 3.10...<3.15 REQUIRED COMPONENTS Development.Module NumPy)
 include(FetchContent)
 FetchContent_Declare(
     pybind11
-    URL https://github.com/pybind/pybind11/archive/refs/tags/v3.0.3.tar.gz
-    URL_HASH SHA256=787459e1e186ee82001759508fefa408373eae8a076ffe0078b126c6f8f0ec5e
+    URL https://github.com/pybind/pybind11/archive/refs/tags/v3.0.4.tar.gz
+    URL_HASH SHA256=74b6a2c2b4573a400cafb6ecbf60c98df300cd3d0041296b913d02b2cbbb2676
     FIND_PACKAGE_ARGS NAMES pybind11
 )
 FetchContent_MakeAvailable(pybind11)

dpnp/backend/extensions/lapack/orgqr.cpp

@@ -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
+        // 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).