TST: test_blas_gemm_thread_safety depends on dev. version of openblas #31708

=================================== FAILURES ===================================
_________________________ test_blas_gemm_thread_safety _________________________
[gw0] linux -- Python 3.12.13 $SRC_DIR_run_env/bin/python
    def test_blas_gemm_thread_safety():
        # gh-31618: concurrently run transpose and no-transpose GEMM variants to
        # exercise possible thread safety issues due to lock sharding between
        # kernels, see OpenBLAS issue #5836.
        num_threads = 8
        num_iters = 10
        M = 512 * 512
    
        rng = np.random.default_rng(0x9e3779b9)
        no_trans = rng.random((M, 4))            # C-contiguous -> NoTrans GEMM
        no_trans_w = rng.random((4, 2))
        trans = rng.random((2, M)).T             # F-contiguous -> Trans GEMM
        trans_w = rng.random((2, 2))
        expected_no_trans = no_trans @ no_trans_w
        expected_trans = trans @ trans_w
    
        mismatches = 0
        lock = threading.Lock()
    
        def closure(i, b):
            nonlocal mismatches
            count = 0
            for _ in range(num_iters):
                b.wait()
                if i % 2:
                    ok = np.array_equal(no_trans @ no_trans_w, expected_no_trans)
                else:
                    ok = np.array_equal(trans @ trans_w, expected_trans)
                if not ok:
                    count += 1
            with lock:
                mismatches += count
    
        run_threaded(closure, num_threads, pass_count=True, pass_barrier=True)
    
        blas_name, blas_version = _detected_blas()
        if mismatches and _openblas_predates_gemm_fix(blas_name, blas_version):
            pytest.xfail(
                f"OpenBLAS version ({blas_version}) predates first OpenBLAS "
                "version with a fix (0.3.33.112)"
            )
    
>       assert mismatches == 0, (
            f"{mismatches} concurrent matmul results were corrupted "
            f"({blas_name} {blas_version})"
        )
E       AssertionError: 1 concurrent matmul results were corrupted (blas 3.9.0)
E       assert 1 == 0
M          = 262144
blas_name  = 'blas'
blas_version = '3.9.0'
closure    = <function test_blas_gemm_thread_safety.<locals>.closure at 0x7f3d07ddd4e0>
expected_no_trans = array([[0.82917678, 0.64631505],
       [1.12787757, 0.61713878],
       [0.58626928, 0.29928036],
       ...,
       [0.30331501, 0.19477312],
       [1.02426229, 0.7088999 ],
       [0.79192141, 0.50717405]], shape=(262144, 2))
expected_trans = array([[1.22946894, 0.72557492],
       [0.37233322, 0.61671449],
       [0.47452178, 0.67690663],
       ...,
       [0.89373242, 0.66193472],
       [0.83352945, 0.72978179],
       [0.78604996, 0.22733695]], shape=(262144, 2))
lock       = <unlocked _thread.lock object at 0x7f3d0e54e540>
mismatches = 1
no_trans   = array([[0.53808463, 0.40540176, 0.97098426, 0.29884582],
       [0.21645565, 0.85018773, 0.82101307, 0.79576803],
    ...0.82990353, 0.84376674, 0.61073338],
       [0.56083043, 0.88714949, 0.28678986, 0.73171138]],
      shape=(262144, 4))
no_trans_w = array([[0.02597612, 0.29835088],
       [0.14194139, 0.17054754],
       [0.57574152, 0.39776106],
       [0.66462428, 0.10178165]])
num_iters  = 10
num_threads = 8
rng        = Generator(PCG64) at 0x7F3D0552F060
trans      = array([[0.52190499, 0.99709558],
       [0.66401165, 0.01960259],
       [0.70724405, 0.10255883],
       ...,
       [0.55080277, 0.62915218],
       [0.65700102, 0.50680001],
       [0.0321854 , 0.80573585]], shape=(262144, 2))
trans_w    = array([[0.5325608 , 0.92152787],
       [0.95429448, 0.24533749]])
../test_run_env/lib/python3.12/site-packages/numpy/_core/tests/test_multithreading.py:149: AssertionError
=============================== warnings summary ===============================