Flaky runtime/test_launch.py::test_memory_leak

[pbchekin]

_______________________________ test_memory_leak _______________________________

device = 'xpu'

    def test_memory_leak(device) -> None:
    
        @triton.jit
        def kernel(in_ptr0, out_ptr0, xnumel, XBLOCK: tl.constexpr):
            xnumel = 10
            xoffset = tl.program_id(0) * XBLOCK
            xindex = xoffset + tl.arange(0, XBLOCK)[:]
            xmask = xindex < xnumel
            x0 = xindex
            tmp0 = tl.load(in_ptr0 + (x0), xmask)
            tl.store(out_ptr0 + (x0 + tl.zeros([XBLOCK], tl.int32)), tmp0, xmask)
    
        tracemalloc.start()
        try:
            inp = torch.randn(10, device=device)
            out = torch.randn(10, device=device)
            kernel[(10, )](inp, out, 10, XBLOCK=16)
            gc.collect()
            begin, _ = tracemalloc.get_traced_memory()
            for _ in range(100):
                kernel[(10, )](inp, out, 10, XBLOCK=16)
            gc.collect()
            end, _ = tracemalloc.get_traced_memory()
>           assert end - begin < 30000
E           assert (79991 - 37617) < 30000

runtime/test_launch.py:69: AssertionError

Examples:

• https://github.com/intel/intel-xpu-backend-for-triton/actions/runs/12292303461/job/34302688686
• https://github.com/intel/intel-xpu-backend-for-triton/actions/runs/12292303461/job/34302689536
• https://github.com/intel/intel-xpu-backend-for-triton/actions/runs/12300835860/job/34330002372

[pbchekin]

Python 3.10, 3.11, 3.12 but looks like there are no cases with Python 3.9.