[PageFault] ElementwiseGlobalRangeKernel cause pagefault

[Stonepia]

🐛 Describe the bug

Affected total of 18 test:

test_stable_sort_against_numpy_xpu_bfloat16
test_stable_sort_against_numpy_xpu_float16
test_stable_sort_against_numpy_xpu_float32
test_stable_sort_against_numpy_xpu_float64
test_stable_sort_against_numpy_xpu_int16
test_stable_sort_against_numpy_xpu_int32
test_stable_sort_against_numpy_xpu_int64
test_stable_sort_against_numpy_xpu_int8
test_stable_sort_against_numpy_xpu_uint8

Run command:

export DisableScratchPages=1 
export NEOReadDebugKeys=1

export PYTORCH_TEST_WITH_SLOW=1 

python -m pytest -v test_ops_fwd_gradients_xpu.py -k test_fn_fwgrad_bwgrad_abs_xpu_float64 

Use the following command for more detail:

export SYCL_PI_TRACE=-1

export ZE_SERIALIZE=2
export OverrideImmediateCmdListSynchronousMode=1

Versions

PyTorch version: 2.5.0a0+gitae708e9
Is debug build: False
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.3 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect
CMake version: version 3.30.0
Libc version: glibc-2.35

Python version: 3.10.14 | packaged by conda-forge | (main, Mar 20 2024, 12:45:18) [GCC 12.3.0] (64-bit runtime)
Python platform: Linux-5.15.47+prerelease24.3.26-x86_64-with-glibc2.35
Is CUDA available: False
CUDA runtime version: No CUDA
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

Versions of relevant libraries:
[pip3] numpy==1.26.4
[pip3] optree==0.12.1
[pip3] torch==2.5.0a0+gitae708e9
[conda] numpy 1.26.4 pypi_0 pypi
[conda] optree 0.12.1 pypi_0 pypi
[conda] torch 2.5.0a0+gitae708e9 dev_0

[Stonepia]

Fixed by
#734
#735