[Break XPU] Triton kernel generated for XPU can not pass runtime_triton_dtype_assert. #141861
Description
🐛 Describe the bug
The PR #141495 introduced runtime_triton_dtype_assert that check the dtype in triton kernel, and XPU can not pass the check in the test case test_torchinductor_opinfo.py:test_comprehensive_masked_cumprod_xpu_float16:
in compile
module = src.make_ir(options, codegen_fns, module_map, context)
File "/home/xinanlin/xinanlin/miniforge3/envs/torch-py3.9/lib/python3.9/site-packages/triton/compiler/compiler.py", line 101
, in make_ir
return ast_to_ttir(self.fn, self, context=context, options=options, codegen_fns=codegen_fns,
triton.compiler.errors.CompileTimeAssertionFailure: at 65:4:
tmp14 = tmp0 * tmp13
tl.static_assert(tmp14.dtype == tl.float32)
tmp15 = tl.broadcast_to(tmp14, [XBLOCK, RBLOCK])
tmp17 = tl.where(rmask, tmp15, 0)
tmp18 = tl.sum(tmp17, 1)[:, None]
tl.static_assert(tmp1.dtype == tl.int32)
tl.static_assert(tmp2.dtype == tl.int64)
tl.static_assert(tmp3.dtype == tl.int1)
tl.static_assert(tmp1.dtype == tl.int32)
tl.static_assert(tmp4.dtype == tl.int64)
tl.static_assert(tmp5.dtype == tl.int1)
tl.static_assert(tmp18.dtype == tl.float16)
^
The above exception was the direct cause of the following exception:
Versions
PyTorch version: 2.6.0a0+gitf83361b
Is debug build: False
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A
OS: Ubuntu 22.04.5 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect
CMake version: version 3.29.4
Libc version: glibc-2.35
Python version: 3.9.20 | packaged by conda-forge | (main, Sep 30 2024, 17:49:10) [GCC 13.3.0] (64-bit runtime)
Python platform: Linux-5.15.47+prerelease23.5.18-x86_64-with-glibc2.35