frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 21:14:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

[inductor] use hint_override in kernel benchmark args (#164207)

Browse Source 
Summary: forward fix T239259207

Test Plan: test_multi_kernel

Differential Revision: D83539263

Pull Request resolved: https://github.com/pytorch/pytorch/pull/164207
Approved by: https://github.com/bobrenjc93, https://github.com/mlazos
This commit is contained in:
Pian Pawakapan
2025-09-30 18:09:29 +00:00
committed by PyTorch MergeBot
parent 719b64ee8b
commit d615f6b935
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
torch/_inductor/codegen/triton.py
View File

@ -4098,7 +4098,9 @@ class TritonKernel(SIMDKernel[TritonCSEVariable]):
f"{var_name} = rand_strided({size}, {stride}, device='{const_tensor.device}', dtype={const_tensor.dtype})" # type: ignore[arg-type] # noqa: B950 line too long
)
elif isinstance(arg_sig, SizeArg):
symval_hint = V.graph.sizevars.size_hint(arg_sig.expr)
symval_hint = V.graph.sizevars.size_hint(
arg_sig.expr, hint_override=self.hint_override
)
# Force the seed_offset to be 0 so calls to the same kernel
# using different seed offset will have the same benchmark harness.