Skip to content

red kernel config heuristics number of args fixed for 2D red kernels #2768

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Oct 30, 2025
Merged

red kernel config heuristics number of args fixed for 2D red kernels #2768

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 16 additions & 16 deletions torch/_inductor/runtime/triton_heuristics.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2586,15 +2586,6 @@ def _persistent_reduction_configs(
)

# defer to more autotuning, initially
if "y" in size_hints:
pass
# TODO(jansel): we should be able to improve these heuristics
if not max_autotune_enabled: # Don't filter if tuning enabled
if reduction_hint == ReductionHint.INNER and rnumel >= 256:
configs = configs[:1]
elif reduction_hint == ReductionHint.OUTER:
configs = configs[-1:]

tiny_configs = [
triton_config_reduction(
size_hints,
Expand All @@ -2603,13 +2594,22 @@ def _persistent_reduction_configs(
)
]

if max_autotune_enabled:
for conf in tiny_configs:
if conf not in configs:
configs.append(conf)
elif reduction_hint == ReductionHint.OUTER_TINY:
configs = tiny_configs

if "y" in size_hints:
pass
# TODO(jansel): we should be able to improve these heuristics
elif not max_autotune_enabled: # Don't filter if tuning enabled
if reduction_hint == ReductionHint.INNER and rnumel >= 256:
configs = configs[:1]
elif reduction_hint == ReductionHint.OUTER:
configs = configs[-1:]
elif reduction_hint == ReductionHint.OUTER_TINY:
configs = tiny_configs
else:
if max_autotune_enabled:
for conf in tiny_configs:
if conf not in configs:
configs.append(conf)

for c in configs:
# we don't need Rn_BLOCK for persistent reduction
for prefix in size_hints:
Expand Down