Skip to content

Add loops for ATen compiler MHA asm gen to reduce instruction count #45

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
booth-algo merged 1 commit into from
May 18, 2026
+312 −20
Merged

Add loops for ATen compiler MHA asm gen to reduce instruction count #45

Show file tree
Hide file tree
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
6 changes: 3 additions & 3 deletions aten/plena/compiler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,9 +37,9 @@ def __init__(self, mlen: int = 64, blen: int = 4, real_data_ratio: float = 1.125
mlen: Matrix tile size (default 64)
blen: Vector tile size (default 4)
real_data_ratio: HBM storage ratio (MXFP8 format = 1.125)
unroll_loops: If True, unroll sub-projection loops at ASM-gen time to
eliminate C_LOOP_START/END overhead. Overridden by the
ATEN_UNROLL env var ("1"=True, "0"=False).
unroll_loops: If True, unroll sub-projection and attention helper loops
at ASM-gen time to eliminate C_LOOP_START/END overhead.
Overridden by the ATEN_UNROLL env var ("1"=True, "0"=False).
"""
_env_unroll = os.environ.get("ATEN_UNROLL", "")
if _env_unroll == "1":
Expand Down
Loading
Loading