Skip to content

[CuteDSL] Explicitly cast for Flash Combine #1925

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
tridao merged 1 commit into from
Oct 7, 2025
Merged

[CuteDSL] Explicitly cast for Flash Combine #1925

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
9 changes: 8 additions & 1 deletion flash_attn/cute/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -359,7 +359,14 @@ def elem_pointer_i64(x: cute.Tensor, coord: cute.Coord, *, loc=None, ip=None) ->
flat_stride = cute.flatten_to_tuple(x.stride)
assert len(flat_coord_i64) == len(flat_stride), "Coordinate and stride must have the same length"
offset = sum(c * s for c, s in zip(flat_coord_i64, flat_stride))
return x.iterator + offset
# HACK: we assume that applying the offset does not change the pointer alignment
byte_offset = offset * x.element_type.width // 8
return cute.make_ptr(
x.element_type,
x.iterator.toint() + byte_offset,
x.memspace,
assumed_align=x.iterator.alignment,
)


@cute.jit
Expand Down