Skip to content
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

reduce for kqmax_new_j is unnecessary #1032

Merged
merged 2 commits into from
Dec 3, 2024
Merged

reduce for kqmax_new_j is unnecessary #1032

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a6792db
kqmax_new_j in every thread within warp is same after operate at line…
Nov 29, 2024
37a5fec
same problem in vec32
Dec 2, 2024
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
3 changes: 2 additions & 1 deletion src/ggml-cuda/fattn-vec-f16.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -220,7 +220,8 @@ static __global__ void flash_attn_vec_ext_f16(
for (int j = 0; j < ncols; ++j) {
half kqmax_new_j = ncols == 1 ? kqmax_new : kqmax_new_arr[j];

kqmax_new_j = warp_reduce_max(kqmax_new_j);
/* kqmax_new_j in every thread within warp is same after operate at line 199,this reduce can be omit */
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Strictly speaking it is not exactly the same since the order in which you sum up floating point values makes a difference for the result but this is negligible. And there is no reason why the maximum value in particular should be preferred.

//kqmax_new_j = warp_reduce_max(kqmax_new_j);
if (threadIdx.x == 0) {
kqmax_shared[j][threadIdx.y] = kqmax_new_j;
}
Expand Down