Skip to content

[SYCL] Support bf16 and quantized type of MUL_MAT#20803

Merged
arthw merged 1 commit intoggml-org:masterfrom
arthw:fix_mul_mat_bf16
Mar 22, 2026
Merged

[SYCL] Support bf16 and quantized type of MUL_MAT#20803
arthw merged 1 commit intoggml-org:masterfrom
arthw:fix_mul_mat_bf16

Conversation

@arthw
Copy link
Copy Markdown
Contributor

@arthw arthw commented Mar 20, 2026

Fix issue: #20713

Enable BF16 and quantized type of MUL_MAT.
All related UT cases are passed on Arc770 and UHD Graphics 770 of 13700K.

@arthw arthw requested a review from a team as a code owner March 20, 2026 14:51
@github-actions github-actions Bot added ggml changes relating to the ggml tensor library for machine learning SYCL https://en.wikipedia.org/wiki/SYCL - GPU programming language labels Mar 20, 2026
@arthw
Copy link
Copy Markdown
Contributor Author

arthw commented Mar 20, 2026

@ggerganov Please help review!

@0cc4m
Copy link
Copy Markdown
Contributor

0cc4m commented Mar 20, 2026

So the actual support was already there and just deactivated? Why?

@arthw
Copy link
Copy Markdown
Contributor Author

arthw commented Mar 21, 2026

So the actual support was already there and just deactivated? Why?

Some broken user cases are fixed, but there is no update for the correct case, I guess.

I will fix the left unsupported cases later.

@arthw arthw merged commit f40a80b into ggml-org:master Mar 22, 2026
85 of 91 checks passed
Seunghhon pushed a commit to Seunghhon/llama.cpp that referenced this pull request Apr 26, 2026
rsenthilkumar6 pushed a commit to rsenthilkumar6/llama.cpp that referenced this pull request May 1, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ggml changes relating to the ggml tensor library for machine learning SYCL https://en.wikipedia.org/wiki/SYCL - GPU programming language

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants