Skip to content

[SYCL] Optimize Q4_0 mul_mat for Arc770, add scripts#22291

Merged
ggerganov merged 8 commits intoggml-org:masterfrom
arthw:opt_arc770_q4_0
Apr 25, 2026
Merged

[SYCL] Optimize Q4_0 mul_mat for Arc770, add scripts#22291
ggerganov merged 8 commits intoggml-org:masterfrom
arthw:opt_arc770_q4_0

Conversation

@arthw
Copy link
Copy Markdown
Contributor

@arthw arthw commented Apr 23, 2026

  1. Optimize Q4_0 mul_mat for Arc770:
LLM Base t/s PR t/s
llama-2-7b.Q4_0.gguf 49 65.24
Qwen3.5-4B-Q4_0.gguf 39.49 45.43

This method only works for Arc 770/750 on Q4_0.
It can't help for other GPU or data types.

  1. Add sycl_hw class to identify the Intel GPU type, arch info. It can help optimize for special Intel GPU model.

  2. Add test scripts for linux and windows to make the test easily.

  3. Update the SYCL.md.

@arthw arthw requested a review from a team as a code owner April 23, 2026 13:34
@arthw arthw requested a review from ggerganov April 23, 2026 13:34
@github-actions github-actions Bot added documentation Improvements or additions to documentation examples ggml changes relating to the ggml tensor library for machine learning SYCL https://en.wikipedia.org/wiki/SYCL - GPU programming language labels Apr 23, 2026
@arthw arthw changed the title [YSCL] Optimize Q4_0 mul_mat for Arc770, add scripts [SYCL] Optimize Q4_0 mul_mat for Arc770, add scripts Apr 24, 2026
Comment thread examples/sycl/win-test.bat Outdated
set "ZES_ENABLE_SYSMAN=1"
%BIN_FILE% -m "%MODEL_FILE%" -no-cnv -p "%INPUT_PROMPT%" -n 200 -e -ngl %NGL% -s %SEED% -c %CONTEXT% %GPUS_SETTING% -lv %LOG_VERBOSE% --mmap

endlocal No newline at end of file
Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

Probably needs to add a newline here for EditorConfig test to pass.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Yes, fix it!

Comment thread examples/sycl/win-start-svr.bat Outdated
set "ZES_ENABLE_SYSMAN=1"
%BIN_FILE% -m "%MODEL_FILE%" -ngl %NGL% -s %SEED% -c %CONTEXT% %GPUS_SETTING% -lv %LOG_VERBOSE% --mmap --host 0.0.0.0 --port 8000

endlocal No newline at end of file
Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

Same here.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Yes, fix it!

Copy link
Copy Markdown
Collaborator

@qnixsynapse qnixsynapse left a comment

Choose a reason for hiding this comment

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

LGTM

@arthw arthw added the merge ready A maintainer can use this label to indicate that they consider the changes final and ready to merge. label Apr 25, 2026
@ggerganov ggerganov merged commit eddd7a1 into ggml-org:master Apr 25, 2026
88 of 93 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

documentation Improvements or additions to documentation examples ggml changes relating to the ggml tensor library for machine learning merge ready A maintainer can use this label to indicate that they consider the changes final and ready to merge. 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