Skip to content

metal : try to unify mul_mv_id kernels#6556

Merged
slaren merged 1 commit intomasterfrom
sl/metal-mvid-tpl
Apr 12, 2024
Merged

metal : try to unify mul_mv_id kernels#6556
slaren merged 1 commit intomasterfrom
sl/metal-mvid-tpl

Conversation

@slaren
Copy link
Copy Markdown
Member

@slaren slaren commented Apr 8, 2024

Prerequisite to #6505

@slaren slaren force-pushed the sl/metal-mvid-tpl branch from a76a3bb to d40b626 Compare April 12, 2024 15:41
@slaren slaren force-pushed the sl/metal-mvid-tpl branch from d40b626 to 39d4427 Compare April 12, 2024 15:41
@slaren slaren marked this pull request as ready for review April 12, 2024 15:42
@slaren slaren requested a review from ggerganov April 12, 2024 15:42
@slaren
Copy link
Copy Markdown
Member Author

slaren commented Apr 12, 2024

@ggerganov these changes add a dummy shared_values parameter to the all the mul_mv kernels, regardless of if they use shared memory, to reduce the number of variations.

Copy link
Copy Markdown
Member

@ggerganov ggerganov left a comment

Choose a reason for hiding this comment

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

Thank you

@slaren
Copy link
Copy Markdown
Member Author

slaren commented Apr 12, 2024

It seems that the ggml-100-m1 runner is not working at the moment. It is also missing from the master runs.

@slaren slaren merged commit fbbc030 into master Apr 12, 2024
@slaren slaren deleted the sl/metal-mvid-tpl branch April 12, 2024 16:13
@ggerganov
Copy link
Copy Markdown
Member

Yeah, the machine is down for some reason - will resolve next week

@ggerganov
Copy link
Copy Markdown
Member

ggml-100-m1 is up again

Seunghhon pushed a commit to Seunghhon/llama.cpp that referenced this pull request Apr 26, 2026
phuongncn pushed a commit to phuongncn/llama.cpp-gx10-dgx-sparks-deepseekv4 that referenced this pull request Apr 28, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants