llama.cpp
ggml-webgpu: Improve the mat-vec and mat-mat of MUL_MAT_ID
#22464

Merged

ggml-webgpu: Improve the mat-vec and mat-mat of MUL_MAT_ID #22464

reeselevine merged 4 commits into ggml-org:master from yomaytk:mul_mat_id_vec

yomaytk requested a review 50 days ago

yomaytk commented on 2026-04-28

CISC approved these changes on 2026-04-28

github-actions added ggml

github-actions added WebGPU

yomaytk force pushed from 3e8ec276 to 7c5cebea 50 days ago

yomaytk commented on 2026-04-28

yomaytk force pushed from f2a55032 to 5466aa82 49 days ago

yomaytk commented on 2026-04-29

Add mat-vec fast path of MUL_MAT_ID.

590fa483

Add shared accumulation vec logic and the other types supports.

afe0c628

yomaytk force pushed from 5466aa82 to 93995420 48 days ago

Add i-quant mat-mat for MUL_MAT_ID and fix some parts

c0700cdf

yomaytk force pushed from 66e74003 to c0700cdf 48 days ago

yomaytk changed the title ~~ggml-webgpu: Improve the mat-vec performance of MUL_MAT_ID~~ ggml-webgpu: Improve the mat-vec and mat-mat of MUL_MAT_ID 48 days ago

reeselevine commented on 2026-04-30

Remove n_experts from shader_lib_context.

a44c0fb9

reeselevine approved these changes on 2026-04-30

reeselevine merged a95a11e5 into master 48 days ago

yomaytk deleted the mul_mat_id_vec branch 47 days ago

Reviewers

reeselevine

CISC

Assignees

No one assigned

Labels

ggml WebGPU

Milestone

No milestone

llama.cpp ggml-webgpu: Improve the mat-vec and mat-mat of MUL_MAT_ID #22464 Merged

ggml-webgpu: Improve the mat-vec and mat-mat of MUL_MAT_ID #22464

llama.cpp
ggml-webgpu: Improve the mat-vec and mat-mat of MUL_MAT_ID
#22464

Merged