llama.cpp
vulkan: ggml src1 precision support
#30656
Merged

vulkan: ggml src1 precision support #30656

0cc4m merged 16 commits into master from 0cc4m/vulkan-ggml-prec-support
0cc4m
0cc4m 0cc4m requested a review from ggerganov ggerganov 35 days ago
0cc4m 0cc4m requested a review 35 days ago
0cc4m 0cc4m changed the title vulkan: ggml prec support vulkan: ggml src1 precision support 35 days ago
github-actions github-actions added testing
github-actions github-actions added Vulkan
github-actions github-actions added ggml
0cc4m 0cc4m force pushed to 57721588 35 days ago
jeffbolznv
jeffbolznv commented on 2026-10-16
jeffbolznv
0cc4m
ggerganov
0cc4m
Green-Sky
ggerganov
0cc4m
ggerganov
0cc4m
jeffbolznv
IMbackK
ggerganov
jeffbolznv
ggerganov
0cc4m
ggerganov
ggerganov
ggerganov
jeffbolznv
jeffbolznv
0cc4m
ggerganov
ggerganov
0cc4m
ggerganov
jeffbolznv
ggerganov
0cc4m
ggerganov
0cc4m vulkan: update accumulator precision selection
ce17c2ca
0cc4m handle fp32 overflows through scaling
5230f70c
0cc4m handle overflows in copy-to-fp16 activations
33395acf
0cc4m add missing cases
ebb477cc
0cc4m simplify code
d7032630
0cc4m add tests
288dd718
0cc4m add ggml prec get helpers, use in Vulkan backend
9a9abc51
0cc4m clean up comments
9adef71e
0cc4m revert comment edit
cec2a992
0cc4m trim tests
622cdd00
0cc4m also handle fp32 src0, fix test scaling src0 values too
89f9fc94
0cc4m don't scale if disabled
9dd57554
0cc4m remove fp16 scaling
c6b27377
0cc4m add check to mul_mat
42826283
0cc4m fix inconsistency
f66f97e7
0cc4m 0cc4m force pushed from a1024c43 to f66f97e7 29 days ago
0cc4m
ggerganov
ggerganov approved these changes on 2026-10-22
jeffbolznv
jeffbolznv approved these changes on 2026-10-22
0cc4m rename to acc_prec
e2d29245
jeffbolznv
jeffbolznv approved these changes on 2026-10-23
0cc4m 0cc4m merged 8e5a58df into master 27 days ago
0cc4m 0cc4m deleted the 0cc4m/vulkan-ggml-prec-support branch 27 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone