llama.cpp
vulkan : dequant q8_0 KV once in coopmat1
#25494
Merged

vulkan : dequant q8_0 KV once in coopmat1 #25494

Nathanw1014
Nathanw1014 vulkan : dequant q8_0 KV once in coopmat1
484ad9ba
Nathanw1014 Nathanw1014 requested a review 95 days ago
thomas9120
github-actions github-actions added Vulkan
github-actions github-actions added ggml
jeffbolznv
Nathanw1014
jeffbolznv
Nathanw1014
jeffbolznv
Nathanw1014
Nathanw1014 vulkan : fall back instead of aborting when FA scratch exceeds maxSto…
b8058343
jeffbolznv
Nathanw1014
En3Tho
Nathanw1014 vulkan : require KV-cache layout in FA dequant path
d7f6bce0
Nathanw1014 vulkan : skip FA dequant path on coopmat2
643e95e7
Nathanw1014 tests : add contiguously-allocated quant K/V FA tests
c3b3bb98
Nathanw1014 Nathanw1014 requested a review from ggerganov ggerganov 83 days ago
github-actions github-actions added testing
Nathanw1014
RandoTTTest
jeffbolznv
0cc4m
0cc4m commented on 2026-08-06
Nathanw1014 vulkan : trim comments
dd39addb
Nathanw1014
jeffbolznv
jeffbolznv commented on 2026-08-09
Nathanw1014 vulkan : tighten permutation checks for FA path
f9f99bf1
Nathanw1014 vulkan : set prealloc_x_need_sync after the FA dispatch
500c8876
0cc4m
rillomas
0cc4m
rillomas
0cc4m
0cc4m requested changes on 2026-08-15
Nathanw1014 Nathanw1014 requested a review from 0cc4m 0cc4m 57 days ago
Nathanw1014
ghost
0cc4m
ghost
RandoTTTest
ghost
rillomas
rillomas commented on 2026-08-18
Nathanw1014 vulkan : exclude Intel Xe1 from FA dequant path
b890946f
Nathanw1014 Nathanw1014 force pushed from 59443262 to b890946f 54 days ago
0cc4m
0cc4m approved these changes on 2026-08-19
jeffbolznv
jeffbolznv approved these changes on 2026-08-19
0cc4m 0cc4m merged dc72703f into master 54 days ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone