llama.cpp
vulkan: fix wrong results when a mul_mat reads a slice of a larger cache
#28956

Merged

vulkan: fix wrong results when a mul_mat reads a slice of a larger cache #28956

0cc4m merged 6 commits into ggml-org:master from ServeurpersoCom:vulkan-batch-stride

ServeurpersoCom requested a review from

ggerganov 49 days ago

ServeurpersoCom requested a review 49 days ago

github-actions added testing

github-actions added Vulkan

github-actions added ggml

jeffbolznv commented on 2026-09-15

vulkan: read the batch stride of an in place src0 from nb[2]

6dd49d05

vulkan: size the in place A and B ranges by their strided extent

6ffa3c25

ServeurpersoCom force pushed to 6ffa3c25 47 days ago

vulkan: address review from jeffbolznv

d83afe48

vulkan: read the batch stride of an in place src0 in mul_mat_vec_id

97317325

vulkan: address review from jeffbolznv

55ef1236

jeffbolznv approved these changes on 2026-09-18

vulkan: keep the A and B ranges exact in mul_mm

e99e2f48

ServeurpersoCom requested a review from

jeffbolznv 37 days ago

0cc4m approved these changes on 2026-09-28

0cc4m merged 52624716 into master 36 days ago

Reviewers

0cc4m

jeffbolznv

ggerganov

Assignees

No one assigned

Labels

testing Vulkan ggml

Milestone

No milestone