llama.cpp
opencl: enable the general fp mm for non-cont input and as a fallback for specialized kqv kernel for adreno
#18970

Merged

opencl: enable the general fp mm for non-cont input and as a fallback for specialized kqv kernel for adreno #18970

max-krasnyansky merged 5 commits into ggml-org:master from qualcomm:lh/copy-to-cont

opencl: add `copy_to_contiguous` and utilize mm kernels

6d0a567b

opencl: only copy to cont for f32 and f16 tensors

b773905f

opencl: use cont mm for fallback when dst is large

861c9815

opencl: use nb local to copy-to-cont

ca8a5064

opencl: use local offset as well

f04a782c

github-actions added ggml

github-actions added OpenCL

lhez marked this pull request as ready for review 7 days ago

lhez requested a review from

max-krasnyansky 7 days ago

max-krasnyansky approved these changes on 2026-01-22

max-krasnyansky merged 9c96465f into master 6 days ago

Reviewers

max-krasnyansky

Assignees

No one assigned

Labels

ggml OpenCL

Milestone

No milestone