]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
opencl: Support broadcast for Adreno MUL_MAT and honor `view_offs` for Adreno Q8_0...
authorHongqiang Wang <redacted>
Mon, 20 Jul 2026 05:48:57 +0000 (22:48 -0700)
committerGitHub <redacted>
Mon, 20 Jul 2026 05:48:57 +0000 (22:48 -0700)
commit178a6c44937154dc4c4eff0d166f4a044c4fceba
tree66f71e3eb90f7166f41c3014de3c671a62c805b2
parent571d0d540df04f25298d0e159e520d9fc62ed121
opencl: Support broadcast for Adreno MUL_MAT and honor `view_offs` for Adreno Q8_0 MUL_MAT for llama-server multi-stream (#25910)

* opencl: handle broadcast for adreno gemm/gemv_noshuffle

* opencl: honor view_offs for adreno noshuffle gemm/gemv

* opencl: general GEMM/GEMV support broadcast

* opencl: remove unnecessary tests

* opencl: remove unnecessary comments

---------

Co-authored-by: Li He <redacted>
ggml/src/ggml-opencl/ggml-opencl.cpp