]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
opencl: transposed gemm/gemv moe kernel with mxfp4,f32 (#16602)
authorShawn Gu <redacted>
Sat, 18 Oct 2025 00:55:32 +0000 (17:55 -0700)
committerGitHub <redacted>
Sat, 18 Oct 2025 00:55:32 +0000 (17:55 -0700)
commit81387858f1fbcc1acedbd308486e1016618ca8f8
tree64a23444d8cf1d94686cb65fb2b66650c14bf8f6
parent66b0dbcb2d462e7b70ba5a69ee8c3899ac2efb1c
opencl: transposed gemm/gemv moe kernel with mxfp4,f32 (#16602)

* opencl: transposed gemm/gemv moe kernel with mxfp4,f32

* add restore kernel for moe transpose

* fix trailing whitespaces

* resolve compilation warnings
ggml/src/ggml-opencl/CMakeLists.txt
ggml/src/ggml-opencl/ggml-opencl.cpp
ggml/src/ggml-opencl/kernels/cvt.cl
ggml/src/ggml-opencl/kernels/gemm_moe_mxfp4_f32.cl [new file with mode: 0644]
ggml/src/ggml-opencl/kernels/gemv_moe_mxfp4_f32.cl [new file with mode: 0644]