]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
opencl: add q5_0 and q5_1 MoE for Adreno (llama/22985)
authorshaofeiqi <redacted>
Wed, 13 May 2026 18:57:31 +0000 (11:57 -0700)
committerGeorgi Gerganov <redacted>
Thu, 14 May 2026 18:26:48 +0000 (21:26 +0300)
commitd4a4d87f0ef6511c1e5fec36a3f84d6710f83c33
tree6a37ebf73a88a24938d314c7431ffb2868f3f29c
parentb19beb6027f45ae5a413554de05ff6896b3a5540
opencl: add q5_0 and q5_1 MoE for Adreno (llama/22985)

* opencl: add q5_0 moe support

* opencl: add q5_1 moe support

* opencl: avoid potential leak

* opencl: suppress unused var warning when building for non-Adreno

---------

Co-authored-by: Li He <redacted>
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_q5_0_f32_ns.cl [new file with mode: 0644]
ggml/src/ggml-opencl/kernels/gemm_moe_q5_1_f32_ns.cl [new file with mode: 0644]
ggml/src/ggml-opencl/kernels/gemv_moe_q5_0_f32_ns.cl [new file with mode: 0644]
ggml/src/ggml-opencl/kernels/gemv_moe_q5_1_f32_ns.cl [new file with mode: 0644]