cuda: fuse MoE expert weighting and reduction (+3% to +9% prefill, MoE models only) - #26160

cuda: fuse MoE expert weighting and reduction (+3% to +9% prefill, MoE models only)#26160

coder543 wants to merge 1 commit into

ggml-org:masterggml-org/llama.cpp:masterfrom

coder543:perf/moe-weighted-reductioncoder543/llama.cpp:perf/moe-weighted-reductionCopy head branch name to clipboard