llama.cpp
opencl: ragged-tile MoE prefill FP16 GEMM optimization (skip padded expert tiles)
#25433
Merged
Merged
opencl: ragged-tile MoE prefill FP16 GEMM optimization (skip padded expert tiles) #25433
max-krasnyansky
merged
3 commits into
ggml-org:master
from
qualcomm:hq/moe-ragged-fp16
opencl: ragged-tile MoE prefill GEMM (skip padded expert tiles)
159ad2c5
opencl: quarter-granularity ragged MoE tile-skip (8-col skip-groups)
f3f59f15
opencl: move ragged moe env var in cl_init
879d0af8
lhez
approved these changes
on 2026-07-08
max-krasnyansky
approved these changes
on 2026-07-08
Login to write a write a comment.
Login via GitHub