llama.cpp
cdc06426 - metal : optimize sparse FA + clean-up (#29377)

Commit
31 days ago
metal : optimize sparse FA + clean-up (#29377) * metal : cache sparse FA indices in shared memory Assisted-by: pi:llama.cpp/DeepSeek-V4-Flash-Vision-Exp * metal : simplify shared memory size calculation Assisted-by: pi:llama.cpp/DeepSeek-V4-Flash-Vision-Exp * pi : update general * metal : unroll sparse index load Assisted-by: pi:llama.cpp/DeepSeek-V4-Flash-Vision-Exp
Author
Parents
Loading