llama.cpp

ggml-cpu: add always_inline to tinyBLAS_PPC accumulator saves

#20791

Merged