llama.cpp

vulkan: implement several ops relevant for ggml_opt

#11769

Merged