llama.cpp

vulkan: workaround MoltenVK compile failure in multi_add

#15506

Merged