llama.cpp

vulkan: Use VK_EXT_shader_64bit_indexing to handle large mat_mul(_id)

#18678

Merged