llama.cpp

vulkan: handle quantize_q8_1 overflowing the max workgroup count

#18515

Merged