llama.cpp

eec1e33a - vulkan: allow graph_optimize for prompt processing workloads (#17475)