Local LLM inference server for AMD Radeon RX 9070 XT — llama.cpp launcher, benchmarks, and configs for various models across Vulkan and ROCm backends