llama.cpp b9571 ships GPU kernel fix for Q4_K quantization

llama.cpp b9571, released by ggml-org, is the latest update to the highly popular open-source LLM inference engine, which has accumulated 116,000 stars and 19,400 forks on GitHub. The core change in this release is a fix to the GPU kernel file mvvq.cu, specifically removing a case for GGML_TYPE_Q4_K. This optimization likely improves the performance and stability of quantized model inference on NVIDIA GPUs, as the mvvq kernel handles matrix-vector multiplication for quantized weights. The commit is signed with GitHub's verified signature, ensuring code integrity.

This release continues llama.cpp's tradition of broad platform support, offering builds for macOS (Apple Silicon with optional KleidiAI, Intel x64, iOS XCFramework), Linux (x64, arm64, s390x with CPU, Vulkan, ROCm 7.2, OpenVINO, SYCL FP32), Android (arm64 CPU), and Windows (x64 and arm64 CPU, CUDA 12.4 and 13.3 DLLs, Vulkan, SYCL, HIP). Some variants are disabled (e.g., macOS Intel KleidiAI, Linux SYCL FP32, Windows SYCL, openEuler). For developers and AI enthusiasts running large language models locally, this update ensures smoother GPU-accelerated inference, especially for those using 4-bit quantized models (Q4_K).