llama.cpp b9466 patches OpenCL warnings for non-Adreno GPUs

llama.cpp, the popular open-source C/C++ LLM inference engine, released version b9466 on June 2, 2024. This maintenance release focuses on fixing compiler warnings in the OpenCL backend, specifically for non-Adreno GPU paths (as noted in pull request #23922). While seemingly minor, addressing these warnings ensures better code hygiene and reduces the risk of undefined behavior on a wider range of GPUs. The release continues llama.cpp's tradition of broad platform support, generating binaries for 22+ build configurations.

The release includes builds for macOS (both Apple Silicon arm64 and Intel x64), iOS (as an XCFramework), Windows (CPU x64/arm64, CUDA 12.4 & 13.3, Vulkan, HIP), Linux (CPU x64/arm64/s390x, Vulkan, ROCm 7.2, OpenVINO, SYCL), and Android arm64. Notable disabled configurations include OpenEuler and some SYCL builds, but the core platforms remain robust. This fix is particularly relevant for developers and researchers using AMD GPUs (non-Adreno) on OpenCL, who may have previously seen compiler warnings. By keeping the codebase clean, the project maintains its reputation as a reliable, high-performance runtime for running large language models locally.