llama.cpp Release b9642 Improves CUDA and Apple Silicon Support

Core summary

One sentence to understand this update

llama.cpp release b9642 updates CUDA support to only F32/F16 for GGML_OP_REPEAT and details various platform support, including macOS Apple Silicon and Linux.

Impact & opportunity

What this could mean

Developers using llama.cpp will benefit from refined CUDA compatibility and clearer guidance on platform support, potentially improving performance and deployment across devices.