llama.cpp b10969 Adds Automated Ubuntu CUDA Release Builds
llama.cpp release b10969 updates its GitHub Actions release pipeline to automatically generate pre-built Ubuntu binaries with CUDA support. The automated builds cover CUDA 12.8 and 13.3 configurations for both x64 and arm64 architectures. This change simplifies deployment for Linux users running NVIDIA GPUs by providing ready-to-use CUDA binaries directly in official release assets. It removes the need for users to manually compile llama.cpp from source to get CUDA hardware acceleration on Ubuntu. The release update configures GCC 14 for CUDA arm64 builds and packages runtime dependencies like `cudart` alongside the binaries. Additionally, support for NCCL (NVIDIA Collective Communications Library) was temporarily removed from this pull request pending license confirmation.
## BACKGROUND
llama.cpp is an open-source C/C++ framework designed for high-performance LLM inference across diverse hardware targets. CUDA is NVIDIA's parallel computing platform that accelerates compute-heavy AI workloads on NVIDIA GPUs. Providing pre-compiled release binaries through CI pipelines allows users to run optimized executables without configuring local toolchains.