llama.cpp v9283 fixes Apple builds and shared library installs
The popular local LLM runtime updates with cross-platform stability fixes...
llama.cpp, the highly popular open-source library for running large language models locally in C/C++, has released version b9283. This maintenance release, tagged on May 22, addresses critical build and installation issues:
First, it fixes a regression from commit bb28c1f that broke the `cmake --install` step when `BUILD_SHARED_LIBS` is enabled. Without this fix, shared libraries like `libllama-server-impl.so` fail to copy, causing runtime errors such as `error while loading shared libraries`. The fix adds proper `install(TARGETS LIBRARY)` directives for all `-impl` libraries.
Second, the release overhauls CI pipelines to support a wide range of platforms and backends: macOS (Apple Silicon arm64, Intel x64, KleidiAI), iOS (fixing XCFramework builds), Linux (Ubuntu CPU x64/arm64/s390x, Vulkan, ROCm 7.2, OpenVINO, SYCL FP32/FP16), Android (arm64 CPU), Windows (x64 CPU, arm64 CPU, CUDA 12/13, Vulkan, SYCL, HIP), and openEuler (x86/aarch64 with ACL Graph).
This release ensures llama.cpp remains stable for the thousands of developers and hobbyists who depend on it for local LLM inference. The shared library fix alone prevents silent failures in custom builds, while the expanded CI guarantees compatibility with new hardware backends like ROCm 7.2 and OpenVINO.
- Fixes cmake install for shared libraries (prevents runtime file-not-found errors)
- Resolves Apple builds: macOS (Intel, Apple Silicon, KleidiAI) and iOS XCFramework
- CI now covers 25+ platform/backend combos including CUDA 13, ROCm 7.2, and openEuler
Why It Matters
Essential patch for llama.cpp users ensures stable local LLM deployments across all major hardware backends.