llama.cpp Release b11349 Adds Vulkan Pipeline Compilation Logging
llama.cpp released build b11349, a minor release introducing enhanced error logging for Vulkan pipeline compilation issues via PR #29794. Updated pre-built binaries have been published across macOS, Linux, Windows, and Android platforms. While a minor patch, improved diagnostic logging helps developers and users troubleshoot shader compilation failures when running large language models via the Vulkan backend across diverse GPU architectures. The patch specifically adds logging to capture failures during Vulkan pipeline creation. Additionally, pre-built macOS binaries with Arm KleidiAI optimizations remain temporarily disabled in this release.
## BACKGROUND
llama.cpp is an open-source C/C++ framework designed for running Large Language Models (LLMs) efficiently on local hardware. Vulkan is a cross-platform 3D graphics and compute API that allows llama.cpp to offload LLM tensor computations to a wide variety of GPUs beyond NVIDIA's CUDA ecosystem.