llama.cpp Releases Version b10516 with Vulkan Null Pointer Fix
llama.cpp has released version b10516, which introduces a minor patch to prevent null pointer issues. Specifically, it adds null checks during the cleanup of Vulkan queue command pools. Although it is a minor patch release, this fix improves the stability of llama.cpp when running on hardware that utilizes the Vulkan API for acceleration. It prevents potential crashes during memory cleanup routines. The update addresses pull request #27353, adding guards against null queue pointers in the `ggml_vk_queue_command_pools_cleanup` function. The release also includes pre-built binaries for various platforms, though some builds like macOS KleidiAI and Ubuntu ROCm are currently disabled.
## BACKGROUND
llama.cpp is an open-source project designed to run LLMs with high performance on consumer hardware using ggml, a tensor library developed by Georgi Gerganov. Vulkan is a cross-platform graphics and compute API that allows llama.cpp to leverage GPU acceleration on a wide range of hardware.