llama.cpp version b9668 implements UMA host-visible memory buffers for better performance on UMA devices, based on 0cc4m's suggestion. The release includes binaries for macOS, Linux, Android, Windows, and openEuler, supporting CPU, Vulkan, ROCm, OpenVINO, SYCL, and HIP, with a dedicated UI package.
llama.cpp release b9668 adds UMA host-visible memory and cross-platform binaries
from English