llama.cpp

80d28f10 - HIP: fix AMDGPU_TARGETS, update documentation (#16803)