The build system for llama.cpp is now CMake, and the structure
of the output has changed slightly.