llama.cpp

f24ed14e - make : pass CPPFLAGS directly to nvcc, not via -Xcompiler (#5598)