llama.cpp

Adding support for the --numa argument for benchmarking.

#7080

Merged