llama.cpp

7f88a1b9 - move it to ggml_backend_reg_i