llama.cpp

ggml : use a simple std::thread in AMX without OpenMP

#20074

Merged