Skip to content
reduce amount of CUDA SM compiled