Since these primitives copy data directly from process to process without queuing, message delivery performance approaches the memory bandwidth of the underlying hardware.

• 因为这些进程间的原语级数据复制不须排队,消息递送的性能接近底层硬件的内存带宽。