Kubernetes之AllReduce任务调度 (CPU+GPU环境) --【D-9/14】Open MPI集群之MPI_Send 和MPI_Recv函数测试

Abstract

The Open MPI cluster is ready. This blog will make use of the MPI_Send() and MPI_Recv() functions in the Open MPI cluster.

1. Blocking vs Non-Blocking Communication

Open MPI supports two types of communication: Blocking and Non-Blocking communication. Specifically, blocking communication has two operations: blocking send and blocking receive operations. Non-Blocking communication has two operations: non-blocking send and non-blocking receive operations. A difference [1] between the blocking send and non-block send operations is that a blocking send operation completes when the message is received by the destination. While a non-blocking send operation completes when the message is sent by the sender.

In Open MPI, blocking communication has primitive implementations [2]: MPI_Send() and MPI_Recv() . Non-bl