Hengfeng Li Hengfeng Li -4 years ago 127
C Question

A problem about using MPI_Send

I'm learning MPI_Send, but I'm confused about this method. I wrote a simple pingpong program which the rank-0 node send the message to rank-1 node, and then the latter one returns a message to the former one.

if (rank == 0) { /* Send Ping, Receive Pong */
dest = 2;
source = 2;
rc = MPI_Send(pingmsg, strlen(pingmsg)+1, MPI_CHAR, dest, tag, MPI_COMM_WORLD);
rc = MPI_Recv(buff, strlen(pongmsg)+1, MPI_CHAR, source, tag, MPI_COMM_WORLD, &Stat);
printf("Rank0 Sent: %s & Received: %s\n", pingmsg, buff);
else if (rank == 2) { /* Receive Ping, Send Pong */
dest = 0;
source = 0;
rc = MPI_Recv(buff, strlen(pingmsg)+1, MPI_CHAR, source, tag, MPI_COMM_WORLD, &Stat);
printf("Rank1 received: %s & Sending: %s\n", buff, pongmsg);
rc = MPI_Send(pongmsg, strlen(pongmsg)+1, MPI_CHAR, dest, tag, MPI_COMM_WORLD);

I run this program on a 3 nodes environment. However, the system displays:

Fatal error in MPI_Send: Other MPI error, error stack:
MPI_Send(173)..............: MPI_Send(buf=0xbffffb90, count=10, MPI_CHAR, dest=2, tag=1, MPI_COMM_WORLD) failed
MPID_nem_tcp_connpoll(1811): Communication error with rank 2: Unknown error 4294967295

I'm wondering why I can send a message from rank-0 node to rank-1 node, but an error occurs when changed from rank-0 node to rank-1 node? Thanks.

Answer Source

Actually have you checked whether strlen(pingmsg) is the same in both MPI_SEND and MPI_RECV

The amount of data sent using MPI_SEND should be less than or equal to the amount of data to be received by MPI_RECV or else it will lead to an error.

Recommended from our users: Dynamic Network Monitoring from WhatsUp Gold from IPSwitch. Free Download