MPI's problem

Bug reports, work arounds and fixes

Moderators: arango, robertson

Post Reply
Message
Author
CMD
Posts: 3
Joined: Tue Jun 01, 2004 1:11 pm
Location: Department of Hydraulic and ocean engineering

MPI's problem

#1 Unread post by CMD »

After update the version on August 11, 2005 , i find that the first node's cup use is about 90 % and the other nodes's cpu use is below 10 % and the first node's time use is twice of the other nodes's time use .

Could anyone tell me what's wrong with that ?
DO i need to setup something ?


Thank you

User avatar
arango
Site Admin
Posts: 1349
Joined: Wed Feb 26, 2003 4:41 pm
Location: DMCS, Rutgers University
Contact:

#2 Unread post by arango »

I don't know why are you having such behavior. The file distribute.F was corrected for a possible parallel bug in non-balanced partitions. This correction only affects IO and it is needed. The communications in the workhorse routine mp_exchange were not modified. This is the routine that is called at every timestep.

You didn't specified how many nodes you are using. It seems that you are having an IO problem and this maybe set-up or archicteture dependent.

Post Reply