LAM/MPI logo

LAM/MPI General User's Mailing List Archives

  |   Home   |   Download   |   Documentation   |   FAQ   |   all just in this list

From: Jeff Squyres (jsquyres_at_[hidden])
Date: 2003-04-14 05:46:45


On Sun, 13 Apr 2003, choy hau yan wrote:

> Thanks for your answer.yes. I have two processor in one node.
>
> If I said there is posible two processors run slower than one processor
> because one become master and just send the data and recv data only and
> another one is just recv data and do calculation,then send back to
> master. No parallel processing happen here.
>
> I assume that the adition time is for sending and recving data. Am I
> right?

167 additional seconds for sending/receiving is a lot of time. This could
be true if you're sending/receiving a *lot* of data, or if your code is
highly synchronized (i.e., doesn't try to be efficient about message
passing implicit synchronization).

Run your code through a profiler and see what functions are taking up a
lot of time.

> p/s:compare with distributed memory, generally, witch one is more
> faster? thanks.

It depends on what you are doing -- there is no good general answer to
this question.

-- 
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/