<div>Agree. As of this moment, I am looking for some way to squeeze the last drop of perofrmance for my application. I do not use Isend or Irecv, and that actually slow down my application in some situation.</div> <div> </div> <div>tan</div> <div><BR><BR><B><I>Jean-Marc Saffroy <saffroy@gmail.com></I></B> wrote:</div> <BLOCKQUOTE class=replbq style="PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #1010ff 2px solid">On Tue, 9 Oct 2007, chong tan wrote:<BR><BR>> having both sender and reciever being the same shared memory may <BR>> require Barriers to be placed to block writing to the mem before it is <BR>> completed consumed. That may actually slow done the run.<BR><BR>MPI_Isend could still benefit from this approach, eg:<BR>- process A does Isend (or Irecv), which simply marks the message to be <BR>sent (or received), then continues with useful computation<BR>- then process B does Irecv (or Isend), which detects the mark and does
<BR>the actual (and single) message copy, then marks the message as sent<BR>- then processes A and B eventually do a Wait or Test<BR><BR>It will probably depend on how much the application allows communications <BR>to overlap with computation (but if it uses Isend/Irecv then it probably <BR>cares already?).<BR><BR><BR>-- <BR>saffroy@gmail.com<BR></BLOCKQUOTE><BR><p> 
<hr size=1>Need a vacation? <a href="http://us.rd.yahoo.com/evt=48256/*http://travel.yahoo.com/;_ylc=X3oDMTFhN2hucjlpBF9TAzk3NDA3NTg5BHBvcwM1BHNlYwNncm91cHMEc2xrA2VtYWlsLW5jbQ--">Get great deals
to amazing places </a>on Yahoo! Travel.