Newsgroups: comp.parallel.mpi
From: guptavi1@cps.msu.edu (Vivek Gupta)
Subject: Re: Clarifications on Ready/Sync modes
Organization: Michigan State University
Date: 12 Aug 1995 16:48:20 GMT
Message-ID: <40im0k$1bum@msunews.cl.msu.edu>

Hi,

> It's not MPI that guarantees the ordering, it's the user that does.
> This is just a way for the user to tell the implementation that he/she
> guarantees the matching receive has been posted before ready send was

Thanks for your response to my queries. It did help me understand this better.

However, do you feel that the ready mode can really result in a better
performance in some case ? Since it is the user's responsibility to guarantee 
posting of matching receive, I feel that it would involve an extra 
send-receive

It might help for long messages or perhaps derived datatypes etc...where
overheads of memory copy etc can be avoided. But can't this also be 
accomplished by a synchronous send (where the protocol guarantees that the 
receive would have been posted before the send completes). In fact, a 
sync-send implementation might use internal protocol messages to achieve the 
same effect as the ready mode !

Do you have some examples where use of a ready mode has been known to produce
better performance. Similarly, I would like to take a look at some examples
for each of the send modes...which result in better performance as compared
to other modes.

Thanks

Vivek

--
****************************************************************************
Dept. of Computer Science, Michigan State University	Ph : (517) 333-4223.
Email : guptavi1@cps.msu.edu    	http://web.cps.msu.edu/~guptavi1/

	  Smile...it makes people wonder what you are thinking...

