Newsgroups: comp.parallel.mpi
From: vgeor@pegasus.unm.edu (Vasilis Georgitsis)
Subject: Implementation question about collective function
Organization: UNM
Date: 24 May 1995 20:08:03 GMT
Message-ID: <3q03n3$sdq@lynx.unm.edu>

Hi there,
I would like to get some information about the implementation of the collective function of MPICH and MPL (if possible).  More specificly

	Barrier
	Broadcast
	Gather
	Scatter
	Reduce

I would like to get information of how those function are implemented, such as the algorithm is used to implement those function.  Is there any close form formula to predict the time for each of those functions?  For example :
  For a PING model the general formula would be :

		t = t_0 + f(n, m)

	where n is the number of processors ( n = 2 for Ping ) and m is the message length.

Is there anything like that for the collective communication functions given above. Any suggestions and/or information would be appriciated. The replys can be addresses at one of the following accounts:

		vasilis@arc.unm.edu
	or
		vgeor@unm.edu

Preferable the first account.

