Newsgroups: comp.parallel.mpi
From: lederman@romano.cs.wisc.edu (Steve Huss-Lederman)
Subject: Re: MPI_Gatherv in f77
Organization: CS Department, University of Wisconsin
Date: 27 Aug 1996 14:38:56 GMT
Message-ID: <LEDERMAN.96Aug27093857@romano.cs.wisc.edu>

This type of problem has been brought to the MPI Forum's attention.
The section on extended collective operations (esp. subsection on
extended all-to-all) in the developing MPI-2 standard hopefully
addresses your needs.  The Forum would be interested if you feel it
does not.

Steve

In article <4vehk7$1jfb@moka.ccr.jussieu.fr> stoclet@ccr.jussieu.fr (Patrick STOCLET) writes:

 > Path: uwvax!uwm.edu!chi-news.cic.net!cs.utexas.edu!math.ohio-state.edu!jussieu.fr!stoclet
 > From: stoclet@ccr.jussieu.fr (Patrick STOCLET)
 > Newsgroups: comp.parallel.mpi
 > Date: 21 Aug 1996 10:33:11 +0200
 > Organization: CCR - Universites Paris VI/VII - Paris - France
 > Lines: 18
 > NNTP-Posting-Host: moka.ccr.jussieu.fr
 > Mime-Version: 1.0
 > Content-Type: text/plain; charset=iso-8859-1
 > Content-Transfer-Encoding: 8bit
 > 
 > 
 > hello, how can i manage to do the following :
 > 
 > 2d processors grid (nxp,nyp)
 > T(nx,ny) on each node coordinates   ix,iy  in processor grid
 > send it at  T_BIG((ix-1)*nx+1:ix*nx,(iy-1)*ny+1:iy*ny) on node 0 
 > 
 >  MPI_Gatherv can provide a stride but not the type necessary at node 0 
 > to define an " array section " 
 > 
 > 
 > 
 > how can i do it ??
 > 
 > thanks for hints ( in f77 )
 > 
 > patrick stoclet 
 > stoclet@cncpst.jussieu.fr

