Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!wuarchive!udel!nigel.ee.udel.edu!mccalpin
From: mccalpin@perelandra.cms.udel.edu (John D. McCalpin)
Newsgroups: comp.lang.fortran
Subject: Re: vectorization question reposed
Message-ID: <MCCALPIN.91Mar30104319@pereland.cms.udel.edu>
Date: 30 Mar 91 15:43:19 GMT
References: <1991Mar30.142903.5225@ariel.unm.edu>
Sender: usenet@ee.udel.edu
Organization: College of Marine Studies, U. Del.
Lines: 54
Nntp-Posting-Host: perelandra.cms.udel.edu
In-reply-to: prentice@triton.unm.edu's message of 30 Mar 91 14:29:03 GMT

>>>>> On 30 Mar 91 14:29:03 GMT, prentice@triton.unm.edu (John Prentice) said:

John> My thanks to those people who responsed to my query about how to
John> collapse a three tier array into one.  Unforunately, none of the
John> responses were quite what I needed (I neglected to specify the
John> constraints: 1) Fortran 77, 2) clean, 3) portable, 4) must
John> calculate i,j,k, and 5) vectorizeable).  The most serious of my
John> ommissions was to neglect to say that I need the i,j,k indices.
John> A solution I came up with that is vectorizeable is: [omitted]

Unless memory is very tight, the following solution is likely to be
*much* faster:

	subroutine foobar
	real a(imax,jmax,kmax)
	real xi(imax,jmax,kmax),xj(imax,jmax,kmax),xk(imax,jmax,kmax)
	logical first
	save xi,xj,xk,first
	data first /.true./

*	initialize once
	if (first) then
	    do k=1,kmax
	        do j=1,jmax
		    do xi=1,imax
		        xi(i,j,k) = float(i)
		        xj(i,j,k) = float(j)
		        xk(i,j,k) = float(k)
		    end do
	        end do
	    end do
	    first = .false.
	endif

*	now do work (presumably many times?)
	do k=1,kmax
	    do j=1,jmax
		do i=1,imax
		    a(i,j,k) = xi(i,j,k) + etc....
		end do
	    end do
	end do
	etc......

John> One other question.  What is the size of the memory banks on the 
John> YMP and hence what strides should you avoid?

The number of banks depends on the machine configuration.  Avoid
powers of 2 in your strides.  2 and 4 are probably OK, but higher
powers of 2 will likely cause performance penalties.
--
John D. McCalpin			mccalpin@perelandra.cms.udel.edu
Assistant Professor			mccalpin@brahms.udel.edu
College of Marine Studies, U. Del.	J.MCCALPIN/OMNET