Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!spool.mu.edu!sdd.hp.com!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!unixhub!slacvm!meb
From: MEB@SLACVM.SLAC.STANFORD.EDU (meb@slacvm.slac.stanford.edu)
Newsgroups: comp.lang.fortran
Subject: Re: Read dimension
Message-ID: <91056.110541MEB@SLACVM.SLAC.STANFORD.EDU>
Date: 25 Feb 91 19:05:41 GMT
References: <61477@eerie.acsu.Buffalo.EDU>
Organization: Stanford Linear Accelerator Center
Lines: 41

In article <61477@eerie.acsu.Buffalo.EDU>, xiaofei@acsu.buffalo.edu (Xiaofei
Wang) says:

(that he wants to use variably dimensioned arrays without declaring
 specific dimensions until run time).

I have little to add to preceding responses, and wouldn't have sent this
this if the originator had not had "buffalo" in his email address.

One of my programming "pleasures" about 20 years ago, when I was working
in the Research Computing Office at Ball State Universiyt, involved the
understanding and maintenance of a rather large program which we called
"Buffalo," but was named something like the Multivariate Analysis of
Variance and Covariance Program Package, distributed by the University
of Buffalo.  It dealt in a quite general fashion with the simultaneous
problems of:

 - no dynamic storage allocation
 - no virtual memory
 - variously sized (memory/partition capacity) target machines
 - various numbers of variously shaped matrices, depending on
     number of cases, number of variables, type of analysis, etc.

The programmer who installed Buffalo was required to provide a single
value before compiling the MAIN (driver) program; that value was the
DIMENSION of a REAL vector (nominally as large as the installation
could accommodate).  The same value also was coded as the first (only?)
value in a COMMON block.

After reading in the necessary problem-specification values, Buffalo
would compute the necessary dimensions, cumulative offsets, etc., and
perform the analyses with one- two-dimensional arrays that were all
allocated within that single vector by calls of the sort:

  (in caller)   CALL SUB(M,N,VEC(IOFF1),VEC(IOFF2),...)

  (in called)   SUBROUTINE SUB(M,N,MAT1,MAT2,...)
                DIMENSION MAT1(M,N),MAT2(N,N)

It wasn't fun to debug, but it didn't need very much.  That program
taught me much of what I know about FORTRAN storage allocation.