NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/comp/lang/fortran/4315

Text File | 1992-11-14 | 1.9 KB | 46 lines

Newsgroups: comp.lang.fortran
Path: sparky!uunet!destroyer!gatech!hubcap!mjs
From: [e-mail] (M. J. Saltzman)
Subject: History question--fortran arrays
Message-ID: <[e-mail]>
Organization: Clemson University, Clemson SC
Date: Fri, 13 Nov 1992 20:53:21 GMT
Lines: 36

On comp.lang.c, another discussion of the relative virtues of Forrtan
vs. C is occurring (relatively civilized so far, compared to the usual
religious wars).  It's not my intention to perpetuate the argument,
but a question came up that might get a more accurate answer here than
the speculation I've seen over there.

Does anyone know the historical justification for having Fortran arrays
stored with the fastest-varying index at the left?

My guess is that this is a "natural" order with respect to some matrix
operations, and that perhaps it has to do with data locality rather
than instruction counts, or else it was something peculiar to the IBM
704.  It's certainly different from the "natural" order dictated by
the C/Pascal/etc. philosophy that a multidimensional array is an array
of arrays (with successive indices appended at the right).

I found a copy of a page from the original FORTRAN manual which contains
the following:

>_Arrangement_of_Arrays_in_Storage_
>
>A 2-dimensioanl array A will, in the object program, be stored
>sequentially in the order A(1,1), A(2,1), ...., A(m,1), A(1,2), ....,
>A(m,2), ...., A(m,n).  Thus it is stored "columnwise", with the first
>of its subscripts varying most rapidly, and the last varying least
>rapidly.  The same is true of 3-dimensional arrays.  1-dimensional
>arrays are of course simply stored sequentially.  All arrays are
>stored backwards in storage; i.e. the above sequence is in the order
>of decreasing absolute location.

But the accompanying article (by Backus) doesn't explain further.

--
        Matthew Saltzman
        Clemson University Math Sciences
        [e-mail]