Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!udel!haven!uvaarpa!murdoch!astsun7.astro.Virginia.EDU!gl8f
From: gl8f@astsun7.astro.Virginia.EDU (Greg Lindahl)
Newsgroups: comp.lang.fortran
Subject: Re: Fortran vs. C for numerical work (SUMMARY)
Message-ID: <1990Nov30.163222.1376@murdoch.acc.Virginia.EDU>
Date: 30 Nov 90 16:32:22 GMT
References: <17680:Nov2806:04:1090@kramden.acf.nyu.edu> <109378@convex.convex.com> <10444:Nov3008:15:4590@kramden.acf.nyu.edu>
Sender: news@murdoch.acc.Virginia.EDU
Organization: Department of Astronomy, University of Virginia
Lines: 16

In article <10444:Nov3008:15:4590@kramden.acf.nyu.edu> brnstnd@kramden.acf.nyu.edu (Dan Bernstein) writes:

>Now instead of adding (i * N * P + k) * sizeof(double) to the base of a,
>the compiler will add i * sizeof(ptr) to the base of b, access that
>value, and add k * sizeof(double) to that.

However, the first expression is a constant. So you only have to do
the multiplication once. One multiplication is often faster than N
memory references.

>But in Fortran you can't do this optimization, because you can't store
>that array of pointers.

The *compiler* isn't can. The compiler could build an array of
pointers and use it inside loops which do "random" accesses. I
suggested this yesterday, I guess you missed it ;-)