Path: utzoo!mnetor!uunet!husc6!bbn!rochester!PT.CS.CMU.EDU!IUS1.CS.CMU.EDU!edw
From: edw@IUS1.CS.CMU.EDU (Eddie Wyatt)
Newsgroups: comp.lang.c
Subject: Re: Conformant Arrays in C
Message-ID: <929@PT.CS.CMU.EDU>
Date: 22 Feb 88 17:03:40 GMT
References: <42529@sun.uucp> <7297@brl-smoke.ARPA> <676@cresswell.quintus.UUCP> <7314@brl-smoke.ARPA>
Sender: netnews@PT.CS.CMU.EDU
Organization: Carnegie-Mellon University, CS/RI
Lines: 28
Keywords: ANSI C Fortran

> >So why is this solution insufficient?
> 
> Because it's a pain in the lower part of the anatomy!
> Or, in other words, just because it is theoretically sufficient
> does not mean that it is convenient.
> This is one of the limitations of C that Fortran programmers
> have difficulty with; they're used to being able to write
> reasonable array manipulation routines.

   How about matrix assignment (a = b).  If you have matrices
contiguously in memory you can do a "single" bcopy to move the data.
The optional representation doesn't allow this.

  Matrix addition  (a = b + c).  while (--size >= 0) *a++ = *b++ + *c++;

    Questionable whether this is actually faster than a[i][j] method
on the non-contiguous matrix.

   A pain that I'm faced with is transferring data between machines.
If the machines are of the same type, I do some heavy optimizations
on data transfer.  In the case of the matrices, I write the whole
matrix at once into a message.  If I where to use the optional 
representation I couldn't perform this optimization.


-- 

Eddie Wyatt 				e-mail: edw@ius1.cs.cmu.edu