Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!mcgill-vision!snorkelwacker!bu.edu!rpi!zaphod.mps.ohio-state.edu!wuarchive!uunet!mcsun!hp4nl!sci.kun.nl!cs.kun.nl!eerke
From: eerke@cs.kun.nl (Eerke Boiten)
Newsgroups: comp.theory
Subject: Re: Procedures with multiple out parameters: why not?
Message-ID: <2118@wn1.sci.kun.nl>
Date: 27 Aug 90 08:29:58 GMT
References: <1990Aug15.231838.5664@zaphod.mps.ohio-state.edu> <4625@uqcspe.cs.uq.oz.au>
Sender: root@sci.kun.nl
Followup-To: comp.lang.misc
Organization: University of Nijmegen, The Netherlands
Lines: 36

In article <4625@uqcspe.cs.uq.oz.au> ianh@batserver.cs.uq.oz.au writes:
[]I suspect the theoretical problem that Dijkstra is referring to is
[]aliasing of variables. For example, consider
[example deleted] 
[]Another more practical example is the following matrix
[]multiplication procedure
[]	procedure MatrixMultiply(var A, B, C : matrix);
[]Here the A and B are passed as var parameters even though they are
[]not being modified by the matrix multiplication procedure. However,
[]if we have a call of the form
[]	MatrixMultiply(M,M,M)
[]we will not get the result that we want as A, B and C will all be
[]aliases for the same matrix M. To avoid this problem we need the
[]header to read
[]	procedure MatrixMultiply( A, B : matrix; var C : matrix)
[]this avoids any aliasing problems and will work correctly for the
[]call
[]	MatrixMultiply(M,M,M).

What I'm going to say is pretty obvious if you're a regular Pascal
programmer, but since this is comp.theory ...

In answer to the question that arises: why would anyone declare *all*
arguments as *var* arguments?
most Pascal implementations put *adresses* of var-arguments on the
stack, whereas *values* of non-var arguments are stacked when a
procedure/function call occurs. Thus, when one has *matrix* (read:
large) arguments, one usually makes them var-parameters.
Note that thus Ian Hayes' example shows that no everywhere-correct &
efficient Pascal matrix multiplication procedure can exist. (Or am I
overlooking something? If I did, we might continue on comp.lang.misc)

Eerke Boiten
Department of Informatics (STOP Project), K.U.Nijmegen
Toernooiveld, 6525 AD Nijmegen, The Netherlands
Tel. +31-80-612236.	Email: eerke@cs.kun.nl