Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!wuarchive!psuvax1!psuvm!odx
From: ODX@PSUVM.BITNET (Tim Larson)
Newsgroups: comp.lang.modula2
Subject: Re: The WITH statement
Message-ID: <89324.084100ODX@PSUVM.BITNET>
Date: 20 Nov 89 13:41:00 GMT
References: <89321.154018ODX@PSUVM.BITNET> <2353@gmu90x.gmu.edu>
Organization: Penn State University
Lines: 25

In article <2353@gmu90x.gmu.edu>, jbaker@gmu90x.gmu.edu (jbaker) says:
>
>I always thought of a WITH statement differently.  It computes the location
>of the record just once, freezes it, and uses that record throughout the
>WITH statement.  Thus, not only does it reduce source code size, it also
>increases program efficiency.
>
This is a good point, it *should* increase efficiency, but it also may not!
A good compiler optimizes code for the programmer, but WITH broke the
optimizer in mine (JPI) and I don't believe it's the compiler's fault.  To
maintain correctness, it had to do the computation for the WITH statement,
then use that computation to reference the record.  When I wrote the same
code and removed the WITH, the code size and execution time was reduced!
The reason is that the compiler was then freed to optimize the record ref.
to registers and reduce the code size.  (Of course, this type of optimization
is specific to PCs.)  The upshot is, the optimization forced on the
programmer by remembering the subtleties of the WITH statement is overshadowed
by the potential of a good optimizing compiler, and the potential cost is
correctness!

It's interesting to note that JPI's manual, when mentioning WITH, mildly
discourages its use.

-Tim Larson
odx@psuvm.bitnet