Xref: news-dnh.mv.net comp.os.msdos.djgpp:1319
Newsgroups: comp.os.msdos.djgpp
Path: news-dnh.mv.net!mv!world!aml
From: aml AT world DOT std DOT com (Andrew M. Langmead)
Subject: Re: ** Comparison between DJGPP V2 & WATCOM C V10 **
Organization: The World @ Software Tool & Die
References: <3vl892$4q AT news DOT irisa DOT fr> <3vlbk8$58b AT odin DOT diku DOT dk> <3vmbge$ekf AT sun001 DOT spd DOT dsccc DOT com>
Date: Wed, 2 Aug 1995 13:34:00 GMT
Lines: 34
To: djgpp AT sun DOT soe DOT clarkson DOT edu
Dj-Gateway: from newsgroup comp.os.msdos.djgpp

jmccarty AT spd DOT dsccc DOT com (Mike McCarty) writes:

>This is about the 2nd most ridiculous post I've seen on such subjects.
>This fellow has hard facts to support a claim. You present nothing but
>arrogance. Present some numbers to back up your claim that with
>"applications" (whatever those are) the code is significantly better than
>other products or at least comparable, and so the case is different from
>what happens with "hello-worlds" (whatever those are), or just shut up
>and go away.

To some extent you are right. The people who were refuting Erwann
Corvellec didn't give reasons why they thought his concerns were
unwarranted. What it comes down to is that the startup code and and
the pieces of the library that got linked in is a fixed size. A bigger
program doesn't get bigger startup code, and printf() doesn't get
bigger the more times it is called. If a 7 line program is almost 25%
bigger, a 30000 line program will not necessarily be 25% bigger. There
aren't many ways a compiler can generate a call to printf(), so the
code generation for function main() can't be that much different
between the two compilers. A simple hello-world program (named because
the book "The C programming Language" written by Kernigan and Ritchie
use a small program that prints "Hello World" as the first example in
their book.) does not make a good test of code generation of a
compiler (something that is important because the bigger a program is,
the more code the compiler generates.) but only shows about the
smallest a program can be. I don't worry about the size of the
smallest programs I write (I assume that the machine that I'm running
on will have the resources to run the smallest programs I write.) just
the bigger ones.

Now if someone compiled a 30000 line program and said that GCC was
almost 25% bigger, then I'd be worried.
-- 
Andrew Langmead