Xref: news-dnh.mv.net comp.os.msdos.djgpp:1375
Path: news-dnh.mv.net!mv!news.sprintlink.net!sunic!sunic.sunet.se!news.lth.se!news.lu.se!ronneby.hk-r.se!news
From: pt93mg AT pt DOT hk-r DOT se (Mats Grahm)
Newsgroups: comp.os.msdos.djgpp
Subject: Re: ** Comparison between DJGPP V2 & WATCOM
Date: 6 Aug 1995 16:31:05 GMT
Organization: College University of Karlskrona-Ronneby
Lines: 41
Distribution: world
References: <3vtbu4$g84 AT news DOT irisa DOT fr>
Reply-To: pt93mg AT pt DOT hk-r DOT se
Nntp-Posting-Host: thebe.pt.hk-r.se
To: djgpp AT sun DOT soe DOT clarkson DOT edu
Dj-Gateway: from newsgroup comp.os.msdos.djgpp

In article g84 AT news DOT irisa DOT fr, ecorvell AT irisa DOT fr (Erwann Corvellec) writes:

>    Well, I didn't intended to be arrogant at all !
>    As I said in my post I was disappointed...

[snip]

>    22 bytes... ;)
>    Maybe you don't know what size optimization means ???
> 
>    Erwann Corvellec.
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>  Student & coder on PC (ASM, C, C++) & fond of LINUX V1.4 ;-D
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Erwann, read what you wrote again. You *are* arrogant, so don't be so surprised
if people get pissed.

And about compilers and their code: When I first started to use gcc on a unix
system, I too was surprised how big some small test program could be, especially
when using C++ iostream instead of printf.

But when I measured speed instead of size I got deeply impressed. Gcc code was 
always much faster than all other compilers I tried. And as others have pointed
out already, the size overhead you have found is constant, and will mean nothing
in a real program (as opposed to hello world ones).

I regard myself as quite good at code optimization and I can tell you that in
9 out of 10 times, speed is the goal, not space. And since I started to use gcc,
I have not have to resort to assembler once, not even in the Doom-like graphical
engine I currently work on. Looking at the assembler output, and maybe re-arrange
things slightly to help the compiler making better optimization choices is almost
always enough to make the compiler eventually come up with code as good as if I
had hand-assembled it.

I have used djgpp only a few weeks now, so I'm far from an expert. But I have no
reason to beleive that djgpp have very different characteristics than unix gcc. 
And that means excellent.

Mats