Mail Archives: opendos/2001/01/25/03:38:49
Hi Bernie!
... Is that why some people report the 16 bit version of DJPEG (not
sure it's exact name) runs *faster* than the 32 bit version??? ;-)
Well, the above is because (IIRC) the 16 bit version outputs to a
different format. ;-)
The point I am making is that it is very hard to make such comparisons,
because you are not comparing "apples with apples" - often there are
functional differences, not to mention different compilers (with different
optimizations). However, as a *general* observation, I do not find that
16 bit applications or drivers run noticeably slower than 32 bit ones.
Indeed, it is my impression that "32 bit" is just an excuse by many
lazy programmers to avoid optimizing their code (NB. I'm not referring
to anyone in particular, here)!
In closing, remember DR-Webspyder? Was it just my imagination,
or did that get significantly slower when it was converted to 32 bit?
Later versions (eg. 2.1b2) were a bit faster, but still they were *not*
any faster than their 16 bit cousin (Arachne ;-) ... ?
Joe.
> -----Original Message-----
> From: Bernie [SMTP:bernie AT mbox302 DOT swipnet DOT se]
> Sent: Thursday, 25 January 2001 17:49
> To: opendos AT delorie DOT com
> Subject: RE: DOS and WIN/98
>
> Joe wrote:
> >Well, firstly I have yet to see ANY 32 bit stuff give much
> >better performance than 16 bit stuff - sure, sometimes there's a
> >little extra performance, but often it's negligible or even negative.
>
> The diffrence for DJPEG was 25-40% (depends on the file used) when I
> recompiled it under DJGPP for Arachne. I wouldn't call that little.
> Other programs might not give the same good improvment but it's often
> better (I disregard small utils here). I haven't done a regular test
> between rar and rar32 but you can notice the diffrence if you (un)compress
> alot of data without timing it. One of these days I'll need to test
> pk(un)zip 2.5 as well with and without DPMI loaded... (And sometime test
> arj vs. arj32).
> //Bernie
- Raw text -