From: Tom St Denis <stdenis AT compmore DOT net>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: (slightly OT) DJGPP Allegro is faster then MingW32 Allegro?
Date: Mon, 29 Jan 2001 18:34:39 GMT
Organization: Deja.com
Lines: 41
Message-ID: <954d3v$772$1@nnrp1.deja.com>
References: <951vr9$9bf$1 AT nnrp1 DOT deja DOT com> <ac2a7t4dqt6nljbuos6iukfll85k32d7sr AT 4ax DOT com> <953o1i$k6g$1 AT nnrp1 DOT deja DOT com> <953s8i$nsc$1 AT nets3 DOT rz DOT RWTH-Aachen DOT DE>
NNTP-Posting-Host: 207.61.93.164
X-Article-Creation-Date: Mon Jan 29 18:34:39 2001 GMT
X-Http-User-Agent: Mozilla/4.75 [en] (X11; U; Linux 2.2.14-5.0 i686)
X-Http-Proxy: 1.0 x68.deja.com:80 (Squid/1.1.22) for client 207.61.93.164
X-MyDeja-Info: XMYDJUIDtomstdenis
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

In article <953s8i$nsc$1 AT nets3 DOT rz DOT RWTH-Aachen DOT DE>,
  Hans-Bernhard Broeker <broeker AT physik DOT rwth-aachen DOT de> wrote:
> [Could you all please reduce quoted content to the part you're
> actually referring to? Thanks.]
>
> Tom St Denis <stdenis AT compmore DOT net> wrote:
> >> On Sun, 28 Jan 2001 20:35:53 GMT, Tom St Denis
<stdenis AT compmore DOT net>
> >> wrote:
> [...]
> >> >I get 850,000 hlines per sec with DJGPP and 400,000 with MingW.
> >> >
> >> >Any clues to why it's slower?
>
> > Well it gets wierder.  My 3d library+test runs at the same fps in
djgpp as
> > ming.  It's only tests/test.exe that differs.
>
> That latter one is not a surprise at all, IMHO. Unless yours is an
> extremely strange 3D library, it's almost certainly going to be
> limited by other factors than execution speed of single API calls. Or
> if it's limited by API call processing, then the API call it
> bottlenecks is hardly ever going to be hline(). 3D textured and
> lighted polygons exercise other functions much more heavily.
>
> Actually, I'd be surprised if a 3D library ever executed a single call
> of the hline() method. Horizontal lines simply don't happen often
> enough, in 3D, to make it worthwile checking for this special case and
> calling the specialized hline() for them.

Um with POLYTYPE_FLAT I thought it always uses hline to draw scanlines,
that would make sense since they are fast to draw.

At anyrate I think it may be either a bug in test.exe or like they said
API overhead.

Tom


Sent via Deja.com
http://www.deja.com/