home *** CD-ROM | disk | FTP | other *** search
- Path: news.eunet.fi!fipnet!kone!jsaarinen
- Newsgroups: comp.sys.amiga.programmer
- X-NewsReader: IntuiNews 1.2b (31.7.94)
- References: <38232459@kone.fipnet.fi> <4gbb5a$104@sunsystem5.informatik.tu-muenchen.de>
- From: "Jyrki Saarinen" <jsaarinen@kone.fipnet.fi>
- Date: Thu, 22 Feb 96 16:00:33 UT
- Comments: Illegal date header - new date added by quicknews
- X-Original-Date: Thu, 22 Feb 96 10:49:49
- MIME-Version: 1.0
- Content-Type: text/plain; charset=iso-8859-1
- Content-Transfer-Encoding: binary
- Distribution: world
- Subject: Re: TMapping again!
- Message-ID: <38232577@kone.fipnet.fi>
-
-
- > well, is it maybe that the lin-interpol-trick works only on triangles ?
- > Then I'd understand the "use triangles" thingy :)
-
- The horizontal adders are constant on triangles when using linear
- interpolation. That is why triangles are good. If flat
- is used, n-sided convex polys are the way to go ofcourse.
-
- > The "descent divide each 16 horiz pix" only works on triangles ?
-
- Perspective mapping is done by interpolating linearly
- u/z, v/z and 1/z over the polygon. Then at each pixel u/z and
- v/z are divided by 1/z to get u,v-coordinates. Now this
- is done at each 16 pixels, but again the reason for
- triangles could be the constant horizontal adders.
-
- On the other hand, with two divides two at each 16 pixels,
- it could be better to use n-sided convex polygons and
- use 3 extra divides per scanline. (u,v,gouraud)
-
- -- _
- a Stellar programmer _ //
- "Amiga - back for the future" \X/
-