Path: chuka.playstation.co.uk!news
From: Andreas Schrattenecker <andreas.schrattenecker@jk.uni-linz.ac.at>
Newsgroups: scee.yaroze.programming.3d_graphics
Subject: HELP:small planar quads
Date: Wed, 14 Jan 1998 15:58:35 +0100
Organization: PlayStation Net Yaroze (SCEE)
Lines: 63
Message-ID: <34BCD29B.5BEA@jk.uni-linz.ac.at>
Reply-To: andreas.schrattenecker@jk.uni-linz.ac.at
NNTP-Posting-Host: ried-remote07.ivnet.co.at
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Mailer: Mozilla 2.02 [de] (OS/2; I)

I have following BIG problem:

Aktualy my scene only contains quads, for performance
reasons. I worked really hard to find a technique for
non-planar(vertices dont lie on same plane) quads, without missing
polygons (quads normally have to be planar).
And it works!, but:

there are missing planar-quads, which _have_to_ work.
What i have found out is:

1) This happens if quads get small 
	a: they are far away
	b: the angle between view-vector and quads gets small

2) No GPU-error -> for missing quads no packets are created.

3) It seems not to be a "to-screen-transformation-error" because
   if increasing projection distance->quads gets bigger->problem
   disappears. Perspective conversion is after transformation ->
   no transformation error.

4) if 1) there are quads displayed although they should be hidden
   (also if negative angle between viewvector and quad gets really big)
   in the same areas where quads disappear (when i change angle) ????

5) useing frontfaced and backfaced quads -> also errors (would anyway
not be 
   a solution-> to slow)

6) triangles are displayed correct (no solution -> to slow, and i 
   would have wasted lots of time for looking on quads).

7) happens with all types of quads: textured,non textured, gouraud,
   gradation (i am not sure for flat-quads now)

8) errors change when define vertices-order different -> but no way
   to define without errors.

9) The only solution i have found is to pull(in realtime) vertices
   in a direction that angle gets bigger or quad gets bigger.
   Possibly missing quads are simply overpainten -> but needs also
   much time.


I really like the playstation-architecture (also GsLib) a lot.
And it is possible to write quality-games also with GsLib which can
compete with n64-games. 
I have now implemented an improved uniform mipmapping, which is
also faster than non-mipmaping (smaller textures when polys gets
smaller)
and my polygon-count is ca. at 2500 displayed-triangles with 64x64
textures
near viewpoint.    

I really try hard to make a high quality game on Net Yaroze,
but this Library-bug makes me crazy.
I have worked ca 500 h (without exaggeration) to solve this problem,
but now i NEED HELP.
It would be fine if someone could help me with this problem,
or if Sony could bring out patches for GsLib to solve such bugs.

Andreas Schrattenecker