home *** CD-ROM | disk | FTP | other *** search
- Path: nntp.teleport.com!sschaem
- From: sschaem@teleport.com (Stephan Schaem)
- Newsgroups: comp.sys.amiga.programmer
- Subject: Re: TMapping again!
- Date: 11 Jan 1996 22:19:30 GMT
- Organization: Teleport - Portland's Public Access (503) 220-1016
- Distribution: world
- Message-ID: <4d42di$9e9@maureen.teleport.com>
- References: <4d0ou6$835@astfgl.idb.hist.no> <Z31Wx*zA0@mkmk.in-chemnitz.de>
- NNTP-Posting-Host: kelly.teleport.com
- X-Newsreader: TIN [version 1.2 PL2]
-
- Andre Weissflog (floh@mkmk.in-chemnitz.de) wrote:
- : In article <4d0ou6$835@astfgl.idb.hist.no>, Lasse Staff Jensen writes:
-
- : > I have been following the discussion on fast tmapping
- : > innerloops, and have a question conserning those
- : > very fast loops (i.e two instructions per pixel).
- : >
- : 2 instructions per pixel for free texture mapping? How
- : do you define 'instruction' ?
-
- I think he was mentiontioning this map code:
-
- REPEAT 32
- 1) move.b (a0,d0.w),(a1)+
- 2) addx.l d1,d0
- ENDR
-
- : > What do you guys do when the delta values are negative?
- : > Clearly an addx.l will not work! Do you use some kind of
-
- : This trick should also work with a negative adder, since the
- : bit that once defined the adders sign is now somewhere in
- : the middle of the longword and doesn't hurt. The real MSB
- : is now only important for producing overflows (or underflows,
- : just as you like it) of the fractional part. It wouldn't
- : work that well if over/underflows of the integer part would
- : happen but since [u,v] have a limited range of [0..txt size]
- : this is unlikely anyway.
-
- In the example above I dont think it will work with neg number.
- Because the msbit of X if neg will have an effect on Y.
-
- On the 'regular' addx methode it does work wihtout problem...
-
- : > symetric tmap, or does the tmap repates outside the frame,
- : > or what? Do you have several innerloops?
-
- : Personally, I have 8 inner loops at the moment.
-
- : But they handle linear, perspective corrected, non-shaded,
- : gouraud-shaded, transparent mapping and all combinations
- : thereof that make sense. I don't use the X-Flag trick however,
- : and the more optimized versions live on the x86 anyway ;-/
-
- the xflag 'trick' is not a bad methode... it enable you
- to preshift value. It would fly on a 150mhz 030 ;)
- people compare to quickly a tmap program on a 66mhz 486 and a
- a1200... to see x86 true code speed they should run it on a
- 12mhz 286.
-
- Stephan
-