NetNews Usenet Archive 1992 #19/NN_1992_19.iso/spool/comp/sys/atari/st/tech/4577

Internet Message Format | 1992-08-26 | 3.3 KB

Path: sparky!uunet!mcsun!corton!loria!loria.crin.fr!eker
From: [e-mail] (Steven Eker)
Newsgroups: comp.sys.atari.st.tech
Subject: Re: Sprite code (Re: Second&last time: Hacking Contest)
Message-ID: <[e-mail]>
Date: 26 Aug 92 19:54:13 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]> <[e-mail]
Sender: [e-mail]
Organization: CRIN (CNRS) Nancy - INRIA Lorraine
Lines: 70

In article <[e-mail]>, [e-mail] (Warwick Allison) writes:

|> >Maybe, maybe not - it depends on your sprite drawing method. The code
|> >I originally posted only read the screen implicity with
|> >       or.w    d0,(a0)+
|>
|> Hardly useful for anything but UNMASKED single-bitplane graphics, which
|> is next-to-useless (at he very least, you usually want to have black
|> and white sprites (like the ST mouse pointer), rather than just black
|> sprites (like the Sun SunView mouse pointer)).  Of course, in the
|> interests of competition, anything is "useful" :-).
|>
Actually I said or.w d0,(a0) _type_ instructions. By this I also mean
and.w d0,(a0), eor.w d0,(a0) ... The code does proper masking.

|> <delight> if you can mail me code-for-code patches, I'll put them into
|> my library (which will be free-to-all one of these days).

I'll do that as soon as I have some time. The main problem is that I use moto
memonics & your code uses MIT memonics (which I can read (to calculate the
timings) but not write (reliably)). Also what regs does gcc (a) use to
communicate with the asm stuff (b) leave free for the asm stuff - I need all
8 data regs!
BTW when I write hybrid C/asm code I always put the asm in a separate .s
file and pull my own arguments off the stack.

|> Actually, I think it was more like 50 at 72Hz (ie. mono).
|>
|> >Still I can't resist the challenge...
|>
|> You didn't post performance stats!  How well does it BURN?

Oops - you didn't say mono - the last routine I posted was for single plane
lores sprites (any plane). Mono should be faster because you don't have
skip over unused planes & can make use of long word ops.

|>
|> Let's make some categories:
|>
|>     1. Preshifted / Unshifted
|>     2. 16x16 / 32x32
|>     3. Masked / Unmasked       (ie. black&white / black)
|>

Sounds good... and when we've run out of sprite possibilities we can move on
to scrolling & parallax scrolling & then sample replay & then polygons & then...

|> Yep.  One better:  with unmasked sprite, why even HAVE sprite data?
|> it uses only 100% more RAM to have the sprite data be the CODE ITSELF,
|> ie. automatically generate the
|>
|>     orl   #....,(a0)     (3 words)
|>     addl  d0,a0          (1 word)
|>
|> instructions from the 2-word sprite-data!
|>
even better:
    or.l    #...,(a0)
    or.l    #...,80(a0)
        or.l    #...,160(a0)
      :      :
saving 4cy/long word. Actually I've considered this "executable data"
idea for 3D vectors but never had the patience to code it up.

|> (Of course, this is only acceptable for HACK CONTESTS.  Anyone who seriously
|> considers this a "programming technique" is a beauxault)
                                                 ^^^^^^^^^ a what?


Steven