🚀 go-pugleaf

d89-rbi@dront.nada.kth.se (Ron Birk) writes:

>It seems that most of the demos doesn't like the cache
>memory!! 

Exactly.  I also have a 500 but I've started realising that you now have to
think about things like this when coding.  At least where it doesn't make 
too much difference in terms of speed.  For competitions the fact is still 
that they are run on a standard 500 and people want to utilise all the speed
they can on them. 

The biggest problem I think is self-modifying code.  

Example:  You want to use a different address in your routine each time. 
	  Say animated bobs for example.

	tst.w	whichpicture
	beq.s	a
	move.l	#pic1,x+2
	bra.s 	b
a:	move.l	#pic2,x+2
b:	...
	...

and in your routine you've got:

x:	move.l	#[anything],$dff050  (blitter source or something)

The correct way is of course:

	tst.w	whichpicture
	beq.s	a
	move.l	#pic1,pointer
	bra.s	b
a:	move.l	#pic2,pointer
b:	
...

pointer: dc.l	0	(reserved space)

x:	move.l	pointer,d0
	move.l	d0,$dff050.

Because I don't think you can  move.l pointer,$dff050 directly, right?

Compare the two:  The second one is much faster on a computer with no cache.
but won't work with cache.

x:	move.l	#[anything],$dff050  (blitter source or something)

x:	move.l	pointer,d0
	move.l	d0,$dff050.

In a loop done several hundred times, this makes a lot of difference.
I have thought about it and most of the time there ARE other ways to do it 
that are almost as fast and work with all computers.  Considering the fact
that groups usually don't want ONLY to win the competition (even if that's a 
big priority) but also to spread their demo to a large group, I think we will
eventually see people learning to code so that it works all over the place.

It's does take more effort, though because it is usually easier to rely on 
various timings on the computer (500) than to make it look good even on a 
faster computer.  What I mean is a lot of people write routines that work
perfectly only on a 500 but look terrible if the processor on a faster 
processor.

Well enough about this, I'm going to go and FTP the howtocode-file now and 
read it since everyone's saying it's so good.

BTW, how many non-swedes read this group?  Is there really any reason for not
typing in Swedish?  After all there are other international amiga groups.


*----------------------------------------------------------------------------*
|            Gunnar Andersson   --  d1gunnar@dtek.chalmers.se                |
|              Chalmers University of Technology, Sweden                     |
*----------------------------------------------------------------------------*

In article <d1gunnar.726756144@dtek.chalmers.se> d1gunnar@dtek.chalmers.se (Gunnar Andersson) writes:

   BTW, how many non-swedes read this group?  Is there really any reason for not
   typing in Swedish?  After all there are other international amiga groups.

Anledningen till att jag skrev på engelska var att jag inte hade lust att
översätta hela texten innan jag postade det i detta möte. Självklart skall
man skriva svenska här.
--
 Ron Birk         CS&CE at Royal Institute of Technology, Stockholm, Sweden.
 Krogarvagen 11   Projects: PlaySID (Amiga), FanGog (Mac), MonkeyBase (Unix)
 14552 Norsborg   E-Mail: d89-rbi@nada.kth.se     - Please, be happy! -