Aw: Re: [stella] Brainstorming: Porting Smithereens/Catapult/Stone Sling

Subject: Aw: Re: [stella] Brainstorming: Porting Smithereens/Catapult/Stone Sling
From: cybergoth@xxxxxxxx
Date: Thu, 24 Jul 2003 09:48:39 +0200 (MEST)
Hi Thomas!

> Manuel wrote:
> > Basically I think this'd work for the stationary part, 
> > though I fear it'd be monochrome.
> You could use the "score mode" here.


I always forget about that one ;-)
> > Or doing the castles with the PF as Rob suggested.
> Yes, probably the best idea. 
> Let's see:
> - 3 reads and writes to PF1 (~19)          
> - 2 reads and writes to GRPx (~16)
> - 2 reads and writes to HMMX, NUSIZx (~32)
> - 1 write to HMOVE (3)
> - minimal loop overhead (5)
> = total: 75 cycles
> Hm, the ball is still missing and then you might want different colors.
> I assume a 2LK kernel is a better idea.

Why a third write to PF1?

I think an interesting thing to keep in mind is, that except for the ball and the PF, no vertical positioning is required, so you could do one top and one bottom kernel:

The top kernel is dead simple, only drawing the upper part of the castle and the ball.

The bottom kernel then does the real work. It's line counter is already indexing all data tables and you never have to answer the "to draw or not to draw" question - you just always draw.

With a game that simple, you can then buffer all your shape/NUSIZX/HMMX data into the RAM.
If, lets say the bottom kernel does 10 2LK lines you can buffer all kernel data in 2*4*10 = 80 bytes, leaving still almost 50 bytes free...


Archives (includes files) at
Unsub & more at

Current Thread