> How about using a single zero page routine which, for delays, farms it
> to an external routine (JSR delayn). The self modification would be to
> WHICH routine. Have a whole bunch of routines delay1 delay2 delay3
> which simply delay that many cycles. OK, you add on 12 for the JSR/RTS,
> you could REALLY hardwire it and have it JMP delayn and JMP back, for 8
> cycles total. The point is, you would only have the single scanline
> routine, it would be quite short, and could fit in zero page.
> The self-modification part would simply be the low byte (if we organised
> properly) of the routine to which we're vectoring. I haven't figured
> feasibility in entirety, but seems eminently possible to me
Well, the problem with that approach is that I need single-cycle precision
with an overhead of *zero*, so that the two objects could be positioned
in the same place (RESP0 and then the *next* instruction must be RESP1.)
But it also has to be flexible enough to allow one cycle between the
stores - this could be done by making the second RESP1 absolute addressing
instead of zeropage. A difference of two is obviously done with NOP, and
so on up the line. But how do you allow all those possibilities with zero
overhead between the stores, except in self-modifying code? It'd also
help if there was zero overhead afer the second RESP, so you could
immediately WSYNC and HMOVE.
and from Bob:
> I've thought about this too, and would love to develop an algorithm to
> do this. I think it can be done without self-modifying code though and I
> have an idea. Unfortunately I have zero free time until Sunday, but I
> will try to get the idea either down in writing or coded to play with.
What's the idea? Give us the general concept and let us brainstorm on it
:)
--
Archives (includes files) at http://www.biglist.com/lists/stella/archives/
Unsub & more at http://www.biglist.com/lists/stella/stella.html
