[hatari-devel] Demo regression between 1.3.1 and current version

npomarede at corp.free.fr npomarede at corp.free.fr
Tue Apr 13 23:15:43 CEST 2010


On Tue, 13 Apr 2010, Eero Tamminen wrote:

> Verified.
>
> Jam-Cols demo seems fine to me too:
> 	http://www.pouet.net/prod.php?which=31452
>
> If there's still a problem, at least I don't see it. :-)

Yes, looks fine, but I don't think it's related to my latest patch, as 
this demo is mainly changing only the background color using the blitter 
(which gives a 8 pixels precision for each color)

>
>
> Btw. the RGBeast demo screen:
> 	http://pouet.net/prod.php?which=29355
>
> Seems to look sometimes partly correct.  Any idea what's needed to
> get it working correctly (info could be added to compatibility page)?

Problem is in blitter timings and the way the blitter can start/work while 
the cpu is still processing an instruction (using some kind of "idle" 
cycles not taken by the cpu).

Hatari can't emulate both blitter and cpu at the same time, it runs them 
one after the other, which will not be enough in the case where blitter is 
used with overscan (at least in this demo)

Very hard to emulate correctly now, this would require to replace the cpu 
core with winuae's one which supports "sub cycle" for each 68000 
instruction (at the cost of a much more cpu intensive emulation)


Nicolas



More information about the hatari-devel mailing list