Re: [hatari-devel] Spec512 screens with higher CPU frequencies

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 02/03/2019 à 18:15, Thomas Huth a écrit :

> Initially, both players start in the same room, and thus both parts of
> the screen use the same palette. You've got to move one of the players
> to a different room to see the problem: When you move one of the two
> players into another room that uses a different palette, each part of
> the screen should get a different palette. However, in 32 MHz, the
> second part of the screen always has the same (or most of the same)
> colors as the first part.
>
> I think the problem might rather be the "CycleEnd >>= nCpuFreqShift" in
> Spec512_EndScanLine() ... that looks like it is causing the update to
> stop too soon. When I comment it out, the split screen in AlphaWaves
> also looks fine again...

I can see the issue after going to another room ; but for me it's more 
an unstable timerb interrupt (due to 32 Mhz) than spec512.c itself, as 
the game in fact only changes the palette at mid-screen

There were some issues at some point when converting the cycle position 
into the current frame into an X/Y position when not using 8 MHz.
I fixed most of them, but that's might be a remaining case, I will look 
into this in more detail (probably the X position is odd instead of even 
and the code in spec512.c doesn't store the color change)

In all case, that's definitely sthg that should work at any cpu freq, 
fixing it in spec512.c might not be the correct fix though, it might be 
in video.c when timerb is handled.

Nicolas