[hatari-devel] Cebit93 demo seems to run well too (no more screen update problems)

[Nicolas Pomarède]

Le 26/01/2011 23:08, Eero Tamminen a écrit :

>> This demo doesn't involve the DSP at all.
>>
>> So, the 16/32 Mhz problem seem to be only related to the 68x000 part of
>> the computer.
>
> Right.  It's an interesting case for the Falcon timings as they're so
> clearly visible, but only in what's shown on screen.

According to the readme with that demo, it also uses blitter a lot, so 
maybe this 16/32 Mhz problem involves blitter too, not just cpu, in 
which case debugging can be a little trickier than just looking at the 
cpu side (even if no dsp is involved). Still, it's interesting.

>
>
>> I think it may comes from Videl emulation missing (wrong VBL, HBL, ...)
>> Do you agree ?
>
> videl.c emulation doesn't do any HBL (line/palette switching stuff), it just
> shows the whole screen at VBL from a memory position given in registers
> and size given in registers with palette coming from suitable STe/TT/Falcon
> registers.
>
> The HBL stuff is all in video.c and AFAIK completely ST/STE specific.
> All that stuff is quite hairy (full of special cases), I think currently
> only Nicolas understands fully how it works, so I think this will need
> his involvement...  It should be fairly well traceable nowadays though.

I don't know videl that much, but as you say, HBL code in video.c is 
really STF/STE specific.
When writing HBL support for videl, better do it from scratch in the 
falcon/ directory, I don't think there's anything to reuse from video.c 
HBL code (it would only complicate things to merge them)

Nicolas