[hatari-devel] EKO System and easter

Laurent Sallafranque laurent.sallafranque at free.fr
Wed Jun 24 23:00:01 CEST 2009


Thomas Huth a écrit :
> On Wed, 24 Jun 2009 09:51:55 +0200
> Laurent Sallafranque <laurent.sallafranque at free.fr> wrote:
>
>   
>> The result is exactly the same than if I use "with DSP"
>> At 16 Mhz, the demo freezes and image and sound are not synchronized
>> At 32 Mhz, everything is synchronized (demo stops when DSP code
>> should start, music continues to play).
>>     
>
> Ok, so this means that the timing problems are not related to the CPU :
> DSP ratio, right?
>   
I agree here.
When I have a doubt with a program, I always try to change the 1:7 
ratio, but it's never involved in the problem.
I think this ratio is the max value that will always let run DSP 
programs and CPU synchronized.

The only thing we could optimize here is to calculate precisely the time 
of each DSP instruction.
But the result would let execute 6 DSP instructions instead of 7 for 1 
CPU instruction.
I don't think that it is the most important code to add for now.
And optimization may not be there. There are some other parts of code to 
optimize into DSP first.

> Have a look at dmaSound.c ... I thing the frequency table should be
> fine...?
>   

They look fine to me too.
>> * The synchro (image/sound) is dependent of the "processor speed".
>>    There must be a timer somewhere which is not correctly computed
>>    (In videl, or dma sound)
>>     
>
> Could be related to the VBL ... as I already mentioned in another mail,
> the Videl emulation does not emulate the screen timings right yet. VBL
> is always triggered in 50 Hz, no matter which refresh frequency is
> programmed into the Videl chip...
> Have I already mentioned that the Videl code needs a major rewrite one
> day? ;-)
>   

Ok for the major rewrite of Videl code. That's again a huge work to add 
to the TODO list.

Regards,

Laurent




More information about the hatari-devel mailing list