Re: [hatari-devel] Falcon cpu/dsp emu issue?

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 28/01/2016 18:51, Eero Tamminen a écrit :

> That was a good catch, just commenting it away got
> programs working.  With it, I guess only interrupt
> / exception handling added CPU cycles count for
> the amount of stuff DSP would be run.
>
>
> > Yes, I rolled back the use of DSP_CyclesGlobalClockCounter, but it seems
> > I forgot to also comment "currcycle = 0" in that case :(
>
> It seems to be in several cores (at least 030 and 040).
>
>
> > I will change this tonight and do more tests.

This is now fixed, dspbench and wilfire are now working in CE mode again.

There was also a small bug when starting hatari with tos 4.04 : hatari 
automatically switched to Falcon mode, but DSP was not enabled (you 
needed to change it in system settings)

I think some cases are missing though : when switching to another TOS 
than 4.xx and nDSPType was DSP_TYPE_EMU, then I think DSP_Disable should 
be called too (unless DSP can work with TOS 3 ?). Else, this will just 
call DSP_Run() for nothing and add a huge slowdown to emulation.

BTW Doug, do you have the result of dspbench when running on a stock 
Falcon ? Could be useful for further reference to check cpu/dsp accuracy 
in Hatari (or maybe this was already posted ?)


Look for more breakages, as I have a work in progress to change the way 
cycles are counted at 16/32 MHz :)

Nicolas