Re: [hatari-devel] Re: 040/060 emulation issue with BadMood

[ Thread Index | Date Index | More lists.tuxfamily.org/hatari-devel Archives ]


So after poking around with the new Hatari in 060 mode, I did find a bunch of oversights and timing gaps in my code - so these have all been fixed. It now runs in Hatari in 060 mode.

However based on tests with real HW, it seems Hatari is somehow outracing real CT60s on the CPU/bus side, leaving the DSP in the dust. Probably to be expected - but I had to set up the sync sites to be extra-pedantic in a way that didn't seem necessary, at least up to 75MHz on a CT60. 

Beyond 75MHz, maybe things start to equalize with Hatari. 

This isn't a bug report - just some extra context for comparison with CT60 timing in case anyone finds it useful..

D.


On 2 March 2015 at 22:30, Eero Tamminen <oak@xxxxxxxxxxxxxx> wrote:
Hi,

On sunnuntai 01 maaliskuu 2015, Eero Tamminen wrote:
> On sunnuntai 01 maaliskuu 2015, Douglas Little wrote:
> > > Both work with higher CPU clock (32Mhz) and TT-RAM, but freeze at
> > > level start when I try Hatari's 040 or 060 emulation.  There's no
> > > Hatari output when that happens (e.g. about bus errors).
> >
> > It looks like a timing mismatch. What were the exact names of the build
> > archives?
>
> bmdoom_v0.33_040_060_test.zip and bmdoom_v0.33_040_060_test3.zip.
>
> > And which build was used to get the disasm?
>
> test3.

And these are the options I was using:
$ hatari --conout 2 --monitor rgb --natfeats yes -s 14 \
  --fast-forward yes --machine falcon --addr24 off --ttram 32 \
  --cpulevel 4 --cpuclock 32 --tos tos404.img ./bmdoom.ttp

Neither monitor, natfeats, cpuclock nor ttram/addr24 options affect the
freeze.  With EmuTOS I was able to check also 040 --mmu option, and that
also freezes in same place.

Btw. using "--cpulevel 4 --mmu yes" makes Hatari non-responsive at TOS boot,
regardless of TOS version.  Hatari then loops doing attached stuff.


        - Eero

> > One of the earlier builds
> > should sync on all transactions so timing issues shouldn't happen
> > unless I missed one. The later ones relax the use of sync in some of
> > the inner loops.
>
> Both behaved the same way.
>
>       - Eero




Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/