Re: [hatari-devel] Lockup demo

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


Hi,


Hatari always shift the data by 2 bits when transfering data from DMA to DSP (eg: Mprg2 player works because of this).

You say 1 bit or two. Can  you be more precise ?

I mean I always shift by 2 bits, but do you mean that in some circumstances, the shift could be only 1 bit ?

In which circumstances ?

Regards


Laurent

Le 06/12/2020 à 21:42, Mariusz Buras a écrit :

If you get noise this might be due to another issue in hatari I've encountered.
When sending data to DSP over DMA in handshake mode data is shifted by
a bit or two. I think hatari didn't emulate that bug properly
properly.

On Sun, 6 Dec 2020 at 20:38, Laurent Sallafranque
<laurent.sallafranque@xxxxxxx> wrote:
  > What about the Lockup demo music?

There were 2 different problems with lockup demo :

- the tridi needed a hostport interruption that was bugged (never tested
as I had never encountered any program that used it)

- the music needs a new interrupt that was never coded until now.

The SCI port is not connected to the Falcon and the Falcon docs says not
to use it.

But Mariusz used the SCI timer to synchronise the music with the SSI and
it works.

So, I have to code a new interrupt into the DSP, and a new timer.

I've nearly finished, but instead of music, I have bad noise for now.

I work on it.

Regards

Laurent


Le 06/12/2020 à 20:10, Eero Tamminen a écrit :
Hi,

On 12/6/20 8:09 PM, Laurent Sallafranque wrote:
No, you're right. I've tested it also with older versions of hatari
and I always get 4 bombs (with or without MMU, ...)
Ok, it's weird that it on some days works for me.


What about the Lockup demo music?

In the Youtube video linked from pouet.net,
music starts once the progress bar at the start
finishes, but in Hatari Lockup demo is silent.


     - Eero

Le 06/12/2020 à 17:29, Eero Tamminen a écrit :
On 12/5/20 5:00 PM, Laurent Sallafranque wrote:
I've comited a more simple version of the fix.

No need to add the "dsp_core.interrupt_state == DSP_INTERRUPT_NONE"
when searching for a new interruption.

The patch is very easy now. I've just moved the 2 interruptions
into the main interruption routine.

Before, the hostport interruption was tested only after a DSP or
host read or write, but the interruption can happen as soon as the
buffers are empty or full.

Now, I just do the test after each DSP instruction.


I've done many tests with demos that I know use DSP interrupts,
games and music programs.

I didn't notice any regression with this patch.

Enjoy Lockup demo now ;)
It works now with DSP and shows 3D graphics, but I don't have any
music.

Do you have music in it?


I tested many programs and only regression I
noticed was in Sidetracked music demo, it doesn't
anymore start, *even with older Hatari version*,
neither with TOS4 or EmuTOS, with or without MMU,
with Hatari default config or my own.

(Dry Egg game, which also worked only with EmuTOS,
does still work.)

I don't understand what could have changed, to
break that demo also with older Hatari versions.

Maybe it's wall clock sensitive?


     - Eero

PS. I'll push update to compatibility list &
release notes after I've done some additional
testing.









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