Re: Sound support (2), using hardware directly

[Julius Schwartzenberg <julius@zgod.cjb.net>]

Stas Sergeev wrote:
>> Yes, I am aware of the MIDI support in Dosemu, but I thought it was 
>> only for General Midi compatibility.
> 
> No. On the software synth side, midid
> have the -m option which will give you
> some basic mt32 support.
> As for the directly passing the midi
> stream to the wavetable - this is also
> possible and described, and will allow
> you to use your AWE synth fully, provided
> it all goes via the mpu-401 in a
> pass-through mode (almost always this is
> the case).
> So you can try and see if there are any
> differences with what you now have - there
> shouldn't be, but you won't need root,
> which is very important in some cases.

Ah, I already suspected this. I'll try this out at some point and see if 
I'm able to hear a difference.

> Well, the primary goal is to design the
> very simple but robust sound subsystem
> from zero, so that it won't depend on
> OSS, but can work also with ALSA and
> anything, and so that it can work with
> many more DOS progs than the current one
> does. But the OPL3, again, is not in the
> list of the primary targets:)
> Of course porting the MAME OPL3 emulator
> to the planned sound system must be easy,
> but we'll see where this all will end up.
> It will probably take some time before
> something can be released though - the
> main problem is that the current code
> also happened to work somehow, and the
> replacement is possible only when the new
> code matches the current one in the
> functionality, which can take long enough.

The current system seems to work pretty well for me, although it seems 
to soundquality could be a little bit higher in certain cases.
While OPL3 emulation would be really nice, wouldn't it be better to 
implement support for real OPL3 chips and emulate and OPL3 at another 
level in the system if one isn't availlable? Then also other 
applications such as FreeSCI and ScummVM would be able to make use of it.

About the new system. If the new code isn't comparable to the current 
code yet, maybe release 1.4 of Dosemu should still contain the old code? 
After that I think the development version could be switched over to the 
new code, even if it's not fully ready yet. Then it would already get 
more testing and it might be easier for development. It shouldn't be a 
big problem if a development version/tree has regressions.

I'm really looking forward to the new stuff.
Thanks,

Julius