MT-32 emulator development: Good news!

noize · Post by **noize** » Mon May 11, 2009 9:05 am

As you can see in cvs, some new features are on-the-fly sawtooth and blitsquare generation. For now this needs a lot of extra cpu power.

But as canadacow wrote himself, one of the main objectives next to better emulation of mt32 sound, is cpu optimalisation.

It will just take some time. I'll bet in a year or so these guys will have this gem running smoothly.

RvLeshrac · Post by **RvLeshrac** » Wed May 13, 2009 6:00 am

*edit* replaced the post here with what I had been trying to post originally (thanks fingolfin!)

-----

It ran *reasonably* well on my laptop's Turion @ 2.1ghz, after I ran it through the Intel compiler. CPU usage is at 50%, which is to say "100% of one core," when playing back some tracks. On other tracks, it barely peaks over 30%. The much-improved audio definitely comes at a price. (for comparison, DOSBox-MT32 doesn't even break 30%)

The new code isn't optimised at all, and a large amount of the audio clipping stems from the Win32 driver *and* the fact that I'm running it on Vista, which no longer has the appropriate level of support for all of the DirectMusic calls in the driver. (can you do lots of math? can you code? there might be a project for you...) I did fix this to a degree by tweaking the buffers, but that broke inline-sysex handling.

As far as multicore, that's a big "no" right now. I'd test it on a big single-core, but my single-core system is dead ATM, and it runs XP. The build I did currently only runs (at all) on Vista (Why? nfi. I don't code). My best guess is that multithreading in MUNT is spotty for several reasons, chief among them being that the waveforms have to be produced in sync and mixed together.

I'm glad people are enjoying the video. Hopefully it reaches someone with the right skillset to contribute to the project.

Longcat · Post by **Longcat** » Wed May 13, 2009 3:13 pm

RvLeshrac wrote:running Vista.

That would explain the lag, lol;)

md5 · Post by **md5** » Wed May 13, 2009 3:32 pm

Jonatan wrote:
RvLeshrac wrote:running Vista.
That would explain the lag, lol;)

er... no

Longcat · Post by **Longcat** » Wed May 13, 2009 4:13 pm

md5 wrote:
Jonatan wrote:
RvLeshrac wrote:running Vista.
That would explain the lag, lol;)
er... no

er... joke? bad day [md5]?:P

RvLeshrac · Post by **RvLeshrac** » Wed May 13, 2009 4:26 pm

It *does* explain the clipping, to a degree. The Win32 driver uses DirectMusic, and DirectMusic's kernel support was removed from Vista.

RvLeshrac · Post by **RvLeshrac** » Thu May 14, 2009 10:42 am

By request, I made an OGG of the Monkey Island (1) Introduction.

MIDI courtesy of http://s-island.mixnmojo.com

You can grab it from:

https://cid-81940f2fc1006a8a.skydrive.l ... spx/Public

LogicDeLuxe · Post by **LogicDeLuxe** » Thu May 14, 2009 11:25 am

RvLeshrac wrote:By request, I made an OGG of the Monkey Island (1) Introduction.

Sounds strange. Like all the volumes in the mix are wrong for some reason. Most of the flutes are barely noticeable for instance.

Another interesting thing I noticed, the engine seems to render the wave at more than 32000 Hz. With original Roland hardware (also true for Soundcanvas), you have mirror frequencies in the spectrum above 16 kHz, but in this recording, it is apparent that instruments were resampled and reach higher frequencies.
This might be a good idea from a hifi point of view, but probably not for accurately emulating the MT-32. And it causes extra CPU usage, I think.

Edit: As a reminder, here is what it sounds on a real MT-32 recorded with a M-Audio Delta 44: http://www.rapidspread.com/file.jsp?id=fy1s3ng8mu

RvLeshrac · Post by **RvLeshrac** » Thu May 14, 2009 5:02 pm

Good catch. 44khz it is, though there's a comment in the code about that being incorrect.

*edit*

Actually, I do get output at 32k now. Though I no longer get any debug messages.

MusicallyInspired · Post by **MusicallyInspired** » Sat May 16, 2009 3:07 am

Sounds good. A lot better than it used to be, anyway. Sounds like they still don't have the reverb properly emulated, though. Still, progress is progress.

RvLeshrac · Post by **RvLeshrac** » Sat May 16, 2009 5:24 am

MusicallyInspired wrote:Sounds good. A lot better than it used to be, anyway. Sounds like they still don't have the reverb properly emulated, though. Still, progress is progress.

Perhaps someone can make a dump of the reverb ROM... *hint*

canadacow · Post by **canadacow** » Thu Jul 23, 2009 9:37 am

We have a dump of the reverb ROM. Thanks!

Right now almost all the math in MUNT has become floating point which allows us to tinker more but places a heavier burden on the CPU.

We were making pretty good progress and then I had to go finish porting X11 and writing a VNC client for my iPhone app. I showed up here as Monkey Island: SE got released for the iPhone tonight and I wanted to see what ScummVM's reaction was. And then I saw this post. I'm glad to see there's still a good bit of interest in MUNT's development.

Reckless · Post by **Reckless** » Thu Jul 23, 2009 5:51 pm

Pretty sure a lot of people would love to have an updated MT32 implementation

Put that iPhone down and get back to it

MusicallyInspired · Post by **MusicallyInspired** » Mon Jul 27, 2009 11:14 pm

I'm eagerly awaiting a fully functional MT-32 emulator and I own an actual MT-32!

RvLeshrac · Post by **RvLeshrac** » Tue Jul 28, 2009 2:00 am

MusicallyInspired wrote:I'm eagerly awaiting a fully functional MT-32 emulator and I own an actual MT-32!

I think most of us who are eager to see a full MT-32 imp. are either owners or former owners. The main reason I'm interested is because it is getting harder and harder to hook up the MT on newer machines, and nigh-impossible on a laptop (which would defeat the portability aspect anyway). Also, the ability to use the MT-32 emu as a virtual instrument will, hopefully, see some action in various music communities.