Sound distortions in Dungeon Master II

Since in the end it seems it wasn't a WHDLoad problem nor a problem with the game itself (it worked in past WinUAE versions) I'm bringing this problem here just in case. My apologies in advance if this one doesn't belong here though...

The original thread is here: http://eab.abime.net/showthread.php?t=33453

[Toni Wilen]

ECS or AGA? If AGA -> not much interest as usual. ECS with proper configuration most likely works fine.

AGA of course. DM2 doesn't work in ECS and it needs to be installed in HD to play.

[viddi]

Originally Posted by Zósite K.S.

AGA of course. DM2 doesn't work in ECS and it needs to be installed in HD to play.

It´s not AGA. E.g. it works on an A2000 with 2 MB ram.

[Retro-Nerd]

There are really two games versions? It's a bit confusing. I've an English manual for the AGA version, Planetemu has an AGA version.

The SPS dump is OCS/ECS, according to HOL/SPS. KG named his WHDLoad/SPS installed game with "AGA", because the readme says "AGA version supported".

Yeah, it seems to work with ECS in standard installation without WHDLoad. But it doesn't matter... the standard installation with ECS chip has the same distortion problem.

[viddi]

As far as I know there is only one version out there.
It is labeled AGA but works on ECS machines.

Primal Rage was labeled as A1200 only as well just because it needs 2 MB!

[Retro-Nerd]

It seems so: A1200/AGA but Workbench 1.2 or higher.

[glue]

buahahaaha A1200 with Workbench 1.2 really good config to operate with hard disk

[meynaf]

I've read that this game requires AGA only because it needs 2MB of chipmem.
But it certainly doesn't use 256 colors.

If you're experiencing sound problems then it may be because the game runs out of memory (I even had random crashes on 2MB).

But the distortion problem makes me think that you've got one or several musics files that aren't ok, and I suggest you reinstall the game without using the buggy unlzx that's provided with the original package.

Ummm... thanks for your suggestions Meynaf, but I've already tried several high memory configurations.

Besides, it cannot come from a damaged music file because the game works perfect with older WinUAE releases as I already said.

[meynaf]

Then I don't know. I've never had this problem. Then again, I've never tried it on winuae
(the game works fine on my 68030, I finished it several times)

[Toni Wilen]

My crystal balls says the "problem" is the usual: too fast chip ram and/or too fast cpu causing some stupid cpu delay loops to run too quickly. I consider these kind of programs (68020+ only) broken and I only "waste" time checking them if I feel interested enough..

Ok, ok... If Toni says so

Shall it be written, shall it be done. I just wonder why it worked not so long ago and not now.

At least I tried.

[meynaf]

Originally Posted by Toni Wilen

My crystal balls says the "problem" is the usual: too fast chip ram and/or too fast cpu causing some stupid cpu delay loops to run too quickly. I consider these kind of programs (68020+ only) broken and I only "waste" time checking them if I feel interested enough..

It is true that DM2 sometimes miss notes even on a 030, but it's not enough to call it distortion...

[Toni Wilen]

(I was interested enough after all..) It worked in old version because audio emulation was broken in other programs. (Maupiti Island for example) but now DM2 (which has quite stupid audio code..) expects something that would not be compatible with MI.. Something is missing..

This is not CPU timing problem, at least.

[meynaf]

Just curious : what does DM2 expect ? (with all the gruesome details, of course )

[killergorilla]

Originally Posted by Retro-Nerd

There are really two games versions? It's a bit confusing. I've an English manual for the AGA version, Planetemu has an AGA version.

The SPS dump is OCS/ECS, according to HOL/SPS. KG named his WHDLoad/SPS installed game with "AGA", because the readme says "AGA version supported".

Nope, I labelled it AGA because when you put it in ECS the whdload version says "This game needs AGA to run"

I started a thread on it a while back:

http://eab.abime.net/showthread.php?...eon+master+aga

[Toni Wilen]

Originally Posted by meynaf

Just curious : what does DM2 expect ? (with all the gruesome details, of course )

DM2: Audio interrupt request cleared, writes to AUDxDAT, start DMA. (DMA should start normally instead of getting in "PIO"-mode audio)

CD32 boot tune: Audio interrupt request cleared, writes first sample to AUDxDAT. Waits for "PIO"-mode interrupt and then starts DMA mode. (Why does it do it this way?)

Maupiti Island speech: Audio interrupt request cleared, writes to AUDxDAT, waits for interrupt, restart until speech ends.

There may be delay between transition from state 0 to state 2 when starting "PIO"-mode audio (HRM says "transfer to main loop occurs immediately after data is written by processer") but I don't see why or how because it would cause other issues and in hardware state transitions checks normally happen every clock cycle "automatically".

Some condition must be missing.. (I have one weird idea which needs testing..)

ADDED: possible missing condition: in "PIO"-mode audio AUDLEN and AUDLC are also copied to backup registers. (Perhaps this condition is "hard-coded" to "set interrupt" condition in hardware? This isn't really described in audio state diagram) This change will be in next beta (if I don't find any programs that fail..)

[meynaf]

Originally Posted by Toni Wilen

DM2: Audio interrupt request cleared, writes to AUDxDAT, start DMA. (DMA should start normally instead of getting in "PIO"-mode audio)

Doesn't look very useful...

But you can just kill the "PIO" audio mode whenever DMA is on and you're done. DMA will always take over and they can't be mixed (not on the same channel, that is) - just my "2 cents" suggestion.

Originally Posted by Toni Wilen

CD32 boot tune: Audio interrupt request cleared, writes first sample to AUDxDAT. Waits for "PIO"-mode interrupt and then starts DMA mode. (Why does it do it this way?)

It is done this way because the author did not want the tune to play correctly on an emulator
Errrh.. seriously, doing so is completely useless IMHO. However, of course, it should be harmless.

Originally Posted by Toni Wilen

Maupiti Island speech: Audio interrupt request cleared, writes to AUDxDAT, waits for interrupt, restart until speech ends.

The normal way of playing a sound with the cpu...

Originally Posted by Toni Wilen

There may be delay between transition from state 0 to state 2 when starting "PIO"-mode audio (HRM says "transfer to main loop occurs immediately after data is written by processer") but I don't see why or how because it would cause other issues and in hardware state transitions checks normally happen every clock cycle "automatically".

Some condition must be missing.. (I have one weird idea which needs testing..)

Programming audio is a pain, so I can't imagine what emulating the hardware accurately must be

Anyway, do programs really depend on audio timings ? Would it add some trouble if you allowed stop-then-restart the audio dma regardless of the time between stop and restart ? (and setting the repeat/replen values after the dma has started)