Audio emulation fidelity on FS-Uae

[Csoft]

Hello folks,

my first message here. I'm developing a little Amiga game in full 68k assembly. I've always coded on real hardware, this time for the first time I cross compiled targeting fs-uae under MacOS.

I'm puzzled with a really weird problem: sound playback. As you know, stopping a sound after you started playing is a bit brutal on Amiga, you have to wait scanlines and then set the audiopointer to a zero buffer in memory (there are other methods involving interrupts but I didn't tried it yet, actually I start audio playback then point to zero buffer the next vbl). I did it in this way and it works perfectly under fs-uae, but it doesn't work under a real Amiga 1200 and under WinUAE.

It's clear that I'm doing something wrong on my implementation (since it doesn't work on real hardware), but the fact that in fs-uae this works misled me to think that it was fine.

This makes me think that fs-uae is not perfectly exact on audio hardware behaviour, because it doesn't show the bug in my code but it make my code work even when it shouldn't.

Obviously I have to fix my code, but I'd love to do it with fs-uae, is there some configuration I can set to make fs-uae more "cycle exact"?

Thank you

Lorenzo

[phx]

Originally Posted by Csoft

have to wait scanlines and then set the audiopointer to a zero buffer in memory

How man scanlines?

I did it in this way and it works perfectly under fs-uae, but it doesn't work under a real Amiga 1200 and under WinUAE.

When I did my ptplayer I calculated using the HRM that 496 CIA ticks should be sufficient for an Audio DMA to occur, even with the lowest playback frequencies.
In reality that failed on some A1200 models. They required up to 550 ticks. No idea why. So maybe you just have to increase your delay for this A1200. I settled on 576 ticks.

[Csoft]

Thank you for your response.

Originally Posted by phx

How man scanlines?

Don't know, I wait for the next vertical blank to point to the zero buffer.

When I did my ptplayer I calculated using the HRM that 496 CIA ticks should be sufficient for an Audio DMA to occur, even with the lowest playback frequencies.
In reality that failed on some A1200 models. They required up to 550 ticks. No idea why. So maybe you just have to increase your delay for this A1200. I settled on 576 ticks.

Ok, I'll give it a try. But the other question remains: why fs-uae makes my (bad) code work even if it shouldn't?

I have always thought that the emulation of fs-uae was perfectly cycle exact!