Strange Paula quirk allows for infinite playback lengths

[Toni Wilen]

Originally Posted by 8bitbubsy

So any LEN divisible by 227 should trigger the trick (together with too low period)?

No. It needs to be 227 divisible but not all values will work because glitch requires both period countdowns to happen in correct order (2->3 vs 3->2). Possibly there is some other condition too.

This is difficult to visualize.. Perhaps it needs some special dma debugger mode to see what really happens, in real time.

[8bitbubsy]

That might explain why period 113 didn't work with len 2270.

[Toni Wilen]

Few days ago I was doing unrelated interrupt line to IPL line change timing tests and noticed there is 2 CCK delay between first AUDxDAT Agnus DMA write and before Paula IPL line changes state.

I thought this might have something to do with this guirk.

And it does, today I did some tests and if I add 2 cycle delay before AUDxLEN counting down is handled in emulation, "quirk range" matches real hardware exactly.

TODO: find some optimal emulation that only does this when it is really needed, to not waste CPU time for something that almost never happens.