strange way to set dma audio

[8bitbubsy]

Originally Posted by leonard

so all these chiptune mods aren't played correctly with all existing amiga players but PT2.3F? Anyone can confirm that?

Sample swapping has been a feature since the very first ProTracker version AFAIK, or at least from very early on... There are plenty of oldskool .MODs out there using the technique.
It works fine in every PT version except the 3.x series (and maybe not 4.x), where they ruined the replayer logic by "fixing" it.
Here's an example demonstrating the effect: https://www.dropbox.com/s/zq6lq3ym1q..._swap.mod?dl=1

EDIT: It's mostly used to simulate PWM by replacing the current sample with a pulse wave of a different duty cycle. The fun part is that the swap is delayed until the current sample has reached the end loop point. This means that the swap will not reset the sampling position in the middle of a sample, so you get no clicks from it (depending on the sample data of course).

How come you mention PT2.3F specifically? Did I mess something up which makes it play back a module differently?

[meynaf]

Originally Posted by leonard

so all these chiptune mods aren't played correctly with all existing amiga players but PT2.3F? Anyone can confirm that?

I have made a few tests and PT3.15 fails where Delitracker2 plays these mods correctly.
As i understand it, the normal behaviour is to change the repeat/replen of currently playing sample.

[no9]

Yep. At the end of the cycle.

I also stressed this feature some months ago in non-chiptune way:
[ Show youtube player ]
Download files

[leonard]

Originally Posted by 8bitbubsy

The fun part is that the swap is delayed until the current sample has reached the end loop point. This means that the swap will not reset the sampling position in the middle of a sample, so you get no clicks from it (depending on the sample data of course).

exactly, and this is a PAULA hardware feature ( the player just "write" the new sample address in PAULA registers and the chip will "really" use this new address at the next loop point

My issue is that players are writing the new sample address during the playback (Audio DMA is still running). And it theory, if you're very unlucky, as the 32bits address write isn't atomic, you may have an audio glitch during 2 samples.

As no one complained for this for years, I guess it's ok in practice. But the
bug is here, in all Amiga players supporting this feature.

[8bitbubsy]

Originally Posted by leonard

exactly, and this is a PAULA hardware feature ( the player just "write" the new sample address in PAULA registers and the chip will "really" use this new address at the next loop point

Yep, I know. Same for the LEN register.

Originally Posted by leonard

My issue is that players are writing the new sample address during the playback (Audio DMA is still running). And it theory, if you're very unlucky, as the 32bits address write isn't atomic, you may have an audio glitch during 2 samples.

As no one complained for this for years, I guess it's ok in practice. But the
bug is here, in all Amiga players supporting this feature.

Sounds like a rare issue, and I don't think I have personally encountered it. It's still worth supporting sample swapping, imo.