sound buffer issue since 3.3 beta 18

[vagrant]

While recording a musicdisk today I noticed some tiny audio 'clicks' in playback. I examine the waveform and audio is actually dropping out intermittently for a very short period of time (between 2 and 10ms).
It happens all the time, not just while recording, and is only audio that is affected - there's no video frame loss.
Anyway, after a lot of testing I can absolutely say it only affect 3.3 beta 18 and the latest public release.
SND does not seem to behave any differently, although I noticed CPU is getting higher spikes since beta 18 (about 10-15%). Could the new API have something to do with it?

[Toni Wilen]

Only change is Logitech LCD. Does same happen if you unplug your LCD keyboard?

[vagrant]

Ok, I plugged in an old ps/2 keyboard - cpu usage still spikes higher than prior to beta 18 but I couldn't hear any audio dropouts.
I plug the logitech in and the audio dropouts are almost immediately noticeable.

[vagrant]

Damn, this morning I realize it happens on beta 17 to, but extremely rare compared to beta 18+.
Sorry I'll keep switching back through betas until I find the exact release it starts.

[Toni Wilen]

Does http://www.winuae.net/files/b/winuae.7z work better? (LCD updating stuff moved to separate thread, if it was the cause for slowdown)

[vagrant]

Yep seems to be fixed now thanks toni
CPU usage no longer spikes & no audio dropouts.
Another thing I forgot to mention was in previous release, the drive access led's on the lcd display was flashing randomly if a disk image was inserted & inactive.. this is also fixed now

Hi folks,

It's been so long since I posted here that I had to recreate my forum account (!), but for what it's worth I had the exact same symptoms with 3.3.0 until I replaced the main executable with that test version. Almost imperceptible gaps in the audio with the release version, smooth as silk with the beta.

I don't have a Logitech keyboard but I do use a Logitech G602 wireless gaming mouse, which will almost certainly share some of the same drivers and/or support software.

Update: just before posting this I updated the Logitech Gaming Software to the latest version (8.83.85) and re-booted. Sound is now perfect in both the beta and release versions of WinUAE, so it definitely looks as though there was some sort of incompatibility there.

I've no idea whether any of this is useful to anyone, but either way I'd like to say thank you for the thread which pointed me in the right direction. Without it there's absolutely no way I'd have suspected mouse support software to be responsible for an audio glitch.

[vagrant]

Toni, long story short the sound problem has still been happening for me.. I did not want to mention until I could accurately determine which beta it has started from.
The above file did seem to fix the problem at first, but later I realize it greatly improved the instance of it happening - but not completely remove it.
It has been hard to determine as with some demos/games it happens rarely, yet more often in others.

So I believe something in beta 9 cause this issue!
I've been running beta 8 for a few weeks now and it has not happened once.. as soon as I switch to beta 9 or newer, it does not take long for the 'dropouts' to return

Beta 9:
- Fixed swapped audio channels if stereo separation was set (b1 cleanup missed code path)
- Fixed uaehf.device SCSI emulation sense data possible crash/wrong status. (trap update related)
- 32M UAE RTG board was always address space aligned. It is wrong in UAE mapping mode.
- Only swallow hardware horizontal and vertical mouse movement when in mouse driver mode, not mouse wheel.
- Improved programmed refresh rate hsync start and end validation to prevent crashes if weird mode is set. (BoggleDop / Sanity)
- Added hardfile virtual bad block support. uaehf.device and hardware SCSI emulation only. Create geometry text file (<name of hardfile.hdf>.geo) in same directory as hardfile, add badblocks=<blocks> line. (<blocks> can be range, 10-100 or single value or multiple ranges/values, for example 10-100,200,1000-20000) Max 8 bad block values/ranges supported. Error is returned if block in badblock range is attempted to be read or written.
- OS4.x UAE trap mode now uses hybrid busy wait mode. Performance has improved again, directory filesystem speed is now fully usable and uaegfx is also noticeably faster.
- UAE indirect trap stability updates.