I have some strange bug on some configurations. Renoise sound lags on relatively high cpu usage (The CPU display says ~36-40%, but in reality the Renoise CPU usage displays something around 50 - 60% in task manager). It sounds kinda zipper sound, as there are left small chunks of silence between chunks of audio. Weird thing is that jack actually doesn’t xrun and other jack hosts don’t stop playing. It feels as if renoise returned jack buffers half filled, or is skipping periods or something similar.
The question is, is Renoise using some other timing methods than Jack for timing audio buffers?
Problem may be related to my laptops faulty APIC implementation. The kernel reports IO-APIC not connected to system timer at startup and when the system is running it appears to be running everything on soft IRQ’s.
Also I noticed that RTC IRQ is climbing quite up in process table as renoise is running so I thought maybe it uses RTC for timing and this causes timer drift with audio clock?
After passing noapic argument to kernel, the problem doesn’t seem to occur in that scale for at least. Anyways, thoughts would be appreciated.
Renoise gets its timing (this will be the same in any audio application) via the soundcard (samplerate and the continues sample buffer “polling” from the soundcard/Jack). We do not use any other HW timers for audio. So to me this looks more like a soundcard/driver/setup problem than a problem in Renoise. For some reason the soundcard seems to skip samples from Jack - if there are really no XRuns with Jack.
You are using Renoise for quite a while now, don’t you? Do you always had this problem? Can you remember at which point this started to happen? You seems to have changed quite a lot of system internals (looking at https://forum.renoise.com/t/linux-tips-on-jack-and-realtime-audio/22897)). Maybe something got broken with your optimizations?
To be sure if its a Renoise problem or not, could you try to replicate this problem with other “Jacked” applications? Maybe this only happens when reaching a certain CPU level…
The problem is then possibly related to linux timers.
Most of instability problems with sound came on upgrade to Ubuntu 8.04. And Linux kernel v2.6.24. Before that there wasn’t any problems with period sizes either.
The optimizations basically removed the problem.
Why am I reporting this is that although Renoise is not only program with similar problems, Renoise gets there somehow faster, on lower CPU loads.
Also, interesting if the sound starts to lag, it gets much much worse when I switch away from Renoise window, to a browser window for example.
I will study this a bit further, and see if I can trace for some reasons behind it.
The only reason why Renoise behaves different here is that you either push your system more to the limits with Renoise (compared to simply running a synth for example) or the Muli Core & XRuns Problem in 1.9. I am not really sure what we could/should do here if this is a general problem.
It seems that the problem may be denormals issue on some certain ladspa plugins instead. Not really sure about it, but I haven’t had time to investigate the issues yet.
Seemed to be denormals issue with one of the LADSPA plugins I use. I recompiled all my plugins to use SSE and now I haven’t had any more issues. Also I have tuned and retuned my computer quite a bit, so I can’t really replicate the problem anymore.
