So this is interesting, I updated from 10.13.6 to 10.14.6. The performance while playing this song is much better now. Max cpu @48khz 8ms is 48%. But still, if I disable PDC, the usage goes down to 34%.
So why is PDC causing such a CPU load with this example song?
EDIT: Or the more correct question is: Why is PDC causing any additional CPU usage at all here? And why it is causing audio stuttering? I think due a bug, it’s is somehow recalculated/realigned all the time.
@taktik I hope that you aware of this pretty drastical performance issue, too? I would be interested here about some technical details, if this is possible for you to share Why is the performance dropping at all here, if PDC is enabled? Since I can’t find a reason why it needs to be calculcated in realtime or so… Is the formula device the problematic part in here?
Finally found the problem here. This crash indeed only happend on Linux and was caused by GCC’s optimizer. Good to have this fixed, as this could have crashes in other situations as well (again, on Linux only).
This Song is a very extreme example - in many regards.
PDC does introduce a little CPU overhead, because it does add delay-lines into various parts of the audio streams in order to actively delay them. Those delay-lines don’t introduce much overhead on thir own, but when there are tons of them this can be noticeably.
No crashes here on win but a significant cpu difference (not in the benefit of 3.5 )
On previous version of Renoise 3.4.2 uses 30%cpu , while the latest 3.5version uses 45-50% +drop-outs, it uses 15-20 % more cpu ??
Weird considering renoise 3.5 had been optimised ,…?
Damn , you must have a maxed out machine
And in the previous version of renoise ?
Here latest renoise consumes a lot more cpu , even when the song is idling
Have had 3.4 uninstalled for ages, but I assume it is less cpu intensive if it is on your part. There have been cpu meter breakdowns on this forum before, I forgot the specifics, but the meter in Renoise isn’t the end all be all.
True , but I can increase the lpb up to 200 wihout dropouts on 3.4.2 , 3.5 can’t cope with it .
So there has to be something why 3.5 performs worse with this specific example file .
Taktik hinted that the huge cpu consumption was due the oversampling and pdc ,
Truth be said , that example song is extreme
Yep, on Linux I believe it also uses quite some more CPU since the 3.5 release…
Keep in mind this is an absurd example. Basically stress-testing the meta modulation chains - look at the instrument, they do actual audio rate (though low rate, like 2-8khz sampling rate or so depending on tpl…) FM synthesis with formula devices, then modulating like gainers or compressor thresholds (can’t remember RN) to make the sounds. It works by running at insane high tick speed rate with cranked up lpb/tpl, and that’s probably what bogs down the renoise engine so much.
Oh I’m a fucking Idiot , pdc was NOT enabled on 3.4.2 while comparing
Enabling it adds 15-20 % of cpu increase
Cpu usage is roughly the same on both , occasional dropouts ( not the latest greatest laptop over here )
Ah I also forgot to disable PDC in new version. Some time since I made the demo. Without it’s better. Still after some time messing with the parameters, renoise crashed hard mode (exception audio slave). Anyone want logs?
Why is the performance dropping at all here, if PDC is enabled? Since I can’t find a reason why it needs to be calculcated in realtime or so… Is the formula device the problematic part in here?
