Go Back   Cockos Incorporated Forums > REAPER Forums > REAPER Bug Reports

Reply
 
Thread Tools Display Modes
Old 08-18-2020, 03:05 AM   #1
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default Master/parent vs. regular sends, RT CPU anomaly

Copying this from this thread: https://forum.cockos.com/showthread.php?t=236759

Test projects attached below.

So, as far as I can tell, the problem SEEMS to me to lie with what the Reaper performance meter calls "RT CPU" and "RT longest block". Not using the master/parent send gives orders of magnitude lower measurements on my system. To be clear, this is NOT related to having "Allow live FX processing" turned on, since on my system it is turned OFF.

I'm ONLY using Reacomp for these tests, with a 30ms pre-comp setting, no other plugins are present.

My latest test run seems to indicate that using layers of nested folders for organizing is not problematic UNLESS you use the intended master/parent sends on child tracks, which seems weird to me. Being able to use the intended master/parent sends would be much more elegant and intuitive, and it would be much better just for workflow reasons.

Using master/parent sends = 120 % RT CPU, 3441 ms RT longest block
NOT using master/parent sends = 5-10 % RT CPU, 1.43 ms RT longest block

"Normal" nested folders, master/parent sends are used





Nested folders, master/parent sends are NOT used, track sends used instead





Not using nested folders, master/parent sends are NOT used


Attached Files
File Type: rpp test routing folders normal.rpp (48.8 KB, 206 views)
File Type: rpp test routing folder organize.rpp (49.3 KB, 201 views)
File Type: rpp test routing no folders.rpp (49.2 KB, 202 views)
ramses is offline   Reply With Quote
Old 08-18-2020, 04:39 AM   #2
beingmf
Human being with feelings
 
beingmf's Avatar
 
Join Date: Jul 2007
Location: Jazz City
Posts: 5,064
Default

Whoa, that's more than impressive!

Is there a script to facilitate the tedious send routine in such a scenario?
Like "Untick nested children folders' master/parent send and create sends to the respective parent track"?
__________________
Windows 10x64 | AMD Ryzen 3700X | ATI FirePro 2100 | Marian Seraph AD2, 4.3.8 | Yamaha Steinberg MR816x
"If I can hear well, then everything I do is right" (Allen Sides)
beingmf is online now   Reply With Quote
Old 08-18-2020, 04:53 AM   #3
Coachz
Human being with feelings
 
Coachz's Avatar
 
Join Date: Oct 2010
Location: Charleston, SC
Posts: 12,762
Default

And what is the Practical use case for nested folders 10 levels down? It's great to chase windmills but it's not worth the developers time if there's no practical application
Coachz is offline   Reply With Quote
Old 08-18-2020, 05:34 AM   #4
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Quote:
Originally Posted by Coachz View Post
And what is the Practical use case for nested folders 10 levels down? It's great to chase windmills but it's not worth the developers time if there's no practical application
If you read the original thread you will see that this is not a matter of "no practical application", it has real life consequences.

And, as I have shown, this is INDEPENDENT of how you choose to organize your projects. This issue relates to ALL master/parent sends (perhaps apart from master/parent sends going straight to the master), regardless of how you choose to organize your projects. Agreed, it is a cumulative effect, but that's beside the point. All CPU consumption effects are cumulative, and this behaviour is worse for those with less than hyper-modern machines, I would reckon.

And, I might add, the current behaviour makes little sense to me. If I can organize project using folders either way, 10 folders down or not, using or not using master/parent sends, it seems like there's no reason for the existing behaviour being like it is. It seems to me like something about how master/parent sends work is simply not right. That's why I posted this here.

Last edited by ramses; 08-18-2020 at 05:49 AM.
ramses is offline   Reply With Quote
Old 08-18-2020, 05:58 AM   #5
Coachz
Human being with feelings
 
Coachz's Avatar
 
Join Date: Oct 2010
Location: Charleston, SC
Posts: 12,762
Default

I routinely use 2 levels of folders for example guitars and below that, folders for acoustic and electric guitars. With 50 to 100 tracks in the project, many plugins and armed tracks it is easy for me to get RT CPU over 50 resulting in skips and clicks. I have ended up freezing every track that has a plugin to get around this. Maybe more investigation could be worthwhile because I found it odd I have to freeze to be reliable.

Windows 10, 12gb ram, i980x CPU, rme 9652 audio card, ssd OS C: and ssd D: sample / project drive
Coachz is offline   Reply With Quote
Old 08-18-2020, 12:26 PM   #6
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Oh, perhaps I should give instructions on how to reproduce.

Put your soundcard at the lowest latency setting. Using the attached project "test routing folders normal.rpp", simply put enough plugins on all tracks to "break" the project. I've used Reacomp with a 30ms lookahead, but you could potentially use any latency prone, cpu-hungry plugin. Make sure audio is really broken, stuttering and becoming really sluggish when stopping/playing for example. Note the plugins you've used and instantiate the same exact plugins on the tracks of the two other projects. Please, go ahead and post your findings, it's not clear that this bug will reproduce across all systems.

I'm on win10, 64 bit.
ramses is offline   Reply With Quote
Old 08-19-2020, 05:15 AM   #7
mozart999uk
Human being with feelings
 
Join Date: Nov 2010
Posts: 1,710
Default

I think this thread here (from post 16 onwards) reports similar issues:

https://vi-control.net/community/thr...-reaper.79077/
mozart999uk is offline   Reply With Quote
Old 08-19-2020, 09:18 AM   #8
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Quote:
Originally Posted by mozart999uk View Post
I think this thread here (from post 16 onwards) reports similar issues:

https://vi-control.net/community/thr...-reaper.79077/
Interesting read. Plugin latency in combination with master/parent sends seem to be the culprit, according to this one poster there:

"After much head scratching, I found that having a plugin with any amount of internal latency on a send that was then routed to its parent via the master-send option caused MASSIVE, unjustified UI performance problems. If I removed the plugin or routed the track to its parent via a discrete send instead of the "master" button, everything worked as expected, even in large projects."
ramses is offline   Reply With Quote
Old 08-19-2020, 10:26 PM   #9
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Here's another thread discussing similar issues:

https://forum.cockos.com/showthread.php?t=188157
ramses is offline   Reply With Quote
Old 08-20-2020, 02:24 AM   #10
mozart999uk
Human being with feelings
 
Join Date: Nov 2010
Posts: 1,710
Default

I've attached my results.

Ryzen 3950x RME AIO at 32 samples. Windows 10 x64.

Interestingly my "folders normal" was fine for the first 2 playbacks of the loop (the loop was about 10 seconds long) but then just broke up with no sound except regular pop on the first beat of every bar.
Attached Images
File Type: png Folders normal playback after two cycles.png (28.6 KB, 218 views)
File Type: png No folders playback.png (36.4 KB, 218 views)
File Type: png Folder organise playback.png (35.9 KB, 196 views)
mozart999uk is offline   Reply With Quote
Old 08-20-2020, 07:03 AM   #11
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Quote:
Originally Posted by mozart999uk View Post
I've attached my results.

Ryzen 3950x RME AIO at 32 samples. Windows 10 x64.

Interestingly my "folders normal" was fine for the first 2 playbacks of the loop (the loop was about 10 seconds long) but then just broke up with no sound except regular pop on the first beat of every bar.
Interesting. Thanks for testing!
ramses is offline   Reply With Quote
Old 08-20-2020, 07:11 AM   #12
mozart999uk
Human being with feelings
 
Join Date: Nov 2010
Posts: 1,710
Default

Pleasure. Can't believe how much lower the usage is without parent sends and folders....
mozart999uk is offline   Reply With Quote
Old 08-20-2020, 09:50 AM   #13
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Quote:
Originally Posted by mozart999uk View Post
Pleasure. Can't believe how much lower the usage is without parent sends and folders....
Yeah, it seems weird right? I'm fairly certain it's not meant to be this way.
ramses is offline   Reply With Quote
Old 01-25-2021, 04:10 AM   #14
mozart999uk
Human being with feelings
 
Join Date: Nov 2010
Posts: 1,710
Default

Just out of interest, have you done any more testing on this?
mozart999uk is offline   Reply With Quote
Old 01-25-2021, 06:36 AM   #15
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Quote:
Originally Posted by mozart999uk View Post
Just out of interest, have you done any more testing on this?
No, but I'm planning to make another test run with the new latency compensation settings. I'll post the results here when I'm done, probably this week.
ramses is offline   Reply With Quote
Old 01-25-2021, 07:00 AM   #16
mozart999uk
Human being with feelings
 
Join Date: Nov 2010
Posts: 1,710
Default

Quote:
Originally Posted by ramses View Post
No, but I'm planning to make another test run with the new latency compensation settings. I'll post the results here when I'm done, probably this week.
Looking forward to seeing what you find out :-)
mozart999uk is offline   Reply With Quote
Old 05-03-2021, 08:13 AM   #17
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Related post: https://forum.cockos.com/showthread....46#post2439846
ramses is offline   Reply With Quote
Old 05-08-2021, 09:49 AM   #18
wallace
Human being with feelings
 
Join Date: Jan 2009
Location: UK
Posts: 594
Default

Thanks for posting this.
At least I now know it’s not just happening with my old clunker of a machine.
wallace is offline   Reply With Quote
Old 09-23-2021, 02:21 PM   #19
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Related thread: https://forum.cockos.com/showthread.php?t=188157
ramses is offline   Reply With Quote
Old 11-27-2021, 08:40 AM   #20
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Possibly related thread: https://forum.cockos.com/showthread.php?t=232901
ramses is offline   Reply With Quote
Old 11-28-2021, 02:54 AM   #21
KuulArt
Human being with feelings
 
Join Date: Apr 2015
Posts: 82
Default

I think I have been affected by this issue as well. (or a very similar one) I am running a decent Ryzen3950x machine and on regular playback it works well.
The issue is massive spikes of RT CPU load on playback start/stop or cursor move. This seems to go away if I disable my mixbus chain. I have made a screen recording showing the issue. It is a regular medium sized project that I would be running that exhibits this issue.

https://www.dropbox.com/s/u6qtn00bjt..._lag.gif?raw=1

In the video you can see that when I move cursor the RT CPU jumps to 400% but when the actual audio is running it goes down to about 30%. And similar things happen when I press play or stop.

This has annoyed me so much that I am considering moving away from REAPER for time being. Its a real workflow killer.

Last edited by KuulArt; 11-28-2021 at 04:16 AM.
KuulArt is offline   Reply With Quote
Old 11-28-2021, 06:26 AM   #22
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

0
Quote:
Originally Posted by KuulArt View Post
I think I have been affected by this issue as well. (or a very similar one) I am running a decent Ryzen3950x machine and on regular playback it works well.
The issue is massive spikes of RT CPU load on playback start/stop or cursor move. This seems to go away if I disable my mixbus chain. I have made a screen recording showing the issue. It is a regular medium sized project that I would be running that exhibits this issue.

https://www.dropbox.com/s/u6qtn00bjt..._lag.gif?raw=1

In the video you can see that when I move cursor the RT CPU jumps to 400% but when the actual audio is running it goes down to about 30%. And similar things happen when I press play or stop.

This has annoyed me so much that I am considering moving away from REAPER for time being. Its a real workflow killer.
If you have the opportunity, could you please try the following. Instead of routing directly to the Master, create a "dummy master" track and route everything there first, using regular track sends instead of master/parent sends. Also, put all your mixbus fx on the dummy master instead. And please report back here if you do.
ramses is offline   Reply With Quote
Old 11-29-2021, 04:03 AM   #23
mozart999uk
Human being with feelings
 
Join Date: Nov 2010
Posts: 1,710
Default

I don't do much high track count audio mixing any more and have never done any in reaper. I use mainly a mix of VSTi's and maybe 5-10 tracks of audio max per project.

Reaper has been rock stable.

HOWEVER, the other day someone brought round their project all in audio (bout 20-30 tracks I think) and after I started adding plugins to various tracks, Reaper was doing this "lag" thing. It was really annoying and embarrassing to be honest.

I looked at the project again after he left and I noticed that I had a few high latency plugins and after I replaced those it became much smoother when seeking.

The problem (like for many users) is it didn't happen repeatedly with the same parameters so I can't really fill in a bug report......

Sometimes opening a plugin gui was fine, other times it was a right royal PITA with a 5-10s wait for reaper to become responsive again.

When I have time I'll change my template to include a submaster track and see what happens.....
mozart999uk is offline   Reply With Quote
Old 01-24-2022, 05:00 AM   #24
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Possibly related thread: https://forum.cockos.com/showthread....10#post2519710
ramses is offline   Reply With Quote
Old 02-07-2022, 03:14 AM   #25
brk303
Human being with feelings
 
Join Date: Nov 2016
Location: Serbia
Posts: 172
Default

Quote:
Originally Posted by ramses View Post
30ms pre-comp setting, no other plugins are present.
Pre-comp causes latency/PDC, can you reproduce the performance difference with no latency ? If you set pre-comp to 0 does it still behave the same ?
brk303 is offline   Reply With Quote
Old 02-07-2022, 10:47 AM   #26
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Quote:
Originally Posted by brk303 View Post
Pre-comp causes latency/PDC, can you reproduce the performance difference with no latency ? If you set pre-comp to 0 does it still behave the same ?
I'm not sure... Perhaps someone could try to reproduce the issue with 0 latency plugins...
ramses is offline   Reply With Quote
Old 02-07-2022, 11:26 AM   #27
brk303
Human being with feelings
 
Join Date: Nov 2016
Location: Serbia
Posts: 172
Default

Quote:
Originally Posted by ramses View Post
I'm not sure... Perhaps someone could try to reproduce the issue with 0 latency plugins...
I did and saw no difference in performance between folders vs sends.
brk303 is offline   Reply With Quote
Old 02-07-2022, 11:07 PM   #28
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Quote:
Originally Posted by brk303 View Post
I did and saw no difference in performance between folders vs sends.
And you managed to bring the system to the breaking point using plugins that don't introduce any latency? Stuttering, glitching out, that sort of thing?
ramses is offline   Reply With Quote
Old 02-08-2022, 06:08 PM   #29
brk303
Human being with feelings
 
Join Date: Nov 2016
Location: Serbia
Posts: 172
Default

Quote:
Originally Posted by ramses View Post
And you managed to bring the system to the breaking point using plugins that don't introduce any latency? Stuttering, glitching out, that sort of thing?
I used a heavier plugin instead of ReaComp (that has zero latency)
I record armed tracks to disable anticipative FX, and got RT CPU to 70%.
And both projects had RT CPU at 70% with same plugins just different routing
brk303 is offline   Reply With Quote
Old 02-09-2022, 12:04 AM   #30
Draiden
Human being with feelings
 
Join Date: Feb 2022
Posts: 2
Default

i got another one. when i move my mastering chain from the master to a regular track in the project, my rt cpu goes to 11000% (no typo, eleven thousand) and ram use is 45gb of 32gb. needless to say that reaper never responded after i did so. i will try rearranging my (very big) project to not use master/parent sends or folders. that might take a couple of days though.
__________________
Worshipping Reaper since 2006.
WIN10, AMD Ryzen 3600, Gigabyte B450M S2H, Crucial Ballistix DDR4 32GB, MSI GTX1050TI, NVMe OS Drive 512GB, RME RayDat PCIe, Focusrite Scarlett 18i20 2nd Gen
Draiden is offline   Reply With Quote
Old 02-09-2022, 01:27 AM   #31
ramses
Human being with feelings
 
Join Date: Jul 2009
Posts: 1,231
Default

Quote:
Originally Posted by brk303 View Post
I used a heavier plugin instead of ReaComp (that has zero latency)
I record armed tracks to disable anticipative FX, and got RT CPU to 70%.
And both projects had RT CPU at 70% with same plugins just different routing
Ok. Yeah, it might not make any difference when using plugins that don't introduce any latency. What if you really push the projects further? And what happens if you enable anticipative FX?
ramses is offline   Reply With Quote
Old 02-10-2022, 06:06 AM   #32
brk303
Human being with feelings
 
Join Date: Nov 2016
Location: Serbia
Posts: 172
Default

Quote:
Originally Posted by ramses View Post
Ok. Yeah, it might not make any difference when using plugins that don't introduce any latency. What if you really push the projects further? And what happens if you enable anticipative FX?
Doesn't matter how much I punish the CPU, it works the same. Only thing that seems to affect performance is with plugins that introduce latency.
brk303 is offline   Reply With Quote
Old 03-20-2022, 05:40 AM   #33
KMJoshi
Human being with feelings
 
Join Date: Sep 2019
Location: Kathmandu, Nepal
Posts: 41
Default Same experience!

Using plugins with heavier latency on folder tracks definitely causes issues. Using the same plugin on individual tracks gives no problem.

A couple of days ago, I was working on a project that had about 6 virtual instruments playing together to create a percussion section. They all had some EQ and other basic stuff individually. However, I added another EQ, compressor, clipper, and reverb on the folder track. That caused a LOT of glitches and crackling. So, I just moved the additional EQ and reverb to the individual tracks, and problem solved.

Clearly, there's something going on with the real-time processing. I really hope we get some sort of solution.
__________________
Articulation mapping, please!
KMJoshi is offline   Reply With Quote
Old 03-23-2022, 05:15 AM   #34
ScuzzyEye
Human being with feelings
 
ScuzzyEye's Avatar
 
Join Date: Apr 2021
Posts: 419
Default

Interestingly this reminded me of a problem I had, that while almost entirely different, was dependent on how much latency was being introduced by a plugin chain.

https://forum.cockos.com/showthread.php?t=262005

My problem was ultimately I had changed thread priority setting, and changing it back to the default solved it. But what seemed to be happening was once the latency was high, plus a certain amount of CPU usage, Reaper stopped handling the user interactive parts of the program in a timely manner.

So if these are related it would seem that once latency gets high, scheduling of some time related threads isn't happening often enough.


A question though. Has anyone experienced this on a Mac? I don't know how much of this scheduling is Reaper selecting when to run things, or the OS's scheduler. The setting I changed I believe was related to how Reaper was tagging processes for the OS to consider running. MacOS will have a completely different scheduler algorithm than Windows. I have access to both Mac and Linux machines. If no one else has tried to reproduce the problem on them, I can give it a shot.

Last edited by ScuzzyEye; 03-23-2022 at 05:23 AM.
ScuzzyEye is offline   Reply With Quote
Old 03-29-2022, 05:30 AM   #35
fricia
Human being with feelings
 
Join Date: Apr 2017
Posts: 119
Default

Quote:
Originally Posted by ramses View Post
Interesting read. Plugin latency in combination with master/parent sends seem to be the culprit, according to this one poster there:

"After much head scratching, I found that having a plugin with any amount of internal latency on a send that was then routed to its parent via the master-send option caused MASSIVE, unjustified UI performance problems. If I removed the plugin or routed the track to its parent via a discrete send instead of the "master" button, everything worked as expected, even in large projects."
I'm observing this exact problem as well! But, as stated in other posts, audio performance is massively affected as well. The more latency a plugin introduces, the more pronounced the effect seems. But it's somehow also tied to how cpu hungry the plugin is and how much load there is on the system in general. To br honest, I can't really make sense of my testing results except that something seems not right! Which we of course already know

Looking at the performance meters I observe the same as others: RT CPU load drops by several orders of magnitued, when tracks are routed to the the parent via a regular send and not via the parent send checkbox.

Would be amazing if this could be looked at by justin or schwa!

Edit: With Anticipative FX Processing turned off there is no difference between regualar send and parent send checkbox.

Last edited by fricia; 03-29-2022 at 06:00 AM.
fricia is online now   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -7. The time now is 02:50 AM.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.