| OvenWerks | wonko: good. I did not put a logging message in to catch this particular problem. You may see extra "device = <your usb device> play: <something> capture: <something> | 00:18 |
|---|---|---|
| wonko | well, at this point zero log entries at all, so I guess that's good then. :) | 00:19 |
| OvenWerks | If you see those in the log file and your setup doesn't quit/lockup/otherwise misbehave, then we have fixed it. | 00:19 |
| wonko | I'll check on it in the morning | 00:19 |
| wonko | I'll keep an eye out for those then | 00:19 |
| wonko | Thanks for fixing that, btw! | 00:20 |
| OvenWerks | your welcome, thanks for pointing it out. | 00:20 |
| === john40 is now known as lyth | ||
| === Unit193 is now known as Guest17120 | ||
| === Guest17120 is now known as Unit193 | ||
| wonko | OvenWerks: so far so good | 11:14 |
| wonko | https://paste.ubuntu.com/p/hXhk3rsrm8/ | 11:14 |
| wonko | oh, that thing in the udevd logs is the wireless mouse receiver, not the audio device. | 11:16 |
| wonko | so unrelated anyway | 11:16 |
| studiobot | KaioHenriquebr was removed by: KaioHenriquebr | 15:28 |
| wonko | OvenWerks: It just happened again. :( | 15:55 |
| OvenWerks | your log from last night shows a lot of xruns | 16:04 |
| OvenWerks | which latency are you using? | 16:05 |
| OvenWerks | and I forget, are you using the lowlatency kernel? | 16:05 |
| OvenWerks | wonko: what does cat /etc/default/rtirq |grep RTIRQ_NAME_LIST show | 16:07 |
| wonko | wonko@deepthought:~/Documents/projects/BattleTech/RTWikiBot$ cat /etc/default/rtirq |grep RTIRQ_NAME_LIST | 16:10 |
| wonko | # RTIRQ_NAME_LIST="rtc snd usb i8042" # old | 16:10 |
| wonko | RTIRQ_NAME_LIST="snd usb i8042" | 16:10 |
| wonko | Linux deepthought 5.0.0-27-lowlatency #28-Ubuntu SMP PREEMPT Tue Aug 20 20:33:37 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux | 16:10 |
| wonko | 128 latency | 16:11 |
| OvenWerks | wonko you use your USB device as jack master? | 16:12 |
| OvenWerks | If so, usb should be before snd | 16:13 |
| OvenWerks | also usb should be changed. | 16:13 |
| OvenWerks | I would guess you have USB mouse and kb? | 16:13 |
| wonko | usb device as jack master | 16:14 |
| wonko | yes on the mouse/kbd | 16:14 |
| OvenWerks | your mouse is at the same (maybe higher priority as your audio. | 16:16 |
| Eickmeyer | That happens a lot if it's a mouse with a high pull rate, like a lot of gaming mice. | 16:17 |
| OvenWerks | wonko: if you type usb-devices do you find that your Sound Device is on the same Bus as you r mouse? | 16:22 |
| OvenWerks | does changing the USB plug you use on your computer change that? (it doesn't on my computer) | 16:22 |
| wonko | mouse: T: Bus=03 Lev=02 Prnt=02 Port=00 Cnt=01 Dev#= 15 Spd=12 MxCh= 0 | 16:24 |
| wonko | audio: T: Bus=03 Lev=01 Prnt=01 Port=01 Cnt=02 Dev#= 13 Spd=480 MxCh= 0 | 16:24 |
| wonko | *everything* is on bus 3 | 16:25 |
| wonko | I only need usb2 for the KA6 | 16:25 |
| wonko | I should see if moving that to one of the usb2 ports would put it on a different bus (currently everything is plugged into the usb3 ports) | 16:25 |
| OvenWerks | Yeah, the new mother boards route all the plugs to one usb bus | 16:25 |
| OvenWerks | That means it is sharing time with whatever else is there. | 16:26 |
| wonko | nope, still on bus 3 | 16:28 |
| wonko | time to go plug the usb pcie card I have in. :) | 16:28 |
| OvenWerks | yeah that may make a difference | 16:28 |
| OvenWerks | I would be interested to know | 16:29 |
| wonko | should I update usb before snd as well? | 16:30 |
| wonko | i still need to find that card and make the time to stick it in | 16:30 |
| wonko | and remember what I did with the screws for the bulkhead | 16:30 |
| OvenWerks | wonko: first you want to get your device on a different usb bus if you can... then do "usb3 snd usb" | 16:31 |
| OvenWerks | assuming usb3 is your device usb...it may not be | 16:31 |
| wonko | it's a usb2 device, so that wouldn't be right | 16:31 |
| OvenWerks | not the same thing. | 16:32 |
| OvenWerks | in this case usb3 is the bus not the type. | 16:32 |
| wonko | oh, ok. | 16:33 |
| * OvenWerks wonders if using a usb 3.0 hub to plug in a USB 2.0 audio device would separate the audio device to another bus. | 16:40 | |
| wonko | I don't think so | 16:45 |
| OvenWerks | A USB3.0 device should get it's own bus... there are 4 buses those other three buses must be there for some reason | 16:47 |
| OvenWerks | A USB3.0 hub should act as a 3.0 to 2.0 device. | 16:48 |
| OvenWerks | that is, it should be a 2.0 host | 16:48 |
| wonko | usb card added | 17:08 |
| wonko | what a colossal pain in the ass | 17:08 |
| wonko | but, now the audio device is on its own bus | 17:08 |
| wonko | OvenWerks: should I tweak rtirq or do you want to see what happens now? | 17:14 |
| OvenWerks | wonko: when you plug your audio device into that plug which usb bus does it end up on? | 17:21 |
| wonko | the new bus5 | 17:24 |
| wonko | http://paste.ubuntu.com/p/M3wK7BT9SW/ | 17:25 |
| wonko | http://paste.ubuntu.com/p/ks3FGGdprb/ | 17:25 |
| OvenWerks | Cool! so change you rtirq line to "usb5 snd usb" | 17:27 |
| wonko | do I need to restart/reboot/etc to get that into effect? | 17:27 |
| OvenWerks | (or whatever just make sure the plain usb is at the end) | 17:27 |
| OvenWerks | yes | 17:27 |
| OvenWerks | you would need to reboot | 17:28 |
| wonko | ok | 17:28 |
| OvenWerks | you can run the rtirq again but it does not remove priority | 17:28 |
| OvenWerks | wonko: what does /etc/init.d/rtirq status | 17:48 |
| OvenWerks | show | 17:48 |
| wonko | lots. :) | 18:03 |
| wonko | http://paste.ubuntu.com/p/KHm2FvvVSj/ | 18:04 |
| OvenWerks | wonko: That still shows snd on top | 18:09 |
| wonko | RTIRQ_NAME_LIST="usb5 snd usb i8042" | 18:09 |
| wonko | is what I put in /etc/default/rtirq | 18:10 |
| OvenWerks | OK, so the usb5 trick doesn't work any more | 18:10 |
| OvenWerks | we need to find out which of those irq goes to your audio device | 18:12 |
| OvenWerks | (and wonder if it will stay there) | 18:13 |
| OvenWerks | Mind ifirq18/38 is your card, that would be ok too | 18:15 |
| OvenWerks | wonko: do you get less xruns? | 18:15 |
| wonko | how do I even check? I don't know without Claudia. :) | 18:15 |
| wonko | just check the log file? | 18:15 |
| wonko | is that the state = Triggered thing? | 18:16 |
| OvenWerks | Huh, something else I should add to controls I guess | 18:16 |
| wonko | that would be nice. :) | 18:16 |
| OvenWerks | it should show client not finished for an xrun | 18:16 |
| OvenWerks | adding xruns would mean a redesign of a good portion of things | 18:17 |
| wonko | http://paste.ubuntu.com/p/bXJKDzZXH3/ | 18:17 |
| OvenWerks | hmm | 18:22 |
| OvenWerks | I would guess we are not quite there | 18:23 |
| OvenWerks | rtirq does nto understand usb5. | 18:23 |
| OvenWerks | wonko: there is another script that does similar to rtirq that may work better for you but it does not come as a package | 18:30 |
| wonko | I'm ok with that | 18:31 |
| wonko | Is any of this still valid: https://wiki.linuxaudio.org/wiki/system_configuration | 18:31 |
| OvenWerks | some of that is out of date. | 18:32 |
| wonko | I ran that script and this is what "failed" | 18:32 |
| wonko | Checking access to the high precision event timer... not readable - not good | 18:33 |
| wonko | Checking access to the real-time clock... not readable - not good | 18:33 |
| wonko | Kernel with Real-Time Preemption... not found - not good | 18:33 |
| wonko | the rest all checks out | 18:33 |
| wonko | anyway, bbiab, need to get the kid from the bus stop | 18:33 |
| OvenWerks | however, it does remind me of something... in -controls on the system tab | 18:33 |
| OvenWerks | (system Tweaks) | 18:33 |
| OvenWerks | make sure CPU governor is set to performance and boost (if it is changable) to off. | 18:34 |
| OvenWerks | https://github.com/jhernberg/udev-rtirq | 18:37 |
| OvenWerks | wonko: ^^ | 18:38 |
| wonko | first thing I did was set performance and boost off | 18:46 |
| wonko | so you guys aren't enabling threadirqs? | 18:47 |
| OvenWerks | I don't know | 18:49 |
| OvenWerks | something to find out | 18:49 |
| wonko | I don't see that you are, so I've added that | 18:49 |
| OvenWerks | cool where? | 18:50 |
| wonko | I changed GRUB_CMDLINE_LINUX_DEFAULT="quiet splash" to GRUB_CMDLINE_LINUX_DEFAULT="threadirqs quiet splash" in /etc/default/grub | 18:50 |
| wonko | so looks like another reboot. :) | 18:50 |
| OvenWerks | thank you | 18:50 |
| OvenWerks | yes | 18:50 |
| wonko | ok | 19:09 |
| wonko | http://paste.ubuntu.com/p/Q4x2BpRqfb/ | 19:10 |
| wonko | no xruns so far | 19:11 |
| wonko | just got a couple | 19:21 |
| OvenWerks | wonko: after a few reboots and so far no sign I actually changed the priority of anything by plugging in a usb device... | 19:28 |
| OvenWerks | I have found that I can turn xhci off in bios after which my usb devices show up on different usb buses | 19:29 |
| wonko | hmm | 19:29 |
| wonko | interesting | 19:29 |
| wonko | so the bus is virtualized by xhci | 19:29 |
| OvenWerks | where everything was on usb3 before, now my mouse is on usb2 and my audio is on usb1 | 19:29 |
| wonko | Ok better | 19:48 |
| wonko | now Quassel-Core won't restart every time I reboot my computer | 19:48 |
| wonko | oh no! | 19:49 |
| wonko | jack server restarted. :( | 19:49 |
| wonko | OvenWerks: Who wrote Carla? you guys? | 20:03 |
| OvenWerks | no falktx did | 20:03 |
| OvenWerks | same as all the kxstudio stuff | 20:03 |
| wonko | It would be nice if it reconnected to jack when it came back. Minor annoyance though, nothing serious | 20:04 |
| OvenWerks | I haven't found any application that does that... | 20:04 |
| OvenWerks | The only way to do that I know of is to keep trying to reconnect and fail till it comes back | 20:06 |
| OvenWerks | bugging jackdbus via dbus may work too, but seems not too rliable. | 20:06 |
| wonko | Ah well, would have been nice, but again, not a major issue | 20:18 |
| wonko | major issue being that jack crashed even with all this new stuff going on. :( | 20:20 |
| wonko | xruns being way down is awesome though | 20:23 |
| OvenWerks | The udev-rtirq doesn't seem to work for me | 20:31 |
| OvenWerks | I was able to more my usb irq by hand: http://paste.ubuntu.com/p/RHGxDRd7s7/ | 20:33 |
| OvenWerks | irq20 is my USB1 and irq23 is sub2 | 20:34 |
| wonko | how do I find out which irq maps to what? | 20:35 |
| wonko | I mean, xruns are *way* down, so that's good | 20:37 |
| OvenWerks | when I did: cat /proc/interrupts I got 20: 2810586 0 30 0 IO-APIC 20-fasteoi ehci_hcd:usb1 | 20:37 |
| wonko | so maybe it's not super important and is "good enough"? | 20:37 |
| wonko | I have 56 cores. That's some ugly output. :) | 20:38 |
| OvenWerks | I guess it depends on what you are doing. | 20:38 |
| OvenWerks | 56 cores? that may be the problem | 20:38 |
| OvenWerks | The low latency tests I have seen tend to favour fewer cores | 20:39 |
| OvenWerks | Also hyperthreading will raise the lowest latency | 20:39 |
| wonko | Well then I'll be happy with the results I have now. :) | 20:39 |
| wonko | I don't even see usb5 in /proc/interrupts | 20:40 |
| OvenWerks | I purposely bought an i5 with 4 cores and no hyperthreads rather than the i7 | 20:40 |
| wonko | irq19 is usb1 and usb3 | 20:40 |
| wonko | irq18 is usb2 | 20:41 |
| OvenWerks | one of the ehci/xhci must be your pcie card | 20:42 |
| OvenWerks | I would say one of the 4* ones | 20:43 |
| OvenWerks | 18 and 38 will be linked and 19/39 too | 20:44 |
| wonko | can I expect zero xruns to be a reasonable goal or is 5 every couple of minutes not terrible | 20:44 |
| OvenWerks | I don't know, try going from 128 to 256 for buffer size | 20:45 |
| OvenWerks | I am pretty sure there some more tricks | 20:47 |
| OvenWerks | I do know that I can do better with what I have. I haven't done a lot of testing lately so I don't know what newer kernels and even this MB are good for. I used to be able to get down to a buffer size 16 with no xruns for hours. | 20:49 |
| OvenWerks | but that was with a pci audio device | 20:49 |
| OvenWerks | You can see that have a whole lot shorter irq list too. All of those megasas and ens1 things you have listed must be one per core. | 20:52 |
| OvenWerks | tying you irq for your usb device along with your audio chain to one core might do wonders too. I know of people who have gotten their buffer size to less than 64 (32/3 I think) even with a usb device | 20:53 |
| OvenWerks | wonko: I wonder if the webcam is causing any odd stuff. | 21:23 |
| wonko | Yeah, let's not get crazy just yet. Mic should be here Monday. Then I'll stop using the webcam | 21:23 |
| wonko | Then we'll see what happens | 21:23 |
| wonko | And leave CPU pining for last. 😁 | 21:24 |
| OvenWerks | maybe the webcam doesn't like 64 for a buffer | 21:24 |
| OvenWerks | cpu pinning is a task and a half I think | 21:24 |
| OvenWerks | And I don't know if it would help anyway. | 21:24 |
| OvenWerks | What I do know is that you seem to have 128 extra irqs that most cpus don't have | 21:25 |
| OvenWerks | I don't know how often they have to tale to each other, but that may be enough to do things. | 21:26 |
| OvenWerks | s/tale/talk/ | 21:26 |
| wonko | The joys of not having desktop CPUs | 21:37 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!