| zequence | len-1304: I was able to upload a lowlatency kernel to my ppa ppa:zequence/kernel-testing | 01:59 |
|---|---|---|
| zequence | It has some stuff disabled | 01:59 |
| zequence | debug stuff mostly | 02:00 |
| zequence | 3.8 for raring | 02:00 |
| zequence | There's no meta, just the image. linux-image-3.8.0-6.4-shaved1-lowlatency | 02:03 |
| === Lump|AFK is now known as Lumpy | ||
| zequence | no headers, so no proprietary drivers | 02:05 |
| zequence | hmm, I think this kernel must really suck for lowlatency | 02:13 |
| Lumpy | heyas | 02:16 |
| zequence | Lumpy: | 02:22 |
| zequence | hi | 02:22 |
| zequence | len-1304: The raring kernel is not performing well for me at all | 02:22 |
| zequence | My "shaved off" version did not have any effect on performance of what I can see. Did cyclictest and some audio testing | 02:23 |
| zequence | This is AMD, so disabling Intel stuff would of course not affect anything | 02:23 |
| zequence | cyclictest shows something like 10x more than one would expect from a well configure realtime kernel | 02:24 |
| zequence | results were far better on Quantal | 02:24 |
| zequence | I need to retest. In fact, I should do this properly and just post the results on ubuntustudio-testing | 02:25 |
| len-1304 | zequence, I am using Linux studio1304 3.8.0-6-lowlatency #4-Ubuntu SMP PREEMPT Tue Feb 12 12:57:58 UTC 2013 i686 i686 i686 GNU/Linux | 05:06 |
| len-1304 | I have not had any problems | 05:06 |
| len-1304 | zequence, I guess I should ask what you mean by performance... | 05:11 |
| len-1304 | To me it means better audio performance. Lower latency without xruns. Greater stability at higher cpu loads. | 05:13 |
| Lumpy | heya len-1304 | 06:16 |
| Lumpy | for what it is worth, linux means better performance | 06:17 |
| Lumpy | :) | 06:17 |
| Lumpy | i get xruns but listeners never seem to notice it | 06:17 |
| Lumpy | nini all | 07:17 |
| === Lumpy is now known as Lump|AFK | ||
| zequence | len-1304: I'm getting really poor results from cyclictest as well as just running jackd with a software | 08:25 |
| zequence | Just running jack is not enough | 08:26 |
| zequence | The goal is to have jack run more or less stable at 32 frames/period with software and cpu load | 08:27 |
| zequence | And more or less perfect at 64 frames/preiod | 08:27 |
| zequence | this is what we had on Natty | 08:27 |
| zequence | With cpu load, I include jerking around with graphics and doing something like decompressing files using all cores | 08:29 |
| zequence | or compiling something | 08:29 |
| zequence | if it's realtime, nothing will budge the process | 08:34 |
| zequence | or should | 08:34 |
| zequence | now, since i've done all too little testing my results are not enough to talk about really. Other than that these were extremely poor | 08:35 |
| zequence | I had better results on Quantal | 08:35 |
| zequence | more than twice as good on cyclictest | 08:35 |
| zequence | hmm, I do have headers. of course I do | 09:29 |
| zequence | I'm gonna try something else next. Use an older release with both it's kernel and newer one | 11:22 |
| zequence | I'm highly suspecting it's not the kernel that causes all of the grief | 11:23 |
| smartboyhw | zequence, promise me one thing: If you are going to use the ~ubuntustudio-core powers to change testcases of Ubuntu Studio in the ISO QA Tracker, INFORM ME | 13:35 |
| * smartboyhw goes to change links at download page of ubuntustudio.org | 15:09 | |
| smartboyhw | It isn't updated.... | 15:09 |
| * smartboyhw is thinking that we should probably have a release checklist..... | 15:09 | |
| smartboyhw | zequence, ^ | 15:09 |
| zequence | smartboyhw: That sounds like a good idea. Actually, I'm starting to think that it might be a good idea also to create a release team | 15:19 |
| smartboyhw | zequence, oh...... | 15:19 |
| smartboyhw | zequence, BTW you got something wrong on the team structure page | 15:19 |
| smartboyhw | For the leaders role (you and scott-work) it is not Quality ISSURANCE it is Quality ASSURANCE :P | 15:20 |
| smartboyhw | There is no such thing as QI:P | 15:20 |
| zequence | smartboyhw: If you find spelling/grammar mistakes, please feel free to correct them | 15:21 |
| smartboyhw | zequence, ok | 15:21 |
| smartboyhw | zequence, I also updated the download links for 12.04.2 in the community wiki and ubuntustudio.org | 15:21 |
| * smartboyhw likes doing typo edits | 15:21 | |
| zequence | Swedish is quite close to English in many ways, and we would use the same word for both | 15:21 |
| zequence | It's easy to make this kind of mistakes | 15:22 |
| smartboyhw | zequence, oh really? | 15:22 |
| * smartboyhw doesn't know that | 15:22 | |
| zequence | The correct grammar there would be "didn't", but I guess that is because Chinese doesn't make a distinction between did or do | 15:24 |
| smartboyhw | zequence, LOL | 15:24 |
| smartboyhw | I have better English than Chinese (srsly) | 15:24 |
| zequence | I speak three languages fluently myself, plus a little German. But, I wouldn't say I'm very good at any of them. For two reasons. I'm not good at remembering words, and I've never put much time into improving my language skills | 15:27 |
| smartboyhw | Uh | 15:27 |
| zequence | My sisters are both speak at least four languages fluently, but they are living abroad | 15:28 |
| zequence | - are | 15:28 |
| smartboyhw | zequence, if you one day set up have a seperate team for releases tell me I might join (of course that depends on you really | 15:28 |
| zequence | smartboyhw: Yeah, I was thinking of that. Since you are the person most active with that anyway | 15:28 |
| zequence | would make sense | 15:29 |
| smartboyhw | zequence, thanks | 15:30 |
| smartboyhw | I seemingly have pleasure doing ISO tests..... (more weirdly, I ranked third in doing 12.04.2 ISO tests, doing 46) | 15:30 |
| zequence | I'll just talk with Scott, so he's in on it. Then I'll create the team. Add you to it, and ask the guy in -release to change it again for us | 15:31 |
| zequence | Wow, that's plenty | 15:31 |
| smartboyhw | zequence, the guy = stgraber. OK | 15:31 |
| smartboyhw | LOL | 15:31 |
| smartboyhw | zequence, there's a typo: I actually only did 44. However the first (pwlars) and the second (psivaa) has got 136 and 129 test results in 12.04.2. How on earth did they accomplish that!? | 15:48 |
| len-1304 | zequence, on my P4 with 32 frames on jack I am pushing 70% DSP, but the cpu% is much higher. I am getting 0 to maybe 3 xruns in a few hours. | 16:01 |
| smartboyhw | Hey len-1304 thank you for testing 12.04.2 | 16:01 |
| len-1304 | The xruns are from SW updates BTW. | 16:01 |
| len-1304 | smartboyhw, NP | 16:01 |
| len-1304 | zequence, I will try to spend some time today pushing things with guitarix and rakarrack. | 16:06 |
| len-1304 | But so far the biggest difference has been turning hyperthreading off. | 16:06 |
| len-1304 | On a multicore machine I don't know that hyperthread is that helpful. | 16:07 |
| * len-1304 thinks hyperthreading was a stop gap before multi-core... or maybe a sales gimmick | 17:16 | |
| Lump|AFK | heya len-1304 | 17:17 |
| Lump|AFK | for what it is worth, I agree with you on the hyper threading but I sure love my atom on my Eee | 17:17 |
| len-1304 | Lump|AFK, how are you. | 17:17 |
| Lump|AFK | i am tired and stresssed atm | 17:18 |
| len-1304 | The atom based MBs seem to be some of the netter ones for audio | 17:18 |
| len-1304 | *better | 17:18 |
| Lump|AFK | been pulled a bit thin recently and I need to hustle some bucks for bills | 17:18 |
| Lump|AFK | but such is the life of one self employed. | 17:18 |
| len-1304 | Been there. This has been a hard year for money for us. | 17:19 |
| Lump|AFK | i am going to try streaming with the Eee in the near future | 17:19 |
| Lump|AFK | some of the local clubs are interested in live remotes | 17:19 |
| Lump|AFK | so i need to decide if i can use the eee or need to build a machine for it | 17:19 |
| len-1304 | I have been looking at linphone with the thought of hacking in jack support. | 17:19 |
| len-1304 | Sounds like fun. | 17:19 |
| Lump|AFK | you have to keep me posted on that one | 17:20 |
| Lump|AFK | now that linux is coming to phones i may get a smart phone down the road | 17:20 |
| Lump|AFK | but after trying windows mobile, i ended up going back to a palm m500 | 17:20 |
| len-1304 | Ya, the problem I see is that the streamer is separate from the rest. I need to see if it is only started when the call connects. | 17:21 |
| Lump|AFK | i just like the palm, jpilot interface much better than micro$oft's outlook and pocket pc | 17:21 |
| Lump|AFK | pluc, imho, the graffiti beats the pants off ms handwriting rec | 17:21 |
| len-1304 | There is a broadcast quality box out that is built to act as a studio end to phone converstions. | 17:21 |
| Lump|AFK | yeah isn't that ironic.. land line providers now specialize in compression and cell phone carries move to fidelity | 17:22 |
| len-1304 | It doesn't use jack, but goes straight to audio or AES3 i/o | 17:22 |
| len-1304 | The problem is that "leased line" is no longer being offered in a lot of places. | 17:23 |
| len-1304 | So AoIP is all that is left. | 17:23 |
| len-1304 | I have been reading a lot of EBU papers on this stuff | 17:24 |
| Lump|AFK | land lines are now optional for carriers outside of business i guess | 17:24 |
| Lump|AFK | which, imho, is rather scary | 17:24 |
| len-1304 | I want to make a SIP client that is standards compliant with the EBU standard for Linux. | 17:25 |
| len-1304 | Landlines (of some form) are still very much in use. The form is changing though. | 17:26 |
| Lump|AFK | ntl, i have to get jiggy with my nick soon | 17:27 |
| len-1304 | Things are going digital right at the house instead of shipping audio all the way the the switch | 17:27 |
| Lump|AFK | you going to be about today? | 17:27 |
| len-1304 | Off and on yes. | 17:27 |
| Lump|AFK | hopefully i can catch ya later | 17:27 |
| len-1304 | K | 17:27 |
| Lump|AFK | btw and fwi, it think i went over 5 hours continuous on the stream last night | 17:28 |
| Lump|AFK | sf was in --> meterbridge ppm --> jack rack with compressor and eq --> meterbridge ppm -- idjc | 17:28 |
| len-1304 | Sounds great. | 17:29 |
| Lump|AFK | largest xrun was just a few tenths over a millisecond | 17:29 |
| len-1304 | zequence, pulse 3.* is now released into quantal seems to fix the starting of jack while streaming so far. | 17:30 |
| len-1304 | Lump|AFK, that is pretty good, was there any noticable glitch? | 17:30 |
| len-1304 | (sound wise?) | 17:30 |
| Lump|AFK | had a good number of listeners in chat throughout the night | 17:31 |
| Lump|AFK | and every time i saw an xrun, i asked | 17:32 |
| Lump|AFK | nobody noticed | 17:32 |
| Lump|AFK | i notice because i monitor the line in and not the compresseedd, eqed stream | 17:32 |
| Lump|AFK | but i think the real time kernel makes the line in skip but not the stream out | 17:32 |
| Lump|AFK | i never can hear it on the finished recording either | 17:33 |
| Lump|AFK | i would like to monitor the stream but the delay ends up making one slur their speach like a drunk | 17:33 |
| Lump|AFK | nonetheless i have to head out or we are going to be a hungry familyu | 17:34 |
| Lump|AFK | er family | 17:34 |
| Lump|AFK | i will be back later today and hopefully about more tomorrow | 17:34 |
| Lump|AFK | i still want to get around to installing the nightly and setting up error reporting and such | 17:35 |
| Lump|AFK | forgive me i have just been buried of the late | 17:35 |
| holstein | on xrun here and there might not be a big deal | 17:35 |
| holstein | i usually go for zero xruns | 17:36 |
| holstein | i would just bump the latency up a little | 17:36 |
| holstein | just the next settings higher | 17:36 |
| Lump|AFK | hey holstein, i have to bolt but i will give that a try | 17:40 |
| Lump|AFK | although really very few xruns... maybe one an hour | 17:40 |
| Lump|AFK | usually it occurs when i switch workspaces | 17:40 |
| holstein | sure, but you likely dont need the low latency | 17:41 |
| holstein | you also could be at like 18ms.. and going to 30 or 40 would really not make that much differece | 17:42 |
| holstein | difference* | 17:42 |
| holstein | for me, anything over about 12ms, and im not able to do live effects or software synths, so it might as well be stable at 80ms | 17:43 |
| Lump|AFK | like i said, i will give it a try | 17:43 |
| holstein | that would literally mean that, on your stream, what you do will happen 80ms later... im sure no one would notice | 17:43 |
| Lump|AFK | ntl, i will play with latency next stream, for now, i need to get to the store | 17:52 |
| Lump|AFK | you have a good day len, holstein and room | 17:52 |
| Lump|AFK | and thanks for the help you always provide | 17:52 |
| Lump|AFK | you being plural | 17:52 |
| holstein | cheers! | 17:53 |
| len-1304 | holstein, Ya, bumping latency up works for most things, but can add extra pauses in when you have contributors that are remote. | 17:56 |
| len-1304 | Though I think at the moment, he is running the remote line in with the mics through a mixer. | 17:57 |
| Lump|AFK | correct len | 17:57 |
| Lump|AFK | so it would be mute | 17:57 |
| Lump|AFK | ntl for real now, my coat is on | 17:57 |
| Lump|AFK | cya | 17:57 |
| len-1304 | holstein, have you tried the difference with hyperthreading turned off or on ? | 17:58 |
| len-1304 | Lump|AFK, bye. | 17:58 |
| holstein | len-1304: i havent, but i can | 17:58 |
| len-1304 | It has dropped me one latency step. | 17:58 |
| holstein | interesting... | 17:58 |
| holstein | you diable that in the bios? | 17:59 |
| len-1304 | Yes ,but it can be done after boot too. | 17:59 |
| len-1304 | There is a syscontrol command to disable some of the cpus. Linux sees hyperthreads as two cpus | 18:00 |
| len-1304 | so disabling every second cpu has the same effect. | 18:00 |
| holstein | i have avlinux on a nice-ish box with ht | 18:01 |
| holstein | i should try it there | 18:01 |
| holstein | its easy to get an xrun | 18:01 |
| len-1304 | holstein, this in on my old P4 box. (single core) | 18:01 |
| holstein | yeah, my main rig is a dual core | 18:02 |
| holstein | iirc the other box with AVlinux is just a nice-ish p4 with ht | 18:02 |
| len-1304 | I don't seem to have any higher cpu% to do the same work either. | 18:03 |
| len-1304 | Lots of things that gave xruns before (like switching workspaces) don't seem to. | 18:04 |
| holstein | that is odd | 18:04 |
| holstein | maybe its the overhead or running HT these days | 18:04 |
| len-1304 | Could be. | 18:05 |
| holstein | i should try that on my daily driver too | 18:05 |
| holstein | my little netbook shows up like it has 4 cores | 18:05 |
| len-1304 | With a RT thread the context switch might happen a lot more often. | 18:05 |
| len-1304 | Ya, linux sees hyperthreading like extra cores... and uses them as such. So it is easy to end up with an RT thread and a non RT thread trying to use the core | 18:07 |
| holstein | i like that email to the list "Hi, I am new to this and expect a lot of support and help from all." | 18:08 |
| len-1304 | Linux doesn't try to limit the time the Non RT thread has on it's "core" | 18:08 |
| len-1304 | :)) | 18:08 |
| holstein | its gotta be some language barrier, or a large misunderstanding of what a mailing list is | 18:08 |
| len-1304 | Or used to comertial lists. | 18:08 |
| len-1304 | commercial. | 18:09 |
| holstein | that too.. | 18:09 |
| len-1304 | I have been working on switching my mode app from tcl/tk to python | 18:12 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!