| len-dt | ailo, ya I've been busy too. I need to package my mode change stuff and add a config tool. I will try to make it open and not set to specific tweaks. | 00:36 |
|---|---|---|
| len-dt | That way a user who has special needs can still do something we haven't thought of. | 00:36 |
| len-dt | Or it can be used for other purposes besides US. | 00:38 |
| len-dt | If nothing else it may make testing easier. | 00:41 |
| scott-work | sorry all, have been sick since monday but i'm feeling much better now | 13:40 |
| ailo | scott-work: Seemed like there might have been a bit of hesitance adding -lowlatency for UKT maintenance :P | 15:06 |
| ailo | I was just thinking about the motivation to add it | 15:06 |
| ailo | Since we will absolutely want it in, no matter which way, and the diff is just a couple of configs, it seems like the best course of action. I guess this is what we have already formulated | 15:07 |
| ailo | Also, the configs are not only there for improving audio performance. They are more or less absolutely needed | 15:09 |
| ailo | I know we all know this, but I wonder what the guys at UKT are thinking about this | 15:10 |
| scott-work | ailo: i spoke with leann late last week on irc after several emails back and forth | 15:18 |
| scott-work | her original concern was adding back another kernel into their responsibility after they had reduced the numbers | 15:18 |
| scott-work | her current concern is validation of usage, to which she submitted an RT ticket to get the download numbers for the studio image | 15:19 |
| scott-work | leann mentioned that one reason the non-pae i386 kernel was dropped was because of low, and dwindling, usage numbers | 15:20 |
| scott-work | however, if the download numbers do not justify UKT maintaining the kernel, leann was very gracious to offer help creating scripts that would minimize the work for updating the lowlatency kernel | 15:21 |
| ailo | scott-work_: Interesting to know download numbers for sure. Has anyone had any info on this before? I can understand wanting to use that as a meter, but from my point of view, the nessecity to have it as the default kernel for Ubuntu Studio, something that other distros do not suffer from (needing a special kernel), makes it a special case. Perhaps if it doesn't end up being maintained by UKT, it will hopefully gain popularity | 19:46 |
| ailo | We should also do more advertising for it. Perhaps highlight it just to make it more visible | 19:48 |
| scott-work_ | ailo: i seem to remember cory or pesia had posted a link a long time ago | 19:56 |
| scott-work_ | i don't remember anything specific, other than perhaps 64bit seemed to double that of 32bit over a period of time | 19:56 |
| ailo | An item we should add to testing is to make some sort of general comparison between -generic and -lowlatency. The cons and pros. Probably the only downside is higher use of power, and it would be good to be able to show how much. | 20:00 |
| ailo | I should start a scratchpad for things like this. | 20:01 |
| ailo | Or add some items to blueprints | 20:09 |
| ailo | Seems like the best way after all :P | 20:10 |
| Len-nb | scott-work_, good to see you back, sorry to hear you were sick. | 20:35 |
| scott-work_ | ailo: definitely add it to the blueprint, that way it might inspire others to think of something else :) | 20:36 |
| scott-work_ | Len-nb: thanks | 20:36 |
| Len-nb | scott-work_, do you have a midi IF? If so what kind? ( looking for non-mpu401 compatible) | 20:38 |
| scott-work_ | no, i don't have any MIDI gear | 20:39 |
| Len-nb | Ok, there have been problems with USB MIDI IFs. I can't test it though. | 20:39 |
| Len-nb | I wanted to see if the problem was wider than just one instance. | 20:40 |
| Len-nb | ailo, I don't think the Low latency kernel would add much power use over the generic. On it's own the change will not make the cpu work harder. That would depend on the software running. "ondemand" still sets the cpu to half speed on idle and the RT stuff only gets used when SW that uses it runs. That SW would have the same effect on generic too. | 20:51 |
| Len-nb | ailo, setting swappiness to 10 instead of 60 would decrease power use. | 20:52 |
| Len-nb | speaking of swappiness, there has been a move in the linux world towards lower numbers. RH now defaults to 40. | 20:53 |
| Len-nb | on my two machines I have swappiness set to 0 on one and 10 on the other. | 20:54 |
| Len-nb | even though they are both low memory machines (1 G) there is no difference in swapping with normal use. | 20:56 |
| ailo | Len-nb: I've heard that -lowlatency might require more power. Would be good to find out | 20:56 |
| Len-nb | in both cases after a days use swap had no use. | 20:56 |
| Len-nb | ailo, figuring a test case might not be too easy. I do have a power meter though. | 20:57 |
| ailo | Jeez, my nvidia card fan is having a real concerto today. | 20:58 |
| ailo | Len-nb: If there is a way, we'll find it | 20:58 |
| Len-nb | have to measure both low use and high use cases. | 20:59 |
| ailo | One could ask the question why does Ubuntu not use a higher resolution by default | 21:00 |
| ailo | For the timer | 21:00 |
| Len-nb | ailo, I had thought of that :-) yes. | 21:01 |
| ailo | Also, what does preempt do, that would be bad for other uses | 21:01 |
| Len-nb | timer | 21:01 |
| Len-nb | the high res timer is there, but not available to userland | 21:02 |
| Len-nb | ailo, I think in due time these changes will become main steam | 21:03 |
| Len-nb | settable on boot anyway. | 21:03 |
| ailo | Actually, that might be something worth trying to hack | 21:04 |
| ailo | Would mean we don't need -lowlatency. Just some boot options | 21:04 |
| Len-nb | ailo, if there is a general trend to command line options, that could be a problem too. There used to be a limit to command line length... | 21:07 |
| Len-nb | so what we want is the kernel config to be something like yes/no/cl | 21:08 |
| ailo | Would be even better if those changes could be done during runtime | 21:10 |
| ailo | I'm not about to try to code it yet anyway :P | 21:10 |
| ailo | Ok, time to hit bed. Work to do tomorrow! | 21:12 |
| Len-nb | ailo, the first thing to find out is if having it running all time really affects anything else. | 21:12 |
| Len-nb | GN | 21:12 |
| ailo | Len-nb: I guess servers like non preempt, but I don't really know why. | 21:13 |
| ailo | As for resolution, no idea | 21:13 |
| ailo | Timer res, that is | 21:13 |
| Len-nb | preempt only happens if SW uses it. | 21:14 |
| Len-nb | timer rez probably have a cpu work load effect | 21:14 |
| Len-nb | preempt in a server could be a security problem. | 21:15 |
| Len-nb | A bad user program could be like a DOS. | 21:15 |
| ailo | All of these things should be written down and confirmed/proved somehow. | 21:16 |
| ailo | I didn't mean to do anything about this yet, since I have other tasks to do first | 21:16 |
| ailo | But, later on.. | 21:16 |
| Len-nb | yup | 21:16 |
| ailo | Sounds like there's some material for a comparison between the kernels anyway | 21:17 |
| ailo | GN Len-nb :) | 21:18 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!