| === JanC is now known as Guest96823 | ||
| === JanC_ is now known as JanC | ||
| om26er | jsalisbury, re bug 1627108, the latest kernel is bad as well. | 15:02 |
|---|---|---|
| ubot5 | bug 1627108 in linux (Ubuntu Yakkety) "X1Carbon comes to a crawl during high CPU usage tasks" [High,In progress] https://launchpad.net/bugs/1627108 | 15:02 |
| jsalisbury | om26er, ack, thanks. I'll build the next one. | 15:02 |
| om26er | jsalisbury, Hi! same results with the new build. Could you maybe build a few versions together to fasten the testing ? | 16:14 |
| apw | om26er, that is a bisect, the next step depends on whether it works or fails with each test | 16:15 |
| jsalisbury | om26er, ack. what apw said :-) | 16:16 |
| om26er | apw, hmm, is that an automated process ? | 16:16 |
| apw | semi-automated is the best description | 16:17 |
| om26er | jsalisbury: that's a good kernel. | 17:40 |
| jsalisbury | om26er, ack, thanks. Next one coming shortly. | 17:41 |
| manjo | rtg, is CONFIG_MEMCG_SWAP_ENABLED is disabled coz of potential high memory consumption ? | 18:19 |
| rtg | manjo, looks like you can enable it on boot command line, "swapaccount=1" | 18:21 |
| manjo | rtg, yes .. but the reason for having it disabled by default is mem consumption ? | 18:22 |
| rtg | manjo, heck if I know. | 18:22 |
| manjo | ☺ | 18:22 |
| === davmor2_ is now known as davmor2 | ||
| manjo | rtg, also arm64 sets CONFIG_NR_CPUS=128 and amd64 sets it to 256, can we make a case to make them match for 17.04 ? ie increase 128 -> 256 for arm64 ? | 18:27 |
| manjo | rtg, I know it is premature atm.. but trying to see if there any opposition | 18:28 |
| rtg | manjo, I actually had CONFIG_NR_CPUS=8192 for awhile, but I think that setting was regressed during Beta2 | 18:28 |
| rtg | 8192 for amd64 | 18:28 |
| manjo | rtg, was it regressed due to bug ? | 18:29 |
| rtg | manjo, no, other reasons | 18:29 |
| manjo | rtg, ic .. May be we can revist bumping arm64 to match amd64 in 17.04 | 18:30 |
| rtg | manjo, I will restore it for 17.04 | 18:30 |
| manjo | rtg, cool | 18:30 |
| manjo | rtg, I will have a list for you for 17.04 wrt arm64 .. | 18:32 |
| om26er | jsalisbury, that is a good kernel as well, but it seems to say that its 4.7-rc5, probably wrong name ? | 18:32 |
| rtg | manjo, perhaps you should start a bug, assign me, and milestone it for 17.04 | 18:32 |
| manjo | yep .. I am doing a review based on a proposal .. will open a bug and assign you | 18:32 |
| om26er | apw, Hello! any update on bug 1622655 ? | 18:33 |
| ubot5 | bug 1622655 in linux-raspi2 (Ubuntu) "Enable CPUset and pids CGroup controllers for RPi kernel" [Undecided,New] https://launchpad.net/bugs/1622655 | 18:33 |
| manjo | rtg, one pita config that is coming up for arm64 is 64k page support.. I know we beat this horse to death last time... but it still is a popular request | 18:34 |
| jsalisbury | om26er, thanks. The bisect bounces around. The naming is done by the mainline-build-one script in kteam-tool. I've never looked at how the naming is done, but I assume it has to do with the closed tag to the commit. | 18:34 |
| jsalisbury | om26er, for example, that commit: | 18:34 |
| jsalisbury | git describe 5048c2af078d5976895d521262a8802ea791f3b0 | 18:34 |
| jsalisbury | v4.7-rc5-679-g5048c2a | 18:34 |
| manjo | rtg, I have 2 people claim perf improvements with 64k pages .. we do have it enabled for ppc64el | 18:35 |
| om26er | I only understood some of that :) | 18:35 |
| manjo | rtg, I hate to bring it up again .. but is this something we should explore again? | 18:36 |
| manjo | rtg, if it is a strong no from your side it is a no .. but if it is a possibility for arm64 I can try to gather some data | 18:39 |
| rtg | manjo, I'm not an expert on that. Perhaps dannf would be a better resource ? | 18:39 |
| manjo | rtg, let me sync with him on other socs we enable and see what we comeup with | 18:40 |
| manjo | rtg, iirc last time we beat on this horse .. he did try to make an argument for 64k pages.. | 18:41 |
| rtg | manjo, I've got no skin in that game. If it works then I'm fine with it. | 18:42 |
| manjo | rtg, ack .. will get back to you after I talk it over with dannf | 18:43 |
| manjo | rtg, target is 17.04 .. so lots of time to gather any proof / data points | 18:44 |
| rtg | yup | 18:44 |
| dannf | rtg, manjo : PoR is not to do 64K pages. there are plenty of negatives - and most of hte positives can be achieved (or improved by) hugepages | 19:04 |
| dannf | we discussed this at a sprint w/ various teams already | 19:05 |
| rtg | dannf, yeah, seems like I turned it off once | 19:05 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!