10:00karolherbst: I got two 8600 GS, somebody interested in one?
12:13RSpliet: If it has interesting performance modes in the VBIOS, sure... but you were probably implying a handover at XDC weren't you?
12:22karolherbst_: RSpliet: yes
12:24RSpliet: If xexaxo is coming and still lives in Cambridge, he might be able to hand it over...
12:28karolherbst_: ohh and by the way, still searching for somebody to share a room with for XDC :(
13:42xexaxo1: RSpliet: I'm planning to, it's a bit borderline though.
13:48mupuf: karolherbst: have you spoken to Ben?
13:54karolherbst: mupuf: yes
19:50NikitaK3: Hello, I am a Computer Engineering student and have very recently started with open source projects. I wish to contribute to nouveau and was just going through the to-dos in DRM tasks. Please guide me on how to start, I was thinking of starting with this entry level to-do task : "Use the kernel tracing infrastructure (latency tracer, irqsoff tracer, http://www.latencytop.org/ , etc.) to find out if Nouveau makes big latencies or disables interrupts for a
19:50NikitaK3: long time. This might be a nice entry level task to get to know the DRM code. "
19:51karolherbst: we have a todo list on the wiki? :D
19:52tobijk: karolherbst: duh
19:52karolherbst: tobijk: seriously, I thought trello is the thing
19:52tobijk: NikitaK3: you may also have a look at https://trello.com/b/ZudRDiTL/nouveau for tasks
19:53NikitaK3: Okay, thanks tobijk
19:54tobijk: karolherbst: imho the other list is to attract newcomers :>
19:58karolherbst: skeggsb: I kind of need to reclock whenever we disconnect the AC adapter on laptops, any idea how to handle this without resuming the GPU?
20:14imirkin_: NikitaK3: what hardware do you have?
20:18airlied: karolherbst: why not just store the reclock jntil next wakeup
20:18airlied: or detect ac status on wakeup
20:18airlied: and reclock appropriately
20:25karolherbst: airlied: well yeah, but skeggsb didn't like the idea in general to mess with runtime pm
20:26karolherbst: airlied: well I could check against the SUSPENDING and SUSPENDED status
20:26karolherbst: I plan to reclock upon resume anyhow later on
20:29karolherbst: airlied: the biggest problem I see with checking the runpm status is, that first you need to detect it is going to suspend, which might be racy and second I need to increase the usage counter because suspending while reclocking is bad as well
20:31airlied: you can get a reference without resuming
20:31airlied: i think its been a while
20:31airlied: runpm is a think i like to forget
20:32karolherbst: well skipping the reclock while suspended is my preference anyway, I just know that skeggsb sees it a bit differently especially for manual triggered reclocks
20:33karolherbst: thing is, the runpm functions don't even check against the SUSPENDING status, which introduces all kind of ugly code
20:33karolherbst: like if you call runtime_pm_suspended it might return false allthough it's set to suspending
20:33karolherbst: and then while increasing the ref counter, you retrigger a resume while suspending
20:33karolherbst: which usually leads to a dead clock somewhere
20:34karolherbst: and I don't see a way to properly check the status without triggering race conditions as well
20:35karolherbst: well except adding some locks inside nouveau itself
20:35airlied: karolherbst: i suppose only manual triggered reclocks are fine to wake up for
20:36karolherbst: would be stupid to wake the GPU up just because the power adapter was disconnected
20:36airlied: bit silly but as long as next resume is at new clocks it shouldbe fine
20:36karolherbst: yeah, that I have already implemented on my branch
20:37karolherbst: even reclock on power disconnection, only thing missing is to actually trigger it on disconnect and not when the therm alarm triggers again
20:39karolherbst: the main reason I want to implement it, was that my GPU seems to be limited to 40W on battery and I don't exactly know why that is, but sounds like a hw feature somewhere
20:40karolherbst: anyhow, I figured it might be a good idea to actually listen to the vbios if it tells us to user lower clocks on battery
20:40karolherbst: or maybe the firmware just doesn't provide enough power anymore
20:45RSpliet: hmm, I think with the current state of nouveau, waking up for a manual reclock might be the best idea. At least the user will have immediate feedback if the thing goes tits up rather than finding out as soon as (s)he launches the workload
20:50karolherbst: but I don't plan to do a manual reclock
20:50karolherbst: more like limiting the clocks when on battery
20:51karolherbst: which is only relevant after the user manually reclocked anyhow
22:25fiftypeople: well mwk you can do it, hahaaa, i just want you to understand guys, cause i do allready, so i'd know comuting is in good hands so i could go ahead and kill people finally
22:29gnarface: wtf was that about?
22:30imirkin_: crazy people ranting
22:30gnarface: he wants you guys to fix a triangle-count-based kernel panic so that it's safe for him to kill people?
22:30gnarface: oh, he must be talking about a video game
22:31imirkin_: trying to reason with insanity is, in itself, insane
22:31gnarface: i wish someone had told me that when i was younger
22:31RSpliet: never too old to learn