06:30 boberfly: Hi all, simple question does a discrete quadro k2000 (kepler NVC0) nouveau support being an output sink normally in $ xrandr --listproviders ? And if the option isn't listed as supported is there something I can do to make it work? Cheers all :)
08:14 Moiman: karolherbst: In function ‘uint8_t nv50_ir::getTEXSMask(uint8_t)’: nv50_ir_emit_gm107.cpp:2739:4: error: control reaches end of non-void function [-Werror=return-type]
10:52 karolherbst: Moiman: thanks... will fix that. Is this with a special compile time option?
12:29 RSpliet: karolherbst: This warning might be suppressed in debug builds (although I wouldn't be happy if it were) due to the assert breaking control flow...
12:30 karolherbst: RSpliet: I even tested automake release
12:31 karolherbst: and I didn't get that error in any build
12:31 karolherbst: I get this though: "warning: control reaches end of non-void function [-Wreturn-type]"
12:31 karolherbst: but not an error
12:32 RSpliet: Ah okay, sounds like in Moiman's set-up this is promoted to an error due to the -Werror=return-type flag. Not sure if that's a default option, whether it's meson vs whatever the other build system is called, whe... either way, the warning is important to acknowledge in source too :-)
12:32 karolherbst: again, it's not hit by the default thing
12:33 karolherbst: RSpliet: for me it was more of a question wheather to push directly or send to the ML
12:34 RSpliet: Ah okay... Idk, but it sounds like a trivial fix of adding a default return value - preferably one that stands out in trace as "there's something wrong" :-P
12:35 karolherbst: RSpliet: I will just send patches for more compile time warnings
12:35 karolherbst: while at it
12:35 RSpliet: That sounds like a good idea :-)
12:36 RSpliet: Btw, does this affect performance measurably, or is it really a matter of generating more compact code and saving registers?
12:37 karolherbst: no idea
12:49 shtrb: What's the Quadro code name ?
12:49 shtrb: *Quadro P2000
13:25 frikinz: Hi everyone. I have a laptop with a Geforce 720M (old, entry level card..). Arch is Fermi (GF117). Codename NVD7. I am only looking for performance not power saving. Any idea/link what I could tweak to get a few more FPS? :) Or would nvidia perform better (which I'd like to avoid usually) ?
13:27 karolherbst: frikinz: we don't support reclocking on fermi yet, there are some WIP patches somewhere though
13:30 frikinz: Thanks karolherbst. Well, given that this arch is stopped since 2012, I do hope your not wasting your time with this arch (priorities priorities!). Do you think that closed source thing would then be better?
13:32 frikinz: I need a new laptop or a new villa to have room for a real PC... but that will take more time eh
13:32 karolherbst: frikinz: yeah, it is very low priority. Nvidia should be better, but dual gpu support on laptops is quite terrible with it
13:32 frikinz: It is, I confirm :)
13:34 frikinz: To be honest nouveau doesn't perform really worse than nvidia driver under windows. Which makes me curious.
13:34 HdkR: Without reclocking support it is sort of impossible to compare
13:34 HdkR: Since Nouveau will always run slow if it can't reclock :P
13:35 frikinz: Sure. But I can say that for this fermi arch, windows support is also far from optimal. The card also is not the best :D
13:36 frikinz: So, I'll test all that. Thanks a lot for your kind support and work.
14:07 RSpliet: I don't think you'll have much luck with the WIP patches on NVD7. They've been generated using data from "older" cards (NVC0-NVC8 or something), there are some known and unknown missing bits for NVD7/NVD9. Which is why they're WIP ;-)
14:11 frikinz: Yes RSpliet I saw the phoronix article mentionning your adventures :) Fine fine, I'm now trying the closed source driver. Just a bit tricky with debian and optimus. Last time resulted in unbootable linux after some dependencies and versionning hell. Let's see this time.
14:14 frikinz: I'll have to find a way to easily change back to nouveau for most of the time. Haven't had a problem until now but now I want to get the most of the card to see if it's worth it.
15:08 linkmauve: Hi, on a 3600m (NV92), writing to /sys/kernel/debug/dri/0/pstate does a write error.
15:08 linkmauve: And setting the kernel parameter to 15 doesn’t seem to make tha fan any louder.
15:10 imirkin: linkmauve: no reclocking on G92
15:10 imirkin: but if you just want the fan to be louder, you can control that through hwmon
15:11 linkmauve: Nah, if it was just a symptom of it not being supported it doesn’t matter.
15:12 linkmauve: Why is it saying that 0f and such also exist in this case?
15:13 imirkin: the pstates are the pstates... it shows you the list
15:13 imirkin: just because they're there doesn't mean you can switch
15:13 linkmauve: Ok.
15:14 linkmauve: Thanks anyway. :)
15:15 imirkin: i think currently reclocking is supported on NV4x/G7x, G96-GT2xx, GKxxx, and GM10x
15:15 linkmauve: The game we wanted to try is too slow on the default pstate, too bad.
15:15 imirkin: consider an amd gpu for your next purchase :)
15:16 linkmauve: It’s not mine.
15:16 imirkin: even better -- get that person to consider an amd gpu for their next purchase!
15:16 linkmauve: No matter how many times I come in this channel, I only have a Nintendo one myself. :p
15:19 HdkR: linkmauve: Tell Nintendo to go with AMD next? :P
15:19 linkmauve: HdkR, the Switch is the only Nvidia GPU I’ve almost never had any issue with, with Nouveau. :p
15:20 HdkR: =o
15:21 HdkR: Yay for released Tegra PMU blobs
15:27 frikinz: Ok got bumblebee and nvidia working. Having tweaked yet but.. no performance difference :D
15:27 frikinz: s/having/haven't/
15:28 frikinz: Off to some other channels. o/
19:46 dasabhi: hey do you guys think contributing to nouveau kills your chances of ever getting hired at nvidia?
19:47 HdkR: No
19:47 loonycyborg: wouldn't it be the other way around?
19:47 dasabhi: i would imagine so too
19:47 dasabhi: but like...
19:47 HdkR: Getting hired at Nvidia kills your chances of contributing to Nouveau
19:47 dasabhi: you are reverse engineering their product
19:47 dasabhi: wait for real?
19:47 HdkR: Cool, you know how the hardware works. Makes engineer bringup easier
19:47 HdkR: ;)
19:47 dasabhi: what if you were an intern?
19:48 HdkR: Interns can be tainted as well. You'll have to explicitly state(and maintain) that you don't want to be tainted to Nvidia IP
19:48 HdkR: and it's hard to guarantee you wouldn't be tainted
19:49 dasabhi: fuk
19:50 HdkR: Until the company changes its open source stance, this isn't going to change
19:51 airlied: get hired at nvidia to work on nouveau :-P
19:51 HdkR: oof
19:55 HdkR: airlied: Just get hired at someplace else to work on radeon ;)
19:55 airlied: just be a goat farmer :-P
19:58 dasabhi: oh yeah
19:58 dasabhi: there is amd
19:58 dasabhi: some of their drivers are open source
19:59 HdkR: The ones that anyone cares about are all open source
20:00 dasabhi: HdkR: i wonder if they have an irc channel
20:00 HdkR: You mean like #radeon?
20:01 dasabhi: HdkR: oh LOOL ty ty
23:41 RSpliet: mwk, pmoreau: Do you happen to know whether there's a second predicate stack dedicated to prebrk and brk? Or do they use a different type of magic to support breaks in nested loops?