00:59Prince781: I'm back. this branch makes quite a difference
01:00Prince781: reclocking works 50% of the time with runpm, and without runpm I don't seem to get any issues.
01:01imirkin: oh yeah... with runpm you have to only reclock when it's on :)
01:01imirkin: so ... run glxgears or something
01:02Prince781: I tested glxgears and unigine-valley. Changing the clock state affects fps noticeably.
01:03Prince781: With "0f" bioshock is actually playable.
01:03imirkin: right, i just mean change clock speed *while* it's running
01:03Prince781: Yeah, I did that.
01:03imirkin: as changing while gpu is off doesn't quite work iirc
01:03imirkin: anyways, glad to hear it semi-works
01:03imirkin: a little curious whether it's any faster than your built-in intel gpu
01:10Prince781: Seems to be about the same with unigine-heaven for both.
01:11Prince781: nouveau's a little faster in certain scenes
01:12Prince781: This is not a scientific observation, though.
06:30lawll: 1.0.12 hasn't been updated for ages
06:30lawll: is a new version expected soon?
06:33dgilmore: skeggsb: http://paste.fedoraproject.org/386555/26836214/
06:38lawll: a new version somewhere, sometime?
11:25karolherbst_work: lawll: what do you mean?
11:52huehner: karolherbst_work: i guess he means is there will be a new nouveau ddx release anytime soon (1.0.12 is last published i think)
12:21karolherbst_work: huehner: yeah, was thinking that, too, but just wanted to make sure
12:43Prince781: hmm, what's the difference between pstate and boost?
15:23karolherbst_work: Prince781: boost limits the max clock and pstate selects the set pstate
15:52vedranm: mesa commit 77ea58ca81eb3f5521d67c4459d3469df6d5ffd8 from tstellar decreased Clover's MAX_GLOBAL_SIZE/MAX_MEM_ALLOC_SIZE due to "a hard limit in older kernels of 256 MB for buffer allocations"
15:52vedranm: does anyone know the commit in the kernel which changed this limit?
15:53vedranm: I tried to grep for TTM, GEM, buffer, bump, increase and combinations of these but I couldn't find anything
15:54hakzsam: vedranm, ask on #radeon
16:04vedranm: hakzsam: tried, will try again
16:39Yoshimo: firmware is stored in the same folder regardless wether it is a 32bit or 64bit system right?
16:47Yoshimo: i get told by nouveau that the firmware file is missing, but the file is there, i also purged and re installed the linux-firmware package and it is still there but said to be missing. How could i figure out why nouveau can't find it and reports error -2?
16:54imirkin_: Yoshimo: chances are the file isn't there
16:54imirkin_: at the time that nouveau tries to load the file
16:54imirkin_: for example if you built nouveau into the kernel, you have to also build the fw into the kernel
16:55imirkin_: if you build nouveau as a module and have an insane setup that loads modules from initrd (as many distros do for some very odd reason), then it has to be in the initrd
16:56Yoshimo: i used the package manager and installed nouveau from there so i assume it is an external module
16:56mlankhorst: not sure why that counts as insane
16:56imirkin_: mlankhorst: because most people don't swap in a high-end raid controller into their laptops without first rebuilding a kernel
16:59Yoshimo: i also updated the kernel recently so all dkms modules were rebuild as well
19:11unraised: Hi, I'm running Fedora 24 and google-chrome keeps freezing my system, forcing a reboot. The last lines of journal output are this:
19:11unraised: Jun 30 12:55:55 localhost.localdomain kernel: nouveau 0000:03:00.0: fifo: DMA_PUSHER - ch 10 [chrome] get 00200dbd40 put 00200dbf8c ib_get 0000020f ib_put 0000026c state 800075f1 (err: INVALID_CMD) push 00
19:12unraised: Does this indicate a problem with nouveau or a hw problem with my gpu?
19:13unraised: (Or google-chrome)
19:19RSpliet: unraised: that sounds like a driver problem...
19:20RSpliet: think you can pull the 4.6.3 kernel from updates-testing and see if that by accident solves anything?
19:20unraised: RSpliet: will do. Can't reproduce it on-demand, but it crashes at least once a day. Only when using google-chrome though.
19:22RSpliet: hmm... I have little knowledge about what might be going wrong. there's also recently been some updates to the DDX driver wrt. potential incomplete pushes
19:24RSpliet: not sure if that fix was pulled into fedora yet - you might have to manually build that
21:31unraised: I'm running google-chrome on Fedora 24, and I'm experiencing crashes that point to nouveau as the culprit. Someone here suggesting updating the kernel to 4.6.3, which I've done, but I'm still getting those crashes.
21:31imirkin_: remind which GPU you have
21:32unraised: Rebooted into new kernel at 13:26 and had a crash at 15:18: "Jun 30 15:18:48 localhost.localdomain kernel: nouveau 0000:03:00.0: fifo: DMA_PUSHER - ch 10 [chrome] get 00200b6cc0 put 00200b6dc4 ib_get 0000016d ib_put 000001d0 state 800075f1 (err: INVALID_CMD) push 00406040"
21:32imirkin_: ah, the 406040 issue - i guess you have a tesla
21:32imirkin_: no clue why it happens, it's a pervasive issue across the tesla line that seems to happen to some people a lot more than others
21:33unraised: 2x "NVIDIA Corporation G86 [Quadro NVS 290] (rev a1) ", but only using 1 (the other is unpopulated)
21:35unraised: Ah ok. So this is a known and reported bug? Maybe there is a way to prevent chrome from invoking that command.
21:36imirkin_: known and reported, yes. no idea why it happens.
21:36imirkin_: it's not a specific command that breaks it ... just breaks sometimes.
21:36imirkin_: the 406040 thing is an error code
21:36imirkin_: not a command
21:37unraised: Oh, gotcha.
21:37imirkin_: remove nouveau_dri.so
21:37imirkin_: i think it's rare that the ddx would hit it
21:37imirkin_: so you still get acceleration, just no GL impl
21:37imirkin_: (at least not hw-accel'd GL)
21:38unraised: Alright, I'll try that. Thanks for the workaround.
21:44lawll: karolherbst: why lol? A newer version of xorg nouveau driver. I've seen the rest getting updates except for nouvau
21:44imirkin_: lawll: is there something in the ddx that you feel is missing?
21:45lawll: oh wait, lawl is my nick, I thought she was laughing at my suggestion :/
21:45lawll: imirkin_: I don't know, may be performance improvements?
21:46imirkin_: lawll: maybe. there are none being worked on though. the ddx is pretty much considered done.
21:46lawll: I see
21:46karolherbst: lawll: also the ddx is a rather "boring" part in the entire stack ;)
21:48lawll: I didn't know this