02:39hakzsam: imirkin, feel free to push this branch when you have time http://cgit.freedesktop.org/~hakzsam/mesa/log/?h=nv50_pipeline_queries thanks!
09:46imirkin_: hakzsam: thanks for fixing up those commit logs :)
09:52hakzsam: imirkin_, and thanks for pushing those commits :)
09:52imirkin_: hakzsam: could i convince you to fix the situation with timestamp_disjoint? iirc nine uses it...
09:52imirkin_: i noticed as i was pushign that you got rid of the query->ready shortcut from nv50_query_ready
09:53imirkin_: (but that probably didn't work since the query was already reset before)
09:54hakzsam: I could try, I'm going to add it to my long todolist ;)
09:56hakzsam: yeah, I don't think q->ready affects timestamp_disjoint since
09:56imirkin_: but i think someone *wanted* it to ;)
09:56imirkin_: they just failed
09:58hakzsam: yeah, and if you tell me how to test if timestamp_disjoint works, I could try to fix it this week end;)
09:59imirkin_: tizbac: iirc you were the one futzing with it?
13:59Berra: How do I go about figuring this out? http://sprunge.us/DZPG
14:00Berra: Computer completely freezes then dies.
14:00imirkin_: was the link supposed to be something more than 2 empty lines?
14:00Berra: Ah, yes. Sorry
14:00Berra: Well... that dump was lost the...
14:01imirkin_: how about you mention what hw you have and what the observable issue is
14:01Berra: echoed it to the file as quickly as I could before it all died but must have not worked.
14:02Berra: The observable issue is trying to start a Qt5 WebEngineView. HW: http://sprunge.us/JCEU
14:02imirkin_: wow, an actual real GF100
14:02imirkin_: two of them, even
14:03imirkin_: i assume you work out of a nuclear power plant?
14:03Berra: The issue is related to something "kernel pushbuf". I got them when I was younger and stupid.
14:03imirkin_: are you using libdrm 2.4.60?
14:04Berra: libdrm-git 18.104.22.168.ga1acffd-1 sorry
14:04Berra: No need to paste
14:04imirkin_: hm ok
14:04imirkin_: seems unlikely that gnurou's change would have any visible effect
14:05Berra: Linux localhost 4.0.4-1-ARCH #1 SMP PREEMPT Mon May 18 06:43:19 CEST 2015 x86_64 GNU/Linux
14:05imirkin_: well, i guess it's some form of new issue. doesn't sound like anything i've heard about
14:06Berra: It would required a low level fix? Or do you think I could investigate it?
14:06imirkin_: there *is* an outstanding issue where supertuxkart 0.9 mega fails that i haven't figured out how to investigate yet
14:06imirkin_: and it does end up in pushbuf list corruption
14:06imirkin_: could well be that you're hitting a similar issue
14:07Berra: If you want a dump I will crash my computer again to get it for you
14:07imirkin_: although in my situation that just causes stk to SIGSEGV
14:07imirkin_: not some sort of horrid hang
14:11Berra: Wouldn't you know it, this time it crashed instantly without dump information..
14:11imirkin_: doh =/
14:11imirkin_: Berra: oh, i assume you're on some sort of recent mesa?
14:12imirkin_: if you can tolerate the annoyance of continually crashing, you could try downgrading
14:12Berra: mesa 10.5.5-1 mesa-libgl 10.5.5-1
14:15Berra: So far I can only produce this crash by starting the QtWebEngineView
14:16Berra: Was it TuxRacer..? I could try and see if it's the same?
14:16imirkin_: supertuxkart 0.9
14:16imirkin_: (mariokart clone)
14:16Berra: Downloading now
14:17imirkin_: if things work the same as for me, you should see a SIGSEGV as soon as you try to go into a level
14:17Berra: In Arch Community Repo : )
14:17Berra: Soon installed, we'll see.
14:18tobijk: imirkin_: wanna rage-push my piglit cull_distance patch? :D
14:19imirkin_: i'm not a rage-pusher
14:19imirkin_: i'm a rage-reverter
14:19imirkin_: hasn't had to come up in mesa yet though :)
14:20tobijk: then do a rage-review, makes me happy as well :D
14:23Berra: Yupp, Crash in tuxkart
14:24Berra:  1387 segmentation fault (core dumped) supertuxkart
14:25imirkin_: k, well that's what i get too :)
14:25imirkin_: and an error in dmesg about invalid pushbufs
14:29Berra: Did you message get through? I got the segfault on supertuxkart
14:29Berra: Then it hung, then crashed
14:30imirkin_: it crashes when i try to start a level for me
14:30Berra: Crashed when I klicked on the singleplayer icon
15:41imirkin_: RSpliet: in future sends, probably just title it "clk/gt215" or whatever. no need for nvkm in subject.
15:43tobijk: why do all nowadays forget the "lists" :D
15:44RSpliet: imirkin: sure
15:44imirkin_: RSpliet: you also appear to have messed up the lists quite badly =/
15:45imirkin_: but seemingly email@example.com also works
15:45imirkin_: which is a bit unexpected
15:45tobijk: i saw that several times lately, did they change somethnig maybe and we are ust getting old?
15:46RSpliet: I take no responsibility in messing up any mailinglist
15:46imirkin_: in the To line of your patches
15:46RSpliet: if git send-email does that, the ML software needs fixing
15:46imirkin_: who else is to blame :p
15:46imirkin_: To: Nouveau@freedesktop.org, "Mailinglist <nouveau"@lists.freedesktop.org,
15:47RSpliet: meh, I told it --to "Nouveau Mailinglist <firstname.lastname@example.org>"
15:47RSpliet: so unless I typo'd, which I'm not going to verify for the sake of argument...
15:48RSpliet: oh btw, that "the fuck" is a copy-paste from gt215. I found it slightly out of place, but I recall that comment being there before I raged through it last summer
15:48imirkin_: don't have to propagate it :po
15:50tobijk: magic 750MHz, ay
15:51RSpliet: meh. anyway, your other comment on 3/9 is somewhat corrected slightly a little in 7/9. Either way, it's irrelevant for now as it mainly targets NVA0 and I explicitly acknowledge that for other cards it needs further investigation
15:51imirkin_: RSpliet: yeah
15:52imirkin_: you got rid of that stupid #define too, which is good
15:53tobijk: will patch 4/9 blow up <GT200 or why is it stated in that way?
15:54RSpliet: tobijk: untested. bails for unknown rammap length or unknown timing entry length
15:55tobijk: ah ok
15:56RSpliet: so quite literally what I sent out is support for NVA0 reclocking with a lot of work done for other cards too. Those cards are explicitly not supported as per patch 9
15:57tobijk: not there (yet)
15:58RSpliet: heh, well, feel free to reply on the ML
15:58RSpliet: I'm off for some well-earned sleep
16:25imirkin_: gr, everywhere i look, bugs :(
16:29tobijk: yeah i just found my piglit tests failing in a full run, while running them on their own they pass just fine :O
16:30tobijk: delays things quite a bit ;-)
16:35Gornle: Hi. I need to fix the resolution used by a linux/64 box with an NVidia 8400 card, and the nouveau driver. At the moment it's using only about 80% of the screen, and the height/width ratio is off. I *THINK* I need to add something like "video=DVI-I-1:1920x1080@xx" to the kernel boot line. But I can't find the docs for that on the wiki (so far). Is there a way from command line that I can *get* the right setting? e.g., DVI-A or DVI-I, what refresh fre
16:35Gornle: q, etc. ?
16:36imirkin_: Gornle: run xrandr --output DVI-I-1 --set 'scaling mode' None
16:36imirkin_: that should disable card scaling
16:36imirkin_: and will let the panel scale
16:37imirkin_: without that, you're stuck on a fixed resolution/frequency no matter what you try to do
16:37tobijk: hope we get to a point where the kernels that do these on default are common...
16:37imirkin_: yeah, i forget when it got flipped... 4.0?
16:37Gornle: imirkin_: Thanks. I need to install xrand (I'm working UP from a 'minimal server' install). Also, how do you know it's "DVI-I-1"? Is that a standard setting?
16:38imirkin_: Gornle: oh, i don't. got it from your comment.
16:38imirkin_: Gornle: just running xrandr should give you the output configuration
16:38imirkin_: i'm assuming you're using X?
16:38imirkin_: if not, you can inspect state with the modeset tool. but it's harder to effect changes since fbcon doesn't like it when you touch things
16:39Gornle: imirkin_: Atm, I've got IceWM to the point it can launch, and Firefox will run. It's pretty minimal. I'd like to KEEP it as minimal as possible, but know I'll have to install 'more'. I'd rather NOT just default to "install all of KDE & X" ...
16:40imirkin_: Gornle: well, there's not a lot of stuff in "X". but xrandr is a pretty basic tool. well worth the ~10KB of disk space
16:40imirkin_: ouch. 56KB =/
16:41Gornle: Sure, not a problem. :-) I do have "/usr/bin/X :0 vt7 -nolisten tcp -seat seat0 -auth /var/lib/kdm/AuthFiles/A:0-4WCHeb" running atm acccording to 'ps ax'.
16:42Gornle: just running 'xrandr', the output atm shows: "Screen 0: minimum 1280 x 1024, current 1280 x 1024, maximum 1280 x 1024; default connected 1280x1024+0+0 0mm x 0mm; 1280x1024 77.00*"
16:42Gornle: It's clearly not seeing the 'whole' monitor
16:42imirkin_: oh, you're not using nouveau
16:43imirkin_: that's vesa or vga driver
16:43imirkin_: pastebin dmesg and xorg log
16:43tobijk: wild guess: install xf86-video-nouveau
16:43imirkin_: ah yes, good point ;)
16:44imirkin_: but... if nouveau were running, vesa would mess things up *bigtime*
16:44imirkin_: so i assume nouveau is also not running
16:44Gornle: the kernel module is loaded. that's different than "nouveau is running"?
16:46tobijk: dmesg | grep nouveau
16:46tobijk: for a quck overview
16:46imirkin_: Gornle: pastebin the full output of dmesg, and the full xorg log file. difficult to help without that.
16:46Gornle: Hm. `xrandr --output DVI-I-1 --set 'scaling mode' None` -> "warning: output DVI-I-1 not found; ignoring; X Error of failed request: BadRROutput (invalid Output parameter)"
16:47Gornle: tobijk: imirkin_ ok. one sec.
16:48Gornle: well, there's NO dmesg output containing 'nouveau' -- except for one line showing that the kernel module is loaded. that's obvioulsy the problem I need to poke at!
16:48Gornle: ok, you got me pointed in the right direction -- off to do some reading up on it. thx!
16:48tobijk: yeah normally you get serveral informative lines
16:49imirkin_: do you have 'modeset=0' or 'nomodeset' somewhere?
16:49Gornle: imirkin_: I currently have "nomodeset drm.modeset=1 nouveau.modeset=1"
16:49imirkin_: nomodeset disables modesetting
16:49imirkin_: remove that
16:50Gornle: it's overridden by tge nouveau.modeset=1, according to the kernel person who said so :-/
16:50imirkin_: (where "modesetting" is a synonym for "modern graphics subsystem", not just the modesetting aspect)
16:50imirkin_: the kernel person who said so apparently lied
16:50imirkin_: viciously and maliciously
16:50Gornle: hrm. k. gonna have to reboot, I suppose.
16:50Gornle: back in a bit. thx.
17:01flipmess: hi, is it possible at all with nouveau to use the vga/dp ports on an laptop with "optimus"? intel card is connected to LVDS and external ports to nvidia GK107GLM [Quadro K2000M].
17:01imirkin_: yes, should work fine
17:02imirkin_: flipmess: see http://nouveau.freedesktop.org/wiki/Optimus/ (esp the using outputs on discrete gpu section)
17:03flipmess: i tried to use that... but when i do xrandr --listproviders i only get 1 provider
17:03imirkin_: flipmess: pastebin dmesg and xorg log
17:07flipmess: http://pastebin.com/FNt7GBSe <- dmesg
17:08flipmess: and with xorg log you mean xorg.0 .1 and .8? sry
17:08imirkin_: probably .0
17:10flipmess: http://pastebin.com/EHyCwX9n <- xorg.0.log
17:11imirkin_: flipmess: do you have an xorg.conf?
17:12imirkin_: flipmess: with a Device "intel" in it?
17:12imirkin_: yeah you do
17:12imirkin_: (**) intel(0): Option "AccelMethod" "sna"
17:12imirkin_: which is the default i think :)
17:12imirkin_: get rid of that file, nouveau should appear
17:13imirkin_: er, i guess i meant Driver "intel"
17:13flipmess: can i just rename it to config.bak?
17:14flipmess: x server restart or reboot?
17:14imirkin_: just X
17:15flipmess: brb.. i hope ;)
17:16flipmess: YES! thx a lot imirkin_
17:17flipmess: do you mind elaborating a bit what was the problem?
17:17imirkin_: X is the worst
17:17imirkin_: it only auto-adds gpu's if you don't configure anything directly
17:18imirkin_: apparently fixing this is deceivingly complicated
17:18flipmess: thx :) now to try the outputs
17:24flipmess: wow.. for the first time in 2 years i can use the vga port... garbled image... but image :D
17:24imirkin_: uhm... "yay"?
17:25imirkin_: what desktop environment are you using?
17:25flipmess: i3wm no de
17:25imirkin_: ah ok
17:25imirkin_: for this to work you need a compositor
17:25imirkin_: not 100% sure if xcompmgr is enough or not
17:26imirkin_: at least... i'm fairly sure
17:26imirkin_: [that one is necessary... def need one for offloading]
17:27flipmess: hm... looks like i'm using compton but i'll try xcompmgr
17:27imirkin_: compton is def good enough
17:27flipmess: because the second screen isn't beeing updated
17:27imirkin_: can you check your dmesg for nouveau errors?
17:28flipmess: no recent nouveau lines
17:31flipmess: i don't think i can use offloading because the powermanagement is still WIP for my card
17:36flipmess: hm.. max resolution is 1280x1024 (says xrandr). the beamer should do HD. is this what xrandr reads from the beamer?
17:37imirkin_: pastebin xrandr output
17:38imirkin_: offloading works fine, you just might not get amazing speeds. although GK107 you can probably reclock
17:38imirkin_: oh yeah, you have DDR3 vram. should almsot certainly work
17:38imirkin_: but you need to make sure nouveau loads with pstate=1
17:38imirkin_: (e.g. by adding nouveau.pstate=1 to the kernel cmdline)
17:43flipmess: http://pastebin.com/YAqHFukH <- xrandr output
17:43imirkin_: ok. and what are you expecting to see?
17:44imirkin_: 1280x800 is HD... "720p"
17:44imirkin_: [well, 1280x720 is 720p, but you get the point]
17:45flipmess: ah.. but it can do 1920x1080
17:45flipmess: so i was expecting that
17:45imirkin_: apparently not
17:46imirkin_: at least not listed in its EDID
17:46imirkin_: or nouveau is rejecting the mode for some reason
17:46imirkin_: you can check in your xorg log
17:51flipmess: (II) NOUVEAU(G0): EDID vendor "SEC", prod id 42247 <- are these lines all the resolutions it got from the beamer? because there is 1920x1080
17:53imirkin_: i'm guessing it's a non-reduced blanking mode or something, and it decides there's not enough bw
17:53imirkin_: or something.
17:53imirkin_: digital is better if it's an option :)
17:54flipmess: only have dp on laptop and hdmi on beamer... need an adapter or something
17:54imirkin_: many DP ports are actually DP+
17:54imirkin_: which means you can just get a round hole/square peg adapter and get HDMI
17:54flipmess: oh... there is ++D on the port.. ^^
17:55imirkin_: Dual-mode DisplayPort (also known as DisplayPort++) can directly output single-link HDMI and DVI signals using a simple passive adapter that adjusts from the different connector and the lower voltages used by DisplayPort
17:55flipmess: nice.. i could probably solder it myself... haha
17:56imirkin_: yeah, coz high-bandwidth links aren't sensitive at all to things like... crap solder
17:56flipmess: hey.. i wouldn't "crap" solder ;)
17:56imirkin_: i'm sure that applies to anything you get in a store
17:57imirkin_: anyways, adapters run $5-$10
17:57imirkin_: but paste the xorg log
17:57imirkin_: we might be able to salvage this
17:58flipmess: thx a lot
17:58imirkin_: it has a non-reduced blanking mode for 1920x1080
17:59flipmess: i'll have to google that
17:59imirkin_: cvt -r 1920 1080
17:59imirkin_: use that to do xrandr --output VGA-1-1 --newmode <stuff>
17:59imirkin_: Modeline "1920x1080R" 138.50 1920 1968 2000 2080 1080 1083 1088 1111 +hsync -vsync
18:00imirkin_: xrandr --output VGA-1-1 --newmode 1920x1080R 138.50 1920 1968 2000 2080 1080 1083 1088 1111 +hsync -vsync
18:00imirkin_: that should work
18:00imirkin_: and then you can flip over to it
18:00imirkin_: [and hope that your device can handle the reduced blanking mode]
18:00imirkin_: oh, crap
18:00imirkin_: xrandr --newmode 1920x1080R 138.50 1920 1968 2000 2080 1080 1083 1088 1111 +hsync -vsync
18:01imirkin_: xrandr --addmode VGA-1-1 1920x1080R
18:03imirkin_: good times?
18:03flipmess: beamer says resolution is 1920x1080 progressive
18:03flipmess: there is a picture ^^
18:03flipmess: i can see part of the laptop screen
18:03imirkin_: what if you clone your desktop
18:04imirkin_: i.e. xrandr --output VGA-1-1 --clone LVDS2
18:04imirkin_: i.e. xrandr --output VGA-1-1 --same-as LVDS2
18:05flipmess: i looked at what arandr says and the screens where overlapping
18:06flipmess: geometry is ok now.. but the colour is funny on the beamer side
18:06flipmess: everything is beige now
18:06imirkin_: anyways, i gtg
18:06flipmess: thx a lot again
18:06imirkin_: might be back later. good luck
18:06imirkin_: try cloning the screens
18:06flipmess: thx cya
18:06flipmess: will do
18:06imirkin_: that should give you a good baseline