00:11kloofy: https://bugs.freedesktop.org/attachment.cgi?bugid=96271&action=viewall appareantly seems so
00:13kloofy: note how the holes are between vm words, probably no USER_SHADER can be marked in two words
00:18kloofy: the packet is itself longer, so yeah this is hacked once again, it can be done
00:19kloofy: if the docs about tag+index+offset appear, we can make the best possible scheduling code, i seem to be talking with myself all the way again
02:41gnurou: skeggsb: btw, do you know how you will address the memory issue that prevented GM204 from probing?
02:42skeggsb: gnurou: yeah, by fixing the mixed-memory detection. i'm shooting off a quick email to you guys today to get some clarification on things
02:42skeggsb: i went to implement it and went "err, it might be this way, but might not be"
02:42gnurou: skeggsb: great!
02:42gnurou: skeggsb: note I will be OOTO for the rest of the week, so be sure to include other people as well
03:00pmoreau: NanoSector: GDDR5 it is
03:00pmoreau: NanoSector: Yours does not reclock with Karol’s patches? :-/
08:52nailyk: hi. I rebuild my kernel with the 64k page size, then i'm trying to build the nouveau module but i got somes errors: http://paste.debian.net/786723/
08:53nailyk: I use the https://github.com/skeggsb/nouveau repo on the 'master' branch. Should i use the 'linux-4.6' branch instead?
08:54nailyk: (my builded kernel is a 4.6.5)
09:04karolherbst_work: nailyk: yes, it has to fit
09:19nailyk: thanks. Switching on the 4.6 brancj right now.
09:28kloofy: http://www.google.com/patents/US20110072213 this one can query all the fields, but i dunno how to do it on amd
09:32kloofy: so on fermi and kepler this idea seems no nuke science, but still looking a way to query the local or effective address by tag on radeon chips
09:33kloofy: i vaguely remember that this ought to function with intel somehow too, allthough not entirely sure, i am in a bit of a time hurry would need some help with this
10:14karolherbst_work: I always wondered, why git uses 4 digits for the patch numbers, because a series with 1000 patches is extremly insane, but then lkml got one today :D
10:15Tom^: it did? O_o
10:15Tom^: some maintainer has a bunch of code to review then heh
10:16urjaman: holy moly
10:17urjaman: but that could actually be a single patch...
10:17urjaman: it seems to be made with a script that just did that one thing and added a patch for every file...
10:25Tom^: seems linus have answered a bit on those patches aswell :p
10:27kloofy: i think i once talked with mareko before he started to act assholish, but infact only reasonable explanation is that index+offset part is either kept in m0 on radeon or in hidden reg
10:53t-ask: hi, which changes should I move where? http://storage2.static.itmages.com/i/16/0802/h_1470182002_2301308_f9c1dde916.png
10:54t-ask: I'm not sure if I interpret the files right
10:54t-ask: BTw this is what I get by using the git mergetool
10:56t-ask: For now I would assume the volt_REMOTE version is the one I take
11:14karolherbst_work: t-ask: never merge upstream branches you don't touch yourself
11:14karolherbst_work: either do a reset --hard or delete your local one and recheckout
11:19t-ask: karolherbst_work: then I better delete everything now and clone https://github.com/karolherbst/nouveau.git again and switch to your "new?" branch?
11:23t-ask: karolherbst_work: or can I just stay on master_4.7 to get all your patches?
12:17Tom^: karolherbst_work: anything new and fancy arrived since your 4.6 branch thats worth upgrading to? :p
12:19karolherbst_work: Tom^: not really
12:19Tom^: il stick around on 4.6 a while longer then.
12:22RSpliet: karolherbst_work: "This is one of the worst patch series submissions in history." - David Miller
12:23karolherbst_work: I saw it
12:23karolherbst_work: but just the thought of it :D
12:23karolherbst_work: like, they guy had to think at least something like that: "mhh, maybe a bit too many patches, maybe I shouldn't do that"
12:23karolherbst_work: but he seemed like to think it is fine
12:24RSpliet: which part of Intel do these patches come from?
12:24karolherbst_work: coffee guys
12:25karolherbst_work: but hey, would make a good entry on the CV
12:25karolherbst_work: 1000+ linux commits :)
12:25Tom^: https://lkml.org/lkml/2016/8/2/2062 sometimes the lkml is quite amusing to read :p
12:27karolherbst_work: the thing is, that guy just annoyed like 1000+ devs… at the same time
12:30RSpliet: devs shouldn't be so combustible ;-)
12:33karolherbst_work: anyway, that would be a patch series I would expect from new students or so
12:34karolherbst_work: thing is, you can't even search for that guy anymore, because google is full of his patches
12:34Tom^: karolherbst_work: perhaps his intentions was precisely that, security through obfuscation.
12:35karolherbst_work: then I call him an amateur
12:52Lekensteyn: wow, somebody thought it would be a good idea to send 1285 patches to a mailing list? You would normally not do that to a single person either..
12:54karolherbst_work: I am sure he created a script which creates one commit for each file changed
13:03RSpliet: I'm sure it was just a gerrit stress-test
13:08mupuf: RSpliet: lol
13:26karolherbst_work: I wouldn't wonder if his email is marked as a spammer so hard, every nigerian prince is envious
13:37RSpliet: maybe superfluously, but Nigerian princes tend to hardly ever be a prince, let alone Nigerian
13:41karolherbst_work: I am sure I got once a message from a real one!
13:50NanoSector: karolherbst_work, have you had a chance to look at the mmio traces?
13:56kloofy: i basically see the radeon method too, but i have no strengths to understand entirely if that would work, needs couple of days
13:56kloofy: they kinda have the CE engine wich can use scatter gather, and use m0 there with GDS
13:58kloofy: but yeah that is wrong channel here, maybe this could also work, when they claim unified cache
13:59karolherbst_work: NanoSector: not yet
14:01kloofy: as i understand they can perhaps send a message when cache is fully empty, they have some sort of counters
18:24karolherbst: my mmiotrace fix commit broke mmiotrace for somebody else......
18:26karolherbst: # echo 0000:00:18.1 > /sys/bus/pci/drivers/intel-lpss/unbind
18:26karolherbst: # echo 0000:00:18.1 > /sys/bus/pci/drivers/intel-lpss/bind
18:26karolherbst: then rcu stall
18:28karolherbst: pq: there?
18:29karolherbst: I have like no clue at all
18:31karolherbst: mupuf: do you have access to a device with lpss ?
18:32Tom^: what is lpss? and do i have it?
18:32karolherbst: low power subsystem or something like that
18:32Tom^: or rather my i7 6700 :p
18:33karolherbst: no frigging clue
18:33Tom^: so just check if /sys/bus/pci/drivers/intel-lpss/ exists when i boot on intel?
18:34Tom^: will do, brb.
18:35mupuf: karolherbst: no idea
18:36karolherbst: that is what I get for fixing code I don't have any clue about :D
18:36karolherbst: but I have an idea... mhh it will be just messy to find out
18:39Tom^: karolherbst: yea i dont have that intel-lpps dir :(
18:39Tom^: guess im for no use then
18:49pq: karolherbst, hi
18:49karolherbst: pq: you didn't get the mmiotrace issue stuff, right?
18:50pq: "get"? what?
18:50karolherbst: my commit broke mmiotrace for somebody
18:52karolherbst: I forwarded the stuff
18:52pq: karolherbst, sorry, I have no idea what any of that stuff means
18:52karolherbst: me neither
18:52pq: what you posted here
18:53karolherbst: mhhh, well, steven talked with the user a bit, but I have also no clue
18:53karolherbst: just saw the message I forwarded
18:58pq: karolherbst, yeah, no idea. You can quote me on that. :-/
18:58karolherbst: crap :/
18:59karolherbst: but I wouldn't be suprised if something is messed up while remove those fault handlers
19:00karolherbst: because that's something which I am sure off, I didn't test, cause I couldn't, like map -> unmap -> map
19:00pq: I don't know any of that kernel stuff anymore, long forgotten
19:01karolherbst: and I never knew them at all
19:01karolherbst: ah well, seems like my next weekend is gone now too :D
19:04gruetzkopf: weekends are an illusion anyways.
19:18Tom^: lol dear god, i just noticed i might be able to destroy my motherboard. it creates a new efi entry for systemd-boot or in the old days gummiboot everytime i reboot.
19:18Tom^: im up to 24 entries right now, what happends when i reach the limit or go above?
19:19Tom^: seems to be some firmware issue, because i simply only have to reboot from uefi to uefi settings, and it creates a new one.
19:35karolherbst: Tom^: do you really expect that the uefi is written well?
19:35Tom^: i halfly did :(
19:36Tom^: since im seemingly already is at newest uefi firmware this vendor has, im gonna have to make some systemd/startup service that simply removes the extra added entries with efibootmgr lol
19:37Yoshimo: if all mainboard vendors would work together on Coreboot it would be better ;)
19:46mjg59: We would certainly have a different set of bugs
19:49karolherbst: pq: yeah, I think I found an issue within unregister_kmmio_probe
19:49karolherbst: "const unsigned long size_lim = p->len + (p->addr & ~PAGE_MASK);" this part looks wrong
19:49karolherbst: because I replaced all PAGE_MASK thingies with page_level_mask
19:49karolherbst: guess I missed that one
19:53karolherbst: mhh, shouldn't change a thing though... odd
21:09NanoSector: does nouveau support the GeForce FX 5200?
21:10urjaman: yes. Does the FX support anything useful? not so much...
21:10NanoSector: as long as it can run lubuntu and internet, it's fine
21:11urjaman: but yeah there is modesetting and accelerated 2D and ... glxgears :P
21:11NanoSector: ooooh glxgears
21:11urjaman: yeah and multi head works
21:11NanoSector: giving away a system with such FX card to someone as an internet PC, nothing fancy
21:12urjaman: the thing is, it doesnt do the GPU accelerated browser stuff, so ... turn of the smooth scrolling etc
21:12NanoSector: will do
21:14urjaman: I'm using an FX5500 PCI (yeah that oldest bus) because i had a problem with my PCIe GFX and yeah ... you can use it, but be prepared to turn off OpenGL things
21:14NanoSector: i'll tell her, no flash
21:15urjaman: and yeah export LIBGL_ALWAYS_SOFTWARE=1 if you want to run anything new opengl ... it'll use totally software and be really slow but that way you can run things if they're simple enough
21:15NanoSector: nice tip, thanks
21:21urjaman: re: video and flash... iirc flash actually runs, it can fallback to software, but it's not very efficient at its drawing, and depending on CPU it might be good at only 480p and/or fullscreen might be a problem (pretty sure my PCI bus is the bottleneck there)
21:21NanoSector: well this is a single core cpu
21:22urjaman: mplayer and such actual video playing things can (atleast with AGP) play 720p with the card okay in my experience, but again, i expect the CPU to be a problem too
21:22urjaman: an FX5500 plugged to an 8-core AMD 3.5Ghz isnt the most normal setup that i have here :P
21:23NanoSector: yeah this is an Athlon 64 single-core with that FX GPU and 512 MB of RAM
21:23NanoSector: karolherbst spontaneously crashed when hearing that
21:25urjaman: An Athlon64 should actually ... atleast as my guess, depending on software, be able to play 720p if the codec is simple enough etc
21:27urjaman: I had a Sempron 2800+ and FX5200 that i used (with Win XP though, so the driver comparison etc is different) to play 720p stuff from youtube with vlc
21:27urjaman: not with the browser or flash, but vlc could do it :P
21:27NanoSector: that's cute
21:33urjaman: oh if i may say, if you can, chuck more RAM into the thing :P
21:33urjaman: the modern web isnt very nice with 512MB... possible i guess, but swappy
21:33NanoSector: i think this eats DDR, which i don't have
21:34urjaman: yeah, i have like 1.5GB of 512MB DDR sticks unused since i upgraded the thing to 2GB with 1GB sticks...
21:34NanoSector: and i'm not buying it for her :) i'm donating this to her, she should be glad, could've given her nothing i guess
21:34urjaman: so yeah okay it works
21:34NanoSector: don't even personally know the woman, dad does
21:34urjaman: for free things are what they are :P
21:35NanoSector: her laptop broken, i'm willing to take a look if needed
21:35NanoSector: in the meantime she can use this thing :P
21:37NanoSector: it's an Athlon 2800+
21:37NanoSector: poor dvd drive, keeps getting stuck
21:50RSpliet: Athlon XP 2800+ isn't capable of 720p h264 in software
21:50urjaman:goes to fetch the sempron ...
21:52RSpliet: and I don't think there's a NVIDIA GPU capable of accelerating h264 in AGP form-factor
21:54urjaman: yeah, but i didnt ever say h264 :P
21:54RSpliet: you did say youtube
21:54RSpliet: and that's either h264 or something more fancy
21:59RSpliet: anyway, fx5200 is notoriously slightly more broken than other FX cards
21:59RSpliet: with nouveau
21:59NanoSector: should i be concerned?
22:00RSpliet: there's one way to find out
22:00NanoSector: and that's waiting for Apt to finish installing a proper desktop...
22:00NanoSector: like geez apt is slow
22:00NanoSector: it's slow on a fast system, it's especially slow on a system like this
22:09urjaman: okay, now to figure out wtf is going on with my coreboot on this K8V-X :P
22:09NanoSector:is still waiting on apt
22:09NanoSector: apparently installing LIGHTWEIGHT *cough* XDE has to take 1.5 hours
22:16JodaZ: i disagree, a athlon 2800 should be able to do 720p in typical bitrates
22:16urjaman: okay i'm not going to get this thing running today, unrelated to anything here :P
22:16NanoSector: apparently installing LXDE also requires that apt rebuilds the boot image thrice
22:16JodaZ: i had a 1.6ghz athlon myself, and it could just barely play full hd video
22:16urjaman: but I have very empirical experience than a Sempron 2800+ does whatever VLC youtube plugin gives (its a 1280x720 video, that i remember :P) in windows XP with an FX5200 or FX5700LE
22:17NanoSector: i actually had XP on here
22:17NanoSector: OMG IT'S DONE
22:18urjaman: obviously not like any bitrate, but youtube isnt the highest bitrate...
22:18NanoSector: would XP be a better choice for web browsing on this than linux?
22:19NanoSector: assuming i actually get display output
22:19NanoSector: oh there we are
22:19urjaman: I dont want to answer really, the answer conflicts with my ideologies :P
22:19NanoSector: let me ask it differently :P
22:19NanoSector: would drivers on Windows XP allow for a better multimedia experience on the internet? :D
22:23urjaman: actually not so sure for just web browsing
22:23NanoSector: 480p on youtube works
22:23urjaman: since the modern browsers will be just bad regardless of the OS
22:23NanoSector: reaaaaaaaaally slowly though
22:23NanoSector: the interface that is
22:23NanoSector: video is fine
22:24NanoSector: firefox is reasonably fast
22:26NanoSector: yeah i dunno
22:26NanoSector: i honestly feel like windows xp was legions faster
23:47Tom^: it probably was hw accelerating stuff at a better rate
23:47Tom^: windows drivers aint so bad
23:47Tom^: the os is a different idiology and debate.