02:33imirkin: hakzsam: oh, fun. that copy-pixels test also fails on llvmpipe.
07:31RSpliet: mupuf: I'm not going to try and implement full preemption on any GPU, my measurements on kepler are all the motivation I need
07:32RSpliet: imirkin_: I'm back in England on the 26th, before that I can't fiddle with the NV34/NV4B that I have
07:37hakzsam: imirkin, oh okay
07:37hakzsam: any ideas why it fails?
08:00mupuf: imirkin_: everything is fine in the end: http://piglit.mupuf.org/results/kepler/html/problems.html
08:20Booti386: imirkin_: Here it is ! :) http://www.megafileupload.com/s8o2/egl-test.tar.gz
08:21Booti386: Crash at main.c:326 in glClear()
08:32hakzsam: mupuf, you can add gk208 too https://people.freedesktop.org/~hakzsam/piglit/kepler/
08:56mupuf: hakzsam: done: http://piglit.mupuf.org/results/kepler/html/problems.html
11:49rhn: I checked out the work I did on Kepler before - it's not part of rnndb for reasons I can't remember, Instead it's parsing
11:50rhn: the diff is relative to an old, unpublished fork (<Sep 2014), but it's small enough to glance over: https://gist.github.com/rhn/1aebb91440a08197aea784e81488a760
11:59karolherbst_work: skeggsb: do you plan to make a second PR with smaller changes?
11:59karolherbst_work: or a second at all? (my stuff should be done in 4.9 anyway, asking about my iccsense patch)
13:37RSpliet: mwk: tl;dr, NVIDIA will use RISC-V to replace their falcons probably
13:38karolherbst_work: "rich OS support"
13:39mwk: ... should I prepare a disassembler?
13:40RSpliet: there should already be a gcc and llvm toolkit from Berkeley available
13:41karolherbst_work: well, will be interessting to know how they secure those
14:16RSpliet: karolherbst_work: probably the same way, there is nothing intrinsic to the ISA to do tuis signed firmeare protection
14:18karolherbst_work: mhh, right
15:05Booti386: imirkin: Here it is ! :) http://www.megafileupload.com/s8o2/egl-test.tar.gz
15:43imirkin_: Booti386: iirc i tried to download that yesterday, and that site didn't work
15:43imirkin_: Booti386: try filebin.ca maybe?
15:43Booti386: Sad :(
15:44Booti386: imirkin_: http://filebin.ca/2o9nPQRUN1Hn/egl-test.tar.gz
15:45Booti386: It crashes in glClear(), but doesn't crash if I remove GL_DEPTH_BUFFER_BIT
15:45imirkin_: thanks. i can't look now, but i'll try tonight or over the weekend
15:45Booti386: Ok, thank you ! :)
15:46imirkin_: Booti386: i recommend filing a bug at bugs.freedesktop.org since (a) others could look and (b) i'm slightly less likely to forget
16:18mwk: RSpliet: actually, the protection *is* intrinsically tied to the falcon ISA
16:19mwk: but given that RISC-V is customizable, that's not a big problem...
19:42Booti386: imirkin: Lazyness :(
19:43imirkin_: ah yes. i know it well.
19:47Lekensteyn: imirkin_: forgot to mention that patch 4 also fixes some laptops that currently do not power off at all, see for example https://bugs.freedesktop.org/show_bug.cgi?id=91408
19:48imirkin_: Lekensteyn: linux kernel has a "no regressions" policy
19:48imirkin_: Lekensteyn: i'm sure what you're doing is moving things into the right direction
19:48imirkin_: but fact is, it'll regress a bunch of people
19:52Lekensteyn: imirkin_: do you have a laptop with HDMI audio too?
19:53Lekensteyn: after the pci/pm patches, runpm will break anyway on the affected devices
19:54Lekensteyn: the alternative is to add a quick pci_d3cold_allow(pdev, false) somewhere (after merging pci/pm) and then replace it by patch 4 after fixing the audio driver
19:54Lekensteyn: uhh, pci_d3cold_disable that is
20:03imirkin_: Lekensteyn: i do not.
20:04imirkin_: Lekensteyn: but tons of people do. afaik it's an extremely common configuration.
20:06imirkin_: and for many of those people, the sole reason they use nouveau is to get it to runpm suspend
20:06karolherbst: Lekensteyn: well, they shouldn't break with either patch :p
20:07imirkin_: karolherbst: does your laptop show the hdmi audio subfunction?
20:07karolherbst: mhh, no idea, I have no port on the nvidia gpu
20:07imirkin_: run "lspci -d 10de:" ?
20:08karolherbst: I have two audio devices, but both on intel
20:08karolherbst: shows only the gpu
20:08imirkin_: hm ok
20:08imirkin_: is it a 3d controller?
20:09imirkin_: or a vga adapter
20:09karolherbst: GA compatible controller: NVIDIA Corporation GK106M
20:09karolherbst: well, the vbios contains a bunch of adapters though
20:09imirkin_: you mean connectors?
20:10karolherbst: it's a mxm card, so things might be a bit "odd"
20:10imirkin_: also there's probably a dock
20:10imirkin_: which would expose those connectors
20:10karolherbst: I don't have one
20:10karolherbst: but I am sure the board might support it
20:11karolherbst: but have like 3 display adapters ports on my intel one already
20:11Lekensteyn: imirkin_: I've been using bbswitch for a long time because I just needed power saving but am moving to nouveau because I sometimes need an external monitor
20:11Lekensteyn: I agree that runpm is important, but it'll be broken if nothing is done now
20:11karolherbst: imirkin_: yeah, 10 connectors
20:11Lekensteyn: if you have no audio function, can you try to get one?
20:12karolherbst: one DCB entry too
20:12karolherbst: Lekensteyn: that's no excuse :p
20:12Lekensteyn: runpm suspend/resume and check with lspci -d10de: -H1
20:12Lekensteyn: karolherbst: yeah, a bit lame :p
20:13karolherbst: Lekensteyn: only shows the gpu
20:13karolherbst: it's a mxm card, so
20:13karolherbst: it's removeable and everything
20:15imirkin_: Lekensteyn: well, if you demonstrate that i'm wrong, and that across generations, having the hdmi adapter show up by default is rare, then i'm all for it
20:15imirkin_: fyi such an adapter is available since GT215
20:15karolherbst: I have a mbp with an hdmi adapter on the nvidia gpu
20:16karolherbst: but it's from work :/
20:16Lekensteyn: I could use the very slow, large Launchpad bug where people uploaded lots of acpidumps, dmidecode output and lspci
20:16imirkin_: previous to that, hdmi audio came in via a sp/dif connector
20:16karolherbst: but I think I might be able to boot a live linux on it
20:34imirkin_: skeggsb: do you know if hansg is still working on the opencl stuff?