00:05mmarc__: Dear All: An opening for position of HPC engineer is available in Saint Petersburg. The job is to develop compute kernels in GCN ISA. Official long-term contract. Please write me for more details.
00:33BootI386: imirkin: It just crashed :)
00:36BootI386: I did this
02:33orbea: imirkin: fyi, it was a xorg bug. https://cgit.freedesktop.org/xorg/xserver/commit/?id=76732f498f1e73fb081841a04faf068660f3d5c7
02:34orbea: colors are still wrong, but it boots
06:38orbea: Anyone mind trying this windows trace via wine with d3d9-nine with nouveau and/or amd to see if this issue is just me or not? http://ks392457.kimsufi.com/orbea/stuff/trace/Tales%20of%20Zestiria.trace.xz Needs a windows version of apitrace. http://ks392457.kimsufi.com/orbea/stuff/trace/Tales%20of%20Zestiria.trace.xz See this issue report for pictures https://github.com/iXit/Mesa-3D/issues/308 and apitrace +
06:38orbea: wine instructions https://github.com/apitrace/apitrace/wiki/WINE
06:39orbea: replaying it with wine opengl is fine
08:14karolherbst: okay... skeggsb https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/commit/?id=3c66c87dc96b3113b5ee84604800c2aabbb48994 this broke backlight on the P50 in dedicated mode
08:16karolherbst: or maybe it was the other commit?... weird, it seems like that issue isn't 100% reproducible, which is kind of painful right now
08:49karolherbst: ohh I see what is going on here... messy
08:52skeggsb: karolherbst: change the "or = nv_encoder->or;" in nva3_get/set_intensity() to "or = ffs(nv_encoder->dcb->or) - 1"
08:53karolherbst: skeggsb: it already fails in backlight_init
08:53skeggsb: yes, it will, but probably for the same reason
08:54karolherbst: I added those acquire/release calls for testing
08:54karolherbst: that was it
08:54skeggsb: nvidia mentioned to me that eDP is a special case for the SOR assignment due to the fixed wiring of GPIOs
08:54skeggsb: it's on the list :P
08:55karolherbst: well right, but now all backlight control is broken :p
08:59karolherbst: skeggsb: something like this? https://gist.github.com/karolherbst/dec34336780088fe20c1d859edf41136
09:24karolherbst: skeggsb, mupuf: I've got a maxwell1 with DDR3 memory :)
09:27karolherbst: wondering how reclocking works on those, because maxwell usually only have GDDR5
09:52Booti386: imirkin: Sorry, http://termbin.com/o6tt
11:01karolherbst: mupuf: I don't know if I should laugh or cry regarding that "new" intel compute thing...
11:41pmoreau: karolherbst: :o
11:42pmoreau: If they are going to upstream all the stuff needed for OpenCL C 2.0 and OpenCL C++ 1.0 in clang and LLVM, then I’m happy.
11:52karolherbst: pmoreau: I am sure they won't
11:53karolherbst: or maybe they do?
11:53karolherbst: who knows
11:53karolherbst: pmoreau: thing is, it isn't from the OTC
11:54pmoreau: The “tag” on Phoronix: “61 Minutes Ago - Intel - BIG + BETTER INTEL OPENCL - 3 Comments” --'
11:57pmoreau: It isn’t from the OTC? Everything seems to point back to 01.org, which is the OTC webpage, isn’t it?
11:59karolherbst: maybe? Maybe not
11:59karolherbst: it doesn't look like something the OTC would do
11:59karolherbst: and my first though was: even if you news site tells me it was the OTC, I am sure it wasn't
12:09pmoreau: The GitHub repos point back to the 01.org webpage for an explanation about this new OpenCL stack.
12:10karolherbst: pmoreau: read carefully ;)
12:11karolherbst: "the Open Source Technology Center at Intel created Beignet* OpenCL driver" vs "In 2016, Intel started"
12:12pmoreau: I have been wondering about that
12:12karolherbst: but yeah, it is no proof
12:12karolherbst: just a feeling I have
12:33mupuf: pmoreau: I doubt too, the people working on that are not OTC.
12:34mupuf: I will not comment any further :D
12:59tomeu:pities whoever has to package that thing
13:00tomeu: they don't have that many changes on top of LLVM though
13:01tomeu: quiet a few more for clang though: https://github.com/intel/opencl-clang/tree/master/clang_patches/releases/4.0.1/patches
13:04mupuf: tomeu: let's hope they get tired of rebasing them ;)
13:04tomeu: they are still at 4.0, so maybe they won't rebase them at all :p
13:48imirkin: BootI386: you probably wanted || instead of &&
13:49imirkin: (silly Boole and his logic...)
13:52RSpliet: I always wondered whether Boole was as lazy as those specific operators...
16:10BootI386: … No, not Boole's fault. Just mine, forgetting my coffee…
16:13BootI386: Well, my midnight coffee actually
16:33Booti386_: nvc0_blit:1573 - Src box (x = 0, y = 0, w = 1920, h = 1080), resource (w = 1920, h = 1080)
16:33Booti386_: nvc0_blit:1574 - Dst box (x = 0, y = -480, w = 1920, h = 1080), resource (w = 600, h = 600)
16:33Booti386_: imirkin: I'm not really sure what I'm supposed to do with that...
16:34imirkin_: flip, usually
16:34imirkin_: also ...
16:34imirkin_: iirc that was supposed to be a negative height
16:34imirkin_: let's see what marek thinks in #dri-devel
20:48otherflow: Hello, I try to compile Linux Kernel from https://github.com/skeggsb/linux but i get this error : "error: code model kernel does not support PIC mode". I've tried to disable PIE mode with the CFLAGS -fno-pie added to the KBUILD_CFLAGS variable on the Makefile but it don't solve
20:48otherflow: my problem
20:49otherflow: Someone can help me ?