01:44 imirkin: Lyude: mesa patches pushed, looking at piglit now
01:44 Lyude: yay
02:17 imirkin: Lyude: going to apply a small patch to your isolines test: https://hastebin.com/vibesidudi.diff
02:17 imirkin: that way the line slants up
02:18 imirkin: Lyude: i'm also going to switch up the GS to take in tris and output points/lines
02:19 Lyude: oh, didn't realize that was actually valid to do when drawing in GL_POINTS mode
02:19 Lyude: anyway, fine by me
02:19 imirkin: well, i'm going to switch the draw primitive to be GL_TRIANGLES
02:19 imirkin: which will have absolutely no effect on the outcome of the GS
02:20 Lyude: ah
02:23 imirkin: also, i'm tabifying your shader_runner.c changes
02:23 Lyude: did my editor do the wrong tabbing?
02:24 Lyude: *indenting
02:24 imirkin: you used 8-space indents instead of tabbed indents
02:25 Lyude: ah, just wanted to know so I could fix my .lvimrc :)
02:25 imirkin: there's a .editorconfig file
02:25 imirkin: and i believe there's a vim plugin to process those too
02:26 Lyude: oh neat
02:26 imirkin: i don't really know the state-of-the-art for vim. last time i used vi, it was elvis ... not sure where vim came from tbh.
02:27 imirkin: [i guess that's not wholly true. i have vim now, and use it for very minor editing tasks, but never anything requiring any sort of proficiency]
02:27 Lyude: someone suggested it to me once and I made the mistake of trying it out
02:27 Lyude: now none of the other editors feel close enough for vim for me to ever switch
02:28 imirkin: kinda like me and aterm
02:28 imirkin: none of the other term emulators are QUITE right
02:29 imirkin: they all fail at one (but usually many) mouse-related interactions, esp related to triple-click, drag, etc.
02:30 imirkin: Lyude: piglit tests pushed. please double-check that the isolines and gs ones still pass.
02:30 Lyude: sure thing
02:38 Lyude: imirkin: everything passes just fine
02:38 imirkin: ok great
02:39 imirkin: will you be tackling any of the other gm200 projects?
02:39 imirkin: esp given your increased familiarity with the tooling?
03:07 Lyude: imirkin: yeah, I was planning on it :)
03:07 imirkin: ok cool
03:07 imirkin: might want to do AMD_vertex_shader_layer & co -- that one's nearly done
03:08 imirkin: there's one final bit i never figured out
03:08 imirkin: but it's tricky without hw access
03:14 Lyude: yeah sure! I'll also probably be working into looking at getting GL_NV_fill_rectangle implemented on other things as well
03:16 imirkin: ah ok. didn't realize you had a deeper purpose than getting your feet wet with it
03:17 Lyude: I wasn't originally but airlied wanted to see if we could implement it on other platforms I'm pretty sure
03:18 imirkin: i don't think so
03:18 imirkin: i assume you saw mareko's reply
03:18 Lyude: right, yeah
03:18 imirkin: other platforms will want a rectangle primitive type of thing
03:18 imirkin: and no interactions with clipping, tess/geom, etc
03:19 Lyude: then nvm then :P
03:59 imirkin: karolherbst: ok, pushing your patches. thanks!
03:59 imirkin: made a minor adjustment to the first one
04:08 karolherbst: imirkin: thanks, but now I am going to sleep :D
04:10 imirkin: nite
09:55 R3d_sky: hello! is there any way to dump signed firmware for gp10x and use them locally?
10:30 karolherbst: R3d_sky: what do you mean by that? the firmware got released already
10:32 R3d_sky: firmware release was only for gp100 i thought?
10:34 karolherbst: R3d_sky: https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/tree/nvidia
10:34 R3d_sky: oh wait i chose history lol
10:34 R3d_sky: so any 3d accel support yet?
10:35 karolherbst: yeah
10:36 R3d_sky: ooh nice
10:37 R3d_sky: if only we had reclocking :( I have another PC with a 750ti but family doesnt like linux.
10:37 karolherbst: mhh, 750ti should work good enough with 4.10
10:37 karolherbst: and recent mesa
10:38 karolherbst: you even get reclocking on it
10:38 R3d_sky: ik, but dad worked at microsoft and as such is generally averse to linux
10:38 karolherbst: ohh, that you mean by family
10:38 R3d_sky: yes
10:38 karolherbst: uhm.... he isn't part of the more costumer related department isn't he? Cause azure runs mostly on Linux servers from what I've heard :O
10:39 R3d_sky: he was back then before vista. He quit slightly before XP release
10:40 karolherbst: mhh, a lot has changed at microsoft and except the manager and customer relation departments most aren't against Linux, that's why this surprises me a little
10:40 R3d_sky: i didnt mean he was at any department
10:40 karolherbst: ohh okay
10:40 R3d_sky: last i heard he was promoting ms SQL and access and vb
10:40 karolherbst: mhh, (ms SQL was/is ported to linux)
10:41 karolherbst: but yeah
10:41 karolherbst: microsoft changed "its" view on linux publicly the last year or so
10:41 R3d_sky: yes
10:41 R3d_sky: but he was there during the ballmer ersa
10:42 karolherbst: uhhh
10:42 R3d_sky: and as such continues
10:43 R3d_sky: and he doesnt like linux since its not a commercial product... idk really
10:44 karolherbst: ...
10:44 karolherbst: okay
10:44 R3d_sky: hes like "OMG code from unknown internet people virus"
10:44 karolherbst: :D
10:44 karolherbst: "unknown"
10:44 R3d_sky: yes ikr
10:46 R3d_sky: he literally didnt even let me use his pc for editing the rpi config.txt (boot config) because "virus"
10:46 R3d_sky: hes mas
10:46 R3d_sky: *mad
10:47 xerpi: if I'm not mistaken it looks like nv50 has: (1 base (primary), 1 cursor and 1 overlay per crtc) * 2 crtcs
10:49 karolherbst: R3d_sky: ..... I am sure he also says MS is mad now, because they run Azure on Linux servers :D
10:49 R3d_sky: yep lol, im gonna tell him that now
10:49 xerpi: and what about that ubuntu thingy on windows 10?
10:50 karolherbst: R3d_sky: or this: :D https://docs.microsoft.com/en-us/azure/virtual-machines/virtual-machines-linux-endorsed-distros?toc=%2fazure%2fvirtual-machines%2flinux%2ftoc.json
10:51 R3d_sky: well he also thinks GMOs are universally bad and believes articles that say the Internet is bad for you so chalk it up to a general lack of reasoning?
10:52 karolherbst: no idea, Microsoft knows they lost on Servers. The only places where you will see MS servers are in coorperations running Exchange and Active Directory... everything else -> meh
10:52 karolherbst: I wouldn't be surprised if they would even port this to Linux as well
10:54 R3d_sky: my CMOS battery drained and hes pegging that on Linux too. I wish they sell medicines for brain malfunctions
10:57 xerpi: looks like skeggsb/linux isn't updated. so which kernel version does skeggsb/nouveau target? 4.11 or 4.12?
11:19 pmoreau: xerpi: If you scan through his commits, you’ll see some that look like: "drm-next 65d1086c44791112188f6aebbdc3a27cab3736d3": that’s the version begin targeted.
11:22 pmoreau: xerpi: Or you can run something like `git log --oneline --grep="drm-next" | sed -e 's/^[a-z0-9]\{8\} drm-next //' | head -1` in skeggsb/nouveau to automatically extract the latest version of drm-next being used.
11:26 xerpi: pmoreau, nice, thanks!
16:05 RSpliet: pmoreau: think you can give my current branch a spin on NVC0?
17:00 pmoreau: RSpliet: Sure! Got some improvements for NVC0? :-)
17:01 RSpliet: Added the default training pattern, so boot should work now
17:01 pmoreau: Cool!
17:01 RSpliet: didn't go through your trace in detail further, so middle perflvl is still a bit fingers-crossed. It'd be nice to have some confirmation of all though
17:02 RSpliet: (nothing changed for the NVC8)
17:08 pmoreau: (I haven’t got the trace either for NVC8… :-/)
17:35 RSpliet: pmoreau: oh actually don't bother testing yet on your NVC0, there's something interesting about it
17:36 RSpliet: on third thought, maybe you could... :-P
17:37 RSpliet: thought training might fail, but should be fine
17:43 pmoreau: RSpliet: So, what’s your final thoughts: should I test it or not? :-p
17:44 RSpliet: just go for it! #yolo
17:54 imirkin: dboyan_: btw, what's the readiness state of your rcp/rsq64 patches?
18:05 karolherbst: this looks right, doesn't it? I am never sure about if I have to negate or not :("slct(true, false, a) -> set(eq, a, 0)"
18:10 ZombieChicken: Is there any way to tell why the nouveau driver seems particuarly slow for a certain chip? I have a (from lspci) GF108 [GeForce GT 620] and 3D programs seem rather slow. also, the fan makes a ton of noise when running this chipset
18:10 karolherbst: ZombieChicken: easy, nouveau can't reclock your GPU
18:10 ZombieChicken: So basically if 3D performance is important, stick to the proprietary driver?
18:11 ZombieChicken: or is there a way to get it to reclock the GPU?
18:12 karolherbst: mhh maybe RSpliet can tell you how you can help with that
18:13 RSpliet: GF108... no not at the moment
18:13 RSpliet: certainly not a DDR3 card, which I suspect the GT620 is
18:15 ZombieChicken: Alright. Thanks
18:32 RSpliet: karolherbst: I literally have this working for one card and one card only. I'm eating my nails for pmoreau's rest results, but unlikely that it's of any use to non-devs at this point
18:32 RSpliet: (and I need a pile of old Fermis delivered to my doorstep...)
18:33 karolherbst: I guess you only look into memory reclocking right now?
18:34 RSpliet: mostly yes
18:35 RSpliet: that is, when I find the energy to push forward again
18:36 karolherbst: well, I guess it won't be too complicated to support more GPUs, if we can do it for one, except that stuff could change by a lot
18:36 karolherbst: but I assume that most differences are due to flags inside the vbios
18:36 RSpliet: that's... mostly precisely what this stuff is all about
18:37 karolherbst: there are still missing bits on kepler+, but I need cards which don't reclock yet...
18:37 RSpliet: took me 5 DDR3 GT215's to get that pretty reliable
18:37 RSpliet: I have one GDDR5 Fermi
18:37 karolherbst: RSpliet: sometimes it is even stuff like: if there is no disp engine, don't touch this reg....
18:38 RSpliet: karolherbst: I know that, but those conditions aren't like... documented publicly
18:38 RSpliet: so without significant amount of samples there's no way to verify that
18:38 karolherbst: true
18:38 RSpliet: hence, if it works for one card, it works for one. If it works for two, it works for three. If you fix it for a fourth, you'll start getting places
18:39 karolherbst: I would go for: push those things with a remark _what_ works and that everything else most likely doesn't and let user report and get you traces or so
18:39 karolherbst: I don't know how many Fermis mupuf has
18:39 RSpliet: at this stage I need traces with faked vbios'es
18:39 RSpliet: remote work on this is nearly impossible, I can't check the display to see how bad I fucked up
18:40 karolherbst: well, but you can check if anything gets rendered at all
18:40 karolherbst: as long as the GPU doesn't hang, it's a step forward
18:41 karolherbst: or ask mupuf to setup a cam and a display :D
18:42 RSpliet: like ye olde coffeecam?
18:42 karolherbst: sure
18:45 imirkin: RSpliet: what if i sent a few DDR3 GF108's your way ... helpful?
19:02 RSpliet: imirkin: depends. I got one here, but can't use it because it only has DP
19:03 RSpliet: on the next Amazon order I'll piggyback a DP->DVI converter
22:40 Horizon_Brave: happy saturday everyone
22:43 Horizon_Brave: So...apparently Broadcom is as if not worse than nVidia when it comes to open source support xD