00:00 ShimmerFairy: I wasn't aware of what silentoldconfig does since I always do 'make oldconfig'. In any case, if you have distro-specific info, I'm sure that'll be more helpful than me coming from a gentoo perspective.
00:01 ShimmerFairy: I only mentioned the oldconfig stuff so you wouldn't have to configure everything manually (and instead just migrate and update your existing config info)
00:02 feliksk: the goal of this is to figure out if the fan noise issue (on my gt 440) has been fixed in the latest kernel
00:02 feliksk: i think its a problem cause in 3.10.x my card was running silently, but when i went to the 4 kernels (like 4.4.23) the fans are really loud
00:03 feliksk: im not running X and when i run the kernel with nomodeset it is silent so i think its an issue with the nouveau kernel module
00:04 ShimmerFairy: that could well be. I'm not a nouveau developer (yet?), so I can't comment much on testing it, I'm just someone who runs Gentoo of all distros :)
00:06 iterati1: feliksk: you have tried the nvidia driver and found that the issue is nouveau in fact ? Or it's a hardware problem ?
00:08 feliksk: i havent tried the nvidia driver (and will never use it due to not being free and having no high res text console support), but i do know that if i install a 3.10.x kernel right now the noise will probably be gone, if that means anything
00:09 feliksk: should i try the nvidia driver?
00:09 iterati1: check also: https://wiki.archlinux.org/index.php/nouveau#Fan_Control
00:10 feliksk: someone here told me that 3.10.x might (dont know for sure) of not even had fan control
00:13 feliksk: lspci | grep VGA says: 01:00.0 VGA compatible controller: NVIDIA Corporation GF108 [GeForce GT 440] (rev a1)
01:38 feliksk: i failed :P my kernel just reboots
01:38 feliksk: im going to try compiling some 3.10.x kernel
02:06 feliksk: 3.10.x didnt compile, crap this is complicated
03:24 imirkin: airlied: so ... based on a brief look at the CTS failures, it looks like TF is somehow busted
03:24 imirkin: in both it uses TF to report the shader outputs. so i'm thinking that's why it fails.
03:25 imirkin: this is most sad, since TF is the thing i understand by far the least
03:28 imirkin: aaaand my personal guess is that this is due to the fact that outputs are no longer sorted
03:28 imirkin: let's see if nha's patch magically fixes all this
03:28 airlied: imirkin: you do appear to pass most of the TF tests
03:34 imirkin: ok ... progress, but still wrong
03:37 imirkin: (gdb) p *pso
03:37 imirkin: $2 = {num_outputs = 1, stride = {1, 0, 0, 0}, output = {{register_index = 255, start_component = 0, num_components = 1,
03:37 imirkin: that's ... probably not great
03:39 airlied: doesn't seem ideal
03:40 imirkin: p outputMapping[30 - VARYING_SLOT_VAR0]
03:40 imirkin: $12 = 0
03:40 imirkin: that seems better
03:42 imirkin: but ... wrong
03:45 imirkin: nm, i broke in the wrong place =/
03:45 imirkin: although still odd it was trying to use one shader's TF settings with another
04:09 imirkin: airlied: i'm like 75% sure it's an insufficient fix, but can you try applying https://patchwork.freedesktop.org/patch/116462/ and see if it helps?
04:09 imirkin: i'm having trouble getting apitrace to pick up my LD_LIBRARY_PATH...
04:09 imirkin: (specifically qapitrace)
04:12 airlied: imirkin: will do when I get near the box again
04:39 imirkin: i don't have high hopes. i'm getting crap when i read stuff out via eglretrace
04:41 gnurou: imirkin: your branch starts, but fonts are rendered strangely: https://drive.google.com/file/d/0B4Wi9SO0F_89V2JXSVBKTENBcmM/view?usp=sharing
04:43 gnurou: imirkin: log: http://pastebin.com/1EYAyyYu
04:44 imirkin: gnurou: ah, that looks familiar....
04:44 imirkin: that happens when you mess up alpha + composite + something
04:44 imirkin: i'll take a look
04:45 gnurou: great, let me know when you need further testing
04:45 imirkin: sadly i'm off to bed now...
04:46 imirkin: airlied: hm, and the geometry shader one you pointed out gets a shader trap. *super*.
04:48 airlied: imirkin: wierd
04:49 imirkin: gnurou: or perhaps the copy engine isn't copying ... dunno.
04:49 imirkin: airlied: gr: SHADER 90000100, sph: 0x000100, stage: 0x10
04:49 imirkin: i don't really remember when this happens =/
04:53 imirkin: airlied: anyways, the deqp gles3 transform feedback works 100%, so ... we're doing something right
04:53 imirkin: airlied: and actually ignore me. nha's patch won't change anything
04:54 imirkin: we populate stuff based on the first/last of the range
04:54 imirkin: in that out array
04:54 imirkin: airlied: so that patch shouldn't fix anything
04:56 imirkin: ooooooooohhhhhhhhhhhhhhhhhhhhhhhhhhh
04:56 imirkin: i think i see what's up.
05:19 imirkin: w00t
05:19 imirkin: take that, sleep!
05:22 airlied_: imirkin: figured it out? I fell off irc for a minute
05:22 imirkin: airlied_: yes. https://patchwork.freedesktop.org/patch/116553/
05:24 airlied_: imirkin: you got one bit of whitespace change in there
05:25 imirkin: yeah, semi-on-purpose
05:26 imirkin: the trailing whitespace was annoying me
05:26 imirkin: and it was adjacent to areas i was changing
05:26 airlied_: imirkin: ah cool
05:27 imirkin: the upshot is that if you don't declare an output, it comes in as 0
05:28 imirkin: er, an input
05:28 imirkin: i wonder what happens if you don't declare an output, does the next stage get a 0?
05:28 imirkin: if so, that'd let me do something semi-close to what mareko was able to do for radeonsi
05:31 imirkin: anyways, pretty happy i was able to notice the missing bits in the shader header
05:31 imirkin: i'll deal with the geom failure one later
15:05 karolherbst: anybody here with a nvac gpu?… except me?
15:05 imirkin: pmoreau does, i believe
15:06 imirkin: and maybe rspliet, if he hasn't thrown it out yet
15:06 karolherbst: good, a user has an issue with a workaround we apply for those, but I couldn't see that issue on mine
15:07 karolherbst: totally forgot mine is a nvac as well...
15:30 imirkin: gnurou: btw, what "thing" generates the menu that you were having issues with? can i just run it as an application, or do i have to install some giant gnome/kde/whatever 10000-package complement as well?
17:44 pmoreau: I do have an NVAC indeed, and I wanted (a long time ago) have a look at Ben’s patch, but never did. But with poma’s email, I should try both.
19:02 imirkin_: does anyone have the game 'batman_arkham_origins'?
19:03 karolherbst: uhh, I might have
19:03 imirkin_: can you check? if so, i might want you to test a patch at some point
19:03 karolherbst: mhh nope, I have the two others
19:03 imirkin_: (based on mareko's recent one)
19:03 karolherbst: arkham city and arkham asylum
19:04 karolherbst: both windows only though
19:04 imirkin_: my theory is that zero-valued exports can be removed
19:05 imirkin_: you have 'the witcher 2' though, right?
19:05 imirkin_: but it like totally doesn't work?
19:05 karolherbst: yes
19:05 karolherbst: well
19:05 karolherbst: it does when I use the gog version
19:05 imirkin_: ah ok
19:05 karolherbst: ohh wrong,
19:05 karolherbst: the with the one disable extension
19:05 imirkin_: presumably still uses the same shaders?
19:05 karolherbst: which isn't any officialy
19:05 karolherbst: let me check again
19:06 imirkin_: i kinda assume it would. it'd be crazy otherwise
19:06 karolherbst: nope, I am sure they have to be different
19:06 karolherbst: maybe same shaders, but compiled differently
19:06 imirkin_: it was only ARB_buffer_storage, right?
19:06 karolherbst: yes
19:06 imirkin_: that has minimal to no effect on how you structure shaders
19:06 imirkin_: it has to do with data management
19:06 karolherbst: I thought they check againt it though
19:07 imirkin_: you can't check against it in a shader
19:07 karolherbst: ahh I see
19:07 karolherbst: ohh, k
19:07 imirkin_: of course you could use diff shaders depending on whether it's available
19:07 imirkin_: but i don't see a great reason why you would
19:07 imirkin_: (doesn't mean there aren't lots of bad ones)
19:09 karolherbst: so, what should I test?
19:10 imirkin_: nothing
19:10 imirkin_: (not yet)
19:10 karolherbst: k
19:10 imirkin_: i just may have a patch in the future
19:11 karolherbst: what is a "zero value export" though? a value you export which is 0 for sure?
19:11 imirkin_: out vec4 foo;
19:11 imirkin_: foo = vec4(0);
19:11 imirkin_: useful, right? :)
19:11 karolherbst: well
19:11 karolherbst: won't that be like 2 instructions top?
19:12 imirkin_: that's a minor point
19:12 imirkin_: the major point is that it's less data being shuffled around
19:12 karolherbst: I see
19:12 hakzsam: imirkin_: yeah, marek did something similar for readeonsi recently
19:12 imirkin_: no clue how big of a difference it makes
19:12 karolherbst: especially because nouveau isn't smart about export layouts anyway, so it would in most cases do 4 movs + 1 export + massive gpr overusage
19:13 imirkin_: hakzsam: the thought is inspired by what mareko did :)
19:13 hakzsam: yep
19:13 imirkin_: hakzsam: i don't think we can have different defaults
19:13 imirkin_: hakzsam: it does seem like undeclared vertex inputs are retrieved as 0's
19:14 imirkin_: one might guess that the same thing happens in other stages.
19:14 hakzsam: yeah, makes sense
19:21 karolherbst: mhh, 4.8 seems broken on my nvac, fun
19:22 karolherbst: ohhh
19:22 karolherbst: didn't install modules
19:50 imirkin_: gnurou: btw, stupid question - did you get any errors in dmesg?
19:50 imirkin_: gnurou: i ran some rendercheck tests this morning on a GM107 and wasn't able identify any obvious issues
21:48 hakzsam: well, no regressions on gm107 and gm206 between mesa-12 and mesa-13-rc1 with piglit, passrate of 99.47% :)
21:49 hakzsam: next step: kepler
23:48 gnurou: imirkin_: nope, dmesg was silent during that time
23:48 gnurou: imirkin_: and the menu is lxde (pretty small)
23:48 imirkin_: ok... i'll see what it takes. i might just try to run some random apps and see what happens
23:49 imirkin_: i don't suppose you have a weirdo 30-bit screen or anything of the sort, right?
23:49 gnurou: just standard 1080p
23:49 imirkin_: k, just checking :)
23:49 gnurou: I've used it with plenty of other cards/boards that did not show this issue
23:50 imirkin_: yeah, just getting some hypothetical oddness out up front, just in case
23:50 imirkin_: i'm sure i just did something dumb
23:51 imirkin_: i only tracked down 1 shader error in my translation from kepler -> maxwell. this seems way too low.
23:51 imirkin_: probably more running around in there
23:52 imirkin_: hakzsam: do you have an actual screen hooked up to your GM107? if so, mind testing my ddx? i've only tested it via x11vnc, which might not give the full picture