18:47 tobijk: somebody already investigated render corruptions with generations after kepler? https://homepages.thm.de/~tjkl80/Screenshot_20170814_204553.png
18:51 karolherbst: tobijk: this is valley, right?
18:51 imirkin_: tobijk: sounds familiar. iirc MESA_DEBUG=flush helps
18:52 pmoreau: imirkin_: Thinking of this one? https://bugs.freedesktop.org/show_bug.cgi?id=100177
18:52 tobijk: imirkin_: karolherbst: yep its valley, but i have other "apps" where corruptions are visible
18:53 imirkin_: pmoreau: yes, i am.
18:53 karolherbst: tobijk: did you check with NV50_PROG_OPTIMIZE=0?
18:53 tobijk: karolherbst: not yet with valley, sec will test
18:53 karolherbst: but this hardly helps, at least for me on kepler
18:53 karolherbst: *rarely
18:55 tobijk: yep the corruptions are still visible with MESA_DEBUG=flush or NV50_PROG_OPTIMIZE=0
18:55 tobijk: so its likely some insn not emmited properly?!
18:56 karolherbst: could be anything
18:56 karolherbst: check dmesg
18:56 tobijk: dmesg is clean
18:57 karolherbst: I could check with my kepler if everything is alright
18:57 karolherbst: just currently doing a CTS full run and it might take some time until I can do anything
18:57 tobijk: karolherbst: its a longer standing issue, having it since switching from kepler to pascal
18:58 tobijk: i would suspect kepler is just fine :)
18:58 karolherbst: most likely
18:59 tobijk: karolherbst: but double checking later on would be nice, its around 20sec in after starting
19:00 karolherbst: I hate those KHR-GL45.arrays_of_arrays_gl.InteractionFunctionCalls* tests
19:00 karolherbst: they take like forever
19:01 karolherbst: well, I think I can simply start valley, should be fine
19:01 imirkin_: tobijk: you know, it feels like one of those saturation issues ;)
19:02 imirkin_: that karol just fixed
19:02 imirkin_: is this against latest master?
19:02 tobijk: imirkin_: yep
19:02 imirkin_: although those would have gone away with NV50_PROG_OPTIMIZE=0
19:02 tobijk: imirkin_: nope its worse with OPTIMZE=0
19:02 karolherbst: uhh
19:02 imirkin_: lol, that's bad!
19:02 karolherbst: tobijk: qhat quality levels?
19:03 karolherbst: well, with OPTIMIZE=0 we also fail more CTS tests, so ;)
19:03 imirkin_: could be a RA issue
19:03 tobijk: karolherbst: Ultra, stereo disabled, 1024x768
19:05 karolherbst: well I am sure it at least used to work on kepler
19:05 tobijk: karolherbst: it did!
19:05 tobijk: does it not?!
19:05 karolherbst: no issue
19:05 karolherbst: well "no". There are some oddities
19:06 imirkin_: btw, i think MESA_DEBUG=flush only works in a debug build
19:06 imirkin_: is this a debug build?
19:06 tobijk: ah heh, nope
19:07 karolherbst: I think there is an issue we also have with heaven
19:08 karolherbst: some kind of "overlay" above distant things, looking like some kind of net or so
19:08 karolherbst: but this could also be super normal
19:08 tobijk: imirkin_: ok, you were right, flush does indeed remove the corruptions
19:08 imirkin_: yay
19:09 karolherbst: tobijk: do you also see this weird thing, especially by the mountains?
19:09 tobijk: purple overlay?
19:09 karolherbst: I will give a screenshot
19:09 imirkin_: tobijk: unfortunately i haven't the faintest clue how to go about debugging stuff like that
19:10 imirkin_: we're clearing missing a flush or something somewhere
19:10 imirkin_: but ... where ... and why
19:10 imirkin_: without proper docs, this is extremely tricky to RE
19:10 imirkin_: (read: time-consuming)
19:10 imirkin_: and i don't have that kind of will power
19:10 tobijk: hmm, already figured
19:10 karolherbst: tobijk: https://i.imgur.com/hKgK8LR.jpg
19:10 karolherbst: tobijk: see those "lines" on the montains
19:11 karolherbst: they are fixed related to the viewport, but if the scene moves, it looks like they are moving down
19:11 karolherbst: could be issues with their blur effect
19:12 tobijk: *still watching*
19:12 tobijk: oh right, there it is
19:12 tobijk: cant remember those issues, but that does not mean too much :D
19:13 karolherbst: they are there forever afaik
19:13 karolherbst: and I think I decided those are totally fine
19:13 karolherbst: but never investigated those
19:13 tobijk:is going to add flushes like everywhere ;-)
19:14 karolherbst: imirkin_: do you know if nouveau/codegen just stucks somewhere with KHR-GL45.arrays_of_arrays_gl.InteractionFunctionCalls?
19:14 karolherbst: KHR-GL45.arrays_of_arrays_gl.InteractionFunctionCalls2
19:14 imirkin_: i haven't debugged it
19:29 karolherbst: I always forget, that some of those tests are not part of 4.5 and only extensions they added there for fun
19:30 imirkin_: AoA was in 4.4 though, no?
19:30 karolherbst: yeah, was more related to other issues
19:30 imirkin_: er no, it was 4.3 :)
20:39 thican: Hello everyone. rhyskidd asked me the other day to report issue if I have some problems with Nouveau driver with my GP107 (GTX 1050 Ti)
20:39 imirkin_: what kind of issue?
20:39 thican: When I played the game Terraria, I noticed huge lags during the play
20:40 imirkin_: did you clock your GPU up?
20:40 thican: imirkin_: I don't think so, I don't remember tweaking anything
20:40 imirkin_: using kernel 4.12+?
20:40 thican: yes.
20:40 imirkin_: cat /sys/kernel/debug/dri/0/pstate
20:41 imirkin_: [pastebin that]
20:41 thican: linux firmware 20170622
20:41 imirkin_: oh wait. GP107.
20:41 imirkin_: not GM107.
20:41 imirkin_: ignore all that.
20:41 thican: No file :-)
20:41 thican: ok
20:41 imirkin_: anyways, you should expect things to be pretty slow
20:41 thican: so, when I switched back to nvidia drivers, the game was okay
20:42 imirkin_: nvidia has not released firmware that enables one to switch to non-boot clocks, which are almost universally very low
20:42 thican: I noticed this message (I didn't copy it): not using EXT_swap_control_tear for Vsync
20:42 thican: with Nvidia, I have this: Using EXT_swap_control_tear VSync!
20:42 imirkin_: yeah, that's not supported on mesa
20:42 thican: ah :/
20:43 thican: so, you are saying it still requires the signed firmware from Nvidia?
20:43 thican: or this is something to be implemented?
20:44 imirkin_: well, signed firmware is required no matter what
20:44 thican: yes, I mean
20:44 thican: "another" one?
20:44 imirkin_: we can develop all the firmware we want, nvidia will never sign it
20:44 imirkin_: it requires firmware for the pmu that nvidia has not released.
20:44 thican: they don't need to sign it if we have the key ;-) (j/k)
20:44 imirkin_: [i believe they have a dummy pmu blob]
20:44 imirkin_: it'd also be impossible to develop without being able to test stuff
20:45 thican: but I understood we got the signed firmware, right?
20:45 thican: So, what blocks us now?
20:46 thican: (FYI, I have huge freeze sporadicly with _nvidia_ drivers, and I would like to move on)
20:46 thican: *freezes
20:46 thican: and with nvidia drivers, I have a low resolution on TTY, while with Nouveau I have nice resolutions for TTY
20:47 imirkin_: if you want well-supported graphics on linux, i'd recommend AMD
20:47 thican: ah? Did it change?
20:47 imirkin_: hm?
20:47 thican: It might be a while, but I remember people telling us to avoid AMD
20:48 imirkin_: from like 6 years ago? yeah
20:48 thican: *ATI
20:48 thican: yes, maybe
20:48 imirkin_: yeah, ATI was a disaster on linux
20:48 imirkin_: although i did get VIVO working on my radeon 7000 back in the day ;)
20:48 thican: I still do the difference between ATI and AMD, the first for GPU the second for CPU :-)
20:49 imirkin_: well, i think r600 was an ATI design. the GCN stuff is well post-acquisition
21:21 karolherbst: imirkin_: okay, no pipeline_statistics_query_tests_ARB fixes, because it's 4.6 :)
21:21 imirkin_: we could stop exposing the ext
21:21 karolherbst: we could
21:25 karolherbst: imirkin_: todo list for 4.4: https://trello.com/c/qtJ5jkZh/13-failing-tests-wip
21:26 imirkin_: you mean 4.5?
21:26 karolherbst: no 4.4
21:26 imirkin_: ok
21:26 karolherbst: those lists are the same
21:26 imirkin_: khr_debug is 4.6 as well
21:26 karolherbst: just the 4.5 has more
21:26 imirkin_: not sure how you're categorizing these
21:26 karolherbst: khr_debug is 4.3
21:27 imirkin_: what's 4.6 then?
21:27 karolherbst: spirv
21:27 imirkin_: oh. khr_no_error.
21:27 imirkin_: basically the same ;)
21:27 karolherbst: right
21:27 karolherbst: :D
21:28 imirkin_: where are all the 3d images tests that fail?
21:28 imirkin_: oh, that's probably the non-layered binding thing
21:28 karolherbst: well everything else passes
21:29 karolherbst: Passed: 6393/6493 (98.5%) without the skips
21:29 karolherbst: and that's for 4.5
21:31 karolherbst: the 4.5 mustpass lists contains all 4.4 tests, makes it easier
21:34 karolherbst: imirkin_: what should we do about "internalError"? just ignore?
21:35 bloblo: hello
21:35 bloblo: need help
21:35 karolherbst: imirkin_: ohh right, I wanted to do that fp64 stuff
21:36 bloblo: i have gtx 770, nouveau set max 5 gt/s
21:36 bloblo: with lspci i have only 5 gt/s
21:36 bloblo: if i blacklist nouveau and activ nvidia then i have 8gts
21:37 bloblo: i have x79 board
21:37 bloblo: 2 pcie3
21:37 karolherbst: bloblo: there is a bug within the vbios parsing I still have to fix
21:37 karolherbst: anyway on a desktop the perf benefit is... not really there
21:38 karolherbst: create a bug report or something and whenever I fix this, you get notified
21:39 bloblo: where i send infos ?
21:39 karolherbst: bugs.freedesktop.org
21:40 bloblo: ok tomorow i try (is night now here)
21:40 bloblo: thanks
21:47 karolherbst: hum, "set $r6 ne u32 $r3 0x7ff"
21:48 imirkin_: yes...
21:48 karolherbst: duh
21:49 karolherbst: there I have to put b32
21:49 karolherbst: and for the other sets, I don't
21:49 imirkin_: predicate vs register
21:49 karolherbst: yeah
21:54 karolherbst: yay
21:54 karolherbst: much better
21:54 karolherbst: only the mods are failing
21:55 imirkin_: cool
21:55 imirkin_: and that's expected
21:55 imirkin_: [for now]
21:55 karolherbst: yeah
21:57 karolherbst: "NotSupported (The test can be run only in ES context)" this makes it easy
21:59 karolherbst: KHR-GL45.conditional_render_inverted.functional
21:59 karolherbst: "The tested mode was GL_QUERY_NO_WAIT_INVERTED. Query was done for target GL_SAMPLES_PASSED, and the test was prepared to pass all fragments."
21:59 karolherbst: and 3 others
21:59 imirkin_: the NO_WAIT stuff may be questionable
21:59 imirkin_: dunno
21:59 imirkin_: could be messed up in conjunction with inverted
21:59 karolherbst: expected 1, resulted 0
22:00 karolherbst: well, intel passes this test
22:00 imirkin_: i meant in nouveau :)
22:00 karolherbst: ohhh
22:00 karolherbst: k
22:04 imirkin_: let's see...
22:05 imirkin_: iirc tobijk implemented this
22:05 karolherbst: ha
22:05 karolherbst: easy fix
22:05 karolherbst: "KHR-GL45.conditional_render_inverted.functional" pass :)
22:05 imirkin_: of course non-inverted now fails :p
22:06 imirkin_: hq->data[1] = 1; /* initial render condition = true */
22:06 karolherbst: uhm
22:06 karolherbst: doubtful
22:06 imirkin_: does that need to change to the condition?
22:06 imirkin_: dunno
22:06 karolherbst: imirkin_: https://github.com/karolherbst/mesa/commit/5654927f55602cfd6e50a7583275da26ebae8e0a
22:06 imirkin_: ahaha
22:07 karolherbst: :p
22:07 imirkin_: so i think in *reality* what needs to happen is
22:07 imirkin_: hrm
22:09 imirkin_: ok
22:09 imirkin_: if hq->state != READY && !wait make it pass
22:09 imirkin_: in nvc0_render_condition
22:10 imirkin_: i.e. cond = NVC0_3D_COND_MODE_ALWAYS
22:10 imirkin_: although
22:10 imirkin_: i really think that
22:10 imirkin_: hq->data[1] = 1; /* initial render condition = true */
22:10 imirkin_: changing that to 0 will fix those cases.
22:11 imirkin_: but unfortunately that information is not available
22:11 imirkin_: at begin query time
22:24 karolherbst: okay, textureGrad is a little broken
22:25 imirkin_: i broke it a while back
22:25 imirkin_: to fix deqp
22:25 karolherbst: :/
22:25 imirkin_: which caused some piglits to regress
22:25 imirkin_: i never got to the bottom fo it
22:25 karolherbst: well, CTS is unhappy now
22:25 imirkin_: but the new code made more logical sense iirc
22:25 imirkin_: this is with cubes right?
22:25 karolherbst: yeah
22:26 imirkin_: i think i might be changing the major axis now
22:26 imirkin_: whereas i'm not supposed to
22:26 imirkin_: dunno
22:26 karolherbst: well, I will check with apitrace
22:26 imirkin_: get blob shader too, could be instrucitonal
22:27 karolherbst: I hope it's an obvious issue
22:27 karolherbst: weird thing is
22:27 karolherbst: with GL_DEPTH_COMPONENT32F it's fine
22:27 karolherbst: GL_RGBA8, GL_RGBA32I, GL_RGBA32UI and GL_STENCIL_INDEX8 broken
22:29 karolherbst: duh...
22:29 karolherbst: apitrace fails me
22:29 karolherbst: crash in strlen
22:30 karolherbst: glGetActiveUniform
22:34 imirkin_: ok yeah
22:34 imirkin_: i think the proper fix is in nvc0_render_condition to check if the query is done, and if it's not, just set the condition to "always"
22:34 imirkin_: well
22:34 karolherbst: well, just don't pass NULL to a gl function
22:34 imirkin_: that's the partial fix
22:34 karolherbst: mhh
22:35 imirkin_: the *proper* fix, is to use a macro which detects it
22:35 imirkin_: [this is in re the inverted + nowait]
22:35 tobijk: hmm, whats up with inverted render cond?
22:35 imirkin_: tobijk: it returns the wrong result if wait==false
22:36 imirkin_: and the query in question has not yet completed
22:37 karolherbst: imirkin_: you remember in which commit you "fixed" the deqp tests?
22:38 imirkin_: let's see
22:38 imirkin_: 6eeb284e4f74a2fe5ae6cba90f97f219935e24df
22:39 karolherbst: mhh, interesting
22:40 karolherbst: this commit fixed 26 sub tests
22:40 karolherbst: without it: 168 fails, with: 144 fails
22:40 karolherbst: *24
22:44 imirkin_: so ... not all bad ;)
22:45 karolherbst: yeah
22:45 imirkin_: anwyays, the textureGrad failures you're seeing now could be different - dunno
22:45 tobijk: mh:
22:45 tobijk: If <mode>
22:45 tobijk: is QUERY_NO_WAIT_INVERTED or QUERY_BY_REGION_NO_WAIT_INVERTED
22:45 tobijk: the GL may choose to unconditionally execute the subsequent rendering
22:46 tobijk: commands without waiting for the query to complete.
22:46 imirkin_: tobijk: right
22:46 imirkin_: tobijk: whereas we're choosing to unconditionally not-execute the subsequent rendering commands
22:46 tobijk: for me "may" sounds like: "do whatever pleases you"
22:47 tobijk: but i'm wrong :D
22:47 imirkin_: well
22:47 imirkin_: the alternative is that the GL impl may choose to wait for the query to complete
22:47 imirkin_: but if it doesn't wait, it has to execute the rendering
23:02 tobijk: so yeah, you wrote it up there already, set condition to _do_it_ :D
23:04 imirkin_: yep.
23:04 karolherbst: that cube bug sounds like something interesting, maybe I figure something out tomorrow
23:04 imirkin_: just need to look at the query state in the !wait case
23:04 imirkin_: karolherbst: ok cool. feel free to ask me how the handleManualTXD stuff works -- it wasn't trivially obvious to me the first 15 times i read it
23:25 tobijk: imirkin_: you are handling the cond patch, am i correct?
23:26 imirkin_: maybe eventually?
23:26 imirkin_: i'm busy with other stuff atm
23:26 imirkin_: let me make a note of it
23:26 tobijk: np
23:26 tobijk: the failing test stays anyway :D
23:27 tobijk: maybe i find the time tomorrow, but now i'm off
23:56 MichaelP: Nvidia GeForce GT 730 20-nouveau.conf ... screen tearing