04:08imirkin: yay, some gles3 xfb tests now pass. some weird failures which are probably just real failures...
06:17imirkin: pmoreau: in case you read scrollback, i have some fixes for your compute commits
06:35imirkin: karolherbst: you're probably the only one who _might_ care ... lmk if this hackery annoys you: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8346
06:35imirkin: [and suggest improvements]
15:24lendasin1: In the age of feeling that my head works the best at 37, when other devils are too tired allready, and live their own violatedly faught life. Heck i even played the best match of table tennis as 36years old to my own suprise, but again i was sure if i climb over or crawl out of the mad stuff started against me, second part of life gonna start to get interesting to me.
15:28lendasin1: I predicted it to some degree, very strong and motivated sturt, a large or long downrun and then a rise again , practically 20years of total misery and bad form, until it started to come back :/
15:31lendasin1: I am in good terms with all the programming too, some sports forms start to come back, however i am waiting for resolution to do other physical work, like when some more power in hands and legs are needed, some heavylifting kind of construction work or such.
15:34lendasin1: sports is like programming , instead of more important charectirisitics about knowing the hw, is to know the properties of your body and technique suitable for this body, so the command center in both cases is the head, the brain.
15:35lendasin1: brain subsitutes wrong memories or false positives with something that you force it to store, when the pressure isn't too high, when they give you a little chanche or room to do that
15:36lendasin1: than one can work this out.
15:39lendasin1: I have isolated two types of depressions on myself, ones that you could even enjoy at times, do not know the severity..this i had when i was young, physical strengths backed it up so good, and the other kind is the one you no longer like, cause something got a hit, but the first one is a dumbass thing, it makes one lazy at the same time.
15:40lendasin1: so in my world, cause i was born to be happy and lucky, mental strenghts without physical strenghts as backup does not exist really, it's a total false concept.
15:46lendasin1: Sometimes back it drove me nuts, that people tend to talk about something consistently that does not seem to exist on it's own like the way they present it to manifest itself.
15:47lendasin1: And as of today, i am a lot better person than the ruthless ego maniac i was when i was young, maybe not totally ego, but i was unstable, lazy, talented and stupid and ruthless to others.
15:49lendasin1: Japanease who developed the medication they give me, i think they want to hear my opinion on it, when the situation demands and the pressure goes too high and one needs to ignore some bad public nasty humiliations against him/her, this med practically works even.
15:50lendasin1: So when i am about to cook in entirely, then this was the only med i was willing to take volunterily, and it is abilify, it has downsides, it can not be mixed only with other apsychiotic kind of meds.
18:03imirkin: lol, dEQP-GLES3.functional.shaders.precision.int.highp_mul_vertex fails at emission time
18:04imirkin: (on nv50)
18:04imirkin: looks like we do a mov instead of an export? oops
18:17imirkin: silly optimizations
18:42imirkin: oh man, i think this is some unfortunate interaction with the expand multiply bs
18:42imirkin: or not. hm. very weird.
19:10RSpliet: imirkin: what's the status of nouveau and hdr?
19:10imirkin: RSpliet: waiting to swap out the G84 for something else
19:10RSpliet: Dare I ask... using Wayland?
19:10imirkin: last i touched it, the next thing on the list was getting 10bpc output working with hdmi
19:11imirkin: i got all the LUT and format bits going
19:11RSpliet: does it work with DP? Which gens of HW are you looking at?
19:11imirkin: i also have a change to supply the "drm" infoframe (aka hdr_output_metadata)
19:11imirkin: RSpliet: only gen of hw i have on-hand which supports hdmi2.0 is pascal
19:12imirkin: RSpliet: however the 10bpc / etc stuff should also work on other ones
19:12imirkin: i think fermi+ or kepler+?
19:12RSpliet: My mileage may vary though, noted
19:12imirkin: unfortunately the only DP hardware i have is fermi, and that doesn't support DP 1.2.
19:12imirkin: well, it might be like gf119+ or something
19:12imirkin: i don't remember
19:13RSpliet: I may have DP hardware... in the Netherlands.
19:13RSpliet: That was a monumental mistake
19:13imirkin: that's a bit of a walk for me too :)
19:13RSpliet: should have seen that pandemic coming back around Christmas 2019
19:13imirkin: coulda should woulda =/
19:13imirkin: next time!
19:14RSpliet: Haha yeah...
19:14imirkin: anyways, for now i'm fixing as much nv50 stuff as i can
19:14imirkin: not sure when the next time i'll have that hw plugged in will be
19:14RSpliet: Anyway, thought I'd ask, given I'm looking at replacing my monitor with something that doesn't suck
19:14imirkin: i like U2415 monitors -- 1920x1200
19:14imirkin: none of this 1920x1080 BS
19:15imirkin: and i rotate them too
19:15imirkin: so ... yeah. pretty ideal for me.
19:15RSpliet: I think on the longer term I'll need a new GPU too, good chance it'll be team Red, but for the time being I have to work with GT640
19:15imirkin: the end result is a giant 4:3 panel
19:15RSpliet: Oh I'm eyeballing LG 27UL600
19:15imirkin: (20:16)
19:16imirkin: so even better than 4:3
19:16RSpliet: Just waiting for after CES, who knows what they'll announce
19:16imirkin: hehe
19:16imirkin: after CES is when you get last year's hw for cheap
19:16imirkin: (or cheap*er*)
19:16RSpliet: exactly!
19:17RSpliet: Because black friday and post-Christmas sale didn't dent the prices enough. Somehow I feel like we're always two weeks removed from a sale...
19:19RSpliet: Anyway, my L1730B has a cluster of dead pixels. So sad, only lasted 16 years
19:20imirkin: that's because there are sales twice a month...
19:20imirkin: ouch
19:20imirkin: such a short life
19:20imirkin: i used a U2407 for ... 12 years. still have it, but don't use it anymore.
19:21imirkin: it just got dimmer over time
19:21RSpliet: just like me
19:21imirkin: eh, maybe
19:21imirkin: could just be self-adjusting memory
19:21imirkin: to make yourself think you were smarter before?
19:22RSpliet: Maybe I'm realising more how little I actually know and understand :-P Anyway, j/k, not going to take that to extremes
19:22imirkin: there's a nice graphic about that
19:22imirkin: https://kmbkteam.wordpress.com/2010/04/26/3-types-of-knowledge/
19:22RSpliet: Was it related to the Dunning Kruger effect?
19:23RSpliet: ah yes
19:24RSpliet: Anyway, thanks for the HDR rundown. Good to know what to expect. I'll hopefully get the new monitor before the end of the month, so if you need some feedback on your patches (provided they touch GK107) HMU
19:24imirkin: will do
19:24RSpliet: It won't be driving the monitor at full 4K... but baby steps
19:24imirkin: note that kepler doesn't support HDMI 2.0
19:25imirkin: but it does support DP 1.2 which should be enough for 4k@60
19:25RSpliet: Oh so it can't do 10bpp either?
19:25imirkin: welllll
19:25imirkin: i think hdmi 1.x can do 10bpc
19:25imirkin: there are gaps in my knowledge as to what's allowed where
19:25RSpliet: Yeah, sadly my card doesn't have a DP out
19:25RSpliet: It has two DVI ports though...
19:25imirkin: hdmi2.0 is def required for 4k@60@yuv444/rgb
19:26imirkin: APPARENTLY the blob drivers manage to get kepler to do 4k@60@yuv420
19:26imirkin: however there's no documented hw support for this
19:26RSpliet: Yeah, already mentally prepared for running it at a lower res until I replace the GPU
19:26imirkin: if you were to RE how they do that, that'd be super-duper
19:26imirkin: also the whole yuv420 story is a bit weak in the kernel in general
19:26imirkin: i haven't checked on the state of the art there in a while
19:26RSpliet: I just fell asleep today during my lunch break, napped for a good hour.
19:27imirkin: re end-user selection of wtf they want, etc
19:27RSpliet: Not sure if I'll be doing any REing anytime soon :-P
19:27imirkin: ok
19:28imirkin: it wouldn't be a huge lift
19:28imirkin: maybe i'll do it ... i'm likely to drop in gk208 or gm107 instead of the g84
19:28imirkin: ugh, but then i have to get blob set up
19:28imirkin: i havne't run blob in ages
19:29RSpliet: Same
19:34imirkin: i guess tbh i'm not sure how to do yuv420 on later gens either
19:35imirkin: and i'm not sure how i'll verify whether i'm actually getting 10bpc or 10bpp image dithered down to 8
19:35imirkin: coz the dithering looks _really_ good
19:36imirkin: i added a test image to modetest which might help, we'll see
19:36imirkin: (8bpp gradient on top, 10bpp gradient on bottom)
20:01imirkin: interesting... all the precision dEQP tests are also failing on nv50
20:01imirkin: will investigate
20:01imirkin: for stuff like float div.
20:18imirkin: karolherbst: and now our favorite issue...
20:18imirkin: Test case 'dEQP-GLES3.functional.shaders.random.all_features.fragment.93'..
20:18imirkin: deqp-gles3: ../src/gallium/drivers/nouveau/codegen/nv50_ir_ra.cpp:1435: void nv50_ir::GCRA::checkInterference(const nv50_ir::GCRA::RIG_Node*, nv50_ir::Graph::EdgeIterator&): Assertion `vB->compound' failed.
20:18karolherbst: ewww
20:19karolherbst: but that's the merge/split thing
20:19imirkin: aka "our favorite issue"
20:19karolherbst: yeah.. wait
20:19karolherbst: I have a patch which could fix it
20:20imirkin: cwabbott made a patch around that i think
20:20karolherbst: yeah
20:20imirkin: but i never invested the time to figure out wtf was going on
20:20karolherbst: I am searching that one
20:20karolherbst: imirkin: https://github.com/karolherbst/mesa/commit/b2a86db447c266df51b4d07e135eb4096a054483
20:20imirkin: will try that in a bit
20:21imirkin: going back to see wtf all the precision stuff is about
20:23imirkin: this is the list of fails i got until that crash: https://hastebin.com/vobalatizi.css
20:23imirkin: i think the cube stuff is about lack of seamless
20:23imirkin: that fragdata one is weird.
20:23imirkin: not to mention all the precision things
20:24imirkin: ERROR: expected compilation to fail, but shaders compiled correctly.
20:24imirkin: gah
20:24imirkin: not-my-problem.
20:24ccr: :P
20:31imirkin: the precision stuff is worrying. although i wonder if it's a texturing issue and not a shader one.
20:31imirkin: coz the outputs are just totally off
20:31imirkin: with something as simple as "abs" - not a lot of opportunity to screw that up
22:50imirkin: karolherbst: heh. that precision stuff? it's some blit issue. all values past 4096 are magically broken
22:50imirkin: sigh
22:51imirkin: didn't i already fix this? apparently not hard enough, or maybe only on nvc0
22:51karolherbst: mhhh
22:51karolherbst: I fixed something for nvc0 at least
22:51imirkin: iirc i made a diff version of your fix
22:51karolherbst: yeah, might be
22:51imirkin: will have to go back and check
22:51imirkin: maybe i/we never pushed?
22:52karolherbst: we pushed something at least
22:53karolherbst: https://gitlab.freedesktop.org/mesa/mesa/-/commit/4d0d9118756325ea83d254515d4c7a410df96f0e that's what I did
22:53karolherbst: but I guess the issue you see is something different
23:00imirkin: no, that's unrelated
23:00imirkin: i'm talking about widths > 1/2 the max fb size
23:01imirkin: 838118462e63745ae70e05b42259f2aa4f81157a
23:01imirkin: that's the one i was thinking of
23:02imirkin: but not this
23:03imirkin: since it's not a msaa situation
23:04Lyude: RSpliet: i'm planning on working on the dp stuff at some point
23:04Lyude: if you'd like to help you're more then welcome