01:07 mwk: !
01:07 mwk: well, I'm certainly going to take a look
01:07 mwk: if only to apply it to Falcon RE
01:09 mwk: RSpliet: is ness some internal nv thing?
04:28 airlied: skeggsb: drivers/gpu/drm/nouveau/nvkm/engine/device/tegra.c:219:2-3: Unneeded semicolon
04:28 airlied: drivers/gpu/drm/nouveau/nvkm/subdev/mmu/base.c:69:2-3: Unneeded semicolon
04:28 airlied: drivers/gpu/drm/nouveau/nvkm/subdev/pci/base.c:90:2-3: Unneeded semicolon
04:29 airlied: drivers/gpu/drm/nouveau/nvkm/subdev/bios/iccsense.c:102:3-4: Unneeded semicolon
04:29 airlied: drivers/gpu/drm/nouveau/nvkm/subdev/bios/iccsense.c:118:3-4: Unneeded semicolon
04:29 skeggsb: airlied: ack, i'll fix those when i package up a couple more things for you
10:16 RSpliet: mwk: I suspect it is... but I'm not too familiar with the hw synthesis lingo
10:51 < Blub\0> weeeee, bye bye S3 patent
16:03 karolherbst: mwk: did you mean the paper?
16:03 karolherbst: ohhh
16:03 karolherbst: the dla stuff
16:06 gfw: imirkin: any news on this bug? Had send you the trace file you requested about a month ago. https://bugs.freedesktop.org/show_bug.cgi?id=102349
16:08 gfw: i think this would also cause ton of bugs on nv40: https://www.phoronix.com/scan.php?page=news_item&px=Shader-Fuzzing-Paper
16:08 gfw: would be nice if someone could help running and fixing that :)
16:08 imirkin_: gfw: haven't had any time, sorry
16:12 gfw: imirkin_: any plans on when you would have time to look at it?
16:13 imirkin_: not really, sorry
16:15 gfw: have spend hours into learning how to debug that, how to compile mesa and so on. would be nice if my work wont be waste of time because this problems are existing for many thousand users out there and many of them are not able to switch from windows to linux because of this errors
16:24 imirkin_: yes, it would be niec.
16:26 imirkin_: the reality is that i'm a volunteer and am subject to the 24-hour-per-day restriction we all have to live with.
16:34 gfw: could skeggsb also look in this issue? maybe this would speed up the things :)
16:45 imirkin_: i don't speak for him, but somehow i'm guessing he doesn't exactly sit around all day doing nothing either =/
16:54 karolherbst: probably not
16:55 karolherbst: gfw: I was fuzzing gl shaders with afl some months ago and also found several issues withing the compiler, but nothing related to generating bad shaders
16:55 karolherbst: except we are already aware of a few issues we might want to fix first
16:55 imirkin_: nv30 backend is full of fail
16:56 karolherbst: not quit sure if I will be able to look at _that_ old hardware soonish :(
16:56 karolherbst: and for that kind of issues
16:57 karolherbst: what I would like to do though is to help mupuf setting up some awesome CI for nouveau, so that we don't make things worse on older hw
16:57 karolherbst: or maybe I simply go ahead and do it myself
16:58 karolherbst: maybe there are enough people who would like to support nouveau by donating a lot of old hardware we could use for that
17:09 imirkin_: i'd rather a lot of people support nouveau by donating their time
17:10 karolherbst: well, having a CI is a good thing regardless
17:10 gfw: karolherbst: i can tell out of my testing, that the plasma issue is there on on nv4x devices. I tested about 8 of them. different quadro and geforce GPUs. All have same error. Donation of just one 1$ card with a fan would be enough to fix all of this issues
17:10 karolherbst: and it would be cool if everybody could "plug'n'play" into our CI system
17:11 karolherbst: gfw: well sure, and time
17:11 karolherbst: hardware costs are mostly irelevant
17:12 gfw: karolherbst: you wrote "maybe there are enough people who would like to support nouveau by donating a lot of old hardware we could use for that"
17:12 gfw: how did you mean it?
17:13 karolherbst: well, if we want to build a CI system, we need hw to run tests on the actual hardware
17:13 karolherbst: but this is just the start
17:59 dcomp: karolherbst: I'd be happy to run a couple of CI tests every night, just give me an easy way to run it. A docker container or something
17:59 karolherbst: dcomp: what we actualy need is something like, dev pushes commit into git tree, CI builds, CI complains about broken stuff automatically
18:00 imirkin_: karolherbst: but random people should be able to host runners
18:00 dcomp: what kind of tests? piglet?
18:01 imirkin_: piglit, cts, deqp
18:01 karolherbst: imirkin_: exactly
18:01 imirkin_: look at what janesma has done... his thing is a lot more controlled obviously
18:01 imirkin_: also this sounds somewhat akin to what the wine guys used to do
18:02 imirkin_: dunno what they do now, tbh
18:02 karolherbst: but I would like something like: "register" runner to CI master, distribute jobs across slaves and merge results
18:02 imirkin_: (for various windows versions)
18:02 karolherbst: I think mupuf actually wanted to include something like this in ezbench or whatever he is currently wokring on
18:02 karolherbst: but I really don't know the current status on this
18:02 mupuf: karolherbst: I pushed another version of this today :)
18:03 karolherbst: :) nice!
18:03 mupuf: but I don't have a "topic" that machines can register to
18:03 mupuf: that will come later
18:03 karolherbst: yeah
18:03 karolherbst: but we should make it as easy as possible for users who really want to help and have unused machines
18:04 mupuf: right now, what I have is a way to push in one command line work onto multiple machines
18:04 mupuf: yes, it will come
18:04 karolherbst: but all that watchdog and turning machines on/off might get complicated
18:04 mupuf: first, we make it work with what we have
18:04 mupuf: and from there, we'll see
18:04 karolherbst: good plan
18:05 karolherbst: I think gfw might be interestedin helping adding another machine to it?
18:05 mupuf: I don't want to hear about it yet, sorry
18:06 karolherbst: mupuf: do you know what would be awesome? If we could manage to get some of those remote thingies ready until FOSDEM
18:06 karolherbst: and we could give it to interested users
18:06 mupuf: oh, right, would be great!
18:06 mupuf: I have the circuits... and the code
18:06 mupuf: but we can't buy CHIPs anymore :o
18:06 mupuf: let's check if they have the new one ready!
18:06 karolherbst: yeah
18:07 karolherbst: I think having like 50 in total would be awesome. I could actually help out here as well
18:07 mupuf: sure
18:07 mupuf: but.... we will need to re-design the circuit around another board that we can buy in a volume :s
18:07 Hooloovo0: that would be an okay idea, I would be fine with running tests on some unused laptops
18:08 karolherbst: laptops are difficult
18:08 Hooloovo0: how so?
18:08 karolherbst: laptops with connectors for docking stations are less difficult, still pain
18:08 karolherbst: Hooloovo0: well, we would need to replace the power button and so on
18:09 Hooloovo0: oh, for that, yeah
18:09 karolherbst: much easier if you only have to replace those on a desktop
18:09 Hooloovo0: hadn't thought about power
18:09 mupuf: yeah, power is a problem
18:09 mupuf: there is always a way
18:09 karolherbst: sure
18:10 mupuf: in the bios, there is often a setting to turn on the machine immediatly when power is applied
18:10 karolherbst: and we could provide information about dock connectors we know, but....
18:10 Hooloovo0: if you had a power switch for the plug, that would work
18:10 Hooloovo0: and possibly be more robust
18:11 karolherbst: well, then you need such a bios
18:11 karolherbst: with laptops there are a lot of ifs, which makes it difficult
18:12 Hooloovo0: yeah :/ I would be fine with modifying the power button on one of these, and have a dock for the otehr
18:22 gfw: mupuf: could you add such functionality to my bios power turn on the machine immediatly when power is applied? That would be great. Thanks! https://github.com/coreboot/coreboot
19:02 mupuf: gfw: ha ha
19:02 mupuf: good one :p
19:18 RSpliet: mupuf: is that a "no"? :-C
19:46 mupuf: RSpliet: well, I don't think any of my mobos would qualify!
19:46 mupuf: ;)
19:46 gfw: mupuf: could you take a look into this issue why it is not working? https://bugs.freedesktop.org/show_bug.cgi?id=102352
19:49 mupuf: gfw: hmm, yes, I saw your ping
19:49 mupuf: I remember indeed fixing the nv4x problem a while ago
19:50 mupuf: gfw: is that a regression? Are you sure my patch was supposed to fix your problem?
19:50 mupuf: there can be many reasons for the fan to be running at 100%
19:50 mupuf: what gpu do you have anyway?
19:51 mupuf: IIRC, I have no gpu with a fan
19:51 mupuf: so I had to fake the bios to make it look like it had
20:03 dcomp: deqp is quite mesmerising...
20:04 imirkin_: might want to run with --deqp-visibility=hidden :)
20:06 dcomp: now time to run it with nouveau and not intel hopefully it works with DRI_PRIME
20:06 imirkin_: should be fine.
20:06 imirkin_: note that there are a few tests that crash nouveau
20:06 imirkin_: i don't think they're in the "mustpass" list
20:07 imirkin_: iirc some test that generates a LOT of primitives from a single vertex with the help of tessellation and geometry shaders.
20:07 imirkin_: something like 4M primitives from a single vertex. that makes the hw unhappy.
20:07 imirkin_: crashes the blob too :) although blob is better at recovery
23:37 duttasankha: where in rnndb I can find PCOUNTER?
23:38 imirkin_: pmu somewhere, probably
23:38 duttasankha: I am using Tesla C2075
23:38 imirkin_: rnndb/pm/pdaemon.xml
23:39 imirkin_: gf100_pcounter.xml
23:42 duttasankha: I didn't see inside pdaemon.xml
23:42 duttasankha: where can I find gf100_pcounter.xml
23:43 imirkin_: rnndb...
23:43 duttasankha: oh I saw it now..that was foolish of me...thanks a ton!!!