2021-07-26 22:56
ChanServ changed the topic of #dri-devel to: <ajax> nothing involved with X should ever be unable to find a bar
00:10
jhli has quit [Remote host closed the connection]
00:11
mlankhorst has quit [Ping timeout: 480 seconds]
00:19
pcercuei has quit [Quit: dodo]
00:37
rasterman has quit [Quit: Gettin' stinky!]
00:42
<
dcbaker >
じゃあね classic drivers!
00:45
<
HdkR >
The implication of じゃあね being something like "See you soon" is scary :P
00:46
<
Sachiel >
yeah, we are all going to die too some day
00:46
<
HdkR >
Classic will come for us in the night
00:47
<
Sachiel >
give it 5 years and there'll be a hollywood reboot for them
00:52
Haaninjo has quit [Quit: Ex-Chat]
00:54
<
dcbaker >
HdkR: I'll probably end up managing the amber branch, so じゃあね felt more appropriate that さようなら
00:54
<
HdkR >
kudos to your sacrifice
00:54
<
jekstrand >
dcbaker: \o/ There will be much rejoicing!
00:54
* jekstrand
pulls main
00:55
<
dcbaker >
Next is removing swr from main
00:55
<
imirkin >
jekstrand: finally, all the copy_image bugs in i965 are solved ;)
00:55
<
dcbaker >
I will delete half of mesa!
00:55
<
imirkin >
dcbaker: this is by far the best way to live the dream of negative LoC contribution to a project.
00:56
<
jekstrand >
dcbaker: Now you just need to find a spare printer driver or two to delete. :)
00:57
<
jekstrand >
imirkin: About time!
00:57
<
bnieuwenhuizen >
wait is swr deprecated now?
00:57
<
jekstrand >
yeah. They asked for it to live in amber
00:58
<
dcbaker >
For my next trick I'll delete all GL drivers except zink
00:58
<
bnieuwenhuizen >
dcbaker: presumably also keeping the d3d12 backend for that other OS?
00:58
<
HdkR >
Oh, if SWR goes to amber, that solves the SSE poisoning problem :P
00:59
<
dcbaker >
bnieuwenhuizen: no, they can just implement vulkan on d3d12, then they too can use zink
00:59
<
dcbaker >
See, the solution is always more layers
01:02
<
jekstrand >
I'll get around to cleaning up the compiler next week.
01:02
mattrope has quit [Ping timeout: 480 seconds]
01:03
<
jenatali >
We can also clean up the non-TLS code for unrelated reasons, now that all platforms are using TLS. Lots of cleanup to be done :)
01:04
<
bnieuwenhuizen >
HdkR: to be fair, "not compiling SWR" and "not including SWR" have the same end result binary wise
01:06
co1umbarius has joined #dri-devel
01:08
columbarius has quit [Ping timeout: 480 seconds]
01:08
<
FLHerne >
Silly question -
*why* is Turnip called Turnip?
01:08
<
jekstrand >
Because chadv likes root vegitables
01:09
<
jekstrand >
The full history of the name is a very old joke that goes back 6-7 years when chadv was still at Intel and has to do with graphics APIs I can't disucss in public.
01:09
<
jekstrand >
But, really, chadv likes root vegitables.
01:11
<
jekstrand >
HdkR: How do you not have tag modifying ability?
01:11
<
jekstrand >
HdkR: Also, I've switched it to SPIR-V+Vulkan
01:11
<
HdkR >
Because I'm not a contributor :P
01:11
<
HdkR >
That should work
01:12
<
jenatali >
dcbaker: After syncing the classic removal, with no reconfiguring of an environ that had no dri drivers, I'm hitting an error because I had the default value of 'auto' for it. Seems the meson error is a little aggressive that it rejects 'auto'
01:13
<
jenatali >
Er, while reconfiguring, without changing any options
01:13
<
gawin >
btw amber is gonna get older (removed earlier) dri drivers?
01:17
<
dcbaker >
gawin: yeah, Amber is going to have classic and swr
01:18
<
dcbaker >
It's going to be a continuation of 21.3
01:26
<
gawin >
ops, my bad, wanted to ask about drivers no longer existing in the tree. (r128, sis, savage, i810 etc) though not sure if we are fine with autotools
01:26
<
anholt >
imirkin: runners poll gitlab.fdo for jobs.
01:27
ngcortes has quit [Remote host closed the connection]
01:31
<
imirkin >
dcbaker: i seem to be getting TGSI dumping now
01:32
<
imirkin >
which i didn't ask for :)
01:32
<
imirkin >
any idea where it's coming from? it's coming in after your changes to drop classic
01:36
<
dcbaker >
I have an idea, I'll look later tonight
01:36
<
dcbaker >
I'm assuming it's from merging mesa common and gallium
01:36
<
imirkin >
dcbaker: weirdly i'm only seeing like one instance of a TGSI shader in the piglit job
01:36
<
imirkin >
but in the deqp job it's constantly happening
01:38
<
imirkin >
dcbaker: i haven't bisected it to a specific change. just noticed it in the jobs (i'm testing some stuff, so i was looking closely)
01:38
<
imirkin >
dcbaker: and none of my changes ... change code :) just CI config changes.
02:04
<
imirkin >
dcbaker: seems like NIR_DEBUG=tgsi somehow. unclear how.
02:05
<
imirkin >
perhaps something's no longer initialized that was before
02:05
columbarius has joined #dri-devel
02:05
<
imirkin >
(same as for the util_detect_cpu thing)
02:06
libv_ has joined #dri-devel
02:06
co1umbarius has quit [Ping timeout: 480 seconds]
02:07
adjtm has quit [Remote host closed the connection]
02:07
adjtm has joined #dri-devel
02:10
jewins has quit [Ping timeout: 480 seconds]
02:11
libv has quit [Ping timeout: 480 seconds]
02:20
Emantor has joined #dri-devel
02:23
<
dcbaker >
That seems quite possible
02:32
nsneck_ has joined #dri-devel
02:38
nsneck has quit [Ping timeout: 480 seconds]
02:42
Company has quit [Quit: Leaving]
03:02
gawin has quit [Ping timeout: 480 seconds]
03:15
gawin has joined #dri-devel
03:20
camus has joined #dri-devel
04:16
<
imirkin >
dcbaker: i wonder if there are now several nir_debug's, and some of them don't get initialized?
04:16
gawin has quit [Ping timeout: 480 seconds]
04:16
<
imirkin >
this would happen if there are duplicate instances of nir.o
04:18
JohnnyonFlame has quit [Ping timeout: 480 seconds]
04:18
<
imirkin >
anyways, i'm going back to before that series for my own tree for now
04:24
<
imirkin >
dcbaker: heh, it wasn't your series
04:24
<
imirkin >
it was happening before already
04:24
<
imirkin >
will track it down
04:24
<
imirkin >
seems like a likely candidate...
04:25
<
imirkin >
... and mareko already tracked it down
04:30
<
imirkin >
jasuarez: --^
04:42
<
dcbaker >
Yay, not me!
04:45
lemonzest has joined #dri-devel
04:48
<
jekstrand >
Not nearly as impressive as dcbaker's diffstat but hooray none the less:
04:48
<
dcbaker >
I need to go rebase my genxml splitting
04:48
* jekstrand
is so glad to get rid of push constant compaction and pull constants!
04:49
<
jekstrand >
dcbaker: Why are we splitting genxml?
04:55
anthk_ has joined #dri-devel
04:58
<
dcbaker >
Because without i965 we always overthink/over build
05:01
<
dcbaker >
*over link
05:13
LexSfX has joined #dri-devel
05:30
fxkamd has joined #dri-devel
05:37
camus has quit [Remote host closed the connection]
05:37
camus has joined #dri-devel
06:09
Duke`` has joined #dri-devel
06:26
camus has joined #dri-devel
06:43
camus has quit [Remote host closed the connection]
06:44
camus has joined #dri-devel
07:02
sdutt has quit [Remote host closed the connection]
07:21
<
jasuarez >
imirkin: forgot to initialize nir_debug in tgsi-to-nir
07:21
<
imirkin >
jasuarez: can't you use the usual macros to initialize it?
07:22
<
imirkin >
or do those not handle what you need?
07:23
<
imirkin >
(tbh i don't actually know how it works offhand, have to look it up each time)
07:23
<
jasuarez >
do you mean the macros like DEBUG_GET_ONCE_FLAGS_OPTION() ?
07:23
<
imirkin >
you seem to use that already
07:24
<
imirkin >
so why do you need the other thing?
07:25
LexSfX has quit [Ping timeout: 480 seconds]
07:25
<
imirkin >
i.e. why does nir_debug_print_shader exist?
07:25
<
imirkin >
it's just asking for trouble
07:25
<
imirkin >
or could it be initialized in some central location
07:25
<
imirkin >
rather than from the various nir users?
07:27
<
imirkin >
or it's because a not-nir thing is trying to use nir_debug that we're running into problems?
07:27
<
jasuarez >
to replace a switch() by an array
07:27
<
imirkin >
anyways, i'm probably not the right person to comment on this, i didn't look over the overall design
07:28
<
jasuarez >
regarding initializing, yes, I was initializing it when you create a nir shader
07:28
<
jasuarez >
as it makes sense to use NIR_DEBUG when you deal with NIR shaders
07:29
<
jasuarez >
but for this TGSI to NIR specific case, didn't realize NIR_DEBUG was used before creating the NIR shader, so it was uninitialized
07:35
alanc has quit [Remote host closed the connection]
07:36
alanc has joined #dri-devel
07:36
<
imirkin >
jasuarez: you could do something snaky with the "default" value provided for NIR_DEBUG
07:36
<
imirkin >
oh hm, no
07:40
mlankhorst has joined #dri-devel
07:45
YuGiOhJCJ has joined #dri-devel
07:45
danvet has joined #dri-devel
07:47
<
jasuarez >
I'll change the default value to 0, and use a static variable to initialize, the same way as the macros do
07:49
<
imirkin >
jasuarez: but how do you ensure that it gets called?
07:49
<
imirkin >
jasuarez: i guess if you require a wrapper to be used to grab nir_debug each time?
07:49
<
jasuarez >
sorry, dind't explain myself
07:50
<
jasuarez >
before I was using the same nir_debug variable to find out if it was initialized
07:51
<
jasuarez >
because the init function is always invoked when you create a NIR shader
07:51
<
jasuarez >
now I still call it, but use a static variable to only do the job once
07:51
<
jasuarez >
this is how the debug macros works
07:53
<
imirkin >
yeah, i already looked over how the macros work
07:53
<
imirkin >
i guess i'm looking for a more foolproof way of dealing with this
07:53
<
imirkin >
so you don't have to remember to call some other function
07:53
<
imirkin >
i guess if that function is only called in the "core" code which cares about the per-shader stuff
07:53
<
imirkin >
then it should be fine i guess?
07:53
<
jasuarez >
I would like to have a way to call this initialization automatically
07:54
<
imirkin >
anyways, i'm off to sleep
07:54
<
jasuarez >
without doing it manually
07:54
<
jasuarez >
yeah, sorry
07:54
<
jasuarez >
I wake up very early to fix this :(
08:03
mclasen has quit [Ping timeout: 480 seconds]
08:03
f11f12 has joined #dri-devel
08:33
Haaninjo has joined #dri-devel
09:20
rasterman has joined #dri-devel
09:30
rasterman has quit [Quit: Gettin' stinky!]
09:33
camus1 has joined #dri-devel
09:35
camus has quit [Read error: Connection reset by peer]
09:35
gouchi has joined #dri-devel
09:37
Company has joined #dri-devel
09:44
yk has quit [Remote host closed the connection]
09:49
YuGiOhJCJ has quit [Quit: YuGiOhJCJ]
09:52
lemonzest has quit [Quit: WeeChat 3.3]
10:04
zsh has joined #dri-devel
10:05
lemonzest has joined #dri-devel
10:31
camus has joined #dri-devel
10:32
camus1 has quit [Read error: Connection reset by peer]
10:54
mlankhorst has quit [Ping timeout: 480 seconds]
10:55
pcercuei has joined #dri-devel
10:59
camus1 has joined #dri-devel
10:59
camus has quit [Ping timeout: 480 seconds]
11:00
mlankhorst has joined #dri-devel
11:03
FireBurn has joined #dri-devel
11:28
oneforall2 has quit [Read error: Connection reset by peer]
11:29
oneforall2 has joined #dri-devel
11:33
camus has joined #dri-devel
11:37
JohnnyonFlame has joined #dri-devel
11:38
camus1 has quit [Ping timeout: 480 seconds]
11:53
anthk_ has joined #dri-devel
11:56
<
anthk_ >
hello, I own an Intel Gemini Lake based NUC [ Mesa Intel(R) UHD Graphics 605 (GLK 3) ]. Will it support Gallium3D?
12:02
<
HdkR >
Gen 9.5 so it should just work
12:03
<
HdkR >
Just a super low end part
12:03
flacks has quit [Quit: Quitter]
12:06
flacks has joined #dri-devel
12:15
<
anthk_ >
HdkR: well, I tried ' MESA_LOADER_DRIVER_OVERRIDE=crocus glxgears -info | less ' and I see the renderer is llvmpipe
12:16
<
HdkR >
anthk_: Crocus is for gen 4-7. You're looking for iris
12:17
<
anthk_ >
ah, iris will work? thanks
12:21
libv_ is now known as libv
12:29
March-123 has joined #dri-devel
12:33
<
anthk_ >
is there an env var to know which DRI file MESA loads?
12:34
<
HdkR >
I usually just strace to see
12:36
anthk_ has quit [Quit: Konversation terminated!]
12:36
<
emersion >
DRI_DEBUG=1 should tell as well
12:41
gawin has joined #dri-devel
12:42
anthk_ has joined #dri-devel
12:47
March-123 has quit [Ping timeout: 480 seconds]
12:59
anthk_ has quit [Read error: No route to host]
12:59
anthk_ has joined #dri-devel
13:15
alyssa has joined #dri-devel
13:16
<
alyssa >
No more mesa classic? Woot woot!!
13:16
<
alyssa >
dcbaker: !!
13:21
JohnnyonFlame has quit [Ping timeout: 480 seconds]
13:28
gawin has quit [Ping timeout: 480 seconds]
13:39
mlankhorst has quit [Ping timeout: 480 seconds]
13:58
danvet has left #dri-devel [Leaving]
14:10
mclasen has joined #dri-devel
14:10
JohnnyonFlame has joined #dri-devel
14:24
offtopic[ON] has joined #dri-devel
14:25
danvet has joined #dri-devel
14:35
mmenzyns has quit []
14:35
gawin has joined #dri-devel
14:35
enilflah has quit []
14:36
mmenzyns has joined #dri-devel
14:41
anthk_ has quit [Ping timeout: 480 seconds]
14:41
March-123 has joined #dri-devel
14:42
ajax has quit [Ping timeout: 480 seconds]
14:50
ajax has joined #dri-devel
14:51
March-123 has quit [Ping timeout: 480 seconds]
15:00
glennk has quit [Remote host closed the connection]
15:01
enilflah has joined #dri-devel
15:11
camus has joined #dri-devel
15:24
jcline has quit [Ping timeout: 480 seconds]
15:26
<
daniels >
wonder how many people are lining up to burn the DRI interfaces
15:28
The_Company has joined #dri-devel
15:34
Company has quit [Ping timeout: 480 seconds]
15:37
mlankhorst has joined #dri-devel
15:37
glennk has joined #dri-devel
16:01
<
jekstrand >
Ooh, can we? Please? Is everything really modesetting these days? That'd be awesome.
16:02
kevintang has joined #dri-devel
16:15
kevintang has left #dri-devel [#dri-devel]
16:16
kevina has joined #dri-devel
16:16
JohnnyonFlame has quit [Ping timeout: 480 seconds]
16:39
kevina has quit [Read error: Connection reset by peer]
16:43
<
imirkin >
jekstrand: nouveau doesn't use modesetting. i also suspect intel gen3 doesn't either.
16:43
<
jekstrand >
But Intel gen3 isn't in the mesa tree anymore. :-P
16:43
<
imirkin >
yes it is
16:43
<
jekstrand >
Oh, right... i915g
16:43
<
imirkin >
and nouveau is too :)
16:45
ella-0_ has joined #dri-devel
16:47
ella-0 has quit [Read error: Connection reset by peer]
17:05
<
bnieuwenhuizen >
... and xf86-video-amdgpu?
17:05
<
imirkin >
that one uses glamor though
17:05
<
imirkin >
so i think it's more like modesetting as far as DRI interfaces are concerned?
17:06
<
imirkin >
but yeah, i doubt modesetting would go over well on r300 gpu's either
17:14
<
vsyrjala >
isn't the dri interface stuff all internal to mesa? ie. why would the ddx matter?
17:14
gawin has quit [Ping timeout: 480 seconds]
17:28
<
bnieuwenhuizen >
I think not? IIRC it is also used by minigbm and AMD OpenCL
17:35
pcercuei has quit [Quit: brb]
17:40
pcercuei has joined #dri-devel
17:42
<
alyssa >
AMD OpenCL is still a thing?
17:43
<
alyssa >
What's all the clover NATIVE stuff for then
17:43
<
FLHerne >
clover doesn't really work yet :p
17:43
<
imirkin >
vsyrjala: presumably in reference to the dri interface that the X DDX uses to query stuff about the driver, and maybe other things?
17:44
<
FLHerne >
well, it lacks support for features that many real-world applications require
17:46
<
alyssa >
FLHerne: that still doesn't explain the NATIVE stuff
17:47
<
alyssa >
admittedly clang -> llvm -> spirv -> nir -> llvm -> amdgpu might be a tad excessive otherwise
17:47
<
bnieuwenhuizen >
not sure if the AMD CL stack for that is still a thing or if it has been replaced by the ROCm based stack
17:59
gawin has joined #dri-devel
18:09
anthk_ has joined #dri-devel
18:19
xx1 has joined #dri-devel
18:25
adjtm has quit [Quit: Leaving]
18:26
mlankhorst has quit [Ping timeout: 480 seconds]
18:26
xx1 has left #dri-devel [Leaving]
18:29
xx1 has joined #dri-devel
18:31
xx1 has quit [Quit: Leaving]
18:42
<
jenatali >
Thanks :) I'll push r-b tags in a minute (once my PC finishes rebooting) and then would love if one of you would merge for me, I don't have piglit dev permissions
18:44
<
Kayden >
I got it :)
18:44
<
Kayden >
you -should- have permissions, though, they should be the same as mesa I thought...
18:45
<
jenatali >
Whoever gave me dev permissions for Mesa did it for the repo, not the project
18:45
<
jenatali >
I saw that most people get it inherited from the overall project though, yeah
18:45
<
jenatali >
Wouldn't complain about getting that fixed :P
18:45
JohnnyonFlame has joined #dri-devel
18:46
<
jekstrand >
jenatali: fixed
18:47
<
alyssa >
6 files changed, 1493 insertions(+), 25 deletions(-)
18:47
<
alyssa >
phoronix in 3... 2...
18:48
<
Kayden >
jenatali: just fixed that :)
18:48
<
jenatali >
Thanks guys :)
18:48
<
Kayden >
definitely ping us if something goes awry, I removed your direct access but granted you the whole project access, it shows you as having access still via the group
18:48
<
jekstrand >
alyssa: Just pulled and didn't see it. Or is an MR?
18:48
<
jenatali >
Cool, will do
18:49
<
alyssa >
jekstrand: MR adding XML + decoding for valhall
18:49
<
alyssa >
Lot of bulk that can be landed without affecting existing Malis at all, sounds like a good way to whittle down my Valhall branch
18:49
<
jekstrand >
alyssa: Do you want a blind ACK or are you oging to just assign marge?
18:50
<
alyssa >
was waiting for bbrezillon
18:50
<
alyssa >
just posted it's a saturday why are y'all working on a saturday
18:50
<
jenatali >
Good question :P
18:51
<
alyssa >
jenatali: I'm a student with minimal spare cycles during the week, what's your excuse :-p
18:51
<
jekstrand >
alyssa: Because we're all lame and don't have real lives. :P
18:51
<
alyssa >
Oh so we're all on the same page then 👍
18:53
fxkamd has joined #dri-devel
18:53
glennk has quit [Remote host closed the connection]
18:54
glennk has joined #dri-devel
18:59
crabbedhaloablut has quit [Ping timeout: 480 seconds]
19:05
eukara has joined #dri-devel
19:05
<
daniels >
imirkin: and yeah, as far as the runners go, they poll a gitlab.fd.o endpoint for any spare jobs every second or two; if they take it they receive the job definition and start pulling in the container image, git tree, artifacts, etc
19:05
<
imirkin >
daniels: makes sense, thanks
19:05
<
daniels >
imirkin: as for htz, I can see tpm trying to debug them in #freedesktop
19:14
mszyprow has joined #dri-devel
19:19
<
dcbaker >
alyssa: don't worry, phoronix can't count :)
19:25
crabbedhaloablut has joined #dri-devel
19:34
LexSfX has joined #dri-devel
19:35
<
daniels >
understanding gives you Phoronix, but we subscribe to the Richard Hamming viewpoint: the goal of computing is insight, not numbers.'
19:39
LexSfX has quit [Remote host closed the connection]
19:39
<
imirkin >
understanding is hard. comparing 2 numbers is easy.
19:39
<
imirkin >
(and just showing 1 number is even easier)
19:40
ppascher has quit [Ping timeout: 480 seconds]
19:50
JohnnyonFlame has quit [Ping timeout: 480 seconds]
19:54
LexSfX has joined #dri-devel
19:59
rasterman has joined #dri-devel
20:23
vyivel has quit [Remote host closed the connection]
20:23
bl4ckb0ne has quit [Remote host closed the connection]
20:23
emersion has quit [Remote host closed the connection]
20:23
emersion has joined #dri-devel
20:25
bl4ckb0ne has joined #dri-devel
20:25
vyivel has joined #dri-devel
20:50
adjtm has joined #dri-devel
20:52
ella-0_ has quit [Read error: Connection reset by peer]
21:04
gouchi has quit [Remote host closed the connection]
21:11
gouchi has joined #dri-devel
21:11
gouchi has quit [Remote host closed the connection]
21:20
ella-0 has joined #dri-devel
21:21
offtopic[ON] has quit []
21:32
gouchi has joined #dri-devel
21:44
<
alyssa >
Fail (Fail)
21:45
<
jekstrand >
Such a helpful message!
21:45
<
alyssa >
jekstrand: TBF that's me running mesa/valhall on my m1 so for some reason nothing is rendered ;-P
21:45
<
imirkin >
yeah, dEQP stands for like double everything's quality... what they meant was "double the quality of error output of the tests"
21:46
<
jenatali >
Huh, switching piglit in CI from glut to waffle (like I have locally) is causing it to hang... that's fun
21:46
<
imirkin >
the KHR tests continue to have zero useful output. very annoying =/
21:47
<
alyssa >
imirkin: oh, yeah... the KHR tests are somehow worse than dEQP for debug...
21:47
<
jekstrand >
alyssa: Psh! Details! They're both ARM, aren't they?
21:47
<
alyssa >
jekstrand: No, only one is an ARM CPU
21:47
<
alyssa >
The other is Apple Architecture(TM)
21:48
<
jekstrand >
Silly me
21:49
<
imirkin >
aarch vs aarch64? :)
21:55
<
alyssa >
jekstrand: pandecode still works with shader-db
21:55
* alyssa
can never remember what other drivers call that
21:55
<
alyssa >
aubinator I think?
21:55
<
imirkin >
that's just intel :)
21:55
<
jekstrand >
We've got like 3 tools
21:55
<
imirkin >
which produces "aub" files i guess, which stands for something clever one might assume
21:56
<
alyssa >
broadcom is clif I guess
21:56
<
jekstrand >
intel_stub_gpu is the thing that doesn't do any dumping
21:56
<
alyssa >
on dEQP-GLES2.functional.shaders.operator.unary_operator.plus.lowp_float_fragment
21:57
<
alyssa >
attribute/varyings are missing there, and the shaders are thoroughly wrong, but.. progress..?
21:57
<
jekstrand >
That's like our aubinator, yeah
21:57
<
imirkin >
looks pretty good
22:08
danvet has quit [Ping timeout: 480 seconds]
22:16
boistordu has joined #dri-devel
22:27
JohnnyonFlame has joined #dri-devel
22:37
The_Company has quit []
22:39
gouchi has quit [Remote host closed the connection]
22:52
Duke`` has quit [Ping timeout: 480 seconds]
23:23
<
alyssa >
did the macOS mesa build just break? :|
23:24
<
alyssa >
or was it already so broken i just need to do the usual dance
23:24
<
alyssa >
okay yes this is the usual SNAFU
23:40
gawin has quit [Ping timeout: 480 seconds]
23:59
robertfoss has joined #dri-devel