00:02
adjtm has quit [Remote host closed the connection]
00:02
adjtm has joined #lima
00:13
<
MoeIcenowy >
anarsoul: maybe we should introduce some app blacklist like panfrost?
00:14
<
MoeIcenowy >
(although I doubt whether the panfrost blacklist should be removed
00:14
<
anarsoul >
I'd prefer to fix the issue in lima
00:14
<
anarsoul >
you can launch it with LIBGL_ALWAYS_SOFTWARE=1 if you want to use it with sw rendering
00:15
<
anarsoul >
MoeIcenowy: btw is your aosc build for pinebook available for download?
00:16
<
MoeIcenowy >
(this should be at America
00:18
<
MoeIcenowy >
BTW I'm using the same version of chromium and AMDGPU driver on my PC
00:18
<
MoeIcenowy >
(my video card changed in the sequence of i965/iris, nouveau, amdgpu in this coronavirus outbreak
00:19
<
MoeIcenowy >
so should not be gallium issue
00:19
<
MoeIcenowy >
or cr issue
00:19
<
anarsoul >
I'm waiting for new thinkpads with ryzen 4000 to update my laptop
00:19
<
anarsoul >
MoeIcenowy: likely some format is not supported, I wonder if chromium now needs float textures
00:20
<
anarsoul >
MoeIcenowy: do you patch lima in order for gnome to work?
00:20
<
anarsoul >
or it's fine as is?
00:20
<
MoeIcenowy >
I used to patch it
00:21
<
MoeIcenowy >
but now seems that patchless is also okay
00:21
<
anarsoul >
gnome from archlinux crashes on lima for me :(
00:21
<
anarsoul >
same on panfrost
00:21
<
MoeIcenowy >
anarsoul: then try to patch TEXTURE_3D ?
00:21
<
MoeIcenowy >
and another crash source is my fix of invalidated resource
00:22
<
MoeIcenowy >
(I mean I fixed the crash source
00:22
<
anarsoul >
yeah, but your fix is already merged into mesa
00:22
<
MoeIcenowy >
BTW I regretted to release the image. I never tested it on Pinebook (only PineTab) and now I start to imagine how it struggles on Pinebook 1080p
00:23
<
MoeIcenowy >
(my own Pinebook runs KDE now
00:23
<
anarsoul >
try wayland session in KDE
00:23
<
anarsoul >
it should work better
00:23
<
anarsoul >
than X11
00:24
<
MoeIcenowy >
anarsoul: I need IM
00:24
<
anarsoul >
input method?
00:24
<
anarsoul >
it's not supported in wayland yet?
00:25
<
MoeIcenowy >
I think so
00:25
<
anarsoul >
it works with cyrillic layouts
00:25
<
anarsoul >
anyway it's worth comparing performance :)
00:26
<
anarsoul >
panfrost is noticeably faster on wayland
00:26
<
anarsoul >
I expect lima to be the same
00:26
<
MoeIcenowy >
Time for a rock'n'roll of Chromium on Panfrost
00:27
<
MoeIcenowy >
IT WORKS! (screaming)
00:27
<
MoeIcenowy >
maybe I need to make a MR to clean the app blacklist
00:27
<
anarsoul >
probably it breaks somewhere else :)
00:27
<
anarsoul >
so I guess it's missing format in lima?
00:30
<
MoeIcenowy >
or maybe I built a desktop GL chromium
00:30
<
MoeIcenowy >
as I seen "2.1" in the version string on Panfr
00:31
<
anarsoul >
that's fine, we're exposing desktop GL
00:31
<
MoeIcenowy >
we lack some desktop features
00:32
<
anarsoul >
yet it shouldn't fail due to some missing format
00:34
<
MoeIcenowy >
maybe I should try to drop ETC?
00:35
<
MoeIcenowy >
not the problem of ETC
00:35
<
anarsoul >
I thought you tried to downgrade chromium and it worked
00:36
<
MoeIcenowy >
anarsoul: run with `--with-gl=egl` works
00:37
<
anarsoul >
try capturing apitrace?
00:37
<
MoeIcenowy >
this enables the usage of GLES
00:37
<
MoeIcenowy >
the trace is totally nonsense
00:37
<
anarsoul >
also I'd suggest trying to add RGBX sRGB formats
00:39
<
MoeIcenowy >
maybe as a distro maker I should try to make it default to GLES on ARM
00:39
<
MoeIcenowy >
(it's also one of the policy now on my distro
00:39
<
MoeIcenowy >
but as a lima developer here
00:39
<
anarsoul >
yet fixing lima is better :)
00:39
<
MoeIcenowy >
should we really fix an issue of our fake GL non-ES?
00:40
<
anarsoul >
we just fake few features
00:40
<
MoeIcenowy >
adding SRGB for RGBX doesn't helps
00:42
<
anarsoul >
try adding traces to lima_format_pixel_supported() and lima_format_texel_supported()?
00:43
<
anarsoul >
nah, nevermind
00:43
<
anarsoul >
just add traces :)
00:49
<
anarsoul >
MoeIcenowy: what gnome version do you use?
00:54
<
anarsoul >
isn't it a bit old?
00:54
<
anarsoul >
or it's 3.34?
00:55
<
MoeIcenowy >
maybe 3.34
00:55
<
MoeIcenowy >
y 3.34
00:56
<
MoeIcenowy >
anarsoul: I added traces for formats not defined in the array
00:56
<
MoeIcenowy >
before the program fails
00:56
<
MoeIcenowy >
it tries to ask for pixel format 68
00:56
<
anarsoul >
and what format is it?
00:56
<
MoeIcenowy >
it's a difficulty to convert it back
00:57
<
anarsoul >
PIPE_FORMAT_R8G8B8_UNORM
00:57
<
anarsoul >
are you sure you're using up-to-date mesa?
00:57
<
MoeIcenowy >
we hit GL_RGB again?
00:57
<
MoeIcenowy >
I am sure
00:58
<
MoeIcenowy >
because it's up-to-date, RGB8 is not defined
00:58
<
anarsoul >
MoeIcenowy: GL_RGB will be promoted to RGBX
00:58
<
MoeIcenowy >
in any case?
00:59
<
anarsoul >
I think so
00:59
<
anarsoul >
MoeIcenowy: maybe they changed their workarounds in recent chromium releases?
01:00
<
MoeIcenowy >
anarsoul: I think we never trigger Cr's workarounds
01:00
<
MoeIcenowy >
it matches Mali blob strings
01:00
<
MoeIcenowy >
Our vendor is not "ARM"
01:00
<
MoeIcenowy >
it's "lima"
01:00
<
anarsoul >
maybe they relaxed it recently?
01:01
<
MoeIcenowy >
maybe our lowering fails in Mesa?
01:01
<
MoeIcenowy >
BTW unknown in this trace means not in lima_format_table
01:01
<
MoeIcenowy >
even those with NONE pixel format is considered known
01:02
<
anarsoul >
MoeIcenowy: probably you should print all the formats
01:02
<
anarsoul >
not only missing
01:04
<
MoeIcenowy >
full log
01:07
<
anarsoul >
70 is PIPE_FORMAT_R8G8B8A8_UNORM
01:07
<
MoeIcenowy >
that sounds quite normal...
01:09
<
anarsoul >
so it failed for some different reason
01:09
<
anarsoul >
did you run out of CMA? :)
01:10
<
MoeIcenowy >
I think no
01:10
<
anarsoul >
try setting break points in lima_resource_create_scanout() and in lima_resource_create_bo()
01:10
<
anarsoul >
at the line where it returns NULL
01:11
<
MoeIcenowy >
breakpoint debugging chromium is nonsense
01:11
<
MoeIcenowy >
it's multi process
01:11
<
anarsoul >
add traces there
01:12
<
anarsoul >
1920x959x4 is quite large so I wouldn't be surprised if you ran out of CMA
01:13
<
anarsoul >
how much CMA you have?
01:14
<
MoeIcenowy >
`Memory: 1789108K/2064384K available (11516K kernel code, 9648K rwdata, 7416K rodata, 896K init, 341K bss, 70476K reserved, 204800K cma-reserved)`
01:15
<
anarsoul >
that should be enough
01:15
<
anarsoul >
I'm out of ideas :)
01:15
<
MoeIcenowy >
anarsoul: I tried to add "return NULL" traces
01:15
<
MoeIcenowy >
no result
01:16
<
MoeIcenowy >
and I think running on GLES succeeds means no CMA issue
01:20
<
anarsoul >
btw 195 is PIPE_FORMAT_R8G8B8A8_SINT
01:21
<
anarsoul >
so I wonder if it needs SINT formats
01:35
<
MoeIcenowy >
SINT is beyond what we can process, right?
01:35
<
MoeIcenowy >
or do we have some lowering facility
01:37
<
anarsoul >
MoeIcenowy: I'm not sure whether it's actually the root cause
01:40
<
anarsoul >
or maybe just switch to firefox? :)
01:59
kaspter has joined #lima
02:12
camus1 has joined #lima
02:13
kaspter has quit [Ping timeout: 258 seconds]
02:13
camus1 is now known as kaspter
03:00
kaspter has quit [Remote host closed the connection]
03:01
kaspter has joined #lima
03:11
camus1 has joined #lima
03:11
kaspter has quit [Ping timeout: 256 seconds]
03:11
camus1 is now known as kaspter
03:21
kaspter has quit [Ping timeout: 256 seconds]
03:22
kaspter has joined #lima
03:38
<
MoeIcenowy >
anarsoul: switch to GLES solves it.
03:38
<
anarsoul >
MoeIcenowy: OK
03:38
<
MoeIcenowy >
but this makes this problem more mysterious
03:50
deesix_ has joined #lima
03:51
dddddd_ has joined #lima
03:52
buzzmarshall has quit [Remote host closed the connection]
03:53
deesix has quit [Ping timeout: 272 seconds]
03:54
dddddd has quit [Ping timeout: 265 seconds]
04:20
kaspter has quit [Ping timeout: 256 seconds]
04:20
kaspter has joined #lima
05:23
_whitelogger has joined #lima
05:35
niceplace has quit [Ping timeout: 272 seconds]
05:37
niceplace has joined #lima
06:02
kaspter has quit [Ping timeout: 256 seconds]
06:02
kaspter has joined #lima
06:54
dddddd_ has quit [Ping timeout: 260 seconds]
06:57
Elpaulo has quit [Quit: Elpaulo]
07:04
kaspter has quit [Ping timeout: 268 seconds]
07:04
kaspter has joined #lima
07:14
camus1 has joined #lima
07:15
kaspter has quit [Ping timeout: 256 seconds]
07:15
camus1 is now known as kaspter
08:42
yann has quit [Ping timeout: 256 seconds]
09:07
<
anarsoul >
rellla: have you tested "[PATCH 0/5] drm/lima: add error debug functionality" patch series?
09:07
<
anarsoul >
if no I'll test it tomorrow. Looks good to me in general
09:08
camus1 has joined #lima
09:10
kaspter has quit [Ping timeout: 240 seconds]
09:10
camus1 is now known as kaspter
09:38
yann has joined #lima
09:55
monstr has joined #lima
10:06
gcl_ has joined #lima
10:07
gcl has quit [Ping timeout: 256 seconds]
10:55
<
rellla >
anarsoul: no, i haven't done that yet ...
10:55
<
rellla >
i can do though
10:56
z3ntu_ has quit [Quit: killed]
10:56
mariogrip has quit [Quit: killed]
11:20
z3ntu_ has joined #lima
11:49
mariogrip has joined #lima
12:09
paulk-leonov has quit [Ping timeout: 272 seconds]
12:10
paulk-leonov has joined #lima
13:32
deesix_ is now known as deesix
14:26
Barada has quit [Quit: Barada]
15:58
tlwoerner is now known as tw-eh
15:58
tw-eh is now known as tlwoerner
16:12
monstr has quit [Remote host closed the connection]
17:11
gcl_ has quit [Ping timeout: 260 seconds]
17:18
gcl has joined #lima
17:20
yann has quit [Ping timeout: 258 seconds]
18:48
yann has joined #lima
20:01
dddddd has joined #lima
20:44
buzzmarshall has joined #lima
21:59
niceplace has quit [Ping timeout: 268 seconds]
22:03
niceplace has joined #lima