<newton688>
Hi All, sorry for the wait. I've raised an issue here with the details of my system, some steps that I used to produce the problem and some kernel logs. If there's any more detail that I can provide, please let me know. Also, workarounds would be great if possible. Thank you. https://gitlab.freedesktop.org/mesa/mesa/-/issues/4389
<alyssa>
newton688: I don't suppose it'd be possible to bisect the issue?
<alyssa>
"Mysterious" faults in dmesg / freezes / hangs are notoriously hard to debug, which is why there are so many of them on the issue tracker.
<alyssa>
(Since the easy issues have been solved :< )
<alyssa>
(More reason to wire up kbase for soak testing to see if mesa really is at fault with stuff)
kaspter has joined #panfrost
<newton688>
alyssa: not sure, what should I bisect? :)
<newton688>
The log seems to be some kind of todo in panfrast.
atler has quit [Killed (orwell.freenode.net (Nickname regained by services))]
atler has joined #panfrost
<italove>
you mentioned that the issue happened after an update recently, so bisecting the mesa looks like a good way to find out which commit caused the issue
<italove>
bisect the mesa repo***
<italove>
newton688: that TODO is there for a long time, I don't think it's related to the regression
jernej has quit [Ping timeout: 240 seconds]
jernej has joined #panfrost
Stary has quit [Ping timeout: 260 seconds]
Stary has joined #panfrost
<HdkR>
I should probably double check gfxbench again, fixed some bugs that might resolve the crash I saw in it
Ashleeee has quit [Ping timeout: 276 seconds]
Stenzek has quit [Ping timeout: 264 seconds]
<newton688>
italove: Ok, thanks. The error does seem highly correlated with the problems I'm seeing. Some have suggested that it could be because KWin is doing fancier things. I'll try to see if switching to Wayland/Gnome helps.
Stenzek has joined #panfrost
<alyssa>
still a panfrost bug...
davidlt has joined #panfrost
<italove>
yep, sorry, I could've phrased that better, the panfrost error messages are correlated with the regression, I was just saying that the "todo" message doesn't really tell us what caused the regression
newton688 has quit [Ping timeout: 240 seconds]
vstehle has joined #panfrost
tgall_fo_ has joined #panfrost
tgall_foo has quit [Ping timeout: 246 seconds]
camus has joined #panfrost
kaspter has quit [Ping timeout: 245 seconds]
camus is now known as kaspter
raster has joined #panfrost
raster has quit [Remote host closed the connection]
raster has joined #panfrost
cwabbott_ has joined #panfrost
narmstrong_ has joined #panfrost
kinkinkijkin_ has joined #panfrost
Werner__ has joined #panfrost
austriancoder_ has joined #panfrost
ric96_ has joined #panfrost
lvrp16_ has joined #panfrost
steev_ has joined #panfrost
phh|new has joined #panfrost
embed-3d has joined #panfrost
milkii_ has joined #panfrost
unoccupied has joined #panfrost
empty_string has joined #panfrost
afaerber_ has joined #panfrost
alpernebbi has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
empty_st- has quit [*.net *.split]
ric96 has quit [*.net *.split]
leah has quit [*.net *.split]
phh has quit [*.net *.split]
Werner has quit [*.net *.split]
milkii has quit [*.net *.split]
afaerber has quit [*.net *.split]
cwabbott has quit [*.net *.split]
kinkinkijkin has quit [*.net *.split]
narmstrong has quit [*.net *.split]
steev has quit [*.net *.split]
lvrp16 has quit [*.net *.split]
austriancoder has quit [*.net *.split]
cwabbott_ is now known as cwabbott
ric96_ is now known as ric96
narmstrong_ is now known as narmstrong
kinkinkijkin_ is now known as kinkinkijkin
austriancoder_ is now known as austriancoder
lvrp16_ is now known as lvrp16
steev_ is now known as steev
raster has joined #panfrost
raster has quit [Client Quit]
phh|new is now known as phh
raster has joined #panfrost
raster has quit [Remote host closed the connection]
<icecream95>
Looks like push constants needs some optimising: 49.93% drawoverhead [.] panfrost_emit_const_buf
raster has joined #panfrost
raster has quit [Client Quit]
rak-zero has joined #panfrost
Werner__ is now known as Werner
stikonas has joined #panfrost
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
raster has joined #panfrost
afaerber_ is now known as afaerber
warpme_ has joined #panfrost
<raster>
o/
<macc24>
o/ raster
<raster>
hooroo
<macc24>
5.10.19 now boots on duet \o/ with fixes stolen from hexdump0815
<bbrezillon>
narmstrong: do you know if it'd be possible to implement ->gem_prime_import_sg_table in the Amlogic driver?
<bbrezillon>
(IOW, is there an IOMMU or is sg-DMA supported by the display controller)
<narmstrong>
bbrezillon: afaik no there is no way (I know) to import via scatter gather and there is no iommu on the display side on the existing socs
<narmstrong>
I’ll dig again in the code to see but amlogic designed the primary plane for « OSD » and 1080p Max Android TV UI
<alyssa>
bbrezillon: what for ooi?
<bbrezillon>
alyssa: looks like the generic vulkan WSI layer creates presentable images using the regular image creation path, and then passes those buffers to the compositor as if they were importable on the KMS side
<bbrezillon>
but I probably missed something
<alyssa>
ah..
<alyssa>
The good news (?) is that compositors are still GL and the display controller doesn't matter for X11/wayland backends, right?
<bbrezillon>
yep
<bbrezillon>
well, it matters for the wayland case (the one I'm testing)
<bbrezillon>
since weston tries to import a GPU buffer in the KMS dev
<tomeu>
bbrezillon: wonder if the v3d vk driver isn't in the same situation