<tomeu>
Lyude: sounds good, I would suggest for us to start sending MRs against the panfrost group, and also hosting there our fork of the kernel driver from ARM
<narmstrong>
tomeu: Lyude: I think it would be a much better base to continue working !
<tomeu>
mmind00: by a chance, do you think the artifacts in the video above could be related to the display?
<tomeu>
narmstrong: tomorrow I will do some dumping
<narmstrong>
I need to re-run kmscube with your fix
<mmind00>
tomeu: hmm, I'm no graphics guru, but at least kmscube via lima works without artifacts on rk3328 (vop with iommu) and rk3188 (vop without iommu)
<mmind00>
tomeu: so at least the prime import stuff seems to work
<mmind00>
tomeu: but then I don't know how your wip kernel driver does that at this point
<tomeu>
mmind00: oh, that's with ARM's driver
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
<mmind00>
tomeu: yep, so I guess it could be related to both somehow competing for display space or something
<mmind00>
tomeu: aka the fbcon and panfrost somehow accessing the same area
<tomeu>
mmind00: rockchip-drm and mali?
<mmind00>
I remember one of Luke's tamil videos where the console also shone through the 3d stuff
<tomeu>
oh, but we don't see the console here, do we?
<alyssa>
So it works well enough to render some stuff but other tiles get wacked before they finish writeout since the hw picks up something's wrong
<alyssa>
I don't think I ever played with kmscube myself, so a rendering bug like that is totally possible
<alyssa>
(I have played with freedreno's ./test-cube and glmark's cubes, but I don't know what the diff is)
chewitt has joined #panfrost
<Lyude>
alyssa: kmscube renders a cube directly to your DRM display device
<Lyude>
With no middle man involved, kmscube makes a perfect test subject for all of your KMS and cube related needs
<alyssa>
Lyude: Hence why I haven't played with it ;)
<Lyude>
Mhm
<alyssa>
What I don't understand is what it could be faulting over
<alyssa>
Since I'm reading the source now and it's not doing anything we haven't done before
<alyssa>
Only thing that's slightly weird is how they're packing the attribs, but I was sure we'd fixed that with glmark a while ago
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
<alyssa>
A panwrap could be enlightening
<alyssa>
Maybe the vbo handling code (panfrost_set_vertex_buffers, panfrost_emit_vertex_data) is wrong
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
<alyssa>
It... looks right?
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
chewitt has quit [Quit: Zzz..]
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
chewitt has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
anarsoul|2 has joined #panfrost
chewitt has quit [Quit: Zzz..]
chewitt has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
pH5 has quit [Quit: bye]
pH5 has joined #panfrost
cwabbott_ has joined #panfrost
cwabbott has quit [Ping timeout: 240 seconds]
cwabbott_ is now known as cwabbott
TheKit has quit [Remote host closed the connection]
TheKit has joined #panfrost
cwabbott has quit [Ping timeout: 250 seconds]
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
cwabbott has joined #panfrost
<daniels>
tomeu: looks tile-shaped - stupid question, but if you do glFinish(); usleep(1sec); before handing to kms, and usleep(1sec) before rendering, does it change?
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
chewitt has quit [Quit: Zzz..]
<alyssa>
..or that
<alyssa>
Also stupid suggestion, what if you set forceFlushFragment to true?