davidlt has quit [Remote host closed the connection]
stikonas has quit [Remote host closed the connection]
icecream95 has quit [Ping timeout: 240 seconds]
guillaume_g has quit [Ping timeout: 246 seconds]
guillaume_g has joined #panfrost
megi has joined #panfrost
guillaume_g has quit [Quit: Konversation terminated!]
raster has joined #panfrost
yann|work has joined #panfrost
<bbrezillon>
robher: we really need http://code.bulix.org/835vlj-1119387, as reported by icecream95 and tlwoerner, if we don't have that one, we keep having GPU page faults
<bbrezillon>
alyssa: looks like we also lack a panfrost_batch_add_bo() http://code.bulix.org/2mo90u-1119485, but that's probably not a problem if panfrost_get_query_result() is always called before panfrost_destroy_query()
<robher>
bbrezillon: okay, I'll apply it.
<bbrezillon>
tlwoerner: BTW, I couldn't reproduce the js faults with this patch applied (tested on mesa master + kernel-5.5+above-patch)
alpernebbi has joined #panfrost
buzzmarshall has joined #panfrost
robmur01 has joined #panfrost
<alyssa>
bbrezillon: occlusion queries aren't used in gles 2.0 anyway
<bbrezillon>
alyssa: ok
<tlwoerner>
bbrezillon: i only see them with chromium, and they're not perpetual, i can run chromium for a bit without any issue, then i'll get a flood of them, then it's good for a while, then a flood, etc
<alyssa>
tlwoerner: chromium shouldn't be using panfrost at all?
<bbrezillon>
tlwoerner: okay. I played with chromium, but not long enough apparently
<alyssa>
I have noticed that chromium will freeze for me and then i need to click the tab to get a repaint
<alyssa>
never looked at dmesg, maybe those are the same page faults?
<tlwoerner>
in my case, visually, chromium is working just fine. i'm playing around on youtube for my tests, is there anything recommended?
guillaume_g has quit [Quit: Konversation terminated!]
raster has quit [Quit: Gettin' stinky!]
yann|work has joined #panfrost
nerdboy has joined #panfrost
raster has joined #panfrost
stikonas has joined #panfrost
<alyssa>
tlwoerner: fwiw, the freezes don't cause any faults on 5.2
<alyssa>
so probably unrelated
<alyssa>
But regardless chromium should be software rendering (which will still work just fine, except you won't have webgl)
<alyssa>
tomeu: IIRC we were underallocating the slices for... mipmapped 3D textures? if my memory serves? since we were mixing up the dimensions, that patch was a ~hack to fix it (and IIRC fixed the test you were asking about)
<alyssa>
MoeIcenowy: OQs are needed for desktop GL and ES3 but not ES2 ;)
<HdkR>
Also ES3 OC is boolean only :P
<tlwoerner>
alyssa: according to the chromium info page, i do have webgl
<alyssa>
....Huh. Did you patch out the blacklist..?
<HdkR>
Or is it using swiftshader?
<alyssa>
HdkR: Also that, yes. We have boolean OQs working... there's supposedly a non-boolean mode too but I've never managed to get it to work reliably.
<HdkR>
neat
<alyssa>
Supposedly = I think I've seen errata for it ;P
<tlwoerner>
i'm using OpenEmbedded to build... and it is using swiftshader (confirmed by looking at build output, regardless of what the comments say)
<tlwoerner>
the faults are from panfrost. e.g.: [ 334.304904] panfrost ff9a0000.gpu: gpu sched timeout, js=0, config=0x3300, status=0x8, head=0xd609
<tlwoerner>
i'm pretty sure it's panfrost. chromium is using GL, and the GL is supplied by mesa/panfrost on this device. the 2nd pic shows this
<tlwoerner>
the first pic says some (most) of the graphics are hardware accelerated
<alyssa>
Alright, yeah
<alyssa>
I.. don't understand how Chromium is slipping past the blacklist.
<alyssa>
But apparently it is and it is known that Chromium+panfrost is pretty broken right now. So business as usual :|
<alyssa>
(If you want to look into debugging Chromium.. PAN_MESA_DEBUG=trace will be your friend, and grep for `fault_pointer`)
<tlwoerner>
awesome. thanks :-)
<alyssa>
Best of luck
<alyssa>
Don't be surprised when fault_pointer doesn't match up with, well, the fault pointer. But it should get you looking in the right thousand lines :p
<tlwoerner>
fire hose!
<alyssa>
hm?
buzzmarshall has quit [Remote host closed the connection]
Depau has joined #panfrost
adjtm_ has joined #panfrost
adjtm has quit [Ping timeout: 240 seconds]
* alyssa
has some "fun" 64-bit disassembly bugs
buzzmarshall has joined #panfrost
alpernebbi has quit [Quit: alpernebbi]
asmallanimal has quit [Remote host closed the connection]
NeuroScr has joined #panfrost
<Stary>
i had to recompile mesa with the blacklist also excluding slack/vscode because they're basically chrome, being electron and all..
<Stary>
also wanted to compile chromium with v4l2 acceleration enabled but that didnt work out
<urjaman>
if you want to run a specific app with llvmpipe instead of panfrost, you can just set LIBGL_ALWAYS_SOFTWARE=1 in the environment for it