Elpaulo has quit [Read error: Connection reset by peer]
Elpaulo has joined #panfrost
megi has quit [Ping timeout: 265 seconds]
vstehle has joined #panfrost
<bbrezillon>
warpme_: I really need a way to reproduce it myself
davidlt has joined #panfrost
<warpme_>
bbrezillon: what about log from apitrace which you can replay in your environment?
<bbrezillon>
warpme_: would work too
<warpme_>
ok - then I'll get apitrace and provide you url to download (it will have also demonstration of another issue: flickering luma/chroma at video playback). what mesa code I need to use?
belgin has joined #panfrost
<bbrezillon>
warpme_: master should be fine
<bbrezillon>
or the last version that was working for you
<bbrezillon>
it shouldn't matter to capture the trace
<warpme_>
bbrezillon: pls find at http://warped.inet2.org/myth-panfrost-trace.xz apitrace from current master of mesa on s912. Trace also shows flickering picture at playback phase. Let me know is this file ok for you to reproduce issues!
megi has joined #panfrost
<bbrezillon>
warpme_: thx
<tomeu>
narmstrong: btw, your runner seems to be down?
raster has joined #panfrost
<alyssa>
bbrezillon: Fixing 78 tests?!
<bbrezillon>
alyssa: yes
<bbrezillon>
warpme_: duh, it works perfectly fine here
<warpme_>
robmur01: so difference between bbrezillon and mine is 5.3-rc8 vs 5.3; bbrezillon can You test with mainline 5.3?
<narmstrong>
warpme_: nop, the main different is t820 vs t860, t820 hasn't been heavily tested since you can only find it on Amlogic S912 and on robmur01's FPGA
<bbrezillon>
warpme_: I can
Depau has joined #panfrost
<warpme_>
aha - so maybe indeed my issue is because current mesa not loves t820?
<narmstrong>
warpme_: nothing has been done/tested for t820 in fact, we use the fact t820 is close to t860
<warpme_>
so i must say support for t820 is pretty good! only issue i have is flickering picture at gles surface rendered video playback....
<tomeu>
narmstrong: I don't see any shared runners with the lava-meson-gxm-khadas-vim2 tag, maybe it's a project-specific runner?
<warpme_>
want->what
<tomeu>
narmstrong: if so, then you need to ask daniels for a token so it's a global shared runner (I think)
<narmstrong>
tomeu: it's shared, but yeah maybe I need a global token
<warpme_>
bbrezillon: you are on t860 right? if so then i think issue is that original commit in question is ok for t860 but not for t820?
<bbrezillon>
warpme_: which would be weird, especially since it's supposed to do the exact same thing (after the missing NULL assignment I provided)
<bbrezillon>
I don't really see why one would work and not the other
<bbrezillon>
I must be missing something obvious
<warpme_>
so lets summarise: we have the same kernel; differences are: hw and kernel patchset. Just for sanity check: may you apply neil's patchset and test?
<tomeu>
t860 and t820 probably have quite different cmd streams
<tomeu>
iirc, t820 is SFBD
<robmur01>
tomeu: pretty sure 820 can do MFBD - it's 720 that's the real weirdo
<warpme_>
ough: so if we have so nicely working t820 - then getting t720 should be easy? (sill have unusable h6 with t720 :-(