alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - Logs https://freenode.irclog.whitequark.org/panfrost - <daniels> avoiding X is a huge feature
NeuroScr has quit [Quit: NeuroScr]
<alyssa> Woo! Passing CI! Onwards!
<alyssa> Wel, not right now, it's late. but hey
adjtm has joined #panfrost
ata2001 has quit [Ping timeout: 276 seconds]
_whitelogger has joined #panfrost
_whitelogger has joined #panfrost
davidlt has joined #panfrost
pH5 has joined #panfrost
megi has quit [Ping timeout: 240 seconds]
davidlt_ has joined #panfrost
davidlt has quit [Remote host closed the connection]
davidlt_ is now known as davidlt
davidlt has quit [Ping timeout: 265 seconds]
_whitelogger has joined #panfrost
<HdkR> Anyone doing Panfrost + Vulkan things atm that I should know about?
yann|work has joined #panfrost
<HdkR> Thinking about doing work on it during the flight to xdc for fun
megi has joined #panfrost
stikonas has joined #panfrost
davidlt has joined #panfrost
stikonas has quit [Remote host closed the connection]
AreaScout_ has joined #panfrost
* alyssa swims in git rebase -i
<alyssa> Down to 9 fixups though
<alyssa> ......or more since .git just got corrupted :V
<alyssa> Down to 4 regardless
<alyssa> fixup.
<alyssa> Onwards!!!
davidlt has quit [Remote host closed the connection]
davidlt has joined #panfrost
<alyssa> shader-db results are starting to look good, but why stop here? :D
<alyssa> ---Wait, still have deqp regressions to sort
* alyssa deflates slightly
<alyssa> Once I finish up this stuff (scalars), next will be finishing writeout opts
<alyssa> Dang. I can't believe my little scheduler (friendship is midgard!) managed to fit 6 instructions in a bundle, filling up every single slot :p
<alyssa> (deqp regressions sorted)
<alyssa> See-the-future writeout opt implemented - shader-db is much happier now
<alyssa> (I.e. all shaders are helped now, afaict)
<alyssa> wat
<alyssa> Why would -bterrain perf go down if it's shaders are clearly helped...?
<davidlt> alyssa, will there be any status updates, i.e. a blog post?
<alyssa> davidlt: Not sure
<alyssa> tomeu: and I are doing a lightning talk on Panfrost at XDC next week
<alyssa> Does that count? :)
<davidlt> alyssa, yes :)
<alyssa> ---also what do you mean XDC starts in 3 days fdjklsjhgfreao
<davidlt> So what are you working on recently? I guess, optimizations now?
<davidlt> I think, I keep hearing that GNOME and even KDE potentially works
<alyssa> Out-of-order scheduling for midgard
<alyssa> And yeah, GNOME works
<alyssa> KDE also works but I haven't tried that myself, ask shadeslayer
<MoeIcenowy> ah recently I also debugged a lot of GNOME on Lima
<davidlt> I am rocking GNOME as KDE doesn't feel right for me
<alyssa> Okay, CI is green with the new writeout stuff... progress is moving well, hah!
<alyssa> Next up, constant combining
<alyssa> This should be the final regression compared to the in-order scheduling
<alyssa> So once this is in, I'll do some final shader-db work and send off the series
<urjaman> progress is progressing
pH5 has quit [Ping timeout: 240 seconds]
pH5 has joined #panfrost
<alyssa> Yup!
<alyssa> Oh, I also need to do the vlut/vadd prioritization thing which might be a small boost
<alyssa> Constant combining allows for another modest win:
<alyssa> total bundles in shared programs: 1819 -> 1802 (-0.93%)
<alyssa> bundles in affected programs: 361 -> 344 (-4.71%)
<alyssa> Guess I might as well fix r26 masking while I'm at it
<alyssa> Also while I'm here, there are some codegen opts that I can do now that didn't make sense before
<alyssa> e.g. inverting conditions
jernej_ has joined #panfrost
jernej has quit [Read error: Connection reset by peer]
jernej_ is now known as jernej
fysa has joined #panfrost
fysa has quit [Read error: Connection timed out]
davidlt has quit [Ping timeout: 240 seconds]
fysa has joined #panfrost
stikonas has joined #panfrost
fysa has quit [Ping timeout: 245 seconds]
fysa has joined #panfrost
popolon has joined #panfrost
<popolon> hi.
<popolon> It seems that changes of this week resolved most faces sorting on RK3288 (extremtuxracer, warzone2100, OpenBVE)
<popolon> I was able to launch last version of flightgear flight simulator, after tuning to lower parameters configuration on an x86 computer and copying to my arm laptop
<popolon> it works at an usable speed, but crash after few seconds to minute due to 4GB RAM only on my computer.
<popolon> and there is still a face priority bug only visible after taking off
<popolon> popolon.org/depots/ArchLinuxARM/mesa-git.panfrost_lima/panfrost_medias/fgfs.20190927.jpg
<popolon> some ground faces go in front of cockpit.
<popolon> In case of, I write a text log of my tests here,I try to update evert time I notice some changes: http://popolon.org/depots/ArchLinuxARM/mesa-git.panfrost_lima/__Working_applications_panfrost.txt
davidlt has joined #panfrost
megi has quit [Quit: WeeChat 2.6]
davidlt_ has joined #panfrost
davidlt has quit [Ping timeout: 240 seconds]
pH5 has quit [Quit: -_-]
marcodiego has joined #panfrost
popolon has quit [Quit: WeeChat 2.6]
davidlt_ has quit [Ping timeout: 240 seconds]
stikonas has quit [Remote host closed the connection]
NeuroScr has joined #panfrost
stikonas has joined #panfrost