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
<icecream95> alyssa: Green, usually
<icecream95> It only happens when s3tc is enabled
<icecream95> It happens with LIBGLALWAYSSOFTWARE=1, so #notourbug
<icecream95> LIBGL_ALWAYS_SOFTWARE*
<alyssa> icecream95: I love #notourbug =D
<alyssa> robclark: I do need to play more stk :p
<robclark> #notourbug or #wehavesamebugasswrast? (I've hit the latter case a few times)
<alyssa> llvmpipe is GL4.5 conformant, how much do we care? :p
<robclark> alyssa: yes you do.. playing^Dtesting is one of the fringe benefits of working on a gfx driver :-P
<alyssa> I don't even like video games!
<robclark> I'm not a big first-person-shooter type (well, I'm probably more of a pacifist).. but big fan of the the mariokart genre ;-)
<alyssa> Definitely not an FPS type myself
<HdkR> I like platformers :>
<alyssa> ~Might be a Quaker thing~
<robclark> somehow first-person-shooter == fps annoys me about as much as digital-rights-managment == drm
<alyssa> Yeah....
<alyssa> robclark: Petition to rename libwidevine to libdrm just to annoy us linux people ;P
<robclark> can I just have nothing to do with widevine instead?
<alyssa> good choice
<icecream95> What annoys me is that swaywm is on page 3 of searching for "sway" on DDG. No, I do not want to "Create visually striking newsletters", Micro$oft
<alyssa> If we rename to Panfro$t, will we get more customers? Does it work like that?
* macc24 laughs in having a dedicated ai that is specifically suited to enclose me in an echo chamber and talking to its creator
<icecream95> alyssa: We have Microsoft code in Panfrost now, right? Micro$oft® Panfro$t for Workgroups 3.11
<alyssa> icecream95: I mean. Technically in NIR, yes.
<macc24> alyssa: wait isn't there an internal panfrost's IR?
<icecream95> macc24:
<macc24> icecream95:
<icecream95> 'mir' and 'bir' for Midgard and Bifrost
<alyssa> macc24:
<icecream95> alyssa:
<macc24> alyssa:
<icecream95> icecream95:
<HdkR> mir,bir,vir
<HdkR> :)
<macc24> HdkR: valhall on panfrost won't happen without a sane dev board running linux
<alyssa> 'bir' is correctly pronounced 'brrrrrgh' 🇨🇦
<macc24> alyssa: bir goes brrrrrgh
<alyssa> icecream95: Passed: 1890/1900 (99.5%)
<macc24> HdkR: and you know who will happily put linux on every chromebook that he touches :D
<alyssa> 10 regressions out of 1900, that's basically correct, let's ship it? ;P
<icecream95> alyssa: It's too hot down here for me to be saying brrr
<alyssa> icecream95: Maybe come back to Bifrost in 6 months?
<icecream95> :)
<macc24> icecream95: at least your part of world isn't on fire
<alyssa> icecream95: https://rosenzweig.io/uh.txt Do you see anything obviously wrong here? Am I being really dense? Is this another hw bug? :|
<HdkR> mediump output losing precision?
<HdkR> while constants are 32bit float that definitely don't fit?
<alyssa> actually it looks like the whole draw might be missing
<alyssa> so maybe the RSD is busted
<alyssa> wait_6/wait_7 stuff
<alyssa> apparently those flags are new to v7? will have to fix for v6 then... wheee.
<alyssa> something about the first clause being ATEST seems to throw it off
<alyssa> WTF!
<alyssa> If I don't have an ATEST instruction (I just do a single clause with BLEND), it's perfectly happy.
stikonas has quit [Remote host closed the connection]
<icecream95> alyssa: There isn't any clause with next_atest
<alyssa> icecream95: That ought to be ok, we start clauses with message passing instructions all the time (usually loads of various types)
<alyssa> and in the BLEND-only one, there's nothing with next_blend
<alyssa> which implies something fundamentally odd about ATEST
mixfix41 has joined #panfrost
dstzd has quit [Quit: ZNC - https://znc.in]
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
dstzd has joined #panfrost
jernej has joined #panfrost
<alyssa> Found an ancient blob trace I have of the blob wriing a constant
<alyssa> It does that pattern (ATEST in the first clause) so maybe some magic bit different
atler has quit [Killed (verne.freenode.net (Nickname regained by services))]
atler has joined #panfrost
vstehle has quit [Ping timeout: 240 seconds]
<alyssa> icecream95: bbrezillon: rebased vulkan w.r.t 5d3fdbc52bf8bee6f0acfd55c87b8d280908b559
raster has quit [Quit: Gettin' stinky!]
xdarklight has quit [Ping timeout: 256 seconds]
xdarklight has joined #panfrost
tchebb has quit [Ping timeout: 256 seconds]
tchebb has joined #panfrost
_whitelogger has joined #panfrost
archetech has quit [Quit: Konversation terminated!]
_whitelogger has joined #panfrost
_whitelogger has joined #panfrost
vstehle has joined #panfrost
ezequielg has quit [Read error: Connection reset by peer]
ezequielg has joined #panfrost
davidlt has joined #panfrost
<icecream95> alyssa: You'll have to make Bifrost about twice as fast for this to avoid GPU timeouts: https://0x0.st/-ioD.png
felipealmeida has quit [Ping timeout: 256 seconds]
felipealmeida has joined #panfrost
davidlt has quit [Ping timeout: 256 seconds]
_whitelogger has joined #panfrost
alpernebbi has joined #panfrost
warpme_ has joined #panfrost
tomboy64 has quit [Ping timeout: 240 seconds]
tomboy64 has joined #panfrost
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
jernej has joined #panfrost
jernej has quit [Read error: Connection reset by peer]
jernej has joined #panfrost
jernej_ has joined #panfrost
dstzd has quit [Quit: ZNC - https://znc.in]
dstzd has joined #panfrost
dstzd has quit [Client Quit]
dstzd has joined #panfrost
jernej- has joined #panfrost
jernej has quit [Ping timeout: 272 seconds]
jernej- is now known as jernej
dstzd has quit [Client Quit]
jernej_ has quit [Ping timeout: 246 seconds]
dstzd has joined #panfrost
dstzd has quit [Client Quit]
dstzd has joined #panfrost
dstzd has quit [Read error: Connection reset by peer]
dstzd_ has joined #panfrost
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
dstzd_ is now known as dstzd
dstzd_ has joined #panfrost
jernej has joined #panfrost
dstzd has quit [Ping timeout: 264 seconds]
dstzd_ is now known as dstzd
davidlt has joined #panfrost
raster has joined #panfrost
stikonas has joined #panfrost
orkid has joined #panfrost
<alyssa> icecream95: You don't like rendering the Mande se?
<orkid> Hi, is this an appropriate place to ask about problems with mesa/panfrost on odroid-n2? I installed debian testing, and using linux 5.10.6, and mesa compiled from git, i am having errors and rendering problems in weston. I'm not sure if this is the right place to ask about this.
<alyssa> weston should work ok on n2, what's up?
<macc24> orkid: does sway work fine?
<orkid> macc24: i don't know, i haven't checked.
<orkid> Unhandled Page fault in AS1 at VA 0x0 in dmesg
<orkid> gpu sched timeouts
<orkid> js fault, js=1
<orkid> in weston, portions of the background flicker black, etc.
<orkid> i can get rid of the flickering by moving a window around the 'clean up' those areas.
<orkid> e.g. i opened a firefox window (MOZ_ENABLE_WAYLAND=1), it's from the latest firefox .deb from untsable, moved it around, and a portion of a terminal in the background flickers white, as i type in irssi in another terminal now.
<orkid> :S
<orkid> i just moved the mouse, and got more unhandled page faults show up in the glitching terminal windows with dmesg -w running within.
<macc24> i can't reproduce this on G72
<orkid> what distro, kernel, etc? maybe i can install something similar and check on the n2.
<macc24> kernel is 5.10.4 with custom patches to support hardware
<macc24> panfrost module is largely unpatches
<macc24> distro is cadmium and it won't run on odroid n2
<orkid> any ideas why this might be happening? the kernel is not totally vanilla perhaps. i installed debian using a bullseye netinst img from tobbter (he works at Hardkernel?). I'm not sure what patches it might have had. Could any such patches be the cause of this?
<macc24> when did you build mesa?
<orkid> .. not sure what else to try. I figured this was quite a minimal install (tobetter netinst, mesa from git, weston, and firefox (happens with both firefox-esr from testing, and firefox from unstable)
<orkid> a few hours ago.
<orkid> git pull says 'up to date.'
<macc24> then i don't know
orkid has quit [Disconnected by services]
orkid has joined #panfrost
<orkid> it seems to have been a kernel issue. installed a different linux-image (with 'panfrost' in the name), and no problems yet.
* orkid knocks on wood
<chewitt> orkid 5.10 is missing patches
<chewitt> the kernels that HK are releasing now have picked the required bits from my tree
<chewitt> the bits are mostly backports from 5.11
<orkid> chewitt: you mean the kernels in the linuxfactory repositories?
<chewitt> I don't use or track the content in the HK repo's .. but from forum posts @tobetter picked everything from my branches
<chewitt> HK like most vendors has so much stuff spread all over the place
<orkid> ok. yeah, somewhat of a hair pulling experience. :) btw, i just ran a git clone of your linux.git amlogic-5.10.y branch. is that the best one to use atm, or is 5.11.y better/ok to use?
<chewitt> I'm using 5.11.y without issues
<chewitt> and by now I've completely forgotten what state I left the 5.10.y branch in :)
<chewitt> beware that I rebase frequently.. so branches are a moving target
<orkid> ok. thanks! i'll have to reclone then, since i just did --single-branch.
<orkid> if i git clone -b amlogic-5.11.y --single-branch, will that cause problems as you rebase, or can i just git pull and all will be swell? :)
<orkid> .. i don't know much about version control systems, i guess i'll have to read about it a bit. sorry for the off topic chat, if it bothered anyone.
alpernebbi has quit [Ping timeout: 256 seconds]
alpernebbi has joined #panfrost
robmur01 has quit [Ping timeout: 256 seconds]
raster has quit [Quit: Gettin' stinky!]
<chewitt> git remote add chewitt https://github.com/chewitt/linux.git
<chewitt> git fetch chewitt amlogic-5.11.y
<chewitt> git reset --hard chewitt/amlogic-5.11.y <= if you didn't make any local changes
<chewitt> git rebase chewitt/amlogic-5.11.y <= if you did make local changes
<orkid> thanks!
robmur01 has joined #panfrost
raster has joined #panfrost
robmur01 has quit [Ping timeout: 240 seconds]
Elpaulo has joined #panfrost
alpernebbi has quit [Quit: alpernebbi]
davidlt has quit [Ping timeout: 256 seconds]
<icecream95> alyssa: Bifrost other than G71 has double the tilebuffer size, so shouldn't pan_internal_cbuf_size be adjusted for that?
<alyssa> icecream95: Probably but my head's in Apple land right now, will look tmrw
<macc24> alyssa: :o without linux on it?
raster has quit [Quit: Gettin' stinky!]
<alyssa> macc24: Yeah, I've learned more about the macOS graphics stack in the past few weeks than I needed to know :p
popolon has joined #panfrost
orkid has quit [Quit: leaving]
<macc24> alyssa: what's inside m1 gpu is it similar to some powervr chips?
tomboy64 has quit [*.net *.split]
<HdkR> You would need to RE some very new PowerVR chips to find out :P
<macc24> HdkR: i was close to checking out if powervr chip on mt8173 can be reversed
<HdkR> At least that one isn't SGX
<HdkR> also don't let the internet's overwhelming noise about Apple's design being PowerVR based just because of the licensing of PVRTC from IMGTec...
<macc24> well i don't plan to buy m1 mac anyway
<HdkR> Not until it runs upstream linux anyway :P
raster has joined #panfrost
<macc24> icecream95: wiykd ability to trace blob on duet be helpful to you?
<macc24> would*
<robclark> tbh, pvr could have been a lot more sane by throwing away img's fw.. which I suspect apple would have done
chrisf has quit [Quit: ZNC - https://znc.in]
chrisf has joined #panfrost
raster has quit [Ping timeout: 256 seconds]
popolon has quit [Ping timeout: 256 seconds]
raster has joined #panfrost
popolon has joined #panfrost
chrisf has quit [Quit: ZNC - https://znc.in]
chrisf has joined #panfrost
raster has quit [Quit: Gettin' stinky!]