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
BenG83 has joined #panfrost
stikonas has quit [Ping timeout: 260 seconds]
raster has quit [Quit: Gettin' stinky!]
icecream95 has joined #panfrost
davidlt has joined #panfrost
<icecream95> BenG83: I have problems with GIMP and other GTK 2 applications on Wayland as well, which cause XWayland to slow to a crawl (even though most of those GTK 2 applications don't use OpenGL at all)
<icecream95> An apitrace on the GIMP won't help as it seems the problem is to do with XWayland.
<icecream95> A workaround is to use Xephyr. Install it (it will likely be in a package called something like xorg-server-xephyr), then run it with Xephyr :5 -resizeable
<icecream95> In another terminal, start an X window manager, for example Openbox: DISPLAY=:5 openbox
<icecream95> Finally, launch the GIMP (or another GTK 2 application with) DISPLAY=:5 gimp
vstehle has quit [Ping timeout: 258 seconds]
<icecream95> When running GIMP, perf top shows panfrost_access_tiled_image_generic.lto_priv.2709 at over 50%
<icecream95> It actually seems a lot better than it was before (but still not useable), but maybe that's because I revert 4fee7b30 (see https://gitlab.freedesktop.org/mesa/mesa/issues/2108)
_whitelogger has joined #panfrost
_whitelogger has joined #panfrost
vstehle has joined #panfrost
_whitelogger has joined #panfrost
_whitelogger has joined #panfrost
davidlt has quit [Ping timeout: 258 seconds]
_whitelogger has joined #panfrost
<tomeu> alyssa: I think fails are totally fine at this stage, but count of crashes and faults should be brought down so running CI becomes feasible
anarsoul|c has quit [Quit: Connection closed for inactivity]
<icecream95> It seems 4fee7b30 has nothing to do with the GIMP problems.
<icecream95> After going to the bother of restarting Sway with 4fee7b30 re-enabled, I decided to do another comparison of before/after that commit.
<icecream95> Although it helps some of the tests such as buffer a lot, refract goes from around 50 fps to 25 fps (and causes Sway to stutter a lot)
anarsoul|c has joined #panfrost
youcai has joined #panfrost
<youcai> hello, did panfrost in kernel 5.5-rc1 require dts changes from 5.5?
<youcai> I had a config that works on 5.5 but got stuck at udev complete init at boot on 5.5 rc1, if not then it could be a bug
sphalerite has quit [Quit: WeeChat 2.6]
sphalerite has joined #panfrost
nerdboy has joined #panfrost
youcai has quit [Ping timeout: 245 seconds]
youcai has joined #panfrost
stikonas has joined #panfrost
warpme_ has quit [Quit: Connection closed for inactivity]
<BenG83> icecream95: thank you for the info, I'll try that
<BenG83> like I said, gtk application on X are redrawing very slow but are usable, doesn't really matter if they even use GL
fysa has quit [Ping timeout: 260 seconds]
warpme_ has joined #panfrost
adjtm has quit [Ping timeout: 258 seconds]
raster has joined #panfrost
abordado has joined #panfrost
_whitelogger has joined #panfrost
_whitelogger has joined #panfrost
icecream95 has quit [Quit: leaving]
BenG83 has quit []
davidlt has joined #panfrost
davidlt has quit [Read error: Connection reset by peer]
BenG83 has joined #panfrost
<BenG83> mmh I see a lot of
<BenG83> [54191.031733] Purging 8388608 bytes
<BenG83> [54191.157452] Purging 8388608 bytes
<BenG83> [54191.194376] Purging 1646592 bytes
<BenG83> which seems to come from
<BenG83> panfrost_gem_shrinker.c: pr_info_ratelimited("Purging %lu bytes\n", freed << PAGE_SHIFT);
<BenG83> what could be the cause for that?
<urjaman> normal operation
<urjaman> (the cause being that that print is too noisy... i think there is a patch for it already too)
<BenG83> ok
<urjaman> not sure if picked up anywhere
youcai has quit [Quit: Leaving.]
<alyssa> That disables tiling so panfrost_access_tiled_image_generic will never be called
<alyssa> So it'll reduce CPU usage, in exchange for slower GPU performance
<alyssa> Depending on workload that may or may not be a good idea ... we may end up specifically blacklisting things like glamor for tiling just based on de facto performance
raster has quit [Quit: Gettin' stinky!]
nerdboy has quit [Ping timeout: 258 seconds]
Depau has quit [Quit: ZNC 1.7.5 - https://znc.in]
Depau has joined #panfrost
Depau has quit [Quit: ZNC 1.7.5 - https://znc.in]
Depau has joined #panfrost
adjtm has joined #panfrost
<urjaman> "Depending on workload that may or may not be a good idea" that seems like a theme with most decision the graphics driver needs to do :P
nerdboy has joined #panfrost
nerdboy has quit [Ping timeout: 260 seconds]
nerdboy has joined #panfrost
<BenG83> mmh, I'm rebuilding cairo to see if I can fix my problem with gtk applications
<BenG83> there are gl/egl/glesv2 and gallium backends for cairo
<BenG83> which one would I need for wayland? egl?
<urjaman> i think the fundamental thing is that gtk2 doesnt do wayland
<urjaman> (gtk3 does)
<BenG83> yeah but something triggers the panfrost crashes
<BenG83> so I'm poking at the gtk libraries atm
<BenG83> currently gtk apps run very slow on my X and really bad on Wayland
<BenG83> everything else but gtk applications seems ok
<urjaman> what mesa are you running? i mean, the one i built yesterday (just to see if something changed) runs my xfce (incl. gimp) fine
<urjaman> (on my C201 so T760 but anyways...)
<urjaman> (in Xorg)
<BenG83> OpenGL version string: 2.1 Mesa 20.0.0-devel (git-2c8742ed85)
<BenG83> from a couple days ago
<BenG83> on 5.4.0
<BenG83> on manjaro-arm
<BenG83> it's maybe just some libray problem
<BenG83> *library
tomeu has quit [Quit: Ping timeout (120 seconds)]
tomeu has joined #panfrost
Wallbraker has quit [Ping timeout: 268 seconds]
BenG83 has quit []
BenG83 has joined #panfrost
abordado_ has joined #panfrost
abordado_ has quit [Client Quit]
abordado has quit [Quit: Leaving]