alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard + Bifrost + Valhall - Logs https://oftc.irclog.whitequark.org/panfrost
megi has quit [Quit: WeeChat 4.2.1]
megi has joined #panfrost
tlwoerner has joined #panfrost
tlwoerner_ has quit [Ping timeout: 480 seconds]
tlwoerner_ has joined #panfrost
tlwoerner has quit [Ping timeout: 480 seconds]
tlwoerner has joined #panfrost
tlwoerner_ has quit [Ping timeout: 480 seconds]
tlwoerner has quit [Remote host closed the connection]
Daanct12 has joined #panfrost
Leopold_ has quit []
Leopold_ has joined #panfrost
KREYREN__ has joined #panfrost
KREYREN_ has quit [Remote host closed the connection]
KREYREN__ has quit [Remote host closed the connection]
Daanct12 has quit [Ping timeout: 480 seconds]
rasterman has joined #panfrost
KetilJohnsen has joined #panfrost
KetilJ has joined #panfrost
KetilJohnsen has quit [Ping timeout: 480 seconds]
Daanct12 has joined #panfrost
warpme has joined #panfrost
warpme has quit []
warpme has joined #panfrost
warpme has quit []
warpme has joined #panfrost
warpme has quit []
simon-perretta-img has quit [Ping timeout: 480 seconds]
simon-perretta-img has joined #panfrost
warpme has joined #panfrost
simon-perretta-img has quit [Ping timeout: 480 seconds]
simon-perretta-img has joined #panfrost
chewitt has joined #panfrost
Daanct12 has quit [Quit: WeeChat 4.2.1]
KetilJ has quit []
simon-perretta-img has quit [Ping timeout: 480 seconds]
simon-perretta-img has joined #panfrost
rasterman has quit [Quit: Gettin' stinky!]
simon-perretta-img has quit [Ping timeout: 480 seconds]
simon-perretta-img has joined #panfrost
warpme has quit []
cphealy has quit [Quit: Leaving]
kazimk has joined #panfrost
cphealy has joined #panfrost
kazimk has quit []
warpme has joined #panfrost
warpme has quit []
simon-perretta-img has quit [Ping timeout: 480 seconds]
simon-perretta-img has joined #panfrost
warpme has joined #panfrost
robmur01 has quit [Read error: Connection reset by peer]
robmur01 has joined #panfrost
warpme has quit []
rasterman has joined #panfrost
warpme has joined #panfrost
warpme has quit []
warpme has joined #panfrost
Leopold_ has quit [Remote host closed the connection]
warpme has quit []
Leopold_ has joined #panfrost
rasterman has quit [Quit: Gettin' stinky!]
warpme has joined #panfrost
warpme has quit []
xdarklight has quit [Quit: ZNC - https://znc.in]
xdarklight has joined #panfrost
<ndufresne> mntirc: also getting a correct driver was the milestone achieve, performance is being looked at now as a second step, would be nice to have some metrics to set the expectation, did you mean A311D or A311D2 ? (cause the second have G52 MP8, which might outperform a G610 MP4)
<mntirc> ndufresne: ah i see, that's what i thought! i ran glmark2-es2-wayland a few hours ago and i get a slightly worse score than on A311D (not D2)
<mntirc> > Mali G52 MP4(6EE)
<ndufresne> with default resolution, you got something under 2K ?
<mntirc> ndufresne: yes
<ndufresne> that's what a user reported recently, and what will be looked at
<mntirc> ndufresne: on a311d i get 1669 and i got like 1200 on rk3588
<mntirc> (will run it again now)
<ndufresne> interesting, my users ws reporting 1900
<ndufresne> but you know, not simple, have your force the freq to its highest ? and CPU governor to performance ?
<ndufresne> letting governor run during benchmark introduce randomness
<ndufresne> but also lower scores
<ndufresne> here I was confused for a moment, on an older valhal (G57) I had much better scores, like 2200, but then notice it has a higher core count ...
<mntirc> ndufresne: i wasn't aware that i can set the gpu freq?
<ndufresne> echo userspace > /sys/class/devfreq/13000000.gpu/governor
<mntirc> ohh
<ndufresne> then you can pick a freq from /sys/class/devfreq/13000000.gpu/available_frequencies
<ndufresne> the simple_ondemand governor is just like his name says, simple
<ndufresne> but the GPU governor will also impact the results, driving GL is always a bit CPU intensive
<mntirc> got a score of 1491 now
<ndufresne> less ?
<mntirc> probably because now i'm using 2.4ghz on the big cores (with new opp tables)
<mntirc> no, i was still running the test before changing the gpu clock, will try that now
<mntirc> but getting a higher score than on 2.2ghz earlier today
<mntirc> ndufresne: i don't have a devfreq driver for the gpu it seems
<mntirc> ndufresne: i.e. /sys/class/devfreq/ doesn't contain 13000000.gpu
<mntirc> ah, it's fb000000.gpu
<mntirc> sorry
<mntirc> ok, testing with 1ghz now
<ndufresne> yeah, this is soc specific, can't test atm on my rk3588, have to rebuild mesa
<mntirc> ndufresne: which soc/gpu are you testing with? imx9?
<ndufresne> I suspect G57 MP5 (880MHz) should outperform G610 MP4 at 1GHz, but I could be wrong
<ndufresne> MT8195, from mediatek
<ndufresne> says MC5, not sure what is the difference betweem MC and MP
<mntirc> ah
<ndufresne> this board will be nearly identical, https://radxa.com/products/nio/12l/
<mntirc> interesting
<mntirc> ok now i'm getting a score of 2102!
<ndufresne> hmm, ok, with 880MHz, and perf gov on the CPU, the G57 here did 4041
<mntirc> ok, did not try perf governor on the cpu yet, but it probably won't get near that score :D
<ndufresne> I'm under the impression that it should in theory be relatively close, unless it has a limitation somewhere else, like RAM
<ndufresne> ram frequence is also something to take into account seriously
<mntirc> makes sense
<ndufresne> Which board to you use ?
Amber_Harmonia_ has joined #panfrost
Amber_Harmonia has quit [Read error: Connection reset by peer]
<mntirc> mntirc: i use firefly icore 3588q with my own adapter for mnt reform
<mntirc> ndufresne: sorry, i meant to write to you
<mntirc> ndufresne: btw even at 1ghz gpu and 2.4ghz cpu, i only get 15fps in firefox on webgl blobs, and google maps is very slow, whereas in chromium both are smooth
<urja> My T760 / RK3288 is running webgl blobs in firefox at 48 ... 60 fps (default settings of 10 blobs, 24^3 resolution)
<urja> yep that's very different (but also, slower), the point being, yep firefox shouldnt be that slow
<CounterPillow> so have you checked yet whether it's actually hitting the GPU or whether Firefox is just blocklisting it for webgl use
<daniels> I think that was verified already
<daniels> nothing structural I’m sure, just that no-one really looked at Firefox yet
<urja> the only things i can think of are this is under wayland/kde and i have layers.acceleration.force-enabled and webgl.force-enabled true (but i'm not sure if those matter or are ancient)