ChanServ changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - Logs https://oftc.irclog.whitequark.org/panfrost - <macc24> i have been here before it was popular
simon-perretta-img has quit [Ping timeout: 480 seconds]
<icecream95>
Ouch, it seems that re-mapping BOs all the time is more expensive than expected... about half the time in memcpy is just handling page faults
* icecream95
tries MAP_POPULATE
atler is now known as Guest192
atler has joined #panfrost
Guest192 has quit [Ping timeout: 480 seconds]
<icecream95>
The unmapping is also part of what makes CPU overhead in Sway so high IIRC
hexdump0815 has joined #panfrost
hexdump01 has quit [Ping timeout: 480 seconds]
<macc24>
i have the dumbest question ever
<macc24>
would it be 3x faster to render everything monochromatic?
Daanct12 has joined #panfrost
Daanct12 has quit [Remote host closed the connection]
Daanct12 has joined #panfrost
Daanct12 has quit [Quit: Leaving]
nlhowell has joined #panfrost
<icecream95>
macc24: If you want to do something like that, why not just use subsampled YUV?
<icecream95>
But that does sound like a stupid enough idea that I might try implementing it
jenneron_ has quit [Remote host closed the connection]
jenneron_ has joined #panfrost
anarsoul|2 has joined #panfrost
<icecream95>
I wonder if my nodearray datastructure would also be useful for tracking BOs...
anarsoul has quit [Read error: Connection reset by peer]
<icecream95>
16777215 BOs ought to be enough for anyone!
tlwoerner has quit [Remote host closed the connection]
tlwoerner has joined #panfrost
nlhowell is now known as Guest219
nlhowell has joined #panfrost
Guest219 has quit [Ping timeout: 480 seconds]
<icecream95>
It might also be better to flip things around, and have a batch_id field in the BO
rasterman has joined #panfrost
nlhowell is now known as Guest222
nlhowell has joined #panfrost
Guest222 has quit [Ping timeout: 480 seconds]
q4a has quit [Remote host closed the connection]
alyssa has quit [Remote host closed the connection]
alyssa has joined #panfrost
nlhowell is now known as Guest226
nlhowell has joined #panfrost
Guest226 has quit [Ping timeout: 480 seconds]
simon-perretta-img has joined #panfrost
rkanwal has joined #panfrost
jenneron__ has joined #panfrost
jenneron_ has quit [Ping timeout: 480 seconds]
icecream95 has quit [Ping timeout: 480 seconds]
jambalaya has quit [Remote host closed the connection]
jambalaya has joined #panfrost
robmur01 has quit [Quit: Leaving]
Danct12 has quit [Quit: Quitting]
nlhowell has quit [Ping timeout: 480 seconds]
nlhowell has joined #panfrost
`join_subline has quit [Remote host closed the connection]
anarsoul has quit [Read error: Connection reset by peer]
q4a has joined #panfrost
nlhowell has quit [Ping timeout: 480 seconds]
anarsoul has joined #panfrost
anarsoul|2 has quit [Read error: Connection reset by peer]
robmur01 has quit [Quit: Leaving]
tlwoerner has quit [Quit: Leaving]
tlwoerner has joined #panfrost
villalvilla has joined #panfrost
greenjustin has joined #panfrost
<villalvilla>
hello
<macc24>
o/
<villalvilla>
I am new in panfrost world, but have quite experience with SoCs, debian and mali support
<villalvilla>
I am trying to decide which is my best option for a Rock Pi 4 (RK3399 chip) for desktop configuration
pi has left #panfrost [#panfrost]
Syonyk has joined #panfrost
Syonyk has quit []
<villalvilla>
my main objective is having browser with hardware acceleration (for the moment). I am trying to organize my thoughts and see what is best: wayland or X11, with panfrost-mesa drivers as a basis. I am working on this for a debian-based distro for education
<macc24>
avoid xorg like a plague
<villalvilla>
got it macc24!
<villalvilla>
I have got no experience with wayland, I have read in many places that the best option is wayfire + mesa 22. Any suggestion on where to start from?
<macc24>
uhh
<villalvilla>
I have a mainline kernel with a vanilla debian's debootstraped version in 15.18rc7 working like a charm, but I would like to know from your experience which are my best options to go with the X server
<CounterPillow>
MOZ_ENAYBLE_WAYLAND=1 firefox and then switching on webrender in about:config should just work
<alyssa>
aha, with the perf counters I have learned this patch to improve ILP does in fact reduce stalls by less than 1% :v
<HdkR>
Yeck yea, every 1% improvement counts :D
<alyssa>
does it?
<villalvilla>
roger that, CounterPillow!
<macc24>
alyssa: 1% by itself doesn't say much
<macc24>
BUT
<macc24>
1% + 1% + 1% + 1% + !%
<macc24>
it's the little things :D
<villalvilla>
any easy-way script to compile mesalib for wayland? (sorry for the newbie questions '-)
<macc24>
you don't need to?
<macc24>
unless you use debian oldstable it's proabbly fine
<villalvilla>
bullseye
`join_subline has quit [Remote host closed the connection]
<villalvilla>
mesa is in version 20.3.5 in bullseye. As far as I know, to unleash all the power of the mali G52 I need mesa 22+
<villalvilla>
(In order to use bifrost, I mean)
<alyssa>
++
robmur01 has joined #panfrost
`join_subline has joined #panfrost
<villalvilla>
I have just upgraded to sid, where mesa is in version 22.0.3
<villalvilla>
thanks
villalvilla has quit [Remote host closed the connection]
icecream95 has joined #panfrost
`join_subline has quit [Remote host closed the connection]
`join_subline has joined #panfrost
camus has joined #panfrost
Danct12 has joined #panfrost
rkanwal has quit [Read error: No route to host]
rkanwal has joined #panfrost
camus1 has quit [Ping timeout: 480 seconds]
`join_subline has quit [Remote host closed the connection]
rkanwal has quit [Read error: Connection reset by peer]
<icecream95>
On the topic of perf counters, we need a way to collect information such as whether we managed to skip a depth/stencil reload better than printf
<icecream95>
Maybe that's perfetto-shaped, but ugh what a lot of code to compile
<icecream95>
But now that my stencil-track branch has depth buffer support, maybe I should also make it skip the resolve when applicable
<alyssa>
\s/
<cphealy>
icecream95: would skipping the resolve in this case save DRAM bandwidth? How about other savings inside the GPU?
<icecream95>
cphealy: Yes, though only by about 20 bytes per tile because of AFBC, so it isn't that significant