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
rasterman has quit [Quit: Gettin' stinky!]
mwebb_ has joined #panfrost
mwebb has quit [Ping timeout: 480 seconds]
macc241 has joined #panfrost
macc24 has quit [Ping timeout: 480 seconds]
macc241 has quit []
macc24 has joined #panfrost
kenzie has quit [Quit: The Lounge - https://thelounge.chat]
kenzie has joined #panfrost
icecream95[m] is now known as icecream95
mwebb has joined #panfrost
mwebb_ has quit [Ping timeout: 480 seconds]
davidlt has joined #panfrost
atler is now known as Guest3260
atler has joined #panfrost
Guest3260 has quit [Ping timeout: 480 seconds]
mwebb_ has joined #panfrost
camus has joined #panfrost
mwebb has quit [Ping timeout: 480 seconds]
camus1 has quit [Remote host closed the connection]
davidlt has quit [Ping timeout: 480 seconds]
mwebb has joined #panfrost
mwebb_ has quit [Ping timeout: 480 seconds]
davidlt has joined #panfrost
rasterman has joined #panfrost
<icecream95> Is there any faster way to get data out of BOs on the CPU side than incredibly slow WC reads?
<icecream95> Probably the only solution is to change the kernel to allow mapping BOs without setting map_wc
<bbrezillon> icecream95: I was considering allowing cached mappings too
warpme_ has joined #panfrost
mwebb has quit [Ping timeout: 480 seconds]
camus1 has joined #panfrost
camus has quit [Remote host closed the connection]
<robmur01> IIRC the main issue with that is that you end up needing userspace to honour the rules of the kernel DMA API about when it can and can't access the buffer, and that's an even sketchier assumption than trusting kernel drivers to get it right :)
<robmur01> alternatively just get a VIM3 (or, in principle, a Juno) and enjoy a cache-coherent GPU already :P
<bbrezillon> robmur01: yep, you need cpu_prep/cpu_fini calls around your CPU access to let the kernel flush the CPU caches
camus has joined #panfrost
camus1 has quit [Ping timeout: 480 seconds]
<robmur01> Do we have a way to indicate the expected usage for individual BOs already? Things that are write-and-forget from the CPU might want to stay non-cacheable for efficiency
<robmur01> (but of course that would make BO caching more "fun")
<robmur01> That could also be handy in terms of mapping things non-shareable at the GPU, but then we can end up having the opposite problem to solve for coherent GPUs, namely making the DMA API do cache maintenance that it thinks it shouldn't need to
<robmur01> I found out last week that the "dma-coherent" fix for Amlogic has broken the DDK in that regard :D
alpernebbi has joined #panfrost
Net147 has quit [Remote host closed the connection]
Net147 has joined #panfrost
atler is now known as Guest3288
atler has joined #panfrost
Guest3288 has quit [Ping timeout: 480 seconds]
camus1 has joined #panfrost
camus has quit [Ping timeout: 480 seconds]
chewitt has joined #panfrost
nlhowell has joined #panfrost
FLHerne has joined #panfrost
f11f12 has joined #panfrost
f11f12 has quit [Quit: Leaving]
hanetzer has quit [Quit: WeeChat 3.2]
<rasterman> alyssa: ping?
<alyssa> rasterman: pong
<rasterman> alyssa: msg..
<alyssa> /accept'd
nlhowell is now known as Guest3301
nlhowell has joined #panfrost
Guest3301 has quit [Ping timeout: 480 seconds]
hanetzer has joined #panfrost
Rathann has joined #panfrost
WoC has quit [Remote host closed the connection]
WoC has joined #panfrost
nlhowell has quit [Ping timeout: 480 seconds]
jambalaya has quit [Ping timeout: 480 seconds]
jambalaya has joined #panfrost
moa is now known as bluebugs
davidlt has quit [Ping timeout: 480 seconds]
atler is now known as Guest3318
atler has joined #panfrost
Guest3318 has quit [Ping timeout: 480 seconds]
rasterman has quit [Quit: Gettin' stinky!]
urja has quit [Read error: Connection reset by peer]
urja has joined #panfrost
alpernebbi has quit [Quit: alpernebbi]
megi has quit [Quit: WeeChat 3.2]
megi has joined #panfrost
Danct12 has quit [Read error: Connection reset by peer]
Danct12 has joined #panfrost
<icecream95> It's a teacher-only day at school today. Do I:
<icecream95> A) Do my homework
<icecream95> 0) Hack on Panfrost
<icecream95> i) Finish a composition task that was due a month ago
<alyssa> #1
<alyssa> or maybe II
<icecream95> Not α?
<macc24> icecream95: optimize panfrost :3
<alyssa> א
<icecream95> macc24: Right, so 1, 0.0, א, A0
<alyssa> Don't forget -0.0
<icecream95> alyssa: You beat me to it...
<alyssa> which is distinct from +0.0
<macc24> Ą
<icecream95> That reminds me of when the maths teacher asked for two non-zero numbers that multiply to give zero...
<alyssa> denorm * 42
atler is now known as Guest3336
atler has joined #panfrost
Guest3336 has quit [Ping timeout: 480 seconds]
Rathann has quit [Remote host closed the connection]
<cphealy> Should this MR land in 21.2.1?
Rathann has joined #panfrost
<alyssa> cphealy: It's `Cc: mesa-stable`, so yes, I expect so
<alyssa> Patch 1 and 3, anyway
<alyssa> Patch 2 is just adding tests for patch 1, doesn't need backport
Rathann has quit [Quit: Leaving]