alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - https://gitlab.freedesktop.org/panfrost - Logs https://freenode.irclog.whitequark.org/panfrost - <daniels> avoiding X is a huge feature
rcf has joined #panfrost
robert_ancell has quit [Ping timeout: 272 seconds]
vstehle has quit [Ping timeout: 246 seconds]
davidlt has joined #panfrost
vstehle has joined #panfrost
Elpaulo has quit [Read error: Connection reset by peer]
Elpaulo has joined #panfrost
hlmjr has joined #panfrost
herbmillerjr has quit [Ping timeout: 268 seconds]
somy has quit [Ping timeout: 252 seconds]
yann has quit [Ping timeout: 272 seconds]
pH5 has joined #panfrost
yann has joined #panfrost
yann has quit [Ping timeout: 246 seconds]
yann has joined #panfrost
yann has quit [Ping timeout: 244 seconds]
adjtm_ has quit [Remote host closed the connection]
adjtm_ has joined #panfrost
adjtm_ has quit [Quit: Leaving]
samueldr has quit [Ping timeout: 245 seconds]
raster has joined #panfrost
samueldr has joined #panfrost
raster has quit [Remote host closed the connection]
bbrezillon has quit [Quit: WeeChat 1.9.1]
afaerber has joined #panfrost
adjtm has joined #panfrost
yann has joined #panfrost
guillaume_g has quit [Quit: Konversation terminated!]
tgall_fo_ has joined #panfrost
tgall_foo has quit [Ping timeout: 246 seconds]
tgall_fo_ is now known as tgall_foo
guillaume_g has joined #panfrost
guillaume_g has quit [Ping timeout: 244 seconds]
guillaume_g has joined #panfrost
<tomeu> alyssa: do you see anything obviously wrong here? http://paste.debian.net/1089319/
<tomeu> by substracting 1 to rt->framebuffer, I stop getting the DATA_INVALID_FAULTs
<tomeu> but of course I get weird rendering
<tomeu> alyssa: oh, everything is fine if I substract 1 to rt->afbc.metadata :p
<tomeu> the hardest problems I have found in panfrost have been solved by changing bits at random
<tomeu> maybe the Zen of panfrost is to shut down your mind and become one with the ever-changing bitstream
<HdkR> Do DATA_INVALID_FAULTS show what address was attempting to get addressed?
<HdkR> accessed*
<tomeu> those aren't faults when trying to access memory, but related to the HW not liking the contents of some descriptors
<tomeu> and it does set an address, but it doesn't point to the descriptor with the fault
<HdkR> Disappointing
<tomeu> but to another descriptor in the same job
<tomeu> though, maybe from Arm's POV there's only one descriptor
<tomeu> guess we cannot tell from our REing
<tomeu> ok, guess it's a good point to finish the day, tomorrow I should be able to send patches for AFBC modifiers in shared buffers
<alyssa> tomeu: Oh, yeah, that +1 is probably wrong.
<alyssa> But the bit about "the hardest problems..." is so accurate, lol
<jernej> tomeu: any luck with T720?
pH5 has quit [Quit: bye]
guillaume_g has quit [Quit: Konversation terminated!]
* alyssa curses CI
pH5 has joined #panfrost
<alyssa> tomeu: I'm sorry, I.... The entire RK3288 CI is a flip-flop.
<alyssa> Can we make the overall green checkmark based on the RK3399 results?
<alyssa> I can't afford to lose perfectly good code because of bugs that might not exist on hardware I can't test on
<alyssa> (Apology prompted by knowingly pushing code failing CI, though it's green on RK3399.. So meh)
<HdkR> Time to flop-flip the flip-flop
stikonas has joined #panfrost
stikonas has quit [Client Quit]
stikonas_ has joined #panfrost
<gtucker> tomeu: there are 2 rk3288 devices in the lab, could you maybe check if the pass/fail results are tied to any of them in particular?
<gtucker> they are both rk3288-veyron-jaq, but still it would be interesting if one was more prone to failing tests than the other
ente has quit [Ping timeout: 258 seconds]
stikonas_ has quit [Ping timeout: 272 seconds]
stikonas_ has joined #panfrost
bbrezillon has joined #panfrost
pH5 has quit [Quit: bye]
* alyssa drowns in abstraction
<alyssa> No, no, this doesn't make sense
* alyssa is trying to figure out how to move transient pools down but this is... ballooning
<alyssa> The goal is so the transient pool is associated wit the panfrost_job, not the context
<alyssa> But this will.... take some design
<alyssa> I think really what we want is a global page pool
<alyssa> s/page/chunk/
<alyssa> And then associated with the job is a set of "borrowed pages"
<alyssa> Low-hanging fruit, I guess
davidlt has quit [Remote host closed the connection]
davidlt has joined #panfrost
<alyssa> OK, I've got transient uploads associated with the batch. That's a start.
<alyssa> Next up, playing games with transient_pool
<alyssa> And by that i mean essnetially rewriting the transient pool logic.
<alyssa> Ahhh the indirection is toxic
BenG83 has joined #panfrost
<sravn> alyssa: thanks for the nice blog post, always interesting to follow the progress and your stories.
Kwiboo has quit [Quit: .]
Kwiboo has joined #panfrost
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
jernej has joined #panfrost
davidlt has quit [Ping timeout: 244 seconds]
<alyssa> sravn: Thank you :)
afaerber has quit [Quit: Leaving]
cwabbott has quit [Read error: Connection reset by peer]
stikonas_ is now known as stikonas