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
vstehle has quit [Ping timeout: 268 seconds]
herbmillerjr has joined #panfrost
davidlt has joined #panfrost
jernej__ has joined #panfrost
vstehle has joined #panfrost
jernej__ has quit [Ping timeout: 244 seconds]
jernej__ has joined #panfrost
<tomeu> weston?
davidlt_ has joined #panfrost
davidlt has quit [Ping timeout: 244 seconds]
jernej__ has quit [Ping timeout: 272 seconds]
yann has quit [Ping timeout: 248 seconds]
BenG83 has quit [Ping timeout: 248 seconds]
chewitt has quit [Quit: Zzz..]
<tomeu> got things working on T720, was a simple matter of flipping the right bits
<tomeu> no weston yet though
<tomeu> but enough for igt :)
<HdkR> :D
<tomeu> narmstrong: do you know when the defconfig changes will make it to a tree that is built by kernelci?
<tomeu> guess we could have our panfrost-for-next branch, but it would be probably not a good use of kernelci resources
stikonas has joined #panfrost
<narmstrong> tomeu: no, i will ask Kevin to apply them on the amlogic tree
stikonas has quit [Remote host closed the connection]
chewitt has joined #panfrost
<tomeu> narmstrong: nice, thanks
chewitt has quit [Quit: Adios!]
davidlt_ is now known as davidlt
yann has joined #panfrost
raster has joined #panfrost
<tomeu> alyssa: do you happen to know how to force a GPU hang with SFBD?
<shadeslayer[m]> I can't think of anything else that needs fixing for now
<tomeu> shadeslayer[m]: you should be clearing up the commit story before asking others to review
<tomeu> otherwise it's more work, and you anyway need to do that
<shadeslayer[m]> Ok, should I just open a MR after?
<tomeu> shadeslayer[m]: well, I think alyssa is not using gitlab's web UI
<shadeslayer[m]> Aw :(
davidlt has quit [Ping timeout: 244 seconds]
<tomeu> shadeslayer[m]: you can just send them to mesa-dev like any other patches
<shadeslayer[m]> I need to figure out how to do that, never had to before :)
<tomeu> awesome :)
unoccupied has joined #panfrost
unoccupied is now known as djp_
djp_ is now known as unoccupied
<shadeslayer[m]> Ack :)
davidlt has joined #panfrost
tgall_foo has joined #panfrost
jolan has quit [Quit: leaving]
jolan has joined #panfrost
davidlt_ has joined #panfrost
davidlt has quit [Ping timeout: 252 seconds]
<alyssa> tomeu: Not sure why you want that ;P
<tomeu> alyssa: we have a test in igt about gpu resets
<alyssa> narmstrong: I ask Kevin lots of things
<alyssa> tomeu: What I'll suggest is taking the ->format field
<alyssa> And just doing a loop of
<tomeu> oh, good idea
<alyssa> for (unsigned i = 0; i < 32; ++i) { fb->format ^= (1 << i); /* render */ }
<alyssa> Eventually you'll break something
<tomeu> alyssa: no luck
<tomeu> trying with clear_flags now
<tomeu> alyssa: I'm only getting DATA_INVALID_FAULTs
jernej__ has joined #panfrost
<tomeu> alyssa: Bingo! 608d
<tomeu> :)
<tomeu> tomorrow will use the value to cause the hang
<anarsoul> alyssa: you should consider switching to gitlab ui, it's convenient
jernej__ has quit [Quit: Konversation terminated!]
<alyssa> tomeu: It scares me that that works.
<alyssa> Just to be clear, setting random bits should *not* cause hangs. But yuknow :P
jernej has joined #panfrost
tlwoerner has quit [Ping timeout: 257 seconds]
tlwoerner has joined #panfrost
yann has quit [Ping timeout: 246 seconds]
BenG83 has joined #panfrost
stikonas_ has joined #panfrost
yann has joined #panfrost
<alyssa> Half of the tests in a section fail, half pass
<alyssa> I set a magic bit
<alyssa> Now the first half all pass, and the second half all fail
<alyssa> One of those days, huh
* alyssa blames mipmapping
raster has quit [Remote host closed the connection]
buzzmarshall has joined #panfrost
davidlt_ is now known as davidlt
raster has joined #panfrost
pH5 has joined #panfrost
buzzmarshall has quit [Quit: Leaving]
buzzmarshall has joined #panfrost
<alyssa> I'm rewriting pantrace stuff
<alyssa> Code is now waaaay simpler and waaaaaaaaaaaaaaaaaaaay faster
<alyssa> As in, fast enough to trace small scenes in real-time now
<anarsoul> cool
<alyssa> And on medium size scenes the bottleneck is just I/O / terminal rendering
<alyssa> (If I pipe the trace to /dev/shm/, it's real-time)
stikonas_ has quit [Remote host closed the connection]
stikonas_ has joined #panfrost
BenG83 has quit [Ping timeout: 258 seconds]
pH5 has quit [Quit: -_-]
davidlt has quit [Ping timeout: 258 seconds]
somy has quit [Ping timeout: 248 seconds]
Elpaulo has quit [Read error: Connection reset by peer]
Elpaulo has joined #panfrost
BenG83 has joined #panfrost
<alyssa> Trying to solve a nasty bug with FBOs
<alyssa> Getting a DATA_INVALID_FAULT; the fault_pointer (which I can now access \ o /) blames the vertex/tiler MFBD
<HdkR> Oh no, buhaboo
<alyssa> HdkR: Hm?
<HdkR> bugaboo*