kaspter has quit [Read error: Connection reset by peer]
kaspter has joined #panfrost
rcf1 has joined #panfrost
camus1 has joined #panfrost
rcf1 has quit [Client Quit]
kaspter has quit [Ping timeout: 272 seconds]
camus1 is now known as kaspter
rcf1 has joined #panfrost
rcf has quit [Quit: WeeChat 2.7.1]
rcf1 is now known as rcf
nerdboy has quit [Ping timeout: 246 seconds]
kaspter has quit [Remote host closed the connection]
kaspter has joined #panfrost
nerdboy has joined #panfrost
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 256 seconds]
camus1 is now known as kaspter
paulk-leonov has quit [Ping timeout: 240 seconds]
paulk-leonov has joined #panfrost
kaspter has quit [Remote host closed the connection]
kaspter has joined #panfrost
guillaume_g has joined #panfrost
yann has joined #panfrost
raster has joined #panfrost
stikonas has joined #panfrost
tomboy64 has quit [Ping timeout: 240 seconds]
kaspter has quit [Ping timeout: 246 seconds]
kaspter has joined #panfrost
nlhowell has quit [Ping timeout: 264 seconds]
<tomeu>
chrisf: I run libmali on the same rootfs where I test panfrost
tomboy64 has joined #panfrost
<tomeu>
you "just" need to find the right kbase and the right libmali
kaspter has quit [Ping timeout: 264 seconds]
kaspter has joined #panfrost
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 256 seconds]
camus1 is now known as kaspter
<cwabbott>
chrisf: the state-of-the-art for comparisons in NIR is to either not lower booleans in NIR, and then you get 1-bit SSA values that you can implement in any way you want in the backend, or implement your own lowering in NIR that produces different instructions
<cwabbott>
there's some code in freedreno to convert "NIR booleans" to 0/1 booleans, but that was from the pre-1-bit-bools era
<cwabbott>
note that for external things like SSBO's, NIR generates a comparison instruction when loading from something with a bool type
adjtm has joined #panfrost
adjtm has quit [Remote host closed the connection]
adjtm has joined #panfrost
<Ke>
looks like debian sid panfrost with linux-5.8 hangs system on firefox zoom screen share
<Ke>
lets see, if LIBGL_ALWAYS_SOFTWARE helps
<Ke>
so far so good
<Ke>
this is on pbp/rk3399
<Ke>
not sure what tech zoom uses for video
adjtm has quit [Read error: Connection reset by peer]
adjtm has joined #panfrost
<Ke>
could try to check serial console
adjtm has quit [Quit: Leaving]
raster has quit [Remote host closed the connection]
raster has joined #panfrost
raster has quit [Client Quit]
raster has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
raster has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
raster has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
guillaume_g has quit [Quit: Konversation terminated!]
raster has joined #panfrost
raster has quit [Remote host closed the connection]
icecream95 has quit [Quit: leaving]
guillaume_g has joined #panfrost
raster has joined #panfrost
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 256 seconds]
camus1 is now known as kaspter
nlhowell has joined #panfrost
nlhowell has quit [Ping timeout: 258 seconds]
<alyssa>
chrisf: bifrost comparisons have a d3d flag which if set does 0/~0
<alyssa>
"frowns at the way the meson script is done for panloader" ...patches welcome? ;P
nlhowell has joined #panfrost
Marex has quit [Quit: leaving]
nlhowell has quit [Ping timeout: 246 seconds]
cwabbott_ has joined #panfrost
cwabbott_ has quit [Client Quit]
cwabbott_ has joined #panfrost
cwabbott has quit [Ping timeout: 272 seconds]
cwabbott_ is now known as cwabbott
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
enunes has quit [Remote host closed the connection]
guillaume_g has quit [Quit: Konversation terminated!]
* tomeu
didn't know we had "insurance" in panfrost ;)
nerdboy has joined #panfrost
enunes has joined #panfrost
<chrisf>
alyssa: aha!
<alyssa>
chrisf: :)
<chrisf>
re d3d/gl flag. i was wondering what that did
<chrisf>
we should probably set it if we're assuming that :)
<alyssa>
I don't think I ever decided which we wanted to assume
<alyssa>
not sure which is more convenient at this point
<chrisf>
bool uniforms go in as 0/~0
<chrisf>
bools in ubos are 0/nonzero, we probably have to emit a comparison to fix them anyway
<chrisf>
but in any case, mixed is no good :)
<alyssa>
the OpenCL blob ends up mixed iirc =p
<alyssa>
It's not entirely clear to me why the hw supports both
<alyssa>
Midgard is always 0/~0 and it's fine there
<chrisf>
as for panloader stuff.. the badness is in assuming that the build tree is inside the source tree
<chrisf>
im not sure i can be bothered untangling it, i just forced the paths to where i built those libs
<Lyude>
chrisf: how's the bf stuff going?
<chrisf>
Lyude: i have a couple of fixes that i think are useful
<chrisf>
Lyude: just working through dEQP-GLES2
<Lyude>
ahh, how far have yall gotten with the scheduler stuff btw orhave you not started on that? just kinda curious
<chrisf>
i havent looked
<alyssa>
Lyude: haven't started, I was away for a little while for various reasons and have been focusing on GLES3.0/Midgard for the very short term
<chrisf>
so -- patches. still mail to mesa-dev list?
<chrisf>
or would you prefer a gitlab MR?
<cwabbott>
chrisf: fwiw, ARM's compiler optimizes -int(bool) to the d3d version
<cwabbott>
it's probably useful for d3d translated shaders to support that optimization
<alyssa>
chrisf: MR, please
<cwabbott>
you can lower booleans to d3d-style ones or gl-style ones in NIR and then have a late algebraic optimization that converts a negate of the gl-style to d3d-style, or vice versa, or you could do something fancier where you analyze the uses/defs to see which is better globally
<cwabbott>
the former will be good enough to make most simple boolean uses optimal
<chrisf>
right
<chrisf>
im not super worried about being optimal yet, i just want to be correct
raster- has joined #panfrost
<chrisf>
alyssa: what do i do with these manual pipelines on the gitlab?
<chrisf>
should they be kicked off now?
<tomeu>
chrisf: one needs to either start them manually, or hack the .gitlab-ci.yml so they are started automatically
<tomeu>
we only cover midgard atm though
<tomeu>
chrisf: for the panfrost jobs, it's enough with starting arm_build
<tomeu>
arm_test is for some fdo jobs that have gitlab runners on qcom hw
<chrisf>
how does the review flow work now? do i still collect r-b lines and amend the patches with them, or is it done a different way?
<tomeu>
yes, submit MR, get r-bs, then re-push with the tags, and assign to Marge
<tomeu>
she will take care of doing one last CI run and merge
raster- has quit [Quit: Gettin' stinky!]
raster has quit [Remote host closed the connection]
raster has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
raster has joined #panfrost
buzzmarshall has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
icecream95 has joined #panfrost
<alyssa>
I wish marge added the tags
<HdkR>
Folder specific tag adding? :)
<HdkR>
Oops, you touched a panfrost folder, now you get the tag
<alyssa>
HdkR: r-b/a-b tags, I mean :)
<HdkR>
ah
<Lyude>
wonder if we could suggest that to gitlab
<alyssa>
Lyude: funny you mention it... Marge supports this exact feature but only on GitLab EE and we're running CE on fd.o because foss
<Lyude>
wat
<Lyude>
that seems, like exactly the kind of thing you'd really only want in open source lol
<alyssa>
Of course we could hack the feature into marge but nobody's bothered
<alyssa>
Marge Bot isn't part of gitlab iirc, but the API it uses for tagging stuff is EE-only
rcf has quit [Quit: WeeChat 2.7]
raster has joined #panfrost
<alyssa>
icecream95: trying to figure out what to do (as a reviewer) wrt fbfetch stuff