<icecream95>
Firefox has been crashing a lot for me today...
<macc24>
icecream95: but not for me
gcl has joined #panfrost
gcl_ has quit [Ping timeout: 240 seconds]
<icecream95>
alyssa: Instead of kill-yanking 96 everywhere, I'm thinking of making a macro for the prefetch size. What header file should the value go in?
<alyssa>
icecream95: ~/mesa/src/panfrost/include/panfrost-job.h I guess? assuming it still exists
stikonas has quit [Ping timeout: 258 seconds]
chrisf has quit [Remote host closed the connection]
chrisf has joined #panfrost
kaspter has joined #panfrost
vstehle has quit [Ping timeout: 246 seconds]
<icecream95>
alyssa: I've updated my shader BO allocation MR
<kinkinkijkin>
he'll yeah es3 on by default
cphealy has quit [Remote host closed the connection]
<alyssa>
icecream95: thank you 👍
<alyssa>
-b
<icecream95>
alyssa: Was that meant to be r-b ?
<alyssa>
`it was, and I'm not supposed to be doing panfrost right now
<alyssa>
You know, gotta take breaks from working on open source drivers for gpus
<alyssa>
(the breaks are working on open source drivers for gpus)
<alyssa>
I have very diverse interests.
<alyssa>
(That was sarcasm.)
<kinkinkijkin>
the greatest diversity is the one where everyone has a gpu driver supporting gl3
<kinkinkijkin>
and wayland
<kinkinkijkin>
x11 has no place in a diverse software portfolio
<alyssa>
kinkinkijkin: is there uh any gpu in wide linux use that isn't mesa supported nowadays?
<HdkR>
Luckily ARM dropped x11 support from their Mali blob right? :)
<HdkR>
alyssa: Nvidia? Lawl
<alyssa>
HdkR: Who?
<HdkR>
Good question
<kinkinkijkin>
alyssa, does every single device supported by mesa support gl3?
<HdkR>
nope
<HdkR>
a2xx and Utgard are two on the mobile side
<kinkinkijkin>
now you have my point
<HdkR>
Old desktop vendor cards as well
<kinkinkijkin>
also some older radeon gpus i know
taowa has joined #panfrost
<kinkinkijkin>
something like radeon 3xxx and prior i think?
<kinkinkijkin>
or 2xxx
<kinkinkijkin>
radeon hd*
<HdkR>
Something around there
<kinkinkijkin>
i remember using a 4170 to play trackmania in wine
<kinkinkijkin>
surprisingly enough, it launched
<kinkinkijkin>
not playable but i still played it
<HdkR>
Then Geforce 7 and older on Nvidia side and....Effectively Intel Ironlake and older
<kinkinkijkin>
trackmania 2 that is, specifically lagoon
<kinkinkijkin>
20fps with nothing rendering correctly and not being able to tell what's due to hardware bugs and what's due to lack of features is fun
<kinkinkijkin>
while trying to run an armhf build of mesa on duet
<kinkinkijkin>
to run x86_32 wine in box86
<kinkinkijkin>
which ive now taken out of the chroot cause the chroot wasn't a good idea
<HdkR>
Chroot was a silly idea. Not sure why they claim you need that
<icecream95>
HdkR: Not all distros have multiarch support
<HdkR>
Guess that's true, just needs the kernel support
kaspter has quit [Ping timeout: 264 seconds]
camus1 has joined #panfrost
<kinkinkijkin>
nobody claimed it
camus1 is now known as kaspter
<kinkinkijkin>
i just did it because i HATE cross-compiling
kaspter has quit [Client Quit]
kaspter has joined #panfrost
kaspter has quit [Ping timeout: 264 seconds]
kaspter has joined #panfrost
<icecream95>
Okay, so that's dual-source blending on Bifrost done, I just need to figure out how to keep it working with DCE enabled
<icecream95>
Alacritty is working well
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 256 seconds]
camus1 is now known as kaspter
<HdkR>
woo
karolherbst has joined #panfrost
cowsay_ has joined #panfrost
cowsay has quit [Ping timeout: 240 seconds]
raster has joined #panfrost
stikonas has joined #panfrost
icecream95 has quit [Ping timeout: 265 seconds]
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 246 seconds]
camus1 is now known as kaspter
<kinkinkijkin>
hmm, armv7 build of mesa inside of chroot worked but couldn't access x server, armv7 build of mesa outside of chroot fails an assert before trying to access x server
<kinkinkijkin>
this is definitely an error on my end and i don't know where
<HdkR>
Right, punching xserver through teh chroot barrier is a PITA. Not worth trying to setup
<HdkR>
what sort of assert?
<HdkR>
You have to be a bit careful since --prefix will hardcode the location for mesa to search for drivers
<HdkR>
It's definitely communicated to the driver at that point, which is half the battle
<kinkinkijkin>
thing is, this didn't happen in the chroot, this doesn't happen with the aarch64 build i have installed as well
<kinkinkijkin>
maybe it's loading the wrong build, or maybe this is a real bug
<kinkinkijkin>
would hate to have found a real bug on this since ive spent the last 6 hours trying to figure out what i did wrong
<kinkinkijkin>
note there is one big difference, the aarch64 build is installed on top of the deb package, this build is not
<HdkR>
Trying to run in the chroot most likely would have failed before even opening the device
<HdkR>
Try running a lighter armhf application like es2gears first
<kinkinkijkin>
never gotten es2gears to work on this setup
<kinkinkijkin>
actually not kidding
<kinkinkijkin>
es2gears segfaults, glxgears runs
<HdkR>
:)
<kinkinkijkin>
hdkr you probably know from when i was active in the dolphin community, it's alien to me that an issue i have with tech might not be either hellmagicks or me being an idiot
<HdkR>
I think you just found a legitimate bug. Probably needs someone with active 32bit libraries to test other applications
raster has quit [Read error: Connection reset by peer]
<macc24>
kinkinkijkin: apt install something-mesa-somethiung:armhf doesn't do the job?
raster has joined #panfrost
kaspter has quit [Ping timeout: 260 seconds]
kaspter has joined #panfrost
<macc24>
networkmanager doesn't work for no reason... enough void for this week
<macc24>
i wonder if gnome now works on bifrost
urjaman has quit [Quit: WeeChat 2.8]
urjaman has joined #panfrost
Venemo has quit [Ping timeout: 256 seconds]
Venemo has joined #panfrost
<macc24>
gnome works surprisingly well on duet
<alyssa>
kinkinkijkin: that sounds like an actual bug, I'd appreciate a capture/trace
<macc24>
if cursor wasn't so laggy i'd use it
<macc24>
and why fractional scaling isn't supported in gnome hmm
stikonas_ has joined #panfrost
stikonas has quit [Ping timeout: 272 seconds]
<macc24>
icecream95: i got iio-sensor-proxy to work correctly on duet
kaspter has quit [Quit: kaspter]
griffinp- has joined #panfrost
listen_ers has joined #panfrost
davidlt has quit [Quit: Leaving]
<italove>
hi everyone, I want to implement multisampled shader images on midgard but there's no test case for it on deqp and I'm not sure how to lookup specific piglit test cases like this. Is there another simple way to test this that I'm missing?
<alyssa>
I'm a bit surprised deqp doesn't have cases
stikonas_ is now known as stikonas
davidlt has joined #panfrost
archetech has joined #panfrost
yann has quit [Ping timeout: 240 seconds]
yann has joined #panfrost
<daniels>
alyssa: re building clover/llvm/llvm-spirv, no you don't need out-of-tree or weird-branch llvm, you just need to pull & build the spirv-llvm-translator branch corresponding to whichever version of llvm you have installed ... see e.g. .gitlab-ci/container/x86_build.sh (grep for SPIRV)
<bbrezillon>
italove: might be worth adding such tests for GLES3.1 (looks 2DMS images are tested in regular GL => external/openglcts/modules/gl/gl4cShaderImageLoadStoreTests.cpp)
<daniels>
italove: there are a few tests for it I found by grepping for image2DMS in piglit/tests/ - I couldn't find any indication that GLES has any support for MSAA images though?
<italove>
daniels hm, maybe it doesn't
<bbrezillon>
oh, that might be it
<daniels>
ESSL reserves {,i,u}image2DMS as keywords but nowhere does it mention support for them, and GL_EXT_gpu_shader5 has a note saying that MSAA image support was removed from ES3.1
<italove>
I guess we just don't implement support for it then? (or at least not in the near feature)
<daniels>
GL_ARB_shader_image_load_store is already exposed (in a GL context, not GLES) in your current branch, right?
<alyssa>
daniels: alright, is that new since oct 2018 maybe?
<daniels>
alyssa: building SPV in CI is 'new' IIRC but Clover itself has always built with vanilla upstream LLVM and in fact builds back to something like 3.8
<alyssa>
italove: If it isn't too much trouble, I'd implement with pigitlit
<italove>
alyssa ok, or maybe doing what bbrezillon suggested of adapting the regular GL 2dms image tests in deqp for GLES3.1
<italove>
daniels I'm not sure, how do I check that?
<alyssa>
daniels: I recall being able o use normal upstream llvm, but needing a weird llvm-spirv build since the one in the distro was too old or something
<daniels>
italove: depends on what you have installed, but mesa/demos will build eglinfo for OpenGL, or you can use glxinfo if you have X11, or wflinfo if you have that installed ...
<daniels>
alyssa: huh, it's worked fine from at least the 9.x release branch
<alyssa>
daniels: 🤷 maybe debian weirdness
<italove>
daniels yeah, it is exposed in regular GL
<daniels>
italove: well there you go, you should be able to use the GL Piglit tests for that ext then :)
<daniels>
I do think it makes sense to finish it off, might as well at this point
<italove>
daniels sure :)
raster has quit [Quit: Gettin' stinky!]
archetech has quit [Remote host closed the connection]
<daniels>
alyssa: yep, scroll down to the bottom and you can see a 13min delta from submit time -> start time; it only took 6 minutes once a device became available
<daniels>
forced a health check and it seems to be fine again