<icecream95>
This one causes the background checkerboard to sometimes appear where it shouldn't in the GIMP
<icecream95>
And I still haven't made an MR for the second CRC bug yet..
raster has joined #panfrost
<icecream95>
The issue doesn't reproduce under Xephyr, so this one will be "fun" to debug
raster- has joined #panfrost
raster has quit [Ping timeout: 246 seconds]
raster- has quit [Client Quit]
raster has joined #panfrost
popolon has quit [Quit: WeeChat 3.1]
mixfix41 has joined #panfrost
raster- has joined #panfrost
raster has quit [Ping timeout: 245 seconds]
raster- has quit [Client Quit]
stikonas has quit [Remote host closed the connection]
atler has quit [Ping timeout: 252 seconds]
atler has joined #panfrost
<icecream95>
...And now shadows in STK are broken again. I miss when I could fix bugs faster than I could find them
<macc24>
is panvk usable for testing?
<icecream95>
d4d3328b952 ("panfrost: Enable 16-bit support on Bifrost") is the first bad commit. Maybe #notourbug then?
karolherbst has quit [Quit: duh 🐧]
karolherbst has joined #panfrost
andrey-konovalov has quit [Read error: Connection reset by peer]
mani_s has quit [Read error: Connection reset by peer]
mani_s has joined #panfrost
<alyssa>
Could go either way.
<alyssa>
There are a _lot_ of buggy apps in the wild that incorrect assume mediump is 32-bit etc.
<HdkR>
At least in GL land it isn't a wrong assumption
<alyssa>
There are also, sadly, a lot of places the Bifrost 16-bit support could be buggy. I'm convinced there are errata in Bifrost I don't know about (yet?) because it's otherwise hard to explain a lot of the mysterious INSTR_INVALID_ENC's.
<alyssa>
(and the peculiar assembly the DDK generates that smellls like workarounds)
<alyssa>
Mali T860, for whatever it's worth, had few "serious" errata.
davidlt has joined #panfrost
<icecream95>
Adding 'precision highp float' to the start of every shader (using MESA_SHADER_{DUMP,READ}_PATH) doesn't fix it...
<alyssa>
Eep.
<alyssa>
PAN_MESA_DEBUG=nofp16 does though?
<icecream95>
yes
<icecream95>
Adding 'precision highp sampler2D;' to the 'sunlightshadow' shader fixes it. Note that the shader does specify highp for sampler2DArray and sampler2DArrayShadow
<icecream95>
The sampler causing the artefacts is just a regular sampler2D, so #notourbug?
vstehle has joined #panfrost
catfella has quit [Remote host closed the connection]
raster has joined #panfrost
davidlt has quit [Ping timeout: 260 seconds]
raster has quit [Quit: Gettin' stinky!]
bbrezillon has quit [Ping timeout: 240 seconds]
raster has joined #panfrost
bbrezillon has joined #panfrost
kaspter has joined #panfrost
kaspter has quit [Excess Flood]
kaspter has joined #panfrost
<danboid>
Oh dear oh dear!
<danboid>
I finally got mesa git built and installed on my g31 board and its worse than the version that was installed previously
<danboid>
A lot worse. Now I gon't even get text in menus and windows
<danboid>
I found a mesa-git PKGBUILD for arm here:
<danboid>
Is this a known error? Should I open a bug report?
<HdkR>
I personally don't know. someone else will know though
<HdkR>
But that is definitely why the rendering is wrong
<danboid>
I'm going to re-install. I was only getting one dmesg panfrost error before and I think it was a different one so I'll have to get back to you
<danboid>
At least I know how to build mesa under Arch ARM now
<danboid>
I'm running Linux 5.11.13 btw
<HdkR>
This is more likely to be a userspace problem
stikonas has joined #panfrost
<danboid>
Within mesa
<HdkR>
right
catfella has joined #panfrost
alpernebbi has joined #panfrost
stikonas has quit [Ping timeout: 260 seconds]
stikonas has joined #panfrost
unoccupied has joined #panfrost
unoccupied is now known as leah
zkrx has quit [Ping timeout: 268 seconds]
thecycoone has quit [Quit: Bridge terminating on SIGTERM]
stikonas has quit [Ping timeout: 268 seconds]
zkrx has joined #panfrost
stikonas has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
leah has quit [Quit: WeeChat 2.8]
unoccupied has joined #panfrost
unoccupied is now known as leah
leah has quit [Client Quit]
Lyude has quit [Quit: WeeChat 3.0.1]
Lyude has joined #panfrost
stikonas has quit [Read error: Connection reset by peer]
stikonas has joined #panfrost
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
warpme_ has quit [Quit: Connection closed for inactivity]
archetech has joined #panfrost
karolherbst has quit [Quit: duh 🐧]
karolherbst has joined #panfrost
kaspter has quit [Quit: kaspter]
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
zkrx has quit [Ping timeout: 260 seconds]
<daniels>
amonakov, robmur01: much of Collabora's Panfrost work is paid, but a lot of it (including e.g. panvk) is indeed just paid for by ourselves - we do a lot more unpaid-for-by-clients work than you might think
<daniels>
danboid: the radeonsi Mesa driver does work on AArch64 systems with PCIE, yeah
<danboid>
daniels, Very nice!
zkrx has joined #panfrost
<daniels>
(we do have physical offices in Cambridge and Montreal, but ~90% were remote even before last year ...)
<amonakov>
danboid, daniels: it seems just having a pcie port is not enough, the pcie controller in the soc should be good enough: https://pipci.jeffgeerling.com/
<amonakov>
istr a recent discussion that involved marcan42 and that link with technical speculation what's broken, but cannot find it again
<daniels>
I did see that, but it sure does seem like a solution in search of a problem
<amonakov>
hm?
<daniels>
I don't entirely know why you'd shackle a powerful GPU to a RPi
<daniels>
(we weren't running it on a Raspberry Pi, that's for sure, but the platform is basically irrelevant - as long as you have AArch64 and working PCIE then radeonsi+amdgpu works well)
danboid has quit [Ping timeout: 245 seconds]
alpernebbi has quit [Quit: alpernebbi]
warpme_ has joined #panfrost
archetech has quit [Quit: Konversation terminated!]
catfella has quit [Remote host closed the connection]
stikonas has quit [Read error: Connection reset by peer]