<icecream95>
the fragment shader compares floating point varying values using ==, they shouldn't assume it's exact
<anarsoul|2>
heh
warpme_ has quit [Quit: Connection closed for inactivity]
vstehle has quit [Ping timeout: 264 seconds]
rhyskidd has quit [Quit: rhyskidd]
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 240 seconds]
camus1 is now known as kaspter
davidlt has joined #panfrost
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 260 seconds]
camus1 is now known as kaspter
vstehle has joined #panfrost
kaspter has quit [Ping timeout: 240 seconds]
kaspter has joined #panfrost
rhyskidd has joined #panfrost
kaspter has quit [Ping timeout: 240 seconds]
kaspter has joined #panfrost
camus1 has joined #panfrost
kaspter has quit [Ping timeout: 258 seconds]
camus1 is now known as kaspter
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
nlhowell has joined #panfrost
<urjaman>
icecream95: they maybe shouldnt (and i know that in general, one shouldnt _compute_ a float and compare it exactly), but now I'm confused as to why is it not exact... it's not like floats have trouble representing the integers from 2 to 10 exactly
guillaume_g has joined #panfrost
buzzmarshall has quit [Remote host closed the connection]
<urjaman>
but given how T860 is exact (apparently atleast in this case) and T760 is exact-but-only-sometimes it's a case of 'this hardware wack, yo' (the VS never does anything but assignment with the mode value, and mostly just passes them onto the FS...)
yann has joined #panfrost
warpme_ has joined #panfrost
<urjaman>
but yeah i'll go and talk to the kicad folks about this, we'll see...
<urjaman>
(once i've like, woken up, tho)
kaspter has quit [Ping timeout: 256 seconds]
kaspter has joined #panfrost
<urjaman>
oh hey
<urjaman>
I tried to open the kicad 3D Viewer (that'd show a 3D rendering of the board), and:
<icecream95>
urjaman: Varying interpolation is done by multiplying the value from each vertex by the barycentric weights of the fragment - if these don't add up to 1, then the value can be slightly off. In this case, it's only off by one - the float is 0x3fffffff (2.000000) instead of 0x40000000 (2.000000)
<urjaman>
interesting (i was expecting it was something to do with the pipeline between the shaders, but i know very little of that)
<urjaman>
also, 5915 does fix the 3D viewer starting and running :) it also has some circles that are 4-sided this time, i'll dump the shaders from that to see what the silly is :P
<urjaman>
... oh. (no shaders...)
<urjaman>
(i was thinking about that, IIRC the 3D Viewer is GL 1.x stuff, predates the accelerated GAL in rest of kicad)
<robmur01>
HdkR, daniels: NXP are shipping Mali-DP500 in LS1028A
<HdkR>
Whoa
<robmur01>
but in general you can guess at how popular Mali Display and Mali Video were by the way they've completely vanished form the "products" section of our website...
<HdkR>
It seems like a natural tie in, but I guess everyone loves reinventing the blocks
<robmur01>
I expect it'll be a machine I'll get to rage against in due course
<alyssa>
[insert Rust noises here]
tomboy65 has quit [Read error: Connection reset by peer]
tomboy65 has joined #panfrost
buzzmarshall has joined #panfrost
raster has joined #panfrost
yann has quit [Ping timeout: 256 seconds]
guillaume_g has quit [Quit: Konversation terminated!]
davidlt_ has joined #panfrost
davidlt_ is now known as davidlt
austriancoder has joined #panfrost
<tomeu>
alyssa: sure, I'm asking because in that case may make sense to take a trace and put it in CI
<alyssa>
tomeu: Oh, that's fair
<tomeu>
hopefully by next week we'll be replaying traces in CI
<alyssa>
ack
narmstrong has quit []
narmstrong has joined #panfrost
karolherbst has quit [Remote host closed the connection]
karolherbst has joined #panfrost
tomboy65 has quit [Remote host closed the connection]
tomboy65 has joined #panfrost
Green has quit [Quit: Ping timeout (120 seconds)]
Green has joined #panfrost
<alyssa>
oh, this bug is... icky
<urjaman>
hmm?
TheKit has quit [Ping timeout: 260 seconds]
TheKit has joined #panfrost
<alyssa>
urjaman: Spilling weirdness in the webgl cts
gcl has quit [Ping timeout: 256 seconds]
davidlt has quit [Ping timeout: 240 seconds]
gcl has joined #panfrost
<urjaman>
I've built the kicad master and wrangled together some hopefully acceptable-looking patches to the shader and made a kicad MR
<urjaman>
It took longer for the C201 to build kicad (on one core tho, i forgot the -j and was in no hurry afterwards lol) than for someone to accept me onto their mailing list (dunno if i'll need that for anything, but I did join anyways... gotta look like a proper dev lol)
<HdkR>
GL 3.3 you say? ooo
cwabbott_ has joined #panfrost
cwabbott has quit [Ping timeout: 260 seconds]
cwabbott_ is now known as cwabbott
<alyssa>
urjaman: \o/
<alyssa>
HdkR: definitely not conformant or complete or anything
* alyssa
wonders why aquarium is stuck at 20fps
<HdkR>
What you're telling me is that I need to try Dolphin on it :P
<alyssa>
HdkR: does Dolphin have an ES3.0 backend?
<HdkR>
yes
<alyssa>
then use that
<alyssa>
and you need to try Dolphin on it
<alyssa>
(:
<HdkR>
haha
<HdkR>
Last I left the GL loader state, it'll choose GL over ES if GL 3.0 is available :D
<alyssa>
HdkR: Only GL 2.1 is available so it's fine
macc24 has quit [Read error: Connection reset by peer]
macc24_ has joined #panfrost
macc24_ is now known as macc24
<alyssa>
aquarium seems heavily fragment bound
<alyssa>
influenced by framebuffer size
raster has quit [Quit: Gettin' stinky!]
<urjaman>
Oh, the Mali T760 is not the first GPU to render circles as triangles (a Radeon R300 done it too, atleast with some drivers...), but kicad people hadnt figured out it was them, the bug report had been closed as a "driver bug, wontfix"
<urjaman>
But the first responder to my MR thanked me and said that the fix looks correct (and linked that old bug report), so yay i think
<HdkR>
glEnable(GL_CIRCLE_STRIP)
<alyssa>
urjaman: Wait, so is it a driver bug or a panfrost bug?
<alyssa>
or both interacting
<urjaman>
alyssa: as far as i know, it's a kicad bug, that kicad had blamed on the drivers (radeon ones, not us, because well ... panfrost just became capable of running kicad :P) before this
<alyssa>
urjaman: what about the stencil/discard bugfix icecream did?
<urjaman>
I didnt look into it more than just reading the diff, but that one seemed like a panfrost bug, works now
<alyssa>
oh, ok
<urjaman>
icecream95 didn't exactly go into detail on how they figured that one out :D
<urjaman>
i dropped the 3D viewer trace and like an hour later *fix*
<alyssa>
(:
<urjaman>
And i'm not enough of an OpenGL whiz yet
<urjaman>
" Merge Request !291 was merged " .... Wow. I'm a kicad contributor now it seems