fenix144 has quit [Remote host closed the connection]
mikelee has quit [Ping timeout: 480 seconds]
mikelee has joined #asahi-gpu
mikelee has quit [Ping timeout: 480 seconds]
fenix144 has joined #asahi-gpu
mikelee has joined #asahi-gpu
<alyssa>
i509vcb: custom border colours in agxv are awkward because they're tied to the textur eofrmat
<alyssa>
format
<alyssa>
which means we can implement VK_EXT_custom_border_color but not customBorderColorWithoutFormat
<i509vcb>
yeah I suspected just wiring up the type in the buffer object would likely not be enough
<alyssa>
well, EXT_custom_border_color has a provision for hardware like AGX, that need to know the format
<alyssa>
they give you the VkFormat
<i509vcb>
Anyways for another day, nausea is not fun
<alyssa>
it's just that there's a feature bit to make it optional
<alyssa>
so we can do the ext without that feature
<i509vcb>
I recall seeing a comment in agx_border about that
<alyssa>
yeah
<alyssa>
The problem is... the two interesting users of that extension (DXVK and VKD3D-Proton) both require customBorderColorWithoutFormat
<alyssa>
so I'm not sure how useful EXT_cbc without that feature is in practice
<alyssa>
Zink would use it. That's probably it
<alyssa>
(I'm not too keen on doing stuff Just For Zink when we have a perfectly working GL driver... OTOH if AGXV+Zink happens to Just Work Well down the line that's a different story.)
<alyssa>
needing to emulate border colours though would be very painful for performance so idk. we'll cross that bridge when we get there. custom borders are rare but occassionally used
<i509vcb>
I imagine it's non trivial to do some sort of lowering for customBorderColorWithoutFormat under the hood in dxvk/vkd3d-proton
mikelee has quit [Ping timeout: 480 seconds]
<i509vcb>
Speaking of agxv I probably need to poke ella about my VK_EXT_physical_device_drm patch
<alyssa>
i509vcb: It's just fundamentally not how d3d works, as I understand
<alyssa>
I don't think they could emulate it any better than we can in the drive
<i509vcb>
Probably best to focus on what is needed for other vulkan things like sway
<alyssa>
meh?
<alyssa>
sway's gles2 renderer is fine
<alyssa>
personally I want Vulkan for workloads that actually need Vulkan
<alyssa>
DXVK/VKD3D being the final boss
mikelee has joined #asahi-gpu
<i509vcb>
I guess we have different priorities
<i509vcb>
A lot of vulkan stuff on linux atm is more "why not?"
<i509vcb>
I guess the most relevant thing I can think of coming up is webgpu implementations really liking vulkan more than the subpar gl implementations
<alyssa>
my Steam bias might be showing ;)
mikelee has quit [Ping timeout: 480 seconds]
<alyssa>
anyway, I sure as hell won't look a gift horse in the mouth if you want to contribute to AGXV =)
ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
mikelee has joined #asahi-gpu
fenix144 has quit [Remote host closed the connection]
fenix144 has joined #asahi-gpu
fenix144 has quit [Ping timeout: 480 seconds]
fenix144 has joined #asahi-gpu
mrhh has joined #asahi-gpu
mrhh has quit [Ping timeout: 480 seconds]
ourdumbfuture has joined #asahi-gpu
mikelee has quit [Ping timeout: 480 seconds]
mikelee has joined #asahi-gpu
nela has joined #asahi-gpu
mikelee has quit [Ping timeout: 480 seconds]
ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
mikelee has joined #asahi-gpu
Z750 has quit [Quit: Ping timeout (120 seconds)]
Z750 has joined #asahi-gpu
c10l5 has joined #asahi-gpu
c10l has quit [Ping timeout: 480 seconds]
crabbedhaloablut has joined #asahi-gpu
fenix144 has quit [Remote host closed the connection]
pyropeter3 has joined #asahi-gpu
pyropeter2 has quit [Ping timeout: 480 seconds]
dylanchapell has quit [Ping timeout: 480 seconds]
mikelee has quit [Ping timeout: 480 seconds]
mikelee has joined #asahi-gpu
fenix144 has joined #asahi-gpu
mikelee has quit [Remote host closed the connection]
mikelee has joined #asahi-gpu
mikelee_ has joined #asahi-gpu
mikelee has quit [Read error: Connection reset by peer]
fenix144 has quit [Ping timeout: 480 seconds]
mikelee_ has quit [Remote host closed the connection]
mikelee has joined #asahi-gpu
tertu2 has quit [Ping timeout: 480 seconds]
mikelee has quit [Remote host closed the connection]
mikelee has joined #asahi-gpu
le0n has quit [Quit: see you later, alligator]
le0n has joined #asahi-gpu
mikelee has quit [Remote host closed the connection]
mikelee has joined #asahi-gpu
tertu has joined #asahi-gpu
jlco has quit [Ping timeout: 480 seconds]
bisko has joined #asahi-gpu
<ella-0>
alyssa: yeah I think that means I'm unblocked so throwing together a repo would be cool :3
cylm has quit [Ping timeout: 480 seconds]
mikelee has quit [Remote host closed the connection]
mikelee has joined #asahi-gpu
user has joined #asahi-gpu
user is now known as Guest6593
Guest6593 is now known as ldev
cr1901 has quit [Ping timeout: 480 seconds]
chadmed has quit [Remote host closed the connection]
<lina>
"gl"scopeclient only has a Vulkan renderer? ^^;;
bisko has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
mikelee has quit [Ping timeout: 480 seconds]
<rowanG337>
Yes they moved to vulkan only when implementing ngscopeclient
<rowanG337>
I asked one of the main devs about it and this is their answer:
<rowanG337>
So, prior to the middle of last year we required opengl 4.3
<rowanG337>
We're moving away from GL due to a combination of poor performance and portability problems, especially on native macos
<rowanG337>
(apple doesnt like open standards apparently)
<rowanG337>
As of now, glscopeclient (which is getting deprecated in the very near future) has vulkan based render and compute backend with a very thin opengl 3.x compositor at the output
<rowanG337>
ngscopeclient, the new GUI, is purely vulkan and does not use GL at all
<rowanG337>
So essentially the name is just a historical artifact at this point.
<lina>
I wonder if it might be a good test case... I can't imagine that kind of program doing anything "super" fancy in terms of graphics API corner cases?
<lina>
So maybe it's worth trying ^^
chadmed has joined #asahi-gpu
<rowanG337>
Yeah that's what I'm expecting. Plus it's always good marketing to get a "real" app working :p
<rosefromthedead>
"sway's gles2 renderer is fine" word on the street is it's actually better because wlr's vk code is still kinda wip
<rosefromthedead>
i think it's just not made very clear because there are a lot of people using vk
mikelee has joined #asahi-gpu
rowanG337 has quit [Quit: Leaving]
mikelee has quit [Ping timeout: 480 seconds]
mikelee has joined #asahi-gpu
i509vcb has quit [Quit: Connection closed for inactivity]
<alyssa>
asahi/mesa:agxv/main will be the new home of agxv
<alyssa>
I'll take care of rebasing that as appropriate, so you don't have to worry about keeping up with the asahi common
<alyssa>
what the driver really needs right now is some tough love working through deqp-vk fails
<alyssa>
I know that's not flashy but you need to eat your dinner before you can have desert
<alyssa>
testing on anything more complicated than a sascha willems demo is going to be counterproductive at this stage, I think
<alyssa>
that goes for that oscilloscope app. when the deqps are passing it'll work. it's not worth trying to bring it up at this point in the mean time
<alyssa>
but yes. you and deqp-vk are going to get real close :}
<alyssa>
good luck
mikelee has joined #asahi-gpu
<alyssa>
12:07 @alyssa | testing on anything more complicated than a sascha willems demo is going to be counterproductive at this stage, I think
<alyssa>
to put this another way -- "I wonder if it might be a good test case." -- deqp-vk has all the test cases that agxv needs for a while
<alyssa>
I know debugging deqps isn't glamorous like showing apps running but it's how high quality drivers are built
<alyssa>
you don't need to look for test cases for agxv. they're right there.
<alyssa>
don't waste your time on !deqp-vk in the meanwhile
<alyssa>
walk before you can run, crawl before you can walk,
<alyssa>
Test case 'dEQP-VK.glsl.conversions.scalar_to_matrix.float_to_mat4_vertex'..
<alyssa>
Fail (Got invalid pixels at sub-case 0)
<alyssa>
just yesterday was the bug of "index draws are broken". way too early to be worrying about real apps to test your work against.
mikelee has quit [Ping timeout: 480 seconds]
<alyssa>
ok, fixed the conversion thing
<alyssa>
(flat shading of matrices was botched)
<alyssa>
re the above though -- I'm not trying to be a downer
<alyssa>
I'm speaking from experience. early panfrost I was more interested in getting apps working than the "boring" CTS work
<alyssa>
and that was a bad idea because it meant everything in the driver was broken and debugging apps was ridiculously hard because where to look? everywhere!
<alyssa>
but debugging CTS cases is straightforward. and then "miraculously" when I got the CTS cases working, the apps worked too.
<alyssa>
no sense in you repeating my mistakes
mikelee has joined #asahi-gpu
hightower3 has quit [Ping timeout: 480 seconds]
mikelee has quit [Ping timeout: 480 seconds]
mikelee has joined #asahi-gpu
hightower2 has joined #asahi-gpu
ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]