00:12
quantum5 has joined #zink
00:40
pendingchaos_ has joined #zink
00:41
pendingchaos has quit [Ping timeout: 480 seconds]
00:43
pendingchaos_ is now known as pendingchaos
02:16
orbea1 has joined #zink
02:17
orbea has quit [Ping timeout: 480 seconds]
02:27
orbea has joined #zink
07:22
caseif has joined #zink
07:24
Plagman has quit [Remote host closed the connection]
12:48
xlei has quit [Ping timeout: 480 seconds]
13:24
xlei has joined #zink
14:16
<
ajax >
Surfaceless platform:
14:16
<
ajax >
zink ERR: 'setupLoaderTermPhysDevs: Failed to detect any valid GPUs in the current config'
14:18
<
ajax >
twisty maze of dlopens here
14:18
<
ajax >
plain vulkaninfo works, 'zink vulkaninfo' fails to find vkCreateInstance
14:19
<
ajax >
this on nvidia so there's probably some glvnd shenanigans involved here
15:08
<
zmike >
ajax: dumb question but I assume you're setting the glvnd var?
15:08
<
zmike >
baryluk: this glmark thing is suuuuper wtf
15:08
<
zmike >
ohhhhhhhhhhh wow
17:40
<
zmike >
baryluk: got it
17:41
<
zmike >
which I think means I've finished all the zink work items for the release \o/
18:48
Plagman has joined #zink
18:49
omegatron has quit [Quit: What happened? You quit!]
18:52
omegatron has joined #zink
19:01
<
baryluk >
zmike: whoh. congrats. take a break for a day dude :)
19:10
<
baryluk >
It is a bit of workaround fix, but if it works then great, it is pretty fast check.
19:46
<
zmike >
I think it only works for radeonsi and iris now because they happen to deallocate in timeline order
19:46
<
baryluk >
trying first items from the bo list
19:46
<
zmike >
zink deallocates in random order
19:46
<
zmike >
so this can never work
19:46
<
zmike >
just been coincidental that it hasn't blown up until now
19:46
<
zmike >
as always, tester prime finds the best bugs
19:47
<
baryluk >
sure. will find some new ones tomorrow, as now I am unblocked ;D
19:47
<
zmike >
just don't update and use main
19:47
<
zmike >
radv is broken for the time being
19:49
<
zmike >
gonna try to get a new wip out in a few...
19:54
<
zmike >
baryluk: new wip up, should have everything all fixed
20:17
<
baryluk >
thx. that is going to be helpful.
20:18
<
zmike >
do you have basemark in your benchmark script?
20:18
<
zmike >
need to figure out what I'm doing wrong in that one at some point
20:20
<
baryluk >
oh. I did not know basemark supports OpenGL. I forgot about it. I will try to add it, but I do not have any data for it yet.
20:20
<
zmike >
it's one of the phoronix benchmarks
20:21
<
zmike >
basically all just buffer allocation and rebinds from what I can see
20:21
<
zmike >
but zink gets dumpstered somehow
20:24
<
baryluk >
I tried to use it, but their license is weird.
20:24
<
zmike >
I just scraped the phoronix benchmark script for all my testing
20:25
<
baryluk >
that is fine.
20:25
<
zmike >
the site and such for that one were a bit odd
20:25
<
baryluk >
I actually had problem with running basemark even on radeonsi. It was crashing my gpu.
20:26
<
zmike >
I'm kinda out of office for the day but if you wanna get a ticket going we can try to figure it out
20:27
<
baryluk >
ok. I see how he launches it in phoronix benchmark script. I will try that. Last time I tried, i had issue with launcher not passing environment variables properly.
20:28
<
zmike >
makes sense
20:38
<
baryluk >
nice. it works good. the generated result report is really nice. it even has all frametimes for each frame.
20:38
<
baryluk >
radeonsi and zink works good.
20:38
<
baryluk >
yes, spending quite a fraction of profile in zink_set_constant_buffer