01:12 gfxstrand[d]: Down to 9 patches
01:13 gfxstrand[d]: I may break up the Z/S patch a bit. Or not. IDK.
04:54 mangodev[d]: airlied[d]: why does this exist 🫠
05:02 mhenning[d]: mangodev[d]: a lot of instructions let you pick whether they sign-extend or zero-extend
05:10 mangodev[d]: ahhhhh okay okay
05:11 mangodev[d]: good to note
05:23 airlied[d]: I think my Z/S patch was one for planes and one for the copy fixes, but I merged it later
12:42 mohamexiety[d]: huh for maint9's queue-less device stuff, is it valid usage to create a queue but... it's just empty
12:58 djdeath3483[d]: queueCount=0 is now valid
13:10 mohamexiety[d]: djdeath3483[d]: yeah but the CTS test is doing queueCount=1 🐸
13:10 mohamexiety[d]: `dEQP-VK.pipeline.no_queues.pipeline_cache.compute`:
13:10 mohamexiety[d]: (gdb) p pCreateInfo->queueCreateInfoCount
13:10 mohamexiety[d]: $1 = 1
13:10 mohamexiety[d]: (gdb) p *pCreateInfo
13:10 mohamexiety[d]: $3 = {sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO, pNext = 0x7fffffffc1f0, flags = 0, queueCreateInfoCount = 1,
13:10 mohamexiety[d]: pQueueCreateInfos = 0x7fffffffc520, enabledLayerCount = 0, ppEnabledLayerNames = 0x0, enabledExtensionCount = 88,
13:10 mohamexiety[d]: ppEnabledExtensionNames = 0x7fffffffb9e0, pEnabledFeatures = 0x0}
13:10 mohamexiety[d]: (gdb) p *pCreateInfo->pQueueCreateInfos
13:10 mohamexiety[d]: $5 = {sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO, pNext = 0x0, flags = 0, queueFamilyIndex = 0, queueCount = 1,
13:10 mohamexiety[d]: pQueuePriorities = 0x7fffffffc4ac}
13:13 djdeath3483[d]: Hmmm... Not what I remember testing on gerrit...
13:13 mohamexiety[d]: it's public now too fwiw https://github.com/KhronosGroup/VK-GL-CTS/commit/b2694325f28586f9150b39b692033e002058a153
13:15 mohamexiety[d]: thing is I am not sure if this is a CTS bug or this is correct
14:53 djdeath3483[d]: Will try to retest
17:08 karolherbst[d]: so the unigine heavens ambient occlusion regression is caused by this: https://gitlab.freedesktop.org/mesa/mesa/-/commit/dae57e184aafdd7da562cb3120d530504a2426fc
17:09 karolherbst[d]: maybe I should check the r600 MRs refrencing it, because we might hit some of the same issues 🙃
18:24 tiredchiku[d]: NVK-git might be busted atm
18:24 tiredchiku[d]: at least on ampere
19:03 mhenning[d]: how so?
19:10 tiredchiku[d]: for me it doesn't bring up SDDM post boot
19:10 tiredchiku[d]: and
19:11 tiredchiku[d]: https://discord.com/channels/1033216351990456371/1033216351990456374/1388534803615187025
19:11 tiredchiku[d]: tbf I can't even swap tty post boot
19:11 tiredchiku[d]: reverting to an older build makes it work though :doomthink:
19:27 mhenning[d]: which older build works?
19:27 mhenning[d]: any chance you could bisect?
19:40 tiredchiku[d]: mhenning[d]: could do tomorrow, yeah
19:41 tiredchiku[d]: mhenning[d]: `f30214c1ef6` at the least
20:10 gfxstrand[d]: It's possible some of the Blackwell patches I landed screwed something up. The full branch passes the CTS, though.
20:11 tiredchiku[d]: I'll try the full branch before bisecting then
20:11 gfxstrand[d]: But if you can get a bisect it's probably an easy fix
20:21 tiredchiku[d]: :meowsalute:
21:05 djdeath3483[d]: mohamexiety[d]: Not sure what you think is wrong in that gdb printout
21:05 mohamexiety[d]: shouldnt it be queueCount = 0?
21:06 mohamexiety[d]: (same for queueInfosCount)
21:07 djdeath3483[d]: but : queueCreateInfoCount: uint32_t = 1
21:08 djdeath3483[d]: I mean the tests calls vkCreateDevice 3 times
21:08 djdeath3483[d]: once with queueCreateInfoCount=1, queueCreateInfoCount=0, queueCreateInfoCount=1
21:09 djdeath3483[d]: the dump you shared doesn't seems to be the one with queueCreateInfoCount=0
21:09 mohamexiety[d]: huh, didnt notice the others. hm
21:09 mohamexiety[d]: the other thing I noticed is that the queue it created is entirely empty, it's not capable of doing anything. is that proper?
21:11 djdeath3483[d]: doesn't seem to upset the validation layers
21:11 mohamexiety[d]: I see
21:11 djdeath3483[d]: I guess 🙂
21:12 mohamexiety[d]: thanks! ❤️