00:03 fdobridge: <g​fxstrand> Well, it survived the synchronization tests. Let's see if it survives the rest of the run.
00:07 fdobridge: <g​fxstrand> ```
00:07 fdobridge: <g​fxstrand> DONE!
00:07 fdobridge: <g​fxstrand>
00:07 fdobridge: <g​fxstrand> Test run totals:
00:07 fdobridge: <g​fxstrand> Passed: 406865/3600883 (11.3%)
00:07 fdobridge: <g​fxstrand> Failed: 31/3600883 (0.0%)
00:07 fdobridge: <g​fxstrand> Not supported: 3193983/3600883 (88.7%)
00:07 fdobridge: <g​fxstrand> Warnings: 4/3600883 (0.0%)
00:07 fdobridge: <g​fxstrand> Waived: 0/3600883 (0.0%)
00:07 fdobridge: <g​fxstrand> ```
00:22 fdobridge: <a​irlied> was that on turing btw?
00:30 fdobridge: <g​fxstrand> Yup
00:33 fdobridge: <g​fxstrand> Here's the fails if anyone's interested
00:33 fdobridge: <g​fxstrand> https://cdn.discordapp.com/attachments/1034184951790305330/1161099206379966524/fails.txt?ex=653710e6&is=65249be6&hm=ac2dd598ed77e57935b09e2da214932515910e32fbde2a4128107ef96d4ff36d&
00:35 fdobridge: <g​fxstrand> I'll take care of all the `dEQP-VK.api.*` tests myself. The two that I'd love it if someone took a crack at are the alpha-to-coverage fails and the depth bias ones.
00:35 fdobridge: <a​irlied> I've looked at the depth bias ones twice now and gotten nowhere, but that was with codegen
00:35 fdobridge: <g​fxstrand> NAK won't change anything
00:35 fdobridge: <g​fxstrand> I've looked at them a few times, too.
00:36 fdobridge: <g​fxstrand> I should be able to knock out the easy ones tomorrow. But those two groups might be annoying.
00:36 fdobridge: <g​fxstrand> The YCbCr fails are us not supporting linear. I think the answer is we disable YCbCr for the CTS run.
00:37 fdobridge: <g​fxstrand> I've got a branch somewhere for linear texturing but it needs some love. Probably not hard but YCbCr is totally optional so I'm fine shutting it off for now.
00:38 fdobridge: <g​fxstrand> @marysaka has a patch for the image qualifiers tests. I gave her some feedback and I'll probably merge that in the morning.
00:39 fdobridge: <a​irlied> are you going to merge NAK once the GS is done or wait to try and pass conformance?
00:39 fdobridge: <a​irlied> or is conformance passing more of an XDC goal?
00:49 fdobridge: <g​fxstrand> More I'll think about merging NAK after XDC and the F2F.
00:49 fdobridge: <g​fxstrand> There's some NIR bits that need to go in and some build system arguing to do.
00:49 fdobridge: <g​fxstrand> I don't want to deal with that in the next 3 weeks.
00:50 fdobridge: <g​fxstrand> Looks like the shader object fails are CTS bugs. Fun!
00:50 fdobridge: <a​irlied> oh no, you update CTS to fix those bugs, and you get a whole bunch of new ones 😛
00:51 fdobridge: <a​irlied> all of my attempts to get newer conformance on llvm/lavapipe recently have ended because I file the CTS bugs and forget about it for 6 months
00:57 fdobridge: <g​fxstrand> Yeah....
00:57 fdobridge: <g​fxstrand> I can also hack Mesa to make it pass. 😅
01:10 fdobridge: <g​fxstrand> Those fails I can hack around, anyway.
01:10 fdobridge: <g​fxstrand> I probably should start working off a real CTS release branch, though. 🤔
01:53 airlied: TimurTabi: any ideas what ROBUST_CHANNEL_PREEMPTIVE_REMOVAL would mean from GSP, it that like something went wrong, so I'm just blowing it all away?
02:39 TimurTabi: airlied: where are you seeing this?
02:45 fdobridge: <g​fxstrand> Well, we're going to try `vulkan-cts-1.3.7.0`
02:45 fdobridge: <g​fxstrand> Who knows what bugs that's going to turn out to have? 🤷🏻‍♀️
02:49 airlied: TimurTabi: when the gsp crashes with those oom errors, it first kills all the channels with error 45
02:49 airlied: which translates into ROBUST_CHANNEL_PREEMPTIVE_REMOVAL
02:49 airlied: https://paste.centos.org/view/6284d415 is the dmesg
02:55 airlied: TimurTabi: I'm going to try and crash things there so I can pull logrm buffer before it fills up with other stuff
03:41 TimurTabi: airlied: also increase the logrm buf to 256k
04:06 fdobridge: <!​[NVK Whacker] Echo (she) 🇱🇹> Do you mean RAM or VRAM?
04:09 fdobridge: <a​irlied> GSP's RAM
04:09 fdobridge: <a​irlied> which is a piece of VRAM I think
04:23 fdobridge: <g​fxstrand> Do we just not have it sized correctly?
04:26 fdobridge: <a​irlied> the last thing Ben did was fix the heap sizing calcs to correspond to openrm
04:28 fdobridge: <g​fxstrand> 🤔
05:28 fdobridge: <a​irlied> what does TIR stand for? I think the TIR regs might be involved in the alpha coverage stuff
05:30 fdobridge: <a​irlied> ah target independent rasterizer
09:33 cwabbott: airlied: there's a vulkan extension that corresponds to TIR in DX land
09:40 cwabbott: ah wait, it's a core Vulkan feature - `variableMultisampleRate`
09:42 cwabbott: basically just the idea of doing whatever multisample count you want (even up to 16 samples) if there are no attachments bound
09:45 fdobridge: <!​[NVK Whacker] Echo (she) 🇱🇹> How easy would that be to implement in NAK/NVK?
11:39 fdobridge: <m​ohamexiety> I was working on this but got heavily sidetracked by something else, sorry. .-.
11:39 fdobridge: <m​ohamexiety>
11:39 fdobridge: <m​ohamexiety> When is the deadline?
13:45 fdobridge: <g​fxstrand> There isn't really a deadline. For now, I'll just shut off ycbcr for the CTS runs.
13:47 fdobridge: <g​fxstrand> In theory, probably not hard. We just need to figure out what registers we're not setting. 🙃
14:51 fdobridge: <m​ohamexiety> got it. I'll try to speed through it these days. sorry again
14:51 fdobridge: <g​fxstrand> No worries. It's fine.
14:51 fdobridge: <g​fxstrand> I'm not super worried about that one.
15:21 fdobridge: <g​fxstrand> Ugh... Of course these draw indirect barrier problems don't exist on Maxwell. 🙄
15:28 fdobridge: <g​fxstrand> Got it. Needed `MME_DMA_SYSMEMBAR`
16:26 fdobridge: <g​eorgeouzou> I tried to implement variableMultisampleRate in https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24621#note_2096567
16:26 fdobridge: <g​eorgeouzou>
16:26 fdobridge: <g​eorgeouzou> Maybe there is a better way to do this correctly?
18:25 fdobridge: <g​fxstrand> Alright... Should be down to just depth bias and alpha-to-coverage.
18:25 fdobridge: <g​fxstrand> Maybe I'll start a run and go buy groceries
18:33 fdobridge: <c​onan_kudo> that's a good plan
18:51 fdobridge: <!​[NVK Whacker] Echo (she) 🇱🇹> Hopefully it isn't like this: https://youtube.com/shorts/PO4k2Svd92s
20:27 fdobridge: <m​arysaka> I tried to poke those earlier today without much success 😅
21:08 fdobridge: <a​irlied> I took a trace of the alpha to coverage and didn't get any enlightenment other than the TIR regs might be involved
22:30 fdobridge: <g​fxstrand> 😫
22:30 fdobridge: <g​fxstrand> I'll poke tonight or tomorrow
22:32 fdobridge: <g​fxstrand> I suspect TIR as well but I've already poked at them some.
22:33 fdobridge: <g​fxstrand> It could also be that we have to emit a dummy color surface with the sample count
22:33 fdobridge: <g​fxstrand> That occurred to me today
22:34 fdobridge: <g​fxstrand> I've got an idea for how to do that with a not-too-insane macro
22:35 fdobridge: <g​fxstrand> For depth bias... I got nothing. 🤷🏻‍♀️
22:51 fdobridge: <a​irlied> Yeah I think NVIDIA emit a dummy I'll upload the trace later in case it helps
23:10 fdobridge: <a​irlied> https://people.freedesktop.org/~airlied/scratch/nv-traces/alpha-coverage-no-color
23:11 fdobridge: <a​irlied> there's the trace in case it helps
23:13 fdobridge: <a​irlied> okay so GSP crash seems to be am MMU fault on a BAR2 NULL ptr, which doesn't seem healthy, not sure what causes it
23:14 fdobridge: <g​fxstrand> ```
23:14 fdobridge: <g​fxstrand> Passed: 401802/3988868 (10.1%)
23:14 fdobridge: <g​fxstrand> Failed: 8/3988868 (0.0%)
23:14 fdobridge: <g​fxstrand> Not supported: 3587054/3988868 (89.9%)
23:14 fdobridge: <g​fxstrand> Warnings: 4/3988868 (0.0%)
23:14 fdobridge: <g​fxstrand> Waived: 0/3988868 (0.0%)
23:14 fdobridge: <g​fxstrand> ```