00:58 fdobridge: <g​fxstrand> So, if I help @marysaka out with 3D, we might be able to pass CTS by XDC...
00:58 fdobridge: <g​fxstrand> IDK if I want to push that hard, though.
01:25 fdobridge: <e​sdrastarsis> Does vkcube work on NAK with the fddx/fddy MR?
03:21 fdobridge: <g​fxstrand> There's more patches needed for it.
03:21 fdobridge: <g​fxstrand> Seriously, what is people's obsession with vkcube? 😂
05:06 fdobridge: <a​irlied> @gfxstrand maybe with some judicious feature and exts disabled
05:06 fdobridge: <a​irlied> There are a few non compiler CTS fails that seem quite tricky
05:12 fdobridge: <g​fxstrand> I fixed 100 or so of those yesterday.
05:13 fdobridge: <g​fxstrand> There's till more so maybe we'll have to disable a few things.
05:13 fdobridge: <g​fxstrand> I'm most worried about XFB, honestly.
05:18 fdobridge: <g​fxstrand> One of these days we need to figure out inverted depth...
05:38 fdobridge: <!​[NVK Whacker] Echo (she) 🇱🇹> 67 crashes 🤔
05:41 fdobridge: <g​eorgeouzou> I could help with that
05:50 fdobridge: <g​fxstrand> Oh, most of those are just that we're missing image atomic compare exchange. There's a guy at Collabora who's supposedly working on that. I need to touch base with him on Monday.
05:51 fdobridge: <g​fxstrand> There's a handful of others that are probably API side issues.
05:51 fdobridge: <a​irlied> If anyone looks at depth stuff, I have unrestricted depth unsolved in an mr as well
05:52 fdobridge: <a​irlied> Which is probably similiar problem
05:53 fdobridge: <a​irlied> I have some traces from nvidia driver but they use a lot of mme macros
05:54 fdobridge: <g​fxstrand> Some of that might be solvable if we used the MME simulator to basically expand them in-place. 🤔
05:54 fdobridge: <a​irlied> I dont have a copy of them, but probably could figure out how to
05:58 fdobridge: <g​fxstrand> At some point, we need to better figure out how to hook things to get more reliable and automatic traces. The thing I typed up is *really* rough.
05:59 fdobridge: <g​fxstrand> Maybe you improved it some? 🤷🏻‍♀️
06:02 fdobridge: <g​fxstrand> A bunch of those are MSAA interpolation issues which I don't feel like trying to fix with codegen. We probably could and arguably should, eventually, but I'd rather figure out how to do it right in NAK and backport than try to sort through the 17 layers of codegen and transitions and GL shader patching and...
06:02 fdobridge: <!​[NVK Whacker] Echo (she) 🇱🇹> What is this secret tone1 emoji? 🐸
06:03 fdobridge: <g​fxstrand> 🤷🏻‍♀️
06:03 fdobridge: <g​fxstrand> I just used my phone keyboard once and now it's in my recent emojis
06:03 fdobridge: <g​fxstrand> `🤷‍♀️:skin-tone-1:`
06:03 fdobridge: <g​fxstrand> Looks like it's that
06:04 fdobridge: <g​fxstrand> Because of course `::` makes sense
07:32 fdobridge: <m​arysaka> We can give it a shot 👍
07:55 fdobridge: <a​irlied> Mine hooks endcommandbuffer works great, just not sure how to get at the mme macros
07:56 fdobridge: <a​irlied> I assume they are uploaded internally and since it's all user command submits it would need to maybe have to work out how to trap on mmio access to the userd page
15:18 fdobridge: <g​fxstrand> I was able to see them by hooking all maps and searching by hand.
15:21 fdobridge: <k​arolherbst🐧🦀> I wonder if there is a way to hook into the userspace command submission stuff
18:58 fdobridge: <g​fxstrand> I'm going to focus next week on the dependency tracker. If you can get the headers reworked, that'll give a base for the rest of it
23:41 fdobridge: <m​aba_kalox> Stupid question, what is MME - I tried to Google it, or figure out what is from a contex (code/discussions in this chat), I was able to understand only that there is some kind of interpreter/VM on nvidia hardware, that can process opcodes, and mme is a set of marcoses, which allows to generate small programs with it?
23:42 fdobridge: <m​aba_kalox> Stupid question, what is the MME - I tried to Google it, or figure out what it is from a contex (code/discussions in this chat), I was able to understand only that there is some kind of interpreter/VM on nvidia hardware, that can process opcodes, and mme is a set of marcoses, which allows to generate small programs with it? (edited)