00:00 mohamexiety[d]: It could be worse but like :SufferingInside:
00:01 steel01[d]: Comments on the gp10b reclocking was like 'this works on my desktop card, right?'. Both of which are kinda tame for phoronix.
00:02 airlied[d]: first rule of phoronix, never click the forums
00:09 steel01[d]: Awww, but it's fun to laugh at. 😛
08:04 karolherbst[d]: what RH conspiracy are they talking about this time?
09:02 asuasuasu[d]: microsoft destroyed people learning about TLBs
09:02 asuasuasu[d]: something something
09:11 marysaka[d]: those comments certainly were something lol
09:30 karolherbst[d]: oh right, all kernel devs are incompetent, because RH made it easy to install Linux smh
10:02 ermine1716[d]: If you use red hat you know redhat, if you write everything yourself you know linux something
10:03 karolherbst[d]: real Linux user uses LFS
10:04 x512[m]: LFS is yet another Linux distro, no?
10:04 ermine1716[d]: It's a book that guides you to build a linux os
10:05 ermine1716[d]: Tbh I did it and it was cool and somewhat enlightening experience
10:06 karolherbst[d]: LFS is like gentoo with the only difference that you type all commands by hand
10:07 karolherbst[d]: build commands I mean
10:09 ermine1716[d]: Yes
10:11 ermine1716[d]: And you have a closer look at the components which comprise a linux os
10:11 ermine1716[d]: Also learned a couple of shell tricks from there
10:15 x512[m]: Do LFS contain distro patches? Or pure vanilla upstream sources?
10:45 ermine1716[d]: It has patches
10:48 ermine1716[d]: Anyway, it's cool for educational purposes, but without package manager it's unusable long term
11:23 tzimmermann: ack on https://patchwork.freedesktop.org/patch/677907/?series=155285&rev=2 ?
14:42 mohamexiety[d]: airlied[d]: notthatclippy dakr: (sorry for ping but in case it's a HW thing)
14:42 mohamexiety[d]: sorry for the ping but going back to my notes there's something I am confused about and curious as I feel I might have misunderstood. what prevents configuring large page sizes for GART at this point? putting compression aside; just large pages. here's this scenario: nouveau gets a bo that is marked as `GART | VRAM`. the size and alignment make it a candidate for 64K pages. what would go wrong
14:42 mohamexiety[d]: here? I am assuming that as long as the bo is in VRAM everything is fine. but then suppose it gets evicted to sysmem/GART. what would make things not work?
18:25 airlied[d]: You would likely need a large page allocation on the CPU side
18:26 airlied[d]: Or at least a contiguous set of CPU pages, but I've idea if the hw supports it
18:26 mohamexiety[d]: doesn't this work transparently on practically all CPUs now? or does it need more explicit management
18:28 mohamexiety[d]: but if we put this aside, what prevents nouveau from downgrading the pages in the page table to smaller ones? I heard that this is how amdgpu does it for instance (not sure if this is part of the general ttm infra or driver specific though)
19:40 airlied[d]: the CPU isn't involved though, the pages have to be allocated properly
19:41 airlied[d]: but I've no idea what is possible here, you'd have to look at openrm maybe
20:23 mohamexiety[d]: hmm I see, thanks. I’ll stick to the original plan then and then poke around in the future
23:05 mangodev[d]: hey gfxstrand[d] (hopefully you don't mind the ping)
23:05 mangodev[d]: what exactly happened to https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25916 ? from the activity, it looked relatively fine, but then got marked as draft and hasn't been touched in a while
23:13 gfxstrand[d]: It's not super important and it turned out to be tricker than expected.