10:22karolherbst: gfxstrand: up for reviewing the spirv ByVal MR I've talked about recently or should I find somebody else? https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29896
11:36zamundaaa[m]: Hazematman: EGL doesn't know the scale of the buffer. If the rects are for the surface coordinate system instead of the buffer, that's a bug in application (GTK) code and not in Mesa
11:37Hazematman: zamundaaa[m]: Thanks I figured it could be that too
11:51karolherbst: DemiMarie: the proper way of doing that would probably be a binary kernel module, but then this opens up the question of if "this blob only does HDMI 2.1 stuff and nothing else" is a violation of an NDA, because reverse enigneering that would be trivial probably
12:47DemiMarie: karolherbst: yup, and since there must be a proper ABI between the binary module and the rest of the code, upstream could plug in an open source replacement without even creating a maintenance burden.
12:48DemiMarie: karolherbst: the reason I wanted it in userspace is security, to protect against malicious displays.
13:27Mis012[m]: I guess the reverse engineering could be done on the windows driver just to be extra safe
13:28Mis012[m]: in any case that's not the problematic part, it could be done if someone cared enough
14:00karolherbst: it is indeed the problematic part
14:00karolherbst: from a legal perspective, what's the difference between putting NDA information in a file decoded by unzip (not shipped by AMD) and a blob decoded by awesome_hdmi21_re_tool (not shipped by AMD) :P
14:01karolherbst: if you already put NDA information into code, you also have to make it not simple to extract (see your NDA for further details)
14:02karolherbst: anyway, the tldr is, if AMD releases something which is way tooo easy to extract those bits from, it becomes an AMD issue
14:03karolherbst: anyway, I'm tired of this discussion anyway. If you think it's that simple, then convince the people in question (AMD) to do it. Anything else is just a waste of time.
14:03Mis012[m]: I mean, we don't need AMD to release anything strictly speaking, if that makes the legal case easier
14:03Mis012[m]: the windows driver could be RE'd
14:04karolherbst: yeah, have fun
14:04Mis012[m]: the real question is whether any patents apply and whether HDMI can claim NDA violation for maintaining code written by someone else
14:05karolherbst: and if that answer is: it's okay to do it, who is gonna do it?
14:05Mis012[m]: I'd be surprised if nobody was interested in doing it
14:06Mis012[m]: but in general it seems useful to answer the question of whether it's legally viable
14:07karolherbst: you can always pay lawyers to figure that out and get a "probably" answer
14:08karolherbst: but anyway, one could just ask AMD, because they have those lawyers and I'm sure they already have the answer
14:08karolherbst: anyway, ask AMD maintainers what they think of it, and then you get your answer
14:09Mis012[m]: I thought they were in this channel anyway
14:10karolherbst: I can only repeat what I just said
14:11karolherbst: if you care that much, then you can talk to them to figure out a solution
14:20DemiMarie: I think future cards will just move the problematic bits into firmware.
14:25DemiMarie: That said, if you can use DisplayPort, that is an excellent workaround.
14:26Mis012[m]: it's not that I personally want to use HDMI 2.x, I just want HDMI to not win
14:27Mis012[m]: and depending on the patent situation they may be overreaching a LOT
14:27DemiMarie: My preferred solution to problems like this is regulatory mandates.
14:27DemiMarie: Sadly they don't exist.
14:28DemiMarie: I am referring to "you must allow open source implementations of protocols" kind of mandates.
14:30Mis012[m]: my pet regulation idea is closing the loophole where you can use signature verification to retain control after sale
14:33Mis012[m]: might be impossible to deal with the loophole of "we'll just put everything important in literal mask rom" though
14:39karolherbst: I mean.. so far the only thing actually tried was AMD to ask HDMI if they can publish the source and HDMI said no. Everything else is just speculation and staying away from legal risks. In any case, I'm sure people can discuss this with AMD maintainers on how to move forward with this. It's just pointless if you ask me
14:39karolherbst: people here can continue to discuss this topic for several hours/days/whatever and the outcome will remain the same. So if you really want something to be done with this, then talk with the relevant people
14:41HdkR: Obviously AMD needs to use a closed source kernel module using a kernel driver shim to implement the features :)
14:43Mis012[m]: that might make it easier to RE, but the question still remains if HDMI has any hope of suing over the RE'd code getting upstreamed
14:46HdkR: Nah, just live with a closed source kernel driver if you want HDMI 2.1
14:46HdkR: EZ
14:47Mis012[m]: closed source kernel drivers should not be allowed to exist under GPL
14:48Mis012[m]: forget nvidia, google is leaning into that shit hard with GKI
14:49FL4SHK[m]: if it's any consolation, I intend on making my custom system not have closed drivers
14:49FL4SHK[m]: and I intend to open source everything
14:51FL4SHK[m]: I can't give source code for hard IP because it's built into the FPGA, though
14:52FL4SHK[m]: ideally I'll be able to get custom ASICs made too some day
14:53FL4SHK[m]: Right now the only options I have for custom ASICs are either too expensive (way beyond my means for now), or worse performance than FPGAs
14:53FL4SHK[m]: at least from what I've seen
14:58Mis012[m]: a fully open ASIC seems like a much harder problem than something you can singlehandedly solve
15:04FL4SHK[m]: right, well it likely couldn't be fully open
15:04FL4SHK[m]: but I could at least release the HDL code
15:04FL4SHK[m]: The main core of it
15:46tnt: Anyone knows from memory if RaptorLake has Tile4/Tile64 like Arc or if it's TileY like AlderLake ? I think I heard newer laptop iGPU are more like Arc now but not sure "when" that happenned.
15:49dj-death: tnt: Tile64 is only DG2+ (so not raptorlake, not alderlake)
15:49dj-death: tnt: the iGPU that is similar to DG2 comes with meteorlake
15:50tnt: dj-death: thanks !