00:38 FL4SHK[m]: does Mesa work with a NetBSD kernel?
00:39 FL4SHK[m]: I'm considering making my FPGA based system use at least a NetBSD kernel
00:39 FL4SHK[m]: might do something like GNU userland, NetBSD kernel
01:21 alyssa: FL4SHK[m]: respectfully, if you're building a system on FPGA, go wild, porting Mesa to your kernel of choice isn't the problem here ;)
01:22 FL4SHK[m]: it is an ambitious project.
01:22 FL4SHK[m]: Still though, some of the work is done.
01:22 FL4SHK[m]: the software porting work is what I think will be where most of the work is...
01:23 FL4SHK[m]: The remaining work, that is
01:25 FL4SHK[m]: I'm pretty much spending all my free time on this stuff though
01:25 FL4SHK[m]: maybe I'll finish in the next five years
01:26 alyssa: i know the feeling.
01:26 alyssa: can't believe we're at 3.5yrs of asahi
01:27 FL4SHK[m]: the reason I wanted to use NetBSD was largely because it is apparently easier to port than Linux
01:29 FL4SHK[m]: I could just start with NetBSD to see if I can boot any OS at all though
01:30 alyssa: FL4SHK[m]: I don't think Linux is especially hard to port? but I don't know the BSDs
01:30 FL4SHK[m]: It's not?
01:30 FL4SHK[m]: hmm
01:30 alyssa: BSDs are easier to upstream stuff into, that does seem to be true..
01:30 alyssa: but not super relevant here
01:30 FL4SHK[m]: right
01:31 FL4SHK[m]: I'll look into porting Linux once I've got my virtual memory set up
01:31 FL4SHK[m]: I did write a GCC port
01:32 FL4SHK[m]: it's functional for a lot of stdlib-less code
01:32 alyssa: oh, right, custom ISA. yeah no idea then
01:32 FL4SHK[m]: ah
01:32 FL4SHK[m]: Well, ideally I'd be running Linux anyway
01:33 alyssa: all my experience here is for arm things, yeah no clue for a new arch bringup
01:33 FL4SHK[m]: If I didn't like the process of designing an instruction set and implementing it I'd be going with RISC-V
04:30 lemonzest: Hi, I posted this bug https://gitlab.freedesktop.org/mesa/mesa/-/issues/11398 about a BGR issue and is there any news on a fix for it?
04:32 alice: if there was it would be posted in the issue
09:19 mlankhorst: Guest11047: ping?
12:41 neitherthat: technically of one were to like pc table straight in the formula, the second algorithm is correct from what i offered it's cause length is smaller when the left side components are larger, so compilation time removes the pcoffset from value+index pair and at runtime you fill in the offset then removing constantdistance and adding a value would eliminate the correct variant too..but there
12:41 neitherthat: is a flaw when you batchadd more than one operand they might confuse the returned index for more than one iowrite, because you subtract those in polderan, they come in as smallest first alike, so you might get the io write wrong cause of incorrect order. That is why i said that only invariant is length. But sure you could still passthrough the length where as you return length+value and
12:41 neitherthat: correct the order by accessing x times correct bank to be removed value from, and that is by filling in offset all values+offsets, and i am sure it will correct the order that way too by forwarding the correct index for ir results, so you just have to sequence the io writes, but none of them can be written in batched mode anyways as they are needed for hw to function properly their
12:41 neitherthat: latency is same for any write. It's also nice algo, now you need no data banks at runtime for code execution. That's a perfect example as to what invariance means imo. It's not complex anymore or is it? But what would we select for the range of pc offsets? And how at runtime to get rid of a true dep, latter first is you add inverse value and subtract the max value you get length, and you
12:41 neitherthat: had max+length in memory so subtract length from it and then the max also and you zeroed out the bank. Now the range of pcoffset is max value+pc, where as we do not design banks or tiled configuration, it's just that the pc is compressed one level deeper than the value itself.
14:06 neitherthat: value is called interchangeably, in fact i would need to call in and out for polderan....but otherwise yes it's working well in decoder encoder, add sub instrinsics for compressed format are ready and ultra grouped parallel execution and data multibankaccess too, so i am trying to capture some form to finish the code. It's lot to type from comipiling however i am very close to my dreams.
14:24 DragoonAethis: dwfreed: your banhammer's assistance would be appreciated
14:41 MrCooper: should already be taken care of
15:56 verybigberry: Heh i guess you can not demonstrate your retardation any better than bannhammer parroting you are pretty much what the myth says about russians, as to how big battleground meat they are, so there is always a new suicidal abortion leftover or brainless pigonoid in your "round defense" system, and that comes along with fraud as to how i did make aggression although before the bomb was
15:56 verybigberry: exploded by your tyrans i made not a single life threat warning to you, you just seem to fancy writing your own destiny in form of digging your own hole for the grave, only crap you have committed to the systems and nothing about common sense like as states of the addition you know, such as one operand is in 1024*2 states other in 1024*2 (*2 due to operand variance bottom and upper and
15:56 verybigberry: 1024 multiplied 3 times due to carry) and result in 1024 possible states, and that gives you maximum number of states as max 32bit number after multiplying those, you are rawly retarded human shit, and even your eastern western beef was what you brought up literally in #osdev channel seeing delusions as how you are better than me and all the eastern people. You are tyran terrorists which
15:56 verybigberry: should be placed to gas chamber and that is exactly what they will do probably or something similar to boot you off for good brainless leeches.
16:07 verybigberry: I am sure my dad long since tries the best for me and has belief in me, but all the invested effort went to trashbin cause of your terrorists he wanted to make my life very good and invested round about million dollars to make sure of that, what is just blown up now, and he is not elon musk to afford such things.
16:08 verybigberry: you are very bad and narcisstic people.
16:09 verybigberry: all the commits best of them at kvm and anything else our eastern team committed and we worked it out , you did nothing but ban people.
16:10 verybigberry: don't show me your clueless pigonoid on my territory bullying me and throwing out from my own hotel.
19:06 jenatali: What's the process for updating blake3? I need a build fix for Windows arm64 that was contributed upstream 2 months ago
19:10 jenatali: Oh I see, it's just vendored, cool
22:51 DavidHeidelberg: jenatali: :D I accept flowers and credit cards
22:52 jenatali: DavidHeidelberg: !29971
22:53 DavidHeidelberg: I see, few commits after the release 😉