00:00 <zf> doesn't really look like it's ready for review, anyway

00:01 <reinist12> The real pipeline has not much to do with extensive coding, cause all the complex parts were done like ossshading guidance, many glsl compilers today, opengl slow mode code does not make much sense neither internally nor in the spec or library for non-programmable state, but to go over doing the real structuresand methods isn't complex, so it was rather nice to still have the library and have it today as reference, programming is job like any other

00:01 <reinist12> one. cause vulkan uses too similar structures and nothing for compressing the data, it's a bogus anyways, not sure why would you want to have zink vulkan on older gl is available in so many projects without modifications needed to start off with.

00:02 <reinist12> The system provided that it works on the slowest mode, is just amazing still.

00:02 <reinist12> I am a big fan of linux based systems, this is my passion and hobby to use them even.

00:02 <zf> also, aren't there threading problems in the kernel, too? I occasionally get GPFs (and more frequently WARNs), it'd be even more nice to fix those

00:04 <reinist12> SO i am so passionate, determned and even obsessive on that lovely sys, that all those accusals against me also do not seem to even make any common sense, that i could be either stupid or criminal, cause i did never nor will ever use my time to be any of those.

00:07 <reinist12> Opengl as ben widawsky also said coherently enough with my understandings is pretty old and vintage stuff that works nice, but fixes to it ain't hard on any of the hw even the vintage kind where it got the wrong implementations from the start, or something initial working ones.

00:14 <reinist12> All that was started on the wrong foot cause hw vendors did not want to expose their isa so effortlessly in the public structures ,but once it went open source, it never made sense anymore to have that spec.

00:15 <reinist12> and also there are ways to lay out bitcode a performant structures and methods without knowing the hw code exactly in the hashes.

00:16 <reinist12> as the time and situation is critical and we may be facing new troubles soon according to predictions of many, then one viewpoint to finnish it off from here, is for you to also shift to investing the time ratter more wisely in that context.

00:19 <reinist12> like betting a lot to extremer security and performance and power save together in between your groups to defend this evil well. If you do not want to cooperate with me, you can do sane things in your team too.

00:22 <reinist12> so the structures to be filled in on the cpu side are and will mostly probably always be coarse grain ones, where the bits of context for any state is minimally presented with 8bits, but the state compiler needs not to honor this , since the hw has fine grain opcode control path.

00:26 <reinist12> as to me this is not anymore complexity for some time, and no stress involved either around implementations, i am allready diving into some of the deep issues and thought controlling as gaia and how to scifily modernize tech, and solve some of the issues which are looking to be real to be solved.

00:28 <reinist12> And there seem to be suprising about of sensefull engineers involved, making also the landscape very enjoyable to handle to anyone, some experienced engineers in that field.

00:30 <reinist12> When you go over some limits you were afraid to be topping out to be maximum, the satisfaction will come for real people also, it will make you calm and motivated to perform work, and that is what engineers with their knowledge really are.

00:30 <reinist12> none bothers anymore heavily about even abusing someone verbally.

00:34 <reinist12> Rather just those formal work hours and mission set up, solutions seen from the start to get to allready easily, and collect your salary and function entirely also, system has no failures then and so it's quite sane not to trigger any conflicts but instead of arguing just do something else, when it comes to explaining best is to talk the truth, if possible.

00:36 <reinist12> so cheers. So memory allocation via syscalls, you might go deeper into thinking what it really does, cause you actually know what it does, how it is implemented you trap all the latest syscalls very easily even with a script

00:36 <reinist12> to be one hundred percent sure that this is exact implementation of new memory allocator

00:37 <reinist12> without any false positives, for instance where the nature of the code works over network like tcp to start with.

00:40 <reinist12> ioct is device memory and mmap and file based, also device but called a host more oftenly, so cpu

00:41 <reinist12> and this kind of start on the chip happens with calling a software interrupt always, same as for socket

00:43 <reinist12> so IOS even with that defense only and shared library cache blobs merged from libraries together into big blob, they long time ago had a program called Joker

00:43 <reinist12> jokers task was to identify all new syscalls that apple made

00:44 <reinist12> that kind of methods should be well understood, if many want to fiddle with the iphone gpus and cpus among your projects

00:44 <reinist12> how to get such systems very effortlessly running

00:51 <HdkR> Need more people with +o in here

00:51 <zmike> I think that was a new record

00:51 <zmike> 45 mins

00:55 <Lightkey> I see six more that could get +o if they wanted or knew they can. :-)

00:57 <Kayden> sorry folks. I forgot I could do anything about that because I didn't have +o on freenode

01:02 <Lightkey> https://www.tomshardware.com/news/zhaoxin-discrete-gpu Also, S3 is, uh.. back?

07:57 <MrCooper> Kayden: FWIW, quiet can be done directly via ChanServ, no need to get ops first

08:00 <Kayden> nice

13:15 <hanetzer> agd5f: heyo. was wondering if there had been any movement on https://gitlab.freedesktop.org/drm/amd/-/issues/1257