07:52llyyr: could a radv maintainer look at https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36400 before 25.2 is out? it'd break libplacebo (mpv dep) otherwise
07:55llyyr: host_image_copy is (mostly) hidden behind RADV_PERFTEST=hic but HOST_IMAGE_TRANSFER_BIT is set unconditionally
08:45freizydark: I have basically two options to choose, commercialize my hack and earn everywhere a lot of money or give it out for free and still earn a lot of money, but also earn respect and untouchable state and status and benefits to wipe out devils from around the worled who terrored me in physical space , i do not consider IRC conflicts or misunderstandings/argues into this mission but only
08:45freizydark: physical space terror. So i already chose to earn respect untouchable state and status to handle devils together with my alliance. So you all get my code and real AI out from my labs instead.
09:05freizydark: I suppose everything like bioengineering and low-powering is allowed, cause the computation throughput is unheard we need to place little restrictions officially and formally on the usage concerning military or mass destruction to get permissions to do that, cause such AI in military robotics can be utilized to commit a mass genocide etc.
09:21freizydark: When i look at those mathematics youtube channels of integral formulas of different kind, from light as wave and particle or area approximations, i actually think about the Steve Jobs who was actually trying to find a silver bullet or golden spoon for presenting fonts way back in recent time history , he researched similar divide and subdivide algorithms for the circle and later on they
09:21freizydark: claimed he got the ideal method to do the needed.
09:30freizydark: I looked at the Asahi project, well earlier macs gained very solid support with a series of hacks done for coreboot by some eastern slavic comitter to arrive to there, apic hack went in for emulatinv vt-x and vt-d by similar comitter from eastern hemisphere likely, libinput finally managed to replace commercial synaptic input drivers, all phones can be broken into nowdays where racepwm and
09:30freizydark: libdimentio intrudes, so to sum up the state, we are in computation heaven and in a line of freedom , should we pay attention in commodity systems for security is big dilemma, cause little change into my codes could land to non-hackable firmware and security safety defense heaven too, which is controversial subject actually, maybe overall we do not want that to happen for commodity
09:30freizydark: systems, and stay on freedom line.
09:38freizydark: netwroking heaven is all more like heaven where we want to be, so all the networking is acclerated towards a lightning fast throughput too, but little interesting story about communication satellites, some privately owned company spinflew a sattellite to orbit on geostationary orbit by not using a rocket engine launcher but some centrifugal force spinner.
09:39freizydark: They have huge impact with their tech into high sattellite cosmology type of global cheap coverage with wireless technology, the security around such things is again very obscure or controversial subject.
11:34Curan: Hello everybody! When I built kernel 6.16, I saw a new CMA option for dma-buf and was wondering if there are recommended settings for CMA with amdgpu/radeonsi? Some quick web searches didn't yield anything, but maybe I missed it? Thanks in advance for your reply!
11:35K900: You don't need that
11:35K900: That's for weird embedded devices
11:56pinchartl: K900: that's a matter of point of view. no later than today I was discussing how weird x86 systems are compared to arm :-)
12:12Curan: K900: thanks for the clarification (indeed most hits were for embedded derivatives of Debian and other distributions ☺️
16:38phryk: MrCooper: you implied there's an alternative to using KMS to put stuff on screen – what would that look like? currently, i'm under the impression that KMS is a hard requirement to set resolution and bind a display to a… i want to say framebuffer?^^
16:39emersion: i don't know of another API apart from legacy fbdev (which is mostly just historical at this point)
16:40emersion: Vulkan's display ext can do KMS for you
16:40emersion: (but doesn't expose a lot of features)
16:40phryk: yeah, i did some awful stuff with uvesafb back in the day to render graphs on my tty ^^;
16:41phryk: okay, that confuses me a bit more about what layer vulkan actually occupies, but probably isn't relevant to me as i want to support at least opengl, too.
16:51phryk: alrighty, gbm bo creation works when running the code as root and i learned that "DRM master" is a thing. I think the picture is finally coming together enough that i might be able to render a single tri with drm/kms+opengl some time next week.
16:53phryk: i think there's basically only three things missing: creating a framebuffer from the gbm buffer, creating an opengl surface from one of those (not sure which) and binding the framebuffer to the display…
17:05pinchartl: phryk: are you aware of https://cgit.freedesktop.org/mesa/kmscube/ ?
17:14mareko: I like the organization of CI jobs: 2 groups of build jobs, no container jobs anymore
17:16MrCooper: the container jobs now only exist when they're really needed, mostly when one of the images is modified
17:43javierm: phryk: besides kmscube that pinchartl suggested, there's also https://gitlab.freedesktop.org/daniels/kms-quads
17:43javierm: the readme says "kms-quads is a simple and well-explained example of how to use the Linux
17:43javierm: kernel's KMS API to drive graphical displays"
17:44javierm: and indeed I learned a lot by reading to the code
17:45javierm: *reading the code
18:57phryk: pinchartl, javierm: yes i have been aware for quite some time, but i never could get them to run and just assumed that was because i run freebsd. turns out they don't work when there's already a drm master – managed to get kmscube working earlier today after i groked that i need to actually stop X/lightdm for it to work. ^^;
19:09pinchartl: phryk: good, there's progress :-)
19:11zmike: mareko: if you have ideas for a better spot to prune buffers I can make that change, but I think otherwise we're now all set