#
✕ nick:i509vcb
#asahi-gpu · 2024-06-05 22:34
i509vcbActually some of the errors were GLES...
#asahi-gpu · 2024-06-05 22:31
i509vcbSeems like some of the wgpu tests do not like the driver and the machine actually hangs lol
#asahi-gpu · 2024-03-20 19:15
i509vcbneeds to fix the mess I made in a local agxv branch after rebasing, although maybe I should rebase again after 276 since it does a lot
#asahi-gpu · 2024-03-20 19:05
i509vcb276 looks very interesting
#asahi-gpu · 2024-03-20 18:56
i509vcbkidplayer666: this I think? https://github.com/orowith2os/mesa-asahi-flatpak
#asahi-gpu · 2024-03-20 18:53
i509vcbI think someone had a hack to make it work?
#asahi-gpu · 2024-03-20 18:53
i509vcbbecause of said unstable kernel uabi
#asahi-gpu · 2024-03-20 18:53
i509vcbI think it's that the flatpak platform image doesn't have asahi inside
#asahi-gpu · 2024-02-27 18:04
i509vcbokay then I guess I'll try to fix it and poke you later
#asahi-gpu · 2024-02-27 18:04
i509vcbpretty much poking about putting the rebase on agxv/main
#asahi-gpu · 2024-02-27 18:03
i509vcbbut does need a bit of fixing still for anything else
#asahi-gpu · 2024-02-27 18:03
i509vcbit does launch vkcube
#asahi-gpu · 2024-02-27 18:03
i509vcbwell I did the rebase like a week ago
#asahi-gpu · 2024-02-27 18:03
i509vcbagxv rebase
#asahi-gpu · 2024-02-27 18:00
i509vcb(rebase-10010 branch on my mesa fork if you forgot)
#asahi-gpu · 2024-02-27 17:56
i509vcbalyssa: poke for agxv rebase (yes it doesn't work on anything other than vkcube but I need to find time to get that dealt with)
#asahi-gpu · 2024-02-18 15:53
i509vcbFor https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_EXT_map_memory_placed.html would anything abi related need to be done or does it just involve editing what agx_bo does?
#asahi-gpu · 2024-02-16 03:58
i509vcbshort of writing a library to speak the wire protocol, but that's honestly more involved than I want to go
#asahi-gpu · 2024-02-16 03:58
i509vcblibwayland-server doesn't give you those and wayland-rs can't expose the buffers without some compat footgun
#asahi-gpu · 2024-02-16 03:57
i509vcbI don't have access to the underlying client buffers unfortunately
#asahi-gpu · 2024-02-16 03:53
i509vcbwayland compositors generally sleep the vast majority of the time anyways
#asahi-gpu · 2024-02-16 03:52
i509vcbI'll probably be able to meet the frame budget for 240Hz easily even with epoll
#asahi-gpu · 2024-02-16 03:52
i509vcbepoll/kqueue is probably fine enough for me
#asahi-gpu · 2024-02-16 03:49
i509vcbchergert: although since I wanted to avoid a hard dep on tokio (smol is actually quite helpful to have around) I have to resort to some hand written future implementations... https://codeberg.org/i509vcb/aerugo/src/branch/master/crates/aerugo/src/lib.rs#L551-L619
#asahi-gpu · 2024-02-16 03:23
i509vcb(I don't know why I like doing these frankly insane projects)
#asahi-gpu · 2024-02-16 03:23
i509vcband then put in the killing blow by doing async rust as well for that...
#asahi-gpu · 2024-02-16 03:21
i509vcbgpu also means I can do frankly dumb things like write a wayland compositor that's almost entirely gpu driven because why not do stuff that makes no sense on the gpu on the gpu
#asahi-gpu · 2024-02-15 23:08
i509vcb(for the clamp value)
#asahi-gpu · 2024-02-15 23:07
i509vcbtelling lowering to output nir_imm_int(b, UINT32_MAX) just results in black, so it's probably not that
#asahi-gpu · 2024-02-15 22:55
i509vcbwith disassembler: https://gist.github.com/i509VCB/c55b4a934dda40e81a1808f73959d2ff
#asahi-gpu · 2024-02-15 22:52
i509vcbsame as depth bias
#asahi-gpu · 2024-02-15 22:52
i509vcbyup that's a memory leak, the number of scissors is never trimmed and it's in a dynarray
#asahi-gpu · 2024-02-15 22:50
i509vcbScissor count seems to increase every few draws
#asahi-gpu · 2024-02-15 22:41
i509vcbunless the example is doing that for some reason
#asahi-gpu · 2024-02-15 22:41
i509vcbno...
#asahi-gpu · 2024-02-15 22:37
i509vcbI have no idea what I should be looking for in agx decode dumps
#asahi-gpu · 2024-02-15 22:37
i509vcbhttps://gist.github.com/i509VCB/6c8e60fe7cdd996f6642469beee3248e triangle example in sascha willem's examples, did set clamp to infinity and output is just black
#asahi-gpu · 2024-02-15 22:20
i509vcbHmm setting clamp to infinity I guess shows more problems still exist in agxv
#asahi-gpu · 2024-02-15 21:58
i509vcbagxv_CmdBindVertexBuffers2 seems like where I'd do that?
#asahi-gpu · 2024-02-15 21:58
i509vcbNot sure where I'd get the values to upload in sysvals to give load_attrib_clamp_agx something to load other than 0
#asahi-gpu · 2024-02-15 15:33
i509vcbfine to push that
#asahi-gpu · 2024-02-15 15:31
i509vcb https://gitlab.freedesktop.org/i509VCB/mesa/-/commits/rebase-10010
#asahi-gpu · 2024-02-15 15:31
i509vcbalyssa: rebased agxv is on my rebase-10010 branch
#asahi-gpu · 2024-02-15 14:07
i509vcbI'll squash the two extra commits in like an hour
#asahi-gpu · 2024-02-15 14:07
i509vcbvkcube works now
#asahi-gpu · 2024-02-15 14:06
i509vcbah no wrong order of ppp fields
#asahi-gpu · 2024-02-15 14:06
i509vcbokay so now just gray output in vkcube
#asahi-gpu · 2024-02-15 02:56
i509vcb(not shown in the branch, setting viewport_count = 1)
#asahi-gpu · 2024-02-15 02:53
i509vcb4 bytes short from gdb
#asahi-gpu · 2024-02-15 02:53
i509vcbClearly I'm not seeing the error, the culprit is probably somewhere here: https://gitlab.freedesktop.org/i509VCB/mesa/-/blob/rebase-10010/src/asahi/vulkan/agxv_cmd_draw.c?ref_type=heads#L657-884

plus ›