16:01 Wally: On libdrm what is the SIMULATE define used for in pushbuf?
16:13 Wally: As far as I can tell it just returns -EINVAL
18:14 Lyude: Wally: I wonder if it's some magic test thing for skeggsb
18:17 Wally: Lyude: No influence on gallium nouveau?
18:17 Lyude: Wally: Not sure tbh, only way to know is with grep me thinks
18:17 Lyude: anything in particular you're trying to do?
18:18 Wally: Something I dont want to do...
18:19 Wally: Implement https://github.com/NVIDIA/open-gpu-kernel-modules/blob/ce3d74ff6b49f7ec0e5e0aa44417f668b0f7189b/kernel-open/nvidia-uvm/uvm_channel.c#L338 on runtime
18:19 Wally: on userspace*
18:20 Lyude: Wally: writing a driver?
18:20 Lyude: fwiw, there is work ongoing to make nouveau an actually competent driver to develop on, we're currently working on trying to get firmware up and running
18:21 Wally: yes
18:22 Lyude: I think skeggsb also has a skeleton vulkan driver we're hoping to submit soon
18:22 Wally: Lyude: Honestly I dont *really* care, I just want my tu160s clocks above 10mhz
18:23 Wally: so I can watch videos on it
18:23 Lyude: oooooooooooh lol
18:23 Wally: Most people need a semi-usable driver not a competent driver
18:24 Sarayan: Wally: isn't that going to require parsing the new bios structures? They're... rather fun
18:24 i509VCB: How does Mesa generally approach skeleton vulkan implementations? Does it mean you need to at least show a triangle or something closer to functional?
18:24 Lyude: Wally: yeah, at least now we'll finally be able to get there with the firmware… btw, I'm not 100% sure but iirc I thought there was a way to boost clocks at least at boot? i will need to poke skeggsb
18:25 Lyude: i509VCB: tbh with nouveau's I'm not sure how far it is right now
18:25 i509VCB: Well I am asking since I was planning to play with vulkan over at asahi
18:25 i509VCB: And nouveau seems to be another driver that probably knows the process
18:27 Wally: Sarayan: Sorry, for the gallium driver?
18:29 Wally: im just writting a gallium driver for ogk
19:32 Sarayan: Wally: for recloking on turing
19:33 Wally: Sarayan: Ah
19:33 Wally: Im not doing that lel, let ogk handle that
19:34 Wally: Is gpfifo
19:35 Wally: How would I operate gpfifo from userland?
19:35 karolherbst: Wally: what do you mean?
19:36 karolherbst: it's not something you want userspace to touch
19:36 Wally: karolherbst: Yet ogk lets me touch it...maybe?
19:36 Wally: no idea
19:37 karolherbst: well.. like full access or just limited feature set?
19:37 Wally: limited feature set
19:37 karolherbst: well.. I also allow userspace to submit jobs and stuff
19:37 Wally: Just enough to do that
19:37 Wally: https://github.com/NVIDIA/open-gpu-doc/blob/master/manuals/turing/tu104/dev_ram.ref.txt#L510
19:38 karolherbst: right
19:38 karolherbst: that's what libdrm abstracts away
19:38 karolherbst: you create your context, bind your channels/subchannels and queue stuff
19:39 Wally: libdrm_nouveau already abstracts that away...ah
19:39 Wally: what does DRM_NOUVEAU_GEM_PUSHBUF_NEW do then
19:41 karolherbst: DRM_NOUVEAU_GEM_PUSHBUF_NEW?
19:41 Wally: something like that...
19:41 karolherbst: you submit command buffers with it
19:42 Wally: isnt that what the open-gpu-doc was reffering to? submitting commands
19:42 karolherbst: not only
19:43 karolherbst: ehh seems like github messed up linking to the correct line for e
19:43 karolherbst: *me
19:43 Wally: ah