19:43imirkin: wtvr, i don't have the mental energy required to worry about this
19:44imirkin: if people are intent on making nouveau suck, i won't stop them
19:44karolherbst: and for 4.6 we probably want to move to nir anyway as even if we would decide on consuming spirv directly, it's so much work that we wouldn't be able to do it anyway
19:44karolherbst: the nir path is quite solid for nvc0
19:44karolherbst: just nv50 is in a worse shape
19:45karolherbst: I can deal with all of this anyway
19:45imirkin: yeah, like i said, i've lost all will to care about this sort of stuff
19:46AndrewR:watches with interest
19:46imirkin: i think gitlab finally finished me off, after a long string of things i hated
19:46karolherbst: imirkin: btw, I might be off on random days, because I still have to deal with the last steps of my relocation, especially finding a flat :D
19:46imirkin: karolherbst: ah ok, you moved?
19:46imirkin: are you still at RH?
19:46karolherbst: currently at my mothers place
19:46imirkin: cool. back to hamburg?
19:46karolherbst: just moving from brno to hamburg
19:47imirkin:is still in nyc
19:47karolherbst: my mother lives in a small city north of hamburg, so the best idea was to just get everything to her and deal with the apartment stuff later :D
19:47karolherbst: but later is now now
19:47imirkin: good luck with the apt stuff
19:48karolherbst: yeah, thanks
19:48imirkin: so for the CL things, what do i have to do to test it?
19:48imirkin: or would i be better off just testing via GLES 3.1?
19:49imirkin: yeah, that might be simpler...
19:49imirkin: will try that.
19:49karolherbst: well, depends on what you want to achieve though
19:49imirkin: i want to make sure the nv50 core works ok
19:49imirkin: opcode emission, buffer support, etc
19:50karolherbst: mhh, so general compute shader support
19:52karolherbst: yeah... I mean there are two parts really: 1. making sure nir to nv50ir is working alright 2. making sure codegen itself is not doing anything stupid. So kind of depends. The nir bits are probably best tested with GL(ES) stuff, but in regards to compute shader support, mhh.. I expect that certain lowering might be missing or something? dunno
19:52karolherbst: it's all disabled by default anyway
19:52karolherbst: one need to set NOUVEAU_ENABLE_CL=1
23:44Lyude: also, I need to figure out if skeggsb has gotten to redoing some of the link training stuff like they wanted to or if I should plan on giving that a shot at some point, since we'll definitely need that for hdr stuff on dp
23:44Lyude: and that might take a while
23:45imirkin: i don't have a HDR DP sink
23:45imirkin: nor do i have a kepler+ GPU with a DP output
23:45imirkin: so ... yeah. i'm pretty much not going to be playing with any useful DP stuff :)
23:46Lyude: also-i bet some of this igt stuff will come in handy for testing various hdr formats on nouveau
23:46imirkin: hdr isn't very easy to test
23:46imirkin: it involves setting infoframe data
23:46imirkin: which is passed through to the kernel as a blob
23:47imirkin: it affects how the colors are interpreted
23:47imirkin: BT.2020 vs BT.701
23:47Lyude: imirkin: I mean-i'm talking about the hdr tests in kms_plane, but also I haven't tried any of those yet (need to update our cairo package in fedora apparently)
23:48imirkin: ah ok. i wonder what those do.
23:48imirkin: where does igt live?
23:48imirkin: nm, found it
23:48Lyude: imirkin: yeah-also I can push my wip branch for nouveau, one sec
23:49imirkin: you mean the stuff in kms_hdr.c presumably?
23:50imirkin: yeah, doesn't seem to be testing a whole lot
23:50Lyude: imirkin: no, kms_plane.c
23:51Lyude: it's got a bunch of tests for the various hdr color formats I believe
23:51imirkin: there's no such thing as "hdr color format"
23:51Lyude: Cairo version too old for XB4H(0x48344258), need 1.17.2, have 1.16.0
23:51Lyude: sorry-this is what I am talking about
23:51imirkin: that's rgba16f
23:51imirkin: which is certainly useful for hdr
23:51Lyude: ah-i think it's just called "hdr formats" somewhere in igt :P
23:52imirkin: but totally works fine with any color scheme
23:52Lyude: but it's probably just mislabeled, meh