16:16 karolherbst: imirkin: still need that mmt trace from me? I can create on for you in around 6 hours or something
17:13 imirkin: karolherbst: i'm all good
18:49 imirkin: karolherbst: trying to get u_blitter going ... it's a bit slow-going. i've modified nouveau to ALWAYS use it, so there are lots of ... complications.
18:50 karolherbst: uff, I see
18:50 imirkin: basically around depth/stencil
18:50 imirkin: since we can't write stencil
18:50 imirkin: so ... it's a forest of if/else's, and will end up with some custom shaders too
18:50 karolherbst: mhh, sounds a bit annoying
18:51 imirkin: yeah, well the current eng3d code is also annoying, for much the same reasons.
18:51 karolherbst: right
18:51 karolherbst: did you saw that suggestion from kayden?
18:51 imirkin: no
18:51 imirkin: i saw a thing from krh, but it won't work
18:51 imirkin: (not as-is)
18:52 karolherbst: ohh, maybe it was from krh
18:53 imirkin: i made a comment about it in #dri-devel yesterday
18:54 imirkin: anyways, all this is making me feel like this isn't worth it. i think we can just steal a shader or two from u_simple_shaders and call it a day
18:55 imirkin: (and give up on the whole triangle approach)
18:55 imirkin: (and treat multisample a little better)
18:55 karolherbst: imirkin: mhh, context recovery will be painful, for the CTS we don't have to implement anything though
18:55 imirkin: unfortunately i think the u_blitter approach won't work at all for pre-GT215 tesla either
18:55 imirkin: dunno.
18:56 imirkin: or i dunno. anyways - going to play with it a bit.
18:56 imirkin: but i'm definitely souring on u_blitter a bit
18:57 karolherbst: uff... the patch for querying the kernel for a dead channel is on my other machine
18:58 karolherbst: I have an approach now where we can detect dead channels, it's just not as nice
18:58 karolherbst: but it works
18:58 imirkin: ok, well i'd definitely like to work out all this blitter bs
18:58 imirkin: it's been like ... 4 years coming
18:58 karolherbst: sure :)
18:59 imirkin: i've hated the cheating we've been doing since i saw the code
18:59 imirkin: (wow ... 2019 ... more than 4 years, probably then)
18:59 karolherbst: and I'd like to workout that MT and recovery stuff
18:59 imirkin: i think i might be able to make substantial progress on it this weekend
19:01 karolherbst: imirkin: do you think you may find some time to look over the general approach on my mt_fixes_take2 branch? still working through the multi threaded EGL deqp tests
19:03 imirkin: tbh, unlikely in the next 2 weeks
19:04 imirkin: there are a few other things i need to do this weekend
19:04 imirkin: and next weekend i'm traveling
19:04 karolherbst: no worries, maybe in two weeks I figured out all the remaining issues :)