19:29karolherbst: imirkin: any ideas why a SPLIT instruction might get removed if the defs are going to be spilled?
19:29karolherbst: I think I am pretty close to have a solution, not a proper but a working one
20:50xerpi: I was looking at nouveau's code and...
20:50xerpi: isn't this line redundant? http://elixir.free-electrons.com/linux/latest/source/drivers/gpu/drm/nouveau/nouveau_bo.c#L970
20:51karolherbst: xerpi: well
20:51karolherbst: it doesn't hurt
20:51xerpi: yeah that's for sure :)
20:51karolherbst: xerpi: ohh actually, look at line 991
20:52xerpi: yea but it's initialized two times before the loop
20:52xerpi: lines 959 and 970
20:52karolherbst: ohh right
20:53karolherbst: yeah, the one in 970 could be removed then I guess
20:54xerpi: cosmetically would be very similar to remove either lines for me, but if you want to reduce the LoC, let's remove the 970 one haha
20:56xerpi: also before reporting that I checked if nouveau_bo_mem_ctxdma modified the reg param, but it only reads from it so maybe it's worth making it const?
20:57Lekensteyn: hi all, do you know whether a HDMI/DP to USB/Ethernet/... exist? something for testing purposes and more portable/affordable than a full monitor?
20:57xerpi: this one is a small function but for big ones having const really helps knowing whether the arg will be modified or not
21:02karolherbst: xerpi: well, feel free to write patches to clean stuff up. It is always welcomed
21:03xerpi: karolherbst, sure! I will gladly do that, it will be my first contribution to the kernel :P
21:05karolherbst: have fun with this then!
21:19karolherbst: imirkin: https://gist.github.com/karolherbst/460c012ae492db73df88d51859c03f42
21:20karolherbst: looks much better now :)
21:20karolherbst: allthough those split spills could be merge, but meh
21:21karolherbst: max gpr count is 7
21:30karolherbst: nice, most of shader-db also still compiles :) guess I will do a piglit/CTS run at home
21:54xerpi: karolherbst, is skeggsb/linux nouveau code in sync with kernel upstream?
21:54xerpi: it looks like it isn't
21:55karolherbst: it depends
21:55karolherbst: but the out of tree repository is the one with all changes
21:55xerpi: if I send patches to the mailing list, which kernel should I base them off?
21:55xerpi: oh I see
22:16xerpi: karolherbst, is there any full kernel with skeggsb/nouveau synced?
22:16karolherbst: uhm, drm-next I think
22:16karolherbst: but there is no guarantee there is any
22:16xerpi: I don't have a nvidia card on my pc so the TTM headers are missing
22:17xerpi: I'll check that
22:35xerpi: karolherbst, how does this look https://hastebin.com/jefujifuno.diff ?
22:36karolherbst: xerpi: looks fine I guess
22:37xerpi: cool, will try to send that to the ml
22:40xerpi: there we go \o/ https://lists.freedesktop.org/archives/nouveau/2017-September/028909.html