08:15karolherbst: pmoreau: I pushed everything I need for a kind of working path here: https://github.com/karolherbst/mesa/commits/nouveau_nir_spirv_opencl
09:35pmoreau: karolherbst: Awesome, thanks! What about the llvm-spirv patches?
10:05karolherbst: pmoreau: nothing useable yet
12:42pmoreau: karolherbst: Unless you’re planning to look into it yourself, I am willing to try to fix it (cause I’ll need to adapt my code as well) and start from wherever you currently are.
12:43karolherbst: ohh, I see
15:45danboyd: has any progress been made on the Firefox resize crash issue? I've been running a bisect but it's slow going.. I can only test about 2 compiles a day.
15:45danboyd: and now my latest bisect isn't compiling...
15:46danboyd: I'm also confused as to why every step of the bisect has been bad so far.
15:47imirkin_: danboyd: someone actually provided a simple repro for a crash with DRI3 + window resizing
15:47imirkin_: i have not had time to repro ye
15:47imirkin_: in large part because i try to avoid crashing my box :)
15:47imirkin_: danboyd: are you using DRI3? if so, try not enabling it and see if it fixes all your woes
15:47danboyd: haha i understand. how do I tell if I'm using DRI3?
15:51pmoreau: danboyd: I think something like `LIBGL_DEBUG=verbose glxgears` should tell you whether it’s DRI2 or 3. It should definitely be mentioned in Xorg.*.log
15:52orbea: Xorg.*.log is not always reliable
15:52orbea: LIBGL_DEBUG=verbose says DRI3 here, xorg says DRI2 :shrug:
15:53danboyd: ok -- yes I am using DRI3
15:54orbea: danboyd: xf86-video-nouveau or modesetting?
16:03danboyd: orbea: not sure...
16:07orbea: DRI3 with the nouveau ddx is problematic for some people
16:07orbea: i'd try either DRI2 + the nouveau ddx or DRI3 + modesetting
16:12danboyd: orbea: I'm attempting to use DRI2 now.. I did 'Option "DRI3" "off"' in xorg.conf, but don't think it took. Xorg.0.log makes reference to DRI2, but glxgears still says DRI3
16:16orbea: make an /etc/X11/xorg.conf.d/nouveau.conf with something like this http://dpaste.com/1P8KCXG
16:17orbea: identifier is w/e you want and should be easy to change nouveau to modesetting and DRI2 to DRI3
16:17orbea: or vica versa
16:17orbea: you will need a new xorg session for them to take effect
16:20orbea: danboyd: also if you are using DRI3 you can disable it for individual programs, for example: LIBGL_DEBUG=verbose LIBGL_DRI3_DISABLE=1 glxgears
16:20orbea: good way to test if something is broken or not with DRI3
16:21orbea: also see this link https://www.mesa3d.org/envvars.html
16:21danboyd: orbea: interesting... will the LIBGL_DRI3_DISABLE=1 trick work with Firefox?
16:21orbea: it should
16:24danboyd: um... stupid question, but it occurs to me that I'm running Wayland... I assume that means the xorg.conf.d files are not relevant for me, correct?
16:26orbea: i'd assume so, but I've never used wayland
16:26danboyd: haha Firefox actually crashes a little faster than normal with LIBGL_DRI3_DISABLE=1
16:26danboyd: oh, well
16:27orbea: you could try xorg instead of wayland, I'm not sure wayland is the most supported with nouveua?
16:27pmoreau: I’d assume they remain relevant for applications running with XWayland, but not for native applications.
16:27pmoreau: *native Wayland applications
16:28pmoreau: danboyd: You might want to try the patches mentioned in https://bugs.freedesktop.org/show_bug.cgi?id=104222
16:29imirkin_: danboyd: are you using nouveau or modesetting?
16:29imirkin_: oh. wayland. nevermind. you're stuck with modesetting and DRI3 afaik.
16:29danboyd: imirkin_: nouveau, I believe
16:29imirkin_: if you're using wayland, the X stuff goes through the equivalent of modesetting
16:30imirkin_: (which has nothign to do with setting modes btw)
16:31danboyd: OK -- in 4.14 + xorg + DRI2, I'm not crashing
16:34danboyd: imirkin_: Thanks for the tip re: patches. I will give that a shot
16:34imirkin_: probably for pmoreau?
16:53danboyd: oh .. right. didn't read that carefully enougy. Thanks, pmoreau
22:19mupuf: karolherbst: check this out http://download.nvidia.com/open-gpu-doc/BIOS-Information-Table/1/BIOS-Information-Table.html#BIT_CLOCK_PTRS_V1
22:20mupuf: this is no wonder we have had issues reclocking on some GPUs, we did not know about any of this :o
22:20karolherbst: "Pointer to a script to run after changing clocks" duh?
22:21karolherbst: but mhh
22:21karolherbst: ohh I see
23:20BootI386: I really don't understand why you wouldn't want to crash your box for testing imirkin :)