01:54karolherbst: robclark: https://gitlab.freedesktop.org/karolherbst/opencl_cts_runner/-/blob/master/clctsrunner.py?ref_type=heads
01:54karolherbst: though for a real CTS run I still use the run_conformance.py file, but I have custom csvs for it
01:55karolherbst: (mostly to deal with spirv and the conformance reqs there)
01:55karolherbst: a full run completes like in 10 hours on my desktop, sooo.. that's managable kinda...
02:29cheako: What do ppl think about https://gitlab.freedesktop.org/mesa/mesa/-/issues/14849 ? There is no one vkd3d project so I kinda don't know where to open an issue, also is that a d3d12 state tracker I see? Why do the debian folks ship it in libgl1-mesa-dri, seems out of place?
15:06alyssa: glehmann: strong work
15:12glehmann: which work? the fp class analysis?
15:31alyssa: yea
15:31alyssa: i'm reviewing now
17:31karolherbst: that reminds me, I wanted to port uub towards 64 bits
18:43mlankhorst: airlied: https://paste.debian.net/hidden/0f099b18 not too evil I hope?
18:45mlankhorst: I'm trying to fix the hardware programming of intel/display on PREEMPT_RT, best I can do is not take any locks during the critical hardware programming part so that interrupts can stay disabled even on PREEMPT_RT, and no locks are used. Hopefully the last part also makes the PREEMPT_RT users happy, since it removes the need for raw spinlocks on xe.
19:32glehmann: nir_alu_srcs_equal returns false for op(+0.0, +0.0), but nir_alu_srcs_negative_equal returns true, amazing
19:33alyssa: glehmann: ÉÉ
19:33alyssa: er
19:33alyssa: glehmann: ??
19:34Hazematman: the curse of Canadian bilingual layout
19:34alyssa: Hazematman: literally
19:35glehmann: well, if the 0s are different defs, nir_alu_srcs_equal returns false, but nir_alu_srcs_negative_equal special cases constants, and uses a floating point compare so -0.0 == +0.0
19:35alyssa: glehmann: thanks I hate it :clown:
19:37glehmann: totally didn't just waste an hour on that
19:37alyssa: you're so real for that though.