01:20 fdobridge: <a​irlied> @georgeouzou pushed a rework to the tess branch, clipping is a lot happier now
01:21 fdobridge: <a​irlied> only dEQP-VK.clipping.clip_volume.depth_clamp.point_list,Fail is failing for me now
01:21 fdobridge: <a​irlied> I had to introduce a bit of an opt loop to fix things
07:42 fdobridge: <g​eorgeouzou> Nice! I will test it out too!
09:37 karolherbst: daxe: I don't think there is much you can do. This warning/error essentially just mean that the kernel decided to kill the GPU context, and it appears it happens because it appears to be stuck
09:37 karolherbst: _but_
09:37 karolherbst: if you are on an older mesa than 22.3 you might want to upgrade
09:37 karolherbst: though mm..
09:38 karolherbst: being on gentoo you are probably on 23.1 or something
09:38 karolherbst: "fifo: fault 00 [VIRT_READ] at 0000000009da3000 engine 40 [gr] client 09 [GPC2/PE_1] reason 02 [PTE] on channel 9 [02bf35f000 Discord[9017]]" is also I think a real error...
10:42 karolherbst: but also super hard to figure out. Mhh, that's on Turing, maybe I can look into some of the random CTS fails we still have, but I think it's mostly clean of invalid data access these days... I might want to look into it tomorrow
21:29 fdobridge: <g​eorgeouzou> It's much better now, but the loop unrolling breaks some tessellation tests:
21:29 fdobridge: <g​eorgeouzou> dEQP-VK.tessellation.user_defined_io.per_vertex.vertex_io_array_size*
21:31 fdobridge: <g​eorgeouzou> It fixes when I remove nir_opt_loop_unroll or nir_opt_trivial_continues.
21:31 fdobridge: <g​eorgeouzou> But removing these makes the clipping tests assert
21:31 fdobridge: <g​eorgeouzou> It fixes when I remove nir_opt_loop_unroll or nir_opt_trivial_continues.
21:31 fdobridge: <g​eorgeouzou> But removing these makes the clipping tests assert again. (edited)
21:37 fdobridge: <a​irlied> okay probably some other ordering issue, I'll see if I can chase it down today
21:37 fdobridge: <g​eorgeouzou> I could also check it tomorrow