02:17 yshui`: is explicit uniform location + uniform initializer broken?
02:17 yshui`: i am getting weird results with radeonsi
06:01 daniels: Company: panvk is being worked on
06:01 daniels: but it won’t support RK3399 for sure
06:15 Company: so it's slowly getting there, but not there yet
12:00 mairacanal: mahkoh, i'm no vgem expert, but i don't think you supose to attach syncobjs to vgem
12:00 mairacanal: i believe you should use its attach and signal ioctls
12:01 mairacanal: maybe Hazematman knows better
12:08 Hazematman: Hey, I'm not an expert in vgem but it doesn't have an ioctl to attach fences. I haven't used it myself so I don't know the details.
12:08 Hazematman: If you're trying to get syncobj from buffers, you can use the generic dmabuf kernel API to return a sync fd for a dmabuf or attach sync fds to a dmabuf.
12:08 Hazematman: it does have***
12:09 Hazematman: using sync fds is not exactly the same as using syncobjs but for the work I've been trying to do (get sync fd support in lavapipe/llvmpipe) it works
16:26 mattst88: mareko: are you still running drawoverhead?
16:30 mattst88: it immediately SIGABRTs with PIGLIT_PLATFORM=surfacelss_egl for me
16:32 mattst88: ah, looks like I need both PIGLIT_PLATFORM=surfaceless_egl and PIGLIT_NO_WINDOW=1 *and* -auto
16:53 mareko: not much anymore, I have a bunch of benchmarks that I run in an automated manner
16:56 mareko: I think we could replace u_simple_shaders with NIR now
17:11 mareko: kusma: didn't you have a MR for replacing some u_simple_shaders with NIR?
17:23 zmike: mareko: do you have any thoughts on enabling variable name storage with lowered io? I plan to try something next week
17:23 zmike: debugging CTS fails without them is basically impossible
17:53 jenatali: zmike: just don't delete the variables?
17:54 zmike: nah that's not an option
17:56 zmike: I'm thinking use a ralloc hashmap stored on the nir and io intrs can have another index for a name field that can be enabled with an env var
18:05 mareko: zmike: what you are describing sounds good
18:15 zmike: cool
18:20 kusma: mareko: I have not gotten far at all, but there's some cleanup patches here: https://gitlab.freedesktop.org/kusma/mesa/-/tree/blitter-cleanups?ref_type=heads