03:52DavidHeidelberg: karolherbst: around? what can be missing, when these two mentioned files are present? Libclc failed to load. Please make sure it is installed and provides spirv-mesa3d-.spv and/or spirv64-mesa3d-.spv
03:52DavidHeidelberg: pkgconfig/libclc.pc ? since all other files libclc generates are headers
04:09airlied: DavidHeidelberg: would need to see the meson build lines, or use strace to see what it's trying to open
04:11DavidHeidelberg: aaah, I just opened rootfs and I copied symlinks.. on my machine it copies content, in CI just symlink..
04:18airlied: that seems suboptimal :-P
04:51DavidHeidelberg: hmm, now the files are there but error msg haven't changed, the path /usr/lib/clc/ is correct, so...
04:54DavidHeidelberg: everytime I touch CI I'm like "one or two rebuilds and we're done"... and then the meme "Two weeks later" comes to the scene
05:32airlied: DavidHeidelberg: fedora or debian image? feodra would be /usr/lib64/clc at least
05:37DavidHeidelberg: deb one, I got it resolved :) now the last bits
06:27DavidHeidelberg: lumag: so... you won a CI for rusticl+freedreno: Pass: 3760, ExpectedFail: 439, Skip: 82, Duration: 1:12, Remaining: 0 https://gitlab.freedesktop.org/dh/mesa/-/jobs/59272254 pretty stable on both Adreno 618 and 630 (well, it's same base...)
06:28DavidHeidelberg: feel free to use commits placed in this branch https://gitlab.freedesktop.org/dh/mesa/-/commits/20240524-redefine-piglit-test for testing your work :)
06:35DavidHeidelberg: since it doesn't make sense to have 630 and 618 because they're almost identical, 618 could be subsituted for Adreno 660
06:37gfxstrand: Lynne: Eventually, probably. Bigger fish to fry at the moment.
06:37gfxstrand: Lynne: That was RE descriptor buffer
07:28stackelheimer: also I am with steady tolerance, some bitch fucker accusing me for herself being a bitch, harasser and extorter and having lucid/vivid underground visions besides harassing random people with ones fuckparade inside their human trash, I am also so famous, there is much bigger chance that all those underground people will be knocked out than me getting victimized again. I am totally
07:28stackelheimer: ignorant of your public figures working against me hence. and I do not finance mob dreaming courtesans anymore. what underground institutions they produce I care none about.
07:42stackelheimer: that I am biochipped I know about there is some surveillance mechanism implanted under my skin, it appears to be very modern ai this time around indeed, whether they cheat defend me or what, is like whatever to me again.
08:59karolherbst: DavidHeidelberg: so, everything figured out?
09:06lumag: DavidHeidelberg, :-D
13:06dj-death: danylo: hey, I'm wondering if when you added json output to u_trace, did you start diffing frames?
13:06dj-death: danylo: in particular for timings of various dispatches etc...
13:11danylo: dj-death https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16914
13:12dj-death: ah yeah I remember now :)
13:12dj-death: thanks
13:12danylo: With recent change to DXVK - it now produces stable results on d3d11 captures
13:13danylo: You'd have to use upstream DXVK with DXVK_CONFIG="d3d11.reproducibleCommandStream = True;"
13:15dj-death: I'm using a different trace format
13:21dj-death: danylo: does your diffing script has a concept of begin/end events?
13:21danylo: dj-death yes
13:21danylo: The compare script has --event-start and --event-end params
13:36dj-death: oh noes, invalid json generated
13:48danylo: dj-death the trouble with json is that if device is created then destroyed, then created again - it would add empty json objects to the output; if app is not properly closed - json also would not be properly closed. Hence compare script could repair json
13:49dj-death: seen 2 end_of_frame back to back
13:50dj-death: how's that even possible
13:51danylo: dj-death mhm, I guess it's because u_trace_context_process is called with eof=true
13:51dj-death: but utctx->start_of_frame at the top should guarantee we generate a new frame
13:52danylo: dj-death try setting eof to false - it should help
14:10dj-death: danylo: yeah
14:44zmike: Hazematman: I pinged you on https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27868 since you seem to have an interest in that space
14:45zmike: I think it would potentially be useful for running compositors? tough to say, but it seems like doing smaller putimage should yield perf gains
14:51Hazematman: Yeah sure I'll take a look at it :) just coming back from vacation right now so it'll have to wait until tomorrow though
14:52glehmann: are opt algebraic patterns with (-)0.0 constant sign sensitive?
14:53glehmann: kind of feels like they aren't, because (('fneu', 'a(is_not_const)', -0.0), ('fneu', 0.0, a)), causes an infinite loop
14:53zmike: Hazematman: I don't think there's any rush :)
15:48mareko: daniels: do you have a pointer to documentation on how to add remote test machines to the Gitlab CI?
15:49mareko: DavidHeidelberg: one interesting topic for CI is adding Coverity validation
15:50zmike: doesn't coverity do its own builds once you hook it up? there's already someone submitting fixes for them
15:51daniels: mareko: hmm, the closest is that https://docs.mesa3d.org/ci/ has links to bare-metal and LAVA; mupuf and eric did a talk about the ci-tron system they've written but I don't think there are any docs on setting your own up as such
15:52daniels: basically bare-metal makes sense only if you want to dedicate machines to GitLab CI and nothing else; LAVA makes sense only if you want to have an external scheduler; ci-tron is somewhere between the two but also extremely container-centric so mostly only practical on high-end hardware
19:41lumag: mripard, what would be the best way to access HDMI's tmds_char_rate from hdmi_codec_ops::hw_params ? Currently msm driver stores pixclock in the global structure, but I think eventually we should get rid of that.
20:14DemiMarie: daniels: how high-end are you referring to?
21:26DemiMarie: If all compositors, X11, and Android supported linux-dma-syncobj-v1, would it be possible to support userspace fences?