07:20 MrCooper: soreau: the timing of when the event is sent isn't well-defined, only that the timestamp in the event corresponds to end of vblank / start of scanout
07:21 MrCooper: Company: I haven't
07:23 soreau: MrCooper: I'd expect it's a bit more precise, considering its purpose
07:23 soreau: maybe it works slightly differently on different hw/drivers
07:24 mripard: pq, emersion: if you trust daniels enough that's fine by me :)
07:25 MrCooper: soreau: the purpose is mostly in the timestamp and signalling completion, not the timing of the event itself
07:26 soreau: well the docs could do with a bit of clarification on that front
10:25 pq: soreau, FYI, Intel at least used to send page-flip complete events slightly ahead of time, meaning that their completion timestamp is in the future. This is expected and correct. The timestamp is well-defined, the event emission is not.
10:26 pq: too bad bjorn quit already
10:57 soreau: pq: where is the documentation outlining this though?
11:29 pq: most likely there isn't, like for kernel UAPI in general
11:35 pq: I cannot even find the definition for the timestamp in docs.kernel.org
11:45 soreau: perhaps the key definition is page-flip
12:17 MrCooper: mattst88: https://gitlab.freedesktop.org/mesa/piglit/-/merge_requests/1215 broke piglit spec@arb_clear_texture@arb_clear_texture-float here with radeonsi on amd64
12:50 MrCooper: daniels: re your post in the "[PATCH RFC 06/25] drm/connector: hdmi: Add VTEM EMP generation", does enabling VRR signalling on the HW link and forcing maximum refresh rate result in measurably higher power draw compared to no VRR signalling with identical mode timings? (Wouldn't have expected that)
13:14 mattst88: MrCooper: thanks for letting me know. I'll have a look today
13:24 MrCooper: ty
16:45 daniels: MrCooper: ah, not with identical timings, but my read of the amdgpu freesync_on_desktop mode and what the NV person was saying is that they’ll go higher - you presumably wouldn’t enable VRR at 60Hz?
18:47 orowith2os: not sure if the bridge is sending my messages over; test?
18:47 orowith2os: aye, we're good
18:47 orowith2os: mkay
18:47 orowith2os: just checking
19:17 bl4ckb0ne: is GL_EXT_window_rectangles missing from features.txt? There's some mentions of implementation in the release notes but thats all I can find
20:43 zmike: lots of stuff is missing from features.txt
21:20 bl4ckb0ne: it looks like its supported on every driver though
21:21 karolherbst: I wished we'd have a way to automatically generate that file
21:21 karolherbst: which... shouldn't be _that_ difficult tho..
21:40 austriancoder: I think that is impossible to generate them, as you need actual hw data from e.g. CI DUTs
21:53 Hazematman: Wasn't there a discussion at one point about generating them from DRM shim drivers?
21:53 Hazematman: I think it can be done but maybe not perfectly still
21:59 austriancoder: that could work