10:52 MrCooper: Tom^: it's supposed to be the same timestamp in both events
10:54 Tom^: MrCooper: yeah was more of can i count that the crtc event comes before the page_flip_handler2 event or are they technically the same events only with more control on which i queue for
10:54 MrCooper: they have different semantics, there's no defined order between them
10:54 Tom^: was thinking of either using the crtc event only or otherwise storing the timestamps from it and then using that timestamp in the page_flip_handler
10:55 Tom^: MrCooper: hm ok
10:55 MrCooper: you need to wait for the page flip event before calling the page flip ioctl again, or it may fail with EBUSY
10:56 Tom^: yeah that i know, was trying to get the more accurate timestamp from crtc into my pageflips heh
11:00 Tom^: i do have it sort of working, but that currently relies that the crtc event happends before the page_flip_handler2 on the same sequence tho
11:30 MrCooper: how do you know it's more accurate? How much difference is there? They should literally come from the same value in the kernel
11:39 Tom^: MrCooper: its being truncated off because of being unsigned int sec, unsigned int usec, crtc provides uint64_t nsec so using the page_flip time its chopping off the last few ns , now how important this is im not entirerly sure. was just interested in using as correct as possible timestamps
11:42 Tom^: but yes its the same timestamps, only one being truncated because of data types
13:15 MrCooper: Tom^: the rounding to µs is unlikely to matter in practice
13:15 MrCooper: the actual refresh cycle varies by microseconds around the nominal value anyway
13:31 Tom^: MrCooper: well isnt deadline evasion in gnome factoring in something like 800us meanwhile this truncating can be up to 999us (worst case scenario), i guess it all depends on what the timestamp here is used for.
13:31 Tom^: but fair point, if the refresh cycle can varie that much.
13:31 MrCooper: the truncation is by less than 1 microsecond
13:34 MrCooper: which makes no difference for the deadline timer, the timer fires up to on the order of a millisecond late anyway :/
13:36 Tom^: oh :|