On Wed, 30 May 2018 18:33:40 +0800 zou lan <nancy.lan....@gmail.com> wrote:
> Hi pekka > > I'm not familiar with kernel. I know part of drm driver is implement by > ourself. I check the vblank callback function, it sends pageflip event > first if there is a frame complete, update vblank timestamp later. > > I don't know why the owner write code like this. but if I swap the code > order, the fps decrease a lot. Hi Nancy, I do not know your hardware, so I cannot say why fps decreases, but the order of sending event first and updating the timestamp later does sound wrong to me. I do not know what would be the right solution for your hardware to get the correct timestamp with the event. What are the exact fps numbers you observe and what is your display refresh rate (the video mode)? > I can sure this impact the weston's page flip timestamp. Does this have > some impacts to weston except timeline? Weston schedules a repaint based on the pageflip timestamp. If the timestamp is in the past by a whole refresh period, the computed new deadline will be in the past as well, so Weston will attempt to repaint as soon as possible. The intended behaviour is that Weston repaints at most once for each display refresh cycle, the deadline being 7 ms (if using the default value of repaint-window) before the next vblank. In other words, there should be a significant delay (say, 50-90% of a refresh period) between receiving the pageflip event and repaint. > I study a little about timeline, I think the picture could still show the > c2p(interval between damage commit and hit screen). Is it right? Thank you. That is practically the definition of c2p, yes, if that is your question. c2p uses the timestamp recorded in the client at commit time, while timeline has the timestamp of the compositor processing the commit, so there can be a small difference. Thanks, pq > 2018-05-24 15:42 GMT+08:00 Pekka Paalanen <ppaala...@gmail.com>: > > > On Thu, 24 May 2018 15:26:23 +0800 > > zou lan <nancy.lan....@gmail.com> wrote: > > > > > Hi pekka > > > > > > >>are the c2p numbers from weston-presentation-shm demo? > > > yes. I use this demo. My clock is CLOCK_MONOTONIC. > > > > > > I check the drm kernel code, I find each vblank event send the last > > > vblank's timestamp. > > > > > > system time sec msec > > > [ 263.000526] vblank_event 262 983 > > > [ 263.017236] vblank_event 263 0 > > > [ 263.033864] vblank_event 263 17 > > > [ 263.050533] vblank_event 263 33 > > > [ 263.067215] vblank_event 263 50 > > > > > > I think that's should be the main reason cause the c2p time weird. I > > don't > > > know why kernel return last frame's timestamp. I am still debuging it. > > > > Indeed, that does not sound right. Maybe contact the driver developers > > about this? > > > > Which kernel display driver are you using? > > > > > > Thanks, > > pq > >
pgpQu83UpxgtI.pgp
Description: OpenPGP digital signature
_______________________________________________ wayland-devel mailing list wayland-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/wayland-devel