<div dir="ltr">Hi pekka<div><br></div><div>
<span style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:14px;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-weight:400;letter-spacing:normal;text-align:start;text-indent:0px;text-transform:none;white-space:normal;word-spacing:0px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">>>are the c2p numbers from weston-presentation-shm demo?</span>
<br></div><div>yes. I use this demo. My clock is
<span style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:14px;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-weight:400;letter-spacing:normal;text-align:start;text-indent:0px;text-transform:none;white-space:normal;word-spacing:0px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">CLOCK_MONOTONIC.</span></div><div><span style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:14px;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-weight:400;letter-spacing:normal;text-align:start;text-indent:0px;text-transform:none;white-space:normal;word-spacing:0px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><br></span></div><div><span style="color:rgb(34,34,34);font-family:arial,sans-serif;font-size:14px;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-weight:400;letter-spacing:normal;text-align:start;text-indent:0px;text-transform:none;white-space:normal;word-spacing:0px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline">I check the drm kernel code, I find each vblank event send the last vblank's timestamp.</span></div><div><br></div><div>system time sec msec</div><div><div>[ 263.000526] vblank_event 262 983</div><div>[ 263.017236] vblank_event 263 0</div><div>[ 263.033864] vblank_event 263 17</div><div>[ 263.050533] vblank_event 263 33</div><div>[ 263.067215] vblank_event 263 50</div></div><div><br></div><div>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.</div><div><br></div><div>Best Regards</div><div>Nancy</div><div><br></div></div><div class="gmail_extra"><br><div class="gmail_quote">2018-05-22 15:51 GMT+08:00 Pekka Paalanen <span dir="ltr"><<a href="mailto:ppaalanen@gmail.com" target="_blank">ppaalanen@gmail.com</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On Mon, 21 May 2018 23:06:12 +0800<br>
<span class="">zou lan <<a href="mailto:nancy.lan.zou@gmail.com">nancy.lan.zou@gmail.com</a>> wrote:<br>
<br>
> Hi pekka<br>
> <br>
</span><span class="">> I debug the presentation time recently. I have checked each user call<br>
> drmAtomicCommit get the corresponding pageflip event.<br>
> But some presentation results (c2p) are a little weird. They are 1~5ms or<br>
> -1 ~ -5 ms. some results are normal(8 ~ 20 ms).<br>
> I used to explain the weird result is because pageflip return last frame's<br>
> timestamp, but I can't explain the normal results.<br>
> <br>
> I suspect whether the kernel and user have the same time start point. It's<br>
> just a guess, I have no evidence.<br>
> <br>
> Do you have some tips about the c2p results? Thank you.<br>
<br>
</span>Hi Nancy,<br>
<br>
are the c2p numbers from weston-presentation-shm demo?<br>
<br>
The presentation-time extension has an event to tell the client which<br>
clock domain is being used for the timestamps. This comes from DRM,<br>
Weston attempts to enable CLOCK_MONOTONIC, but if your kernel is<br>
ancient or otherwise weird, the default in DRM is CLOCK_REALTIME.<br>
<br>
Check your weston log for these lines:<br>
presentation clock: CLOCK_MONOTONIC, id 1<br>
presentation clock resolution: 0.000000001 s<br>
<br>
If the resolution is much coarser than this, you have a kernel build<br>
configuration issue.<br>
<br>
If all that checks out and you see the problem even with unmodified<br>
weston, then I would suspect a driver bug, but it's hard to tell.<br>
<br>
You may want to grab a timeline recording and post that compressed. We<br>
can then use wesgr to visualize it:<br>
<a href="https://github.com/ppaalanen/wesgr" rel="noreferrer" target="_blank">https://github.com/ppaalanen/<wbr>wesgr</a><br>
You can start and stop the recording by typing Mod+Shift+Space, 't'. It<br>
writes a file to weston's current working directory and you see it<br>
mentioned in the log as well.<br>
<br>
<br>
The c2p times should really not vary much. On an intel GPU, with 60 Hz<br>
display, the default mode of weston-presentation-shm gives me a steady<br>
23-24 ms c2p. This is the default repaint window setting. If I use the<br>
low latency mode (-p), I get c2p 17-18 ms. (The intel driver can<br>
actually send the pageflip event a little earlier than the timestamp it<br>
carries, because it predicts the vblank when it's sure the presentation<br>
will happen.)<br>
<br>
I also found a crash bug in Weston just now when playing with<br>
weston-presentation-shm, probably related to a patch I recently landed.<br>
<br>
<br>
Thanks,<br>
pq<br>
</blockquote></div><br></div>