Bug 74573 – [chromium] DelayBasedTimeSource should not change its timebase on late ticks

• Status: RESOLVED
• Resolution: FIXED
• Component: New Bugs

[Nat Duca]

[chromium] DelayBasedTimeSource should not change its timebase on late ticks

[Nat Duca]

Created attachment 119353 [details]
Patch

[Nat Duca]

Grace, we should cherry pick this when it lands or whenever you feel you want it. It dramatically improves smoothness.

[James Robinson]

Comment on attachment 119353 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=119353&action=review

logic looks reasonable, math could be tighter. i think you want something like

if (newTickTarget <= now)
  newTickTarget += ceil((now - newTickTarget)/m_intervalMs) * m_intervalMs;

although i haven't tested that.

R=me on idea

> Source/WebCore/platform/graphics/chromium/cc/CCDelayBasedTimeSource.cpp:128
>  // Here we realize we're more than a tick late. We adjust newTarget to be 16.667 from now, and post a task off that new
>  // target.

i _think_ you should update this bit of the comment as well

> Source/WebCore/platform/graphics/chromium/cc/CCDelayBasedTimeSource.cpp:143
> +    while (newTickTarget <= now)
> +        newTickTarget = newTickTarget + m_intervalMs;

this should be a single multiplication instead of a loop

this might give us really short delays (like 2ms). should we have a floor here, below which we push out to the next estimated vsync?

> Source/WebCore/platform/graphics/chromium/cc/CCDelayBasedTimeSource.cpp:148
> +    if (!delay) {
> +        newTickTarget = newTickTarget + m_intervalMs;
> +        delay = static_cast<long long>(newTickTarget - now);
> +    }

i think you can get rid of this by being more careful with the math earlier

[Nat Duca]

Comment on attachment 119353 [details]
Patch

Incidentally, if I had cancellable tasks from trchen then I could post the task, then cancel it after running the tick to avoid pointless ticks. :>

[James Robinson]

I think you're joking, but I don't see what improvement that would bring - the task would still fire and it'd check a bit to see what to do (nothing). TANSTAAFL.

[Nat Duca]

Created attachment 119541 [details]
Patch for landing

[WebKit Review Bot]

Comment on attachment 119541 [details]
Patch for landing

Rejecting attachment 119541 [details] from commit-queue.

Failed to run "['/mnt/git/webkit-commit-queue/Tools/Scripts/webkit-patch', '--status-host=queues.webkit.org', '-..." exit_code: 1

ERROR: /mnt/git/webkit-commit-queue/Source/WebCore/ChangeLog neither lists a valid reviewer nor contains the string "Unreviewed" or "Rubber stamp" (case insensitive).

Full output: http://queues.webkit.org/results/10902604

[Nat Duca]

Committed r103070: <http://trac.webkit.org/changeset/103070>