[Webkit-unassigned] [Bug 240605] MotionMark's multiply suite may generate incorrect score when requestAnimationFrame runs at 120hz

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=240605

Scott Violet <sky at chromium.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Increase maximum complexity |MotionMark's multiply suite
                   |of multiply Suite in        |may generate incorrect
                   |MotionMark                  |score when
                   |                            |requestAnimationFrame runs
                   |                            |at 120hz

--- Comment #2 from Scott Violet <sky at chromium.org> ---
Change the description to better reflect the problem as there are numerous possible solutions.

I've attached a patch that increases the totalRows. totalRows directly influences the maximum complexity. In other words, this patch effectively increases the maximum complexity multiply runs to. This solution gives more headroom, and effectively fixes the issue. It may make sense to use a slightly higher number, possibly 60.

As the scoring generally assumes 60hz, another possibility is to use Math.max(1000/60, frame_rate) when determining frame-rate for scoring. This is a bit more invasive, but is another possibility. This has the result of smoothing out the frame rates, so it may have other side effects.

A final possibility is to run a number of initial frames to determine rAF, and then score based on how well that frame rate is maintained. This would mean you can not compare scores between Safari and Chrome or Firefox, which is not ideal.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20220518/c587c1cf/attachment.htm>