Bug 184499 - MotionMark gives wrong results after spectre mitigations of performance.now()
Summary: MotionMark gives wrong results after spectre mitigations of performance.now()
Status: RESOLVED INVALID
Alias: None
Product: WebKit
Classification: Unclassified
Component: Tools / Tests (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Nobody
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2018-04-11 08:00 PDT by Carlos Alberto Lopez Perez
Modified: 2018-04-11 16:26 PDT (History)
6 users (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Carlos Alberto Lopez Perez 2018-04-11 08:00:30 PDT
After spectre mitigations performance.now() not longer a hight resolution timer. This causes that the default settings of motionmark to give wrong results.

Motion mark supports 3 methods to measure performance as you can see here at the bottom right http://browserbench.org/MotionMark/developer.html

The default should be changed from performance.now() to one of the other methods.
Comment 1 Simon Fraser (smfr) 2018-04-11 09:21:24 PDT
perf.now() and rAF timestamp should be quantized the same way, and I think Date.now() will also be.
Comment 2 Carlos Alberto Lopez Perez 2018-04-11 16:22:50 PDT
(In reply to Simon Fraser (smfr) from comment #1)
> perf.now() and rAF timestamp should be quantized the same way, and I think
> Date.now() will also be.

That makes sense, actually.

I have been testing motionmark with WPE trunk and I have been getting sometimes some weird results with the default settings (like 1 points on some tests with a +-3000% percent of variance).

So, at first I thought it could be related to the spectre mitigations of performance.now() because I couldn't reproduce with the other methods of the benchmark like rAF.

But I'm not able to reproduce this with a webkit build of last trunk on Mac. So it should be something else...

I'm closing this as invalid, then.

Thanks for the comments.