851699 - (browsermark) [meta] Investigate performance on Browsermark

Status: NEW

(Core :: General, defect)

Description

[Jeff Walden [:Waldo]]

It's a whole-browser-engine benchmark that various people are starting to look at.  Relevant technical claims about it from http://www.rightware.com/rightware-unveils-the-industrys-most-disruptive-benchmark-browsermark-2-0/ are:

"""
Browsermark measures browser performance by executing a variety of carefully studied tests in the browser. Browsermark scores are not influenced by Internet connection speed or latency.

Browsermark takes 4 minutes to run during which time it evaluate the browser’s performance with a rendering test, WebGl test, HTML5 video test, HTML5 web worker test, data manipulation tests, DOM operations test and text parsing tests.

Browsermark scores are in operations per second. When a browser (A) scores twice as much as another browser (B) on the same device it means that for a given test either:

* Browser A ran twice as many operations as browser B in the same time period
* Or browser A ran the test in half the time as browser B
"""

I only heard about it today, don't know anything about it or its quality, just heard claims that we're slower on it than some other browsers/engines in some situations.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Note https://twitter.com/gsnedders/status/312196446445244418

But yes, it would be good to analyze what's actually going on here.  In particular, note http://www.tomshardware.com/gallery/Browsermark,0101-376243-0-2-3-0-png-.html (from http://www.tomshardware.com/reviews/web-browser-chrome-25-firefox-19,3459-6.html ).

Comment 2

[Wellington Torrejais da Silva]

Very slow WebGL render: https://webcompat.com/issues/24724
Making everything freeze, browser and OS slow during the WebGL test.

Thanks!