add ability to test order-of-magnitude performance in layout tests

[Ojan Vafai]

add ability to test order-of-magnitude in layout tests

[Ojan Vafai]

Created attachment 64389 [details]
Patch

[Dirk Pranke]

I'm glad these tests work they way they do, but the ChangeLog gives little insight into how or why you're doing these tests. Is it your intent to have the magnitude script be used by various perf tests to test expected performance? Should there be a page on webkit.org or something explaining this? 

Maybe at least a reference to magnitude-perf.js should be included into the ChangeLog?

[Ojan Vafai]

(In reply to comment #2)
> Is it your intent to have the magnitude script be used by various perf tests to test expected performance?

Yes. There's precedent for this kind of testing with dump-as-markup.js.

> Should there be a page on webkit.org or something explaining this? 

Eventually yes. For now, I want to get it reviewed and into the code base and have a few people other than me write some tests to work out the kinks. Then I'll write something up and email webkit-dev.

[Dirk Pranke]

groovy :)

[Ojan Vafai]

Committed r65614: <http://trac.webkit.org/changeset/65614>

[noel gordon]

perf/array-reverse.html seems flaky
  http://test-results.appspot.com/dashboards/flakiness_dashboard.html#group=%40ToT%20-%20webkit.org&tests=perf%2Farray-reverse.html

[Tony Chang]

Marked flaky in http://trac.webkit.org/changeset/104177 .