Test optionating

[deepak1556]

providing options for users to choose test to run on start. Also started work with workers, implemented gc load in workers.

[jdarpinian]

I like the idea of having a way to run just one test for development purposes, but I don't think it's a good idea to have checkboxes. When users run the test, they should always run every sub-test so the total score is consistent. Only developers should be able to choose which sub-test to run.

Instead of checkboxes, how about a query parameter in the URL to specify a specific test to run? Something like this:
http://localhost:5578/latency-benchmark.html?singleTest=2
That would run just the native reference test by itself (tests[2]).

Also, I would prefer to avoid adding a dependency on jquery.

[deepak1556]

that was stupid of me to forget url parsing.. Changed it now! can now use http://localhost:5578/latency-benchmark.html?name=1sep2sep3 ,name and sep doesnt matter..

[jdarpinian]

I don't think this page needs a list of tests. I prefer not to have the list of tests in two places. I'll definitely forget to update one or the other.

[jdarpinian]

Sorry for the slow response, this is my 20% project so I can't spend that much time on it at work. I'm interested in the worker GC test. Have you tried it out and gotten any interesting results? Are there any differences between browsers? In the past I've found it difficult to make a test that will reliably cause long garbage collection events to happen.

[deepak1556]

i will make those changes, for the worker gc test in my 32-bit system the css jank shows a constant number on all browsers but js jank is of this order firefox (1.7) > opera (1.4) = chrome while scroll jank is more in opera (4.5) > firefox (3.4) > chrome (2.3) . I was planning to work on the todo list and complete them, is that fine ? also some guidelines on how you were trying to carry out them before would be useful for me. thanks @jdarpinian !

[jdarpinian]

Those jank numbers are pretty low. Ideally each test should cause more than 10 frames of jank in at least one browser. If all browsers score well, then it's not really worth running the test.

Are you sure the gcload function is actually causing long GC events? In Chrome you can see the length of GC events in the developer tools or in chrome://tracing.

Going down the TODO list sounds good! I think the most interesting items might be image resizing, HTML parsing, and layout. The longer the jank you can cause, the better the test!

[rubenmachuca]

only those who have all the permissions thanks