Friday, June 19, 2015

Re: GWT 2.7 performance issue

Any known regressions? We are seeing a small but significant performance regression of about 20%.

Our standard performance monitoring is showing a clear performance regression when running Firefox (ESR)/Linux.

I get the 20% figure from using chrome profiling on Win7. A complex UiBinder/Editor/RequestFactory takes 900ms in 2.6 but 1100ms in 2.7. This is when the browser/app is "warm".

We also think we are seeing delays of up to 400ms with Scheduler.get().scheduleEntry(action)... A lot of our webdriver tests were failing as they would click something then fail due to the delay. A known issue or local weirdness?

Sorry if I've not done my research properly. I'm a bit lost how the move of the issues is going.

Cheers

Sam


On Wednesday, January 21, 2015 at 5:31:05 PM UTC, Jens wrote:
Much to our surprise the test show a quite large performance hit. For example a simple test running three for loops inside each other took almost 50% longer with 2.7 then with 2.6.1.

What work are you doing inside these for loops? If these loops are empty or only do local, useless work it might be a bug in code pruning, e.g. GWT 2.6.1 has removed the for loops from the final output while GWT 2.7 keeps them. You can check that by comparing the final JS output (compile with mode PRETTY).

Can't remember any commit to GWT that might cause a 50% performance drop, especially as GWT now has a new benchmark system in place to detect performance regressions.

-- J.

--
You received this message because you are subscribed to the Google Groups "Google Web Toolkit" group.
To unsubscribe from this group and stop receiving emails from it, send an email to google-web-toolkit+unsubscribe@googlegroups.com.
To post to this group, send email to google-web-toolkit@googlegroups.com.
Visit this group at http://groups.google.com/group/google-web-toolkit.
For more options, visit https://groups.google.com/d/optout.

No comments:

Post a Comment