Performance of WebView in JavaFX

Daniel picture Daniel · May 26, 2012 · Viewed 23.9k times · Source

I have a HTML5 UI and a Java backend and want to avoid rebuilding the HTML ui in plain java, so my idea was to run a local webserver and use a webview to render it in a "native" window. The solution seems to be to use a JavaFX WebView which can be embedded in swing. On the paper it sounds great (especially since they claim to use WebKit which has a significantly better performance for my UI in Chrome / Safari).

It works, but: the performance is INCREDIBLY BAD. Orders of magnitude slower than the same thing in Chrome, Safari (or even the slower Firefox). It is literally unusable (my UI makes quite heavy use of JQuery / JS). Btw, I am using the code from this post.

The question is:

  1. Can anybody second my experience? Am I simply doing something wrong or is this caused by JavaFX and thus "normal"?

  2. Any better idea how to achieve this? I am currently simply firing up the system browser which works but is not as nice (does not look as integrated). 

Answer

jewelsea picture jewelsea · May 26, 2012

I have used WebView quite a bit and usually the performance was perfectly fine and very usable.

  • Html5 compliance is good.
  • JavaScript performance varies but I found it about one third the speed of a recent version of Chrome according to Google's V8 benchmark (which Chrome is presumably tuned against).
  • Rendering performance didn't seem to be much of an issue.
  • Very intensive HTML Webapps such as some in the Chrome experiment library did not start up as quick as in some other browsers.
  • WebGL is not supported so sites which fall back to software rendering of graphics from WebGL are much slower.

The biggest issue I had with WebView is that it is not quite as stable for cutting edge features and intensive use as other browsers, but not really any serious performance issues.

Here are a few benchmark stats (WebView version used was sourced from JavaFX 2.2 build 9):

Compliance

Running a html5 test to test html5 compliance (scores out of 500):

Chrome 19      402 + 13 bonus points
Firefox 12     345 +  9 bonus points
WebView 2.2b9  296 +  7 bonus points 
IE 9.0.6       138 +  5 bonus points

Running an acid3 test, webview scores 100/100 same as the other test browsers, but, like IE9, the final rendering has a slight imperfection.

Javascript

Sunspider Javascript benchmark (lower is better):

IE 9.0.6       146.7ms 
Chrome 19      151.5ms
Firefox 12     185.8ms
WebView 2.2b9  199.5ms 

Google V8 Javascript benchmark (higher is better):

Chrome 19      15323
Firefox 12     9557
WebView 2.2b9  5145
IE 9.0.6       3661

Mozilla Kraken Javascript benchmark (lower is better):

Chrome 19      2416.8ms
Firefox 12     2112.9ms
WebView 2.2b9  7988.9ms
IE 9.0.6       9403.0ms

Intensive Canvas

Spinning 3D buddha (higher is better):

Chrome 19      60fps
Firefox 12     43fps
IE 9.0.6       16fps
WebView 2.2b9  7fps

JQuery

JQuery test suite execution (lower is better):

Chrome 19      21826ms
WebView 2.2b9  22742ms
Firefox 12     23554ms
IE 9.0.6       28247ms

Based on the above benchmarks (run on my Windows 7 desktop), as long as WebView is stable and functional enough for you, then performance of WebView vs other browsers should not be an issue (as long as your app does not feature a lot of 3D spinning buddhas . . . :-).

Update

As uta's answer states, the JavaScript JIT compiler is switched on for the Win 32 bit build of JavaFX 2.2 and off for the Win 64 bit build of JavaFX 2.2. This means that WebView JavaScript benchmarks are significantly slower (typically 4 to 5 times slower) when running the 64 bit JavaFX version vs the 32 bit JavaFX version.