Nashorn performance
83 views
Skip to first unread message
Tal Liron
Mar 12, 2014, 8:56:37 AM3/12/14
to prudence-...@googlegroups.com
Is it worth running Sincerity and Prudence on JVM 8? What performance benefits can you expect for JavaScript on Nashorn?
The answer is mixed, but actually very promising.
As it stands, the official first release of OpenJDK 8 will offer very bad overall performance for JavaScript in the Three Crickets stack. The reason for this is an inherent limitation in the way that Nashorn caches code in that version, resulting in a lot of recompilation of already-compiled code. And compilation in Nashorn is very slow. It's not impossibly slow for production environments: if you're doing a lot of page caching, as you should, the speed of JS code execution will have a minimal overall effect. Still, why would you choose Nashorn if Rhino performs so much better?
But there's some very good news up the line. OpenJDK 8u20 will fix this issue, finally truly opening up the power of Nashorn for us. The fix has already been committed: I've tested a custom build of Nashorn, and found performance to be twice as fast as Rhino for the same code.
So, the future is bright! But also the present: if you are curious, you can build your own nashorn.jar from source to see these benefits.
I will do my best to convince the Nashorn team at Oracle that this is an important enough bug fix to include in the official release of OpenJDK 8. Wish me luck. :)
The answer is mixed, but actually very promising.
As it stands, the official first release of OpenJDK 8 will offer very bad overall performance for JavaScript in the Three Crickets stack. The reason for this is an inherent limitation in the way that Nashorn caches code in that version, resulting in a lot of recompilation of already-compiled code. And compilation in Nashorn is very slow. It's not impossibly slow for production environments: if you're doing a lot of page caching, as you should, the speed of JS code execution will have a minimal overall effect. Still, why would you choose Nashorn if Rhino performs so much better?
But there's some very good news up the line. OpenJDK 8u20 will fix this issue, finally truly opening up the power of Nashorn for us. The fix has already been committed: I've tested a custom build of Nashorn, and found performance to be twice as fast as Rhino for the same code.
So, the future is bright! But also the present: if you are curious, you can build your own nashorn.jar from source to see these benefits.
I will do my best to convince the Nashorn team at Oracle that this is an important enough bug fix to include in the official release of OpenJDK 8. Wish me luck. :)
Tal Liron
Mar 12, 2014, 10:58:03 AM3/12/14
to prudence-...@googlegroups.com
Good news: OpenJDK 8u20 should be out in the third quarter of 2014, so that's not very far away.
And even before that, there should be preview releases on jdk8.java.net.
And even before that, there should be preview releases on jdk8.java.net.
Reply all
Reply to author
Forward
0 new messages