This is the mail archive of the java@gcc.gnu.org mailing list for the Java project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Memory residence


Anthony Green wrote:

On Wed, 2005-05-25 at 09:39 -0400, Bryce McKinlay wrote:


Unfortunately, Eclipse startup isn't all that good in our experience - as you noticed, it is significantly slower than Sun's JVM. I think we can do better. A lot of the startup time is spent in ld.so symbol resolution, which should be drastically reduced once remaining ABI issues are resolved and we can switch to using mostly private symbols.



Eclipse startup was _faster_ with gcj than Sun or IBM JVMs prior to gcj
4.0.



But also prior to Eclipse 3.0, right?


I suspect the performance regression is related to our new class loading
scheme.



Yeah, there may well be room for optimization here. But Eclipse is now also doing much more work during startup, since it is now loading its own bytecode and send that to ClassLoader.defineClass() before it gets the native class back. This is fairly memory intensive, plus if there are bottlenecks in the code eclipse is using to do this loading (java.nio?) it could be another source of slowness.


Bryce


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]