I do agree with you that if you are testing a single real-world use case, the performance difference would be hard to measure, but when a put a bunch a use cases together, as it might happen in certain situations, it becomes noticeable.
>From my tests I have found that CgLib is a very good compromise - it is much faster and flexible (you can provide different types of callbacks depending on the situation) than JDK proxies and much simpler than Javassist. -Harish On 5/23/05, Achim Huegen <[EMAIL PROTECTED]> wrote: > I personally don't think, that the performance differences can be > described as some constant factors. It heavily depends on your environment. > Maybe a raw call of a interceptor and service if both do nothing may > vary by a factor of even some hundred percents. > But as soon as your service and interceptor perform some reasonable > business logic, the infrastructure overhead won't be noticeable. > > The main difference between JDK proxies and JavaAssist seems to be > the use of reflection in the first case. I once made some tests > comparing the performance of direct calls with that of reflection calls > and wasn't even able to measure any significant difference in a real > world scenario. > > A hot spot vm is doing a lot of optimization here. > But look here for some other numbers: > http://www-106.ibm.com/developerworks/java/library/j-dyn0603/ > > Achim Huegen > > > Michael Mattox wrote: > > Greetings, > > > > I'm interested in using HiveMind for dependency injection and I'm very > > interested in using the interceptors. I've been researching the > > performance aspects and so far what I've seen has been pretty vague. I've > > read that with JDK 1.5 the JDK dynamic proxy performance is about 1.5 > > times slower. That's a big hit for my application if I'm using these > > interceptors for all calls to all my services. I then read that CGLIB has > > a drop-in replacement for the JDK dynamic proxies but I cannot find any > > performance measurements, especially for JDK 1.5. Finally I read in the > > HiveMind documentation that JavAssist gives a "slight" perf advantage but > > in several instances it was stated that it greatly increases the > > development complexity. I'm curious what the performance increase is > > exactly. If Javassist is only 1.4 times slower, then I think it's not > > worth it. 1.1 times and it may be worth it. I'm hoping to get some > > feedback on this because creating a little prototype will take some time > > although I'm starting to think it may be worth it. > > > > Thanks > > Michael > > > > > > > > -- > > This E-mail is confidential. It may also be legally privileged. If you are > > not the addressee you may not copy, forward, disclose or use any part of it. > > If you have received this message in error, please delete it and all copies > > from your system and notify the sender immediately by return E-mail. > > Internet communications cannot be guaranteed to be timely, secure, error or > > virus-free. The sender does not accept liability for any errors or > > omissions. > > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
