Proxy.isProxyClass scalability (original) (raw)

Peter Levart peter.levart at gmail.com
Wed Apr 24 08:11:26 UTC 2013

On 04/24/2013 08:58 AM, Peter Levart wrote:

In your revisions, you optimize for 0-interface and 1-interface proxy class. What I hacked up earlier was just to use Class<?>[] as the key (need to make a copy of the array to prevent that being mutated during runtime) that is a simpler and straightforward implementation. I didn't measure the footprint and compare the performance of your versions. *Have you seen any performance difference which led you to make the recent changes?* I developed two different approaches: 1. Key made of WeakReference-s to interface Class objects. Strong points: - no key aliasing, validation can be pushed entirely to slow-path - quick and scalable - less memory usage than original code for 0 and 1-interface proxies

That's the sole reason for optimized: WekaReferece -> WeakReference -> ... -> WeakReference linked list instead of the WeakReference[] array approach. And it's not using any recursion for equals/hashCode, so the peformance is comparable to array approach and it doesn't suffer from StackOverflowExceptions for lots of interfaces.

Weak points: - more memory usage for 2+ -interface proxies Latest webrev: http://dl.dropboxusercontent.com/u/101777488/jdk8-tl/proxy-wc-wi/webrev.02/index.html Comments: I like this one. If 2+ -interface proxies are relatively rare and you don't mind extra space consumption for them, I would go with this approach.

2. Key made of interned Strings (names of interfaces) Strong points: - quick and scalable - much less memory usage than original code for all variations of interface counts and in particular for 0, 1 and 2-interface proxies

The special-cased keys for 1 and 2-interface proxies (the most frequent) make for additional space savings.

Weak points: - key is aliased, so the validation of interfaces has to be done - I tried to do it post-festum with intf.isAssignableFrom(proxyClass) but you say that is not reliable. If the validation is performed on fast-path before proxy class is obtained, using Class.forName, the slow-down is huge. Latest webrev: http://dl.dropboxusercontent.com/u/101777488/jdk8-tl/proxy-wc/webrev.03/index.html Comments: This is the best space-saving approach. With tricks for single and two-interface keys, the savings are noticable.

Regards, Peter

More information about the core-libs-dev mailing list