On 10/19/2012 07:31 PM, Mark Roos wrote: > I see an interesting discussion here on value types. I assume that at > least part of the > intent is to solve the performance hit by using boxed integers. As > such I was thinking > about how that would affect my Smalltalk implementation. > > A simple case for me is the boxing of longs. I use longs everywhere I > need an integer > so both Character and Integer objects use a long to hold values. My > encapsulating > object is a instance of a Java object called RtObject which holds a > reference to a method lookup > array and the value ( a long ). > > What I wonder is how my GWT would work if I use value types? My > current GWT test compares the > lookup array with the assumption that if the lookups are equal (==) > then the method resolution is as well. > It seems like every GWT first needs to test for value types and if > true then to use a shared lookup array. > I can see how Java classes could be handled invisibly but how would I > use them? I would need a way > to get a class tag of some sort that I could use for my own lookup. > Or would I create a Java class for > each value type I use and use the Java class as the test? In any case > it seems like a fast way to check > if its a value type is a minimum.
The idea is that you should change nothing of your GWT code. Just uses the native method that tag your objects or classes to declare them as a value objects/classes, the JIT will do the job. The main issue in your case is that your boxed class should be not mutable. > > regards > mark Rémi _______________________________________________ mlvm-dev mailing list mlvm-dev@openjdk.java.net http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
