Mandy pushed utility methods for substitutability:
http://hg.openjdk.java.net/valhalla/valhalla/rev/a0ee83b59f56
Karen Kinnear:
"Given that we already have existing code using acmp bytecodes that deal with Object/Interface, and given that longer term we will want the vm to handle fastpaths quickly - would it make sense to leave javac alone, put fastpath logic in the JIT and call isSubstitutable for slowpath?"
We should either remove the existing "new" acmp implementation or guard it with a flag.
http://hg.openjdk.java.net/valhalla/valhalla/rev/a0ee83b59f56
Karen Kinnear:
"Given that we already have existing code using acmp bytecodes that deal with Object/Interface, and given that longer term we will want the vm to handle fastpaths quickly - would it make sense to leave javac alone, put fastpath logic in the JIT and call isSubstitutable for slowpath?"
We should either remove the existing "new" acmp implementation or guard it with a flag.
- relates to
-
JDK-8215501 [lworld] javac should accept op== on both refs and values
-
- Resolved
-
-
-
- Resolved
-
-
-
- Resolved
-
-
-
- Resolved
-
-
JDK-8210763 [lworld] Add API to test substitutability for value types
-
- Resolved
-