On 11/04/2013 11:56 PM, John Rose wrote:
On Nov 4, 2013, at 1:37 PM, Remi Forax <fo...@univ-mlv.fr
<mailto:fo...@univ-mlv.fr>> wrote:
Note that while this fix is ok now,
it introduced a performance regression, doing reflection on a
functional method of an inner class
is faster than doing reflection on the functional method of a lambda
proxy.
This particular fix didn't introduce a regression, it turned a throw
(of NoClassDefFoundError) into correct but slow execution. Some
keeping score might call that a speedup.
:)
But you are right that there is a performance pothole in the
interoperation between lambdas and reflection.
This implementation RFE, to use method handles instead of code
spinning, would take care of that particular problem:
https://bugs.openjdk.java.net/browse/JDK-6824466
I remember writing a code like that a long time ago, to see if it was
possible,
as far as I remember, the main issue was to be able to say, please compiler
compile this method handle chain into a blob that I can reuse.
— John
regards,
Rémi
