4.0.0 JNI/reflection invocation 2/4/8 times as slow as in 3.3.3

[Jost Boekemeier]

Hi,

I have a test which compares the execution time of a pure java procedure
vs. JNI/reflection calls.

Under certain conditions the JNI/reflection overhead doubles for each
invocation of the test program since gcc 4.0.0.  This problem only
appears if I call ~200000 reflection methods in each round; when calling
~60000 methods the initial overhead is the same as in 3.3.3 and it
doesn't increase with each invocation.

Does anybody know which change since 3.3.3(*) could cause this odd
behaviour?  


Jost

*I have tried to compare with gcc 3.4.2 (the one that ships with FC3),
but this version crashes in CallAnyMethod or something when I ask a hash
for more elements through its iterator interface.  This bug is fixed in
4.0.0, so I didn't care tracking down this particular problem.