3) We have to check that we are not (de)serializing an AVL each time we
need to manipulate a tree. I may be wrong, but this may be the case (another
reason why I need to review the code)
No problem let's work it together - should be fun.
Not sure that we can get a lot of uimprovment in this area :
performances are just excellent.
I did some profiling today, and even if global perfs for bbb are 30%
slower than those we got on trunk (3220 req/s compared to 5000 req/s),
there are many places where we can easily make it faster. For instance,
we are cloning 3 times an entry when only once is needed (it cost a hell
lot of time). We have many other potential waste of CPU in the new code.
So, bottom line : great job Kiran !
--
--
cordialement, regards,
Emmanuel Lécharny
www.iktek.com
directory.apache.org
