Name: rm29839 Date: 10/29/97


On a multi-PE Solaris machine with native threads
it is about *TEN* times slower to call a
synchronized method that calling a regular method.
Even if no two threads compete at all.

For example, on a 4 PE machine, you can therefore
expect a speedup of 4 divided by 10. It is hence
faster to use a 1 PE machine with green threads.
That cannot be true!

My assumption is that you use regular Solaris
OS mechanisms for the locking. Then you have
at least two entries/extis from the kernel for
each sync method.
(Review ID: 14697)
======================================================================