Restrictions for lock coarsening?

[David Holmes - Sun Microsystems]

Christian Thalinger said the following on 01/09/09 05:52:
> On Thu, 2009-01-08 at 10:59 -0800, Vladimir Kozlov wrote:
>> It was volano benchmark and startup benchmarks set.
>> For example, the regression for NetBeans startup was about 50%.
> 
> Wow!  50% is a lot.  I wonder what happens that the slowdown is that
> big.

During startup a lot of locks get touched by many threads so a lot of 
bias revocation occurs and that requires a safepoint each time. So 
basically the system spends a lot of time at safepoints and that adds a 
lot of overhead.

Also note, as I understand from work I did late in the Java 6 cycle, 
once an object has its biased revoked it remains revoked ie it will not 
participate in biased-locking again. This was done to prevent 
"thrashing" and it also makes things easier to reason about.

David Holmes