Why is Linux thread locking so slow?

[Jeff Sturm]

Tom Tromey wrote:
> I guess it would be possible to extend libgcj so that you can specify
> spinlock or pthread mutex on a per-object basis (via some nonportable
> hack).  I doubt this is really worthwhile.

Or, if we discover some useful (fast, MP-aware) locking code to use in
place of pthread_mutex_lock, we can make it a configurable option for
libgcj, no?

It'd be easier than getting it into the pthread library...

The framework is in place for supporting alternative thread systems
already, so we could (just for example) create a Solaris threads port,
as well, if there were any compelling reason to do so.

-- 
Jeff Sturm
jsturm@sigma6.com