Synchronization is expensive!Data is collected with run.tprof and visualized by VPA.
Wednesday, November 26, 2008
How slow is CMPXCHG?
In order to understand this, I profile one of the amino lock-free data structure and get following result:
Synchronization is expensive!
Data is collected with run.tprof and visualized by VPA.
Synchronization is expensive!Data is collected with run.tprof and visualized by VPA.
Subscribe to:
Post Comments (Atom)
3 comments:
Yes, I have same experience. What can we do about that?
We can try our best to reduce the number of synchronization.
There are some successful examples:
1. Biased locking in JVM
2. FIFO Queues by Edya Ladan-Mozes and Nir Shavit, 2004
Thanks! these works are interesting.
Post a Comment