Orthogonal persistence implemented with non-volatile memory (NVM) allows the programmers to easily create persistent containers, which are container data structure preserved even after the process terminations due to a system crash. However, all such implementations use heavy-wait write barriers.

We propose techniques to reduce the overheads arising from these write barriers. Our proposed techniques include two approaches: (i) light-weight writer barriers: we avoid the expensive \MFENCE instructions in the hot path of the write barrier, and instead, we let the thread attempting to make an object persistent, wait till this attempt is visible to all the threads by handshaking with them. (ii) eliding redundant overheads: we use a static analysis supported scheme to elide the overheads in the first approach, based on an intuition that the overheads are necessary only when an object being made persistent is accessible by other threads. We implemented each of these approaches in RBP (replication based persistency) implemented in the HotSpot VM of OpenJDK. Our evaluation results showed that these approaches lead to significant gains.