Studying Garbage Collection for Software Transactional Memory

Participant: Nhan Nguyen

Home Institution:
Chalmers University of Technology

Home Country: Sweden

Host: Prof. Marc Shapiro

Host Institution: INRIA / LIP6
Host Country: France

Start Date:

End Date: 2013-03-30

Garbage collection (GC) and transactional memory, two technologies that make programming easier, share several similarities. Garbage collection implementation can leverage transactional memory techniques and vice versa [1]. Garbage collection also plays an important role in software transactional memory. As they often produce a lot of garbage, their performance is influenced by the performance of the collection of transactional garbage. Mutual awareness between GC and software transactional memory implementations can benefit both.
My objective in this STSM is to address the scalability issue of Java garbage collection, in the multicore environment, and to study transactional garbage in Java transactional memory systems. I have previously been working the topic of garbage collection, particularly in old generation collection in Java Virtual Machines as part of my PhD thesis. Meanwhile, the Regal team has been studying young generation garbage collection. Our mutual expertise are complement and we aim at improving the performance and scalability of garbage collectors in NUMA machines. Within the scope of this STSM, I will study the behavior of current old generation collectors in NUMA machines. Together with the Regal team, we propose a NUMA-aware old generation collector and integrate it with the developed young generation NUMA-aware collector. The integrated garbage collector can provide us a base for studying transactional garbage in order to understand the influence of garbage collection on software transactional memory (STM) systems.

Report: here