Memory Models: A Case for Rethinking Parallel Languages and Hardware by Sarita V. Adve and Hans-J. Boehm
This is a pre-print of the actual version.
The era of parallel computing for the masses is here, but writing correct parallel programs remains far more difficult than writing sequential programs. Aside from a few domains, most parallel programs are written using a shared-memory approach. The memory model, which specifies the meaning of shared variables, is at the heart of this programming model. Unfortunately, it has involved a tradeoff between programmability and performance, and has arguably been one of the most challenging and contentious areas in both hardware architecture and programming language specification. Recent broad community-scale efforts have finally led to a convergence in this debate, with popular languages such as Java and C++ and most hardware vendors publishing compatible memory model specifications. Although this convergence is a dramatic improvement, it has exposed fundamental shortcomings in current popular languages and systems that prevent achieving the vision of structured and safe parallel programming.
This paper discusses the path to the above convergence, the hard lessons learned, and their implications. A cornerstone of this convergence has been the view that the memory model should be a contract between the programmer and the system - if the programmer writes disciplined (data-race-free) programs, the system will provide high programmability (sequential consistency) and performance. We discuss why this view is the best we can do with current popular languages, and why it is inadequate moving forward. We then discuss research directions that eliminate much of the concern about the memory model, but require rethinking popular parallel languages and hardware. In particular, we argue that parallel languages should not only promote high-level disciplined models, but they should also enforce the discipline. Further, for scalable and efficient performance, hardware should be co-designed to take advantage of and support such disciplined models. The inadequacies of the state-of-the-art and the research agenda we outline have deep implications for the practice, research, and teaching of many computer science sub-disciplines, spanning theory, software, and hardware.