CACM publishes informed rebuttal: "Why STM Can Be More than a Research Toy"
Communications of the ACM is publishing this coming month the article "Why STM Can Be More than a Research Toy" by A. Dragojevic, P. Felber, V. Gramoli and R. Guerraoui. This article demonstrates that STM finally wins its spurs by outperforming sequential applications with only four CPU cores. The publication responds to a paper invited about two years ago in the same venuethat suggested the confinement of STM to a research toy by questioning its ability to leverage multicore architectures.
The authors, from EPFL and University of Neuchatel, revisit these conclusions through the most to date extensive comparison of STM performance to sequential code. They evaluate an STM system on a wide range of benchmarks and two different multicore systems, and dissect the inherent costs of synchronization as well as the overheads of compiler instrumentation and transparent privatization.
The results show that an STM with manually instrumented benchmarks and explicit privatization outperforms sequential code by up to 29 times on SPARC with 64 concurrent threads and by up to 9 times on x86 with 16 concurrent threads. Indeed the overheads of compiler instrumentation and transparent privatization are substantial, yet they do not prevent STM from generally outperforming sequential code.
For more information, visit: http://cacm.acm.org/magazines/2011/4/106585-why-stm-can-be-more-than-a-research-toy/comments