The Elephant and the Mice: The Role of Non-Strict Fine-Grain Synchronization for Modern Many-Core

AbstractThe Cray XMT architecture has incited curiosity among computer architects and systemsoftware designers for its architecture support of ne-grain in-memory synchronization. Al-though such discussion go back thirty years, there is a lack of practical experimental platformsthat can evaluate major technological trends, such as ne-grain in-memory synchronization.The need for these platforms becomes apparent when dealing with new massive many-coredesigns and applications.This paper studies the feasibility, usefulness and trade-o s of ne-grain in-memorysynchronization support in a real-world large-scale many-core chip (IBM Cyclops-64). Weextended the original Cyclops-64 architecture design at gate level to support the ne-grainin-memory synchronization feature. We performed an in-depth study of a well-known kernelcode: the wavefront computation. Several versions of the kernel were used to test the e ectsof di erent synchronization constructs using our chip emulation framework. Furthermore,we tested selected OpenMP kernel loops against existing software-based synchronizationapproaches.In our wavefront benchmark study, the combination of ne-grain dataow-like in-memorysynchronization with non-strict scheduling methods yields a thirty percent improvement overthe best optimized traditional synchronization method provided by the original Cyclops-64design. For the OpenMP kernel loops, we achieved speeds of three to fourteen times thespeed of software-based synchronization methods.