II-New: A Computational Infrastructure for Scalable Transactional Memory Abstractions in Managed Languages

点击翻译按钮获取中文摘要

This proposal supports the acquisition of an Azul Vega 3 compute appliancefor research on developing an open-source general-purpose transactionalinfrastructure for high-level managed languages. This appliance is equippedwith hundreds of processor cores, substantial memory, and bandwidthavailability. The cores compute in a flat SMP configuration, simplifyingprogrammability and management. The appliance has a number of novelfeatures including hardware support for concurrent garbage collection, andoptimistic lock elision that provide critical building blocks for efficienttransactional memory (TM) support. The project's research agenda is toexplore new approaches to scalable parallel program construction thatexploits underlying hardware primitives that mask or mitigate criticalbottlenecks that have thus far hampered effective scalable softwaretransactional memory systems.The project will use the novel features of the Azul appliance to (i) explorenew language abstractions with special focus on scalability of transactionalapplications; (ii) understand how primitive support for garbage collectionand transactions influence algorithm and application development built usingthese abstractions; (iii) examine the impact of these mechanisms oncompiler and runtime design; and (iv) develop a large number of benchmarksand workloads to better understand TM performance characteristics.This project is part of a larger NSF-funded effort to catalyze TM research,which in turn is a key technique to improving software development forgeneral-purpose parallel computers. Widespread adoption of new scalableconcurrency abstractions such as transactional memory is an essentialcomponent in the technology roadmap of major industry players, and the focusof much attention in academia.