Performance Pathologies in Hardware Transactional Memory

Abstract

Hardware Transactional Memory (HTM) systems reflect choices from three key design dimensions: conflict detection, version management, and conflict resolution. Previously proposed HTMs represent three points in this design space: lazy conflict detection, lazy version management, committer wins (LL); eager conflict detection, lazy version management, requester wins (EL); and eager conflict detection, eager version management, and requester stalls with conservative deadlock avoidance (EE). To isolate the effects of these high-level design decisions, we develop a common framework that abstracts away differences in cache write policies, interconnects, and ISA to compare these three design points. Not surprisingly, the relative performance of these systems depends on the workload. Under light transactional loads they perform similarly, but under heavy loads they differ by up to 80%. None of the systems performs best on all of our benchmarks. We identify seven performance pathologies-interactions between workload and system that degrade performance-as the root cause of many performance differences: FRIENDLYFIRE, STARVINGWRITER, SERIALIZEDCOMMIT, FUTILESTALL, STARVIN-GELDER, RESTARTCONVOY, and DUELINGUPGRADES. We discuss when and on which systems these pathologies can occur and show that they actually manifest within TM workloads. The insight provided by these pathologies motivated four enhanced systems that often significantly reduce transactional memory overhead. Importantly, by avoiding transaction pathologies, each enhanced system performs well across our suite of benchmarks.

Emerging systems theory-based pathologies for governance of complex systems

“…First, it provides a means to diverge from the traditional medical formulation of pathology (Dietel and Schäfer, 2008;van den Tweel and Taylor, 2010) commonly associated with disease etiology, disease pathogenesis, cell morphological changes and health change consequences (Kumar et al, 2010) to a more contemporary formulation that considers health of inanimate systems including computer systems (Bobba et al, 2007), complex organisations (Barnard, 1946;Beer, 1984;Ríos, 2012) and social systems (Beer, 1984;Yolles, 2007). Second, this extended view of pathology provides a platform for viewing pathology in terms of factors and issues that act to reduce expected system performance based on an inadequacy in (or violation of) functions and their underlying principles.…”

Directory-Based Conflict Detection in Hardware Transactional Memory

“…The directory not only is able to detect the conflict in one hop, but it can also take action without having to wait until the conflicting block is in a base state (unlike the base approach that relies on forwarded coherence traffic), contributing to even faster detection/action. Aborting conflicting transactions earlier reduces the effect of pathological execution patterns such as futile stall and other conflicting interactions that affect eager CD systems like LogTM-SE [2]. The remarkable reductions in the execution time of vacation and genome are due to this quicker and more effective response.…”

Analysis of the Influence of the Conflict Management Policy of the Transactional Memory on the System Performance and Bus Traffic

“…[5] . 트랜잭셔널 메모리는 프로세서들 간의 경합 정도와 충돌 관리 정책에 따라서 트랜잭 션 수행 속도 및 시스템 버스의 트래픽 정도가 크 게 달라진다 [6] . 특히 시스템 버스는 공유메모리 멀 티프로세서 시스템의 성능과 규모를 결정짓는 매우 중요한 요소이며, 시스템 버스의 성능은 시스템 버 스의 트래픽 관리 방법에 의해 크게 영향을 받는다 [7] .…”