Lazy release consistency for software distributed shared memory

Abstract

Relaxed memory consistency models, such a s release consistency, w ere introduced in order to reduce the impact of remote memory access latency in both software and hardware distributed shared memory (DSM). However, in a software DSM, it is also important to reduce the number of messages and the amount o f d a t a e xchanged for remote memory access. Lazy release consistency is a new algorithm for implementing release consistency that lazily pulls modi cations across the interconnect only when necessary. T race-driven simulation using the SPLASH benchmarks indicates that lazy release consistency reduces both the number of messages and the amount of data transferred between processors. These reductions are especially signi cant f o r programs that exhibit false sharing and make extensive use of locks.

A compiler abstraction for machine independent parallel communication generation

“…Consistency models have been used to express how a shared memory changes state [2,5,18]. The consistency model is the mechanism by which the programmer and the compiler agree on when memory updates take place, and as such is a source language rather than a compiler concept like IRONMAN.…”

Finding a suitable system scale to optimize program performance on software DSM systems

“…For the sake of programming convenience, multiple consistency models, such as sequential consistency [9] and release consistency [2,8], etc., are exploited in the implementation of shared address space. In sequential consistency model, multiple concurrent read accesses are allowed to the same memory page.…”

“…In DSM systems, a virtual shared address space [9] is provided among the processors loosely coupled in the network. Data are exchanged and kept consistent [1,7,8] implicitly via the network among the processors to generate a shared memory abstraction for use by the programmers. Therefore, programmers are able to write parallel programs in much the same way as they program in shared memory multiprocessor systems [3,13].…”

“…However, network speed and the amount of network traffic have great effect on the performance of programs in a DSM system. Consequently, most of the researches in the past focused on the design of performance-enhancing technologies such as relaxed consistency protocols [2,7,8,16,23], multithreading [6,17,22], load balance [24][25][26], communication minimization [19], etc.…”

A Platform-Independent Distributed Runtime for Standard Multithreaded Java

“…(10) The new JMM is expected to be based on the principles of Lazy Release Consistency (LRC). (11) Therefore, we employ a DSM protocol that complies with LRC. Since there is a natural mapping for Java volatile variables to the release-acquire semantics of LRC in the revised JMM, we encapsulate accesses to volatile variables with acquire-release blocks.…”