Many modern computer systems and most multicore chips (chip multiprocessors) support shared memory in hardware. In a shared memory system, each of the processor cores may read and write to a single shared address space. For a shared memory machine, the memory consistency model defines the architecturally visible behavior of its memory system. Consistency definitions provide rules about loads and stores (or memory reads and writes) and how they act upon memory. As part of supporting a memory consistency model, many machines also provide cache coherence protocols that ensure that multiple cached copies of data are kept up-to-date. The goal of this primer is to provide readers with a basic understanding of consistency and coherence. This understanding includes both the issues that must be solved as well as a variety of solutions. We present both highlevel concepts as well as specific, concrete examples from real-world systems. Table of Contents: Preface / Introduction to Consistency and Coherence / Coherence Basics / Memory Consistency Motivation and Sequential Consistency / Total Store Order and the x86 Memory Model / Relaxed Memory Consistency / Coherence Protocols / Snooping Coherence Protocols / Directory Coherence Protocols / Advanced Topics in Coherence / Author Biographies

The performance of three directory-based cache- coherence protocols; strict request-response, intervention forwarding and reply forwarding are evaluated via simulation on the SOME-Bus, which is a fiber-optic interconnection network supporting DSM. The simulated system contains 64 nodes, each of which has a processor, cache controller, directory controller and output channel. Simulations have been conducted for each protocol to measure average processor utilization and average network latency for varying values of DSM parameters such as the ratio of the mean channel service time to mean thread run time (T/R), probability of a cache block being in modified state {P(M)}, the fraction of write misses {P(W)} and under different traffic patterns. The results reveal that the performance of all protocols decreases under all traffic patterns as P(W), P(M) or T/R increases. The effect of P(W) on the performance of the protocols reduces as P(M) increases. Reply forwarding performs the best for high P(M) values, intervention forwarding yields the best performance for low P(M) and high P(W) values and strict request-response is the best protocol under hot-region (HR) traffic.

Many modern computer systems, including homogeneous and heterogeneous architectures, support shared memory in hardware. In a shared memory system, each of the processor cores may read and write to a single shared address space. For a shared memory machine, the memory consistency model defines the architecturally visible behavior of its memory system. Consistency definitions provide rules about loads and stores (or memory reads and writes) and how they act upon memory. As part of supporting a memory consistency model, many machines also provide cache coherence protocols that ensure that multiple cached copies of data are kept up-to-date. The goal of this primer is to provide readers with a basic understanding of consistency and coherence. This understanding includes both the issues that must be solved as well as a variety of solutions. We present both high-level concepts as well as specific, concrete examples from real-world systems. This second edition reflects a decade of advancements since the first edition and includes, among other more modest changes, two new chapters: one on consistency and coherence for non-CPU accelerators (with a focus on GPUs) and one that points to formal work and tools on consistency and coherence.

Caches help reduce the effect of long-latency memory requests, by providing a high speed data-path to local memory. However, in multi-processor systems utilizing shared memory, cache coherence protocols are necessary to ensure sequential consistency. Of the multiple coherence protocols developed, the scalability of directory-based schemes makes them ideal for RAW's architecture [1]. Although one such system has been demonstrated as a proof-of-concept, it lacks the ability to meet the requirements of load-intensive, high performance applications. It further provides the application developer with no programming constructs to easily leverage the system. This thesis further develops shared memory support for RAW, by bringing greater programmability and performance to shared memory applications. In doing so, it reveals that shared memory is a practical programming paradigm for developing parallel applications on the RAW architecture.