Design Principle of Massively Parallel Distributed-Memory Multiprocessor Architecture
نویسندگان
چکیده
In this papery we discuss the design principles of massively parallel distributed-memory multiprocessor architecture and propose the Datarol-H architecture. We present the architecture of the massively parallel Datarol-H machine and show a Datarol-H processor design, including communication protocol and handling mechanisms of remote memory access and remote process/procedure invocation. Last, we show several evaluations of the Datarol-H processor from the viewpoint of effect on thread execution, effect of implicit register loading, and swapping cost by software simulation.
منابع مشابه
A Design Principle of Massively Parallel Distributed-Memory Multiprocessor Architecture
Distributed-memory multiprocessor architecture is essential in developing massively parallel machines. One of the most important design issues in such a distributed-memory multiprocessor architecture is a latency problem which is caused by remote procedure invocation and remote memory access. Remote memory access and remote procedure invocation occur so often in massively parallel execution, an...
متن کاملDatarol: A Parallel Machine Architecture for Fine-Grain Multithreading
In this paper, we discuss the design principle of massively parallel distributed-memory multiprocessor architecture, and introduce the Datarol-II machine architecture. We present the Datarol-II processor design, including communication protocol and handling mechanisms of remote memory access remote process/procedure invocation. Several evaluation data of the Datarol-II processor are shown from ...
متن کاملImplementing Ml on the Fujitsu Ap1000
The CAP ML project seeks to develop a version of ML that is suitable for use on a distributed memory multiprocessor architecture such as the Fujitsu AP1000. Language extensions are proposed that have been developed in conjunction with a programmming methodology that is appropriate to that of a massively parallel computer whilst retaining a functional style. The implementation, which is based on...
متن کاملPerformance Prediction of Benchmark Programs for Massively Parallel Architectures
A performance prediction method is presented, which accurately predicts the runtime of a parallel application using the messagepassing model and the asynchronous task programming paradigm. The introduced model also considers the complete memory hierarchy of a multiprocessor node architecture. We study the optimizied parallel Linpack application as an example benchmark program. The predicted per...
متن کاملA Scaleable Multiprocessor Architecture with Multiple Read-Write Memory Model
This paper presents a scalable multiprocessor architecture with multiple access memories and multi-way busses. This parallel architecture with more intelligent memory model and efficient multi-way interconnection network organization is called as CRrCW (Concurrent Read and restricted Concurrent Write) scaleable multiprocessor system. The memory and network model provides concurrent memory acces...
متن کامل