UPC Collectives Library 2.0
نویسندگان
چکیده
Collective communication has been a part of the UPC standard since having been introduced in 2005 with the UPC Specification version 1.2. However, unlike MPI collectives, UPC collectives have never caught on and are rarely used. In this paper we identify and discuss several fundamental limitations and important missing features in the design of the existing UPC collectives that make them inconvenient to use and unsuitable for performance optimization. Next, we propose a new, consistent, portable and high performance collectives library that is aimed to augment UPC with a full complement of the collectives used by MPI. Ours is a pure library based approach; we change none of the functions in the existing UPC specification. We discuss the implementation requirements for this new UPC collectives library, and how our design attempts to minimize the implementation effort by enabling the reuse of existing collective implementations.
منابع مشابه
A Parallel Numerical Library for UPC
Unified Parallel C (UPC) is a Partitioned Global Address Space (PGAS) language that exhibits high performance and portability on a broad class of shared and distributed memory parallel architectures. This paper describes the design and implementation of a parallel numerical library for UPC built on top of the sequential BLAS routines. The developed library exploits the particularities of the PG...
متن کاملEvaluation of Optimized Barrier Algorithms for SCI Networks with Different MPI Implementations
The SCI Collectives Library is a new software package which implements optimized collective communication operations on SCI networks. It is designed to be coupled to different higher-level communication libraries (especially MPI implementations) by adapter modules, thereby giving them access to these optimized collectives. In this work, we present the design of the SCI Collectives Library and o...
متن کاملUPCBLAS: a library for parallel matrix computations in Unified Parallel C
The popularity of Partitioned Global Address Space (PGAS) languages has increased during the last years thanks to their high programmability and performance through an efficient exploitation of data locality, especially on hierarchical architectures such as multicore clusters. This paper describes UPCBLAS, a parallel numerical library for dense matrix computations using the PGAS Unified Paralle...
متن کاملLibPhotonNBC: An RDMA Aware Collective Library on Photon
Collectives are a widely utilized programming construct for assembling parallel communication in High Performance Computing (HPC) applications. They are categorized into either non-blocking and blocking invocation, depending on whether the control is returned immediately to a user or not. In particular, non-blocking variety allows applications to perform other useful computations while hiding t...
متن کاملProbability Collectives - A Distributed Multi-agent System Approach for Optimization
Introducing a new hobby for other people may inspire them to join with you. Reading, as one of mutual hobby, is considered as the very easy hobby to do. But, many people are not interested in this hobby. Why? Boring is the reason of why. However, this feel actually can deal with the book and time of you reading. Yeah, one that we will refer to break the boredom in reading is choosing probabilit...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2011