Intone — Tools and Environments for OpenMP on Clusters of SMPs1
نویسنده
چکیده
Clusters of small-scale SMP computers are becoming more and more common as high-performance computing needs have arised, not only in national scientific laboratories, but also in enterprises of various kinds. An SMP-cluster represents a sweet-spot of cost-efficiency compared to a larger SMP system or to a cluster with smaller nodes. With the emergance of OpenMP, shared memory computing has also become mainstream in the parallel computing world. Unfortunately, clusters of SMPs do not directly support a shared memory programming model in hardware and software distributed shared memory (DSM) techniques must be adopted in order to run OpenMP on a cluster of SMP as well as on, e.g., a high-end Sun Enterprise server. This abstract describes a research project which has been initiated to alleviate some of the problems of running OpenMP applications on clusters of SMPs. The acronym Intone stands for Innovative Tools for Non-Experts. Although OpenMP is not stated in the project title, it is an integral part of the approach taken in the project. Some of the goals of Intone are to: • Develop a graphical performance analysis system for OpenMP (both Fortran and C/C++) applications • To investigate the possibility of extending OpenMP with explicit message-passing primitives for performance critical communication • To advance the state-of-the-art of implementations of OpenMP for clusters of SMPs with software DSM • To evaluate the performance analysis system, the OpenMP extensions and the OpenMP-specific software DSM system with production parallel applications I will here briefly describe the activities in the project for the first three goals. Figure 1 shows the relationship between the software components in the envisioned system. The parts that Intone will deal with are underlined. Parallel applications are developed in a cyclic fashion. From a sequential program a parallel version is created either manually or automatically by parallelising tools such as the MIPSpro compiler from SGI or the KAP toolset [3, 7]. This code will be compiled with OpenMP compilers developed within the 1. INTONE is a project proposal tentatively financed by the European Commission under Framework Programme 5/IST. Contract negotiations are being carried out.
منابع مشابه
OpenMP Performance Analysis Approach in the INTONE Project
In this paper we present the general approach adopted in the INTONE project for performance analysis and optimization of OpenMP applications. The approach considers the following components: runtime interface (instrumentation and threading support) and its library implementation, compilation environments for Fortran90 and C/C++, and an extension of the VAMPIR graphical tool. The paper also incl...
متن کاملParallel computing using MPI and OpenMP on self-configured platform, UMZHPC.
Parallel computing is a topic of interest for a broad scientific community since it facilitates many time-consuming algorithms in different application domains.In this paper, we introduce a novel platform for parallel computing by using MPI and OpenMP programming languages based on set of networked PCs. UMZHPC is a free Linux-based parallel computing infrastructure that has been developed to cr...
متن کاملAccelerating high-order WENO schemes using two heterogeneous GPUs
A double-GPU code is developed to accelerate WENO schemes. The test problem is a compressible viscous flow. The convective terms are discretized using third- to ninth-order WENO schemes and the viscous terms are discretized by the standard fourth-order central scheme. The code written in CUDA programming language is developed by modifying a single-GPU code. The OpenMP library is used for parall...
متن کاملPerformance Characteristics of Hybrid MPI/OpenMP Implementations of NAS Parallel Benchmarks SP and BT on Large-Scale Multicore Clusters
The NAS Parallel Benchmarks (NPB) are well-known applications with the fixed algorithms for evaluating parallel systems and tools. Multicore clusters provide a natural programming paradigm for hybrid programs, whereby OpenMP can be used with the data sharing with the multicores that comprise a node and MPI can be used with the communication between nodes. In this paper, we use SP and BT benchma...
متن کاملTrasgo 2.0: Code generation for parallel distributed- and shared-memory hierarchical systems
1 Extended Abstract Current multicomputers are typically built as interconnected clusters of shared-memory multicore computers. A common programming approach for these clusters is to simply use a message-passing paradigm, launching as many processes as cores available. Nevertheless, to better exploit the scalability of these clusters and highly-parallel multicore systems, it is needed to effici...
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 2000