Strategies And Experiences Using High Performance Fortran

Since its beginnings in the 1950's, the Fortran language has been the language of choice for most scientific and engineering programming. Compilers, seizing on the simplicity of the language, have historically generated highly-optimized machine code. High performance Fortran (HPF) is a relative new addition to the Fortran dialect It is an attempt to provide an efficient high-level Fortran parallel programming language for the latest generation of been debatable. By operating at a high level, the HPF standard does not provide some low-level detail required to achieve maximum performance in a multiprocessor system. Message passing using highly-tuned libraries, such as the message passing interface (MPI), will more often than not require less wall clock time than a comparable HPF code. However, the HPF language and its compilers continue to mature and improve. HPF provides a convenient way to represent parallelism for those most comfortable with data parallel methodologies. As such, it can require a shorter time to solution and provide an acceptable level of efficiency. This report discusses our experiences with the language, as well as coding strategies and vendor-specific "hooks" that can be used to boost performance.

High Performance Computing is an integrated computing environment for solving large-scale computational demanding problems in science, engineering and business. Newly emerging areas of HPC applications include medical sciences, transportation, financial operations and advanced human-computer interface such as virtual reality. High performance computing includes computer hardware, software, algorithms, programming tools and environments, plus visualization. The book addresses several of these key components of high performance technology and contains descriptions of the state-of-the-art computer architectures, programming and software tools and innovative applications of parallel computers. In addition, the book includes papers on heterogeneous network-based computing systems and scalability of parallel systems. The reader will find information and data relative to the two main thrusts of high performance computing: the absolute computational performance and that of providing the most cost effective and affordable computing for science, industry and business. The book is recommended for technical as well as management oriented individuals.

Software -- Programming Languages.

The aim of this book is to explain to high-performance computing (HPC) developers how to utilize the Intel® Xeon PhiTM series products efficiently. To that end, it introduces some computing grammar, programming technology and optimization methods for using many-integrated-core (MIC) platforms and also offers tips and tricks for actual use, based on the authors’ first-hand optimization experience. The material is organized in three sections. The first section, “Basics of MIC”, introduces the fundamentals of MIC architecture and programming, including the specific Intel MIC programming environment. Next, the section on “Performance Optimization” explains general MIC optimization techniques, which are then illustrated step-by-step using the classical parallel programming example of matrix multiplication. Finally, “Project development” presents a set of practical and experience-driven methods for using parallel computing in application projects, including how to determine if a serial or parallel CPU program is suitable for MIC and how to transplant a program onto MIC. This book appeals to two main audiences: First, software developers for HPC applications – it will enable them to fully exploit the MIC architecture and thus achieve the extreme performance usually required in biological genetics, medical imaging, aerospace, meteorology and other areas of HPC. Second, students and researchers engaged in parallel and high-performance computing – it will guide them on how to push the limits of system performance for HPC applications.

Over the past decade high performance computing has demonstrated the ability to model and predict accurately a wide range of physical properties and phenomena. Many of these have had an important impact in contributing to wealth creation and improving the quality of life through the development of new products and processes with greater efficacy, efficiency or reduced harmful side effects, and in contributing to our ability to understand and describe the world around us. Following a survey ofthe U.K.'s urgent need for a supercomputingfacility for aca demic research (see next chapter), a 256-processor T3D system from Cray Research Inc. went into operation at the University of Edinburgh in the summer of 1994. The High Performance Computing Initiative, HPCI, was established in November 1994 to support and ensure the efficient and effective exploitation of the T3D (and future gen erations of HPC systems) by a number of consortia working in the "frontier" areas of computational research. The Cray T3D, now comprising 512 processors and total of 32 CB memory, represented a very significant increase in computing power, allowing simulations to move forward on a number offronts. The three-fold aims of the HPCI may be summarised as follows; (1) to seek and maintain a world class position incomputational scienceand engineering, (2) to support and promote exploitation of HPC in industry, commerce and business, and (3) to support education and training in HPC and its application.

Modern Fortran teaches you to develop fast, efficient parallel applications using twenty-first-century Fortran. In this guide, you’ll dive into Fortran by creating fun apps, including a tsunami simulator and a stock price analyzer. Filled with real-world use cases, insightful illustrations, and hands-on exercises, Modern Fortran helps you see this classic language in a whole new light. Summary Using Fortran, early and accurate forecasts for hurricanes and other major storms have saved thousands of lives. Better designs for ships, planes, and automobiles have made travel safer, more efficient, and less expensive than ever before. Using Fortran, low-level machine learning and deep learning libraries provide incredibly easy, fast, and insightful analysis of massive data. Fortran is an amazingly powerful and flexible programming language that forms the foundation of high performance computing for research, science, and industry. And it's come a long, long way since starting life on IBM mainframes in 1956. Modern Fortran is natively parallel, so it's uniquely suited for efficiently handling problems like complex simulations, long-range predictions, and ultra-precise designs. If you're working on tasks where speed, accuracy, and efficiency matter, it's time to discover—or re-discover—Fortran.. About the technology For over 60 years Fortran has been powering mission-critical scientific applications, and it isn't slowing down yet! Rock-solid reliability and new support for parallel programming make Fortran an essential language for next-generation high-performance computing. Simply put, the future is in parallel, and Fortran is already there. Purchase of the print book includes a free eBook in PDF, Kindle, and ePub formats from Manning Publications. About the book Modern Fortran teaches you to develop fast, efficient parallel applications using twenty-first-century Fortran. In this guide, you'll dive into Fortran by creating fun apps, including a tsunami simulator and a stock price analyzer. Filled with real-world use cases, insightful illustrations, and hands-on exercises, Modern Fortran helps you see this classic language in a whole new light. What's inside Fortran's place in the modern world Working with variables, arrays, and functions Module development Parallelism with coarrays, teams, and events Interoperating Fortran with C About the reader For developers and computational scientists. No experience with Fortran required. About the author Milan Curcic is a meteorologist, oceanographer, and author of several general-purpose Fortran libraries and applications. Table of Contents PART 1 - GETTING STARTED WITH MODERN FORTRAN 1 Introducing Fortran 2 Getting started: Minimal working app PART 2 - CORE ELEMENTS OF FORTRAN 3 Writing reusable code with functions and subroutines 4 Organizing your Fortran code using modules 5 Analyzing time series data with arrays 6 Reading, writing, and formatting your data PART 3 - ADVANCED FORTRAN USE 7 Going parallel with Fortan coarrays 8 Working with abstract data using derived types 9 Generic procedures and operators for any data type 10 User-defined operators for derived types PART 4 - THE FINAL STRETCH 11 Interoperability with C: Exposing your app to the web 12 Advanced parallelism with teams, events, and collectives

This book constitutes the refereed proceedings of the 4th International Symposium on High Performance Computing, ISHPC 2002, held in Kansai Science City, Japan, in May 2002 together with the two workshops WOMPEI 2002 and HPF/HiWEP 2002. The 51 revised papers presented were carefully reviewed and selected for inclusion in the proceedings. The book is organized in topical sections on networks, architectures, HPC systems, Earth Simulator, OpenMP-WOMPEI 2002, and HPF-HiWEP 2002.

CUDA Fortran for Scientists and Engineers: Best Practices for Efficient CUDA Fortran Programming shows how high-performance application developers can leverage the power of GPUs using Fortran, the familiar language of scientific computing and supercomputer performance benchmarking. The authors presume no prior parallel computing experience, and cover the basics along with best practices for efficient GPU computing using CUDA Fortran. In order to add CUDA Fortran to existing Fortran codes, they explain how to understand the target GPU architecture, identify computationally-intensive parts of the code, and modify the code to manage the data and parallelism and optimize performance – all in Fortran, without having to rewrite in another language. Each concept is illustrated with actual examples so you can immediately evaluate the performance of your code in comparison. This second edition provides much needed updates on how to efficiently program GPUs in CUDA Fortran. It can be used either as a tutorial on GPU programming in CUDA Fortran as well as a reference text. Presents optimization strategies for current hardware, including Hopper generation GPUs Includes discussions of new language and hardware features, including managed memory, tensor cores, shuffle instructions, new multi-GPU paradigms Offers resources and strategies for porting large codes to GPUs, including language features as well as library use