Accelerator Programming Using Directives

Name: Accelerator Programming Using Directives | 4th International Workshop, WACCPD 2017, Held in Conjunction with the International Conference for High Performance Computing, Networking, Storage and Analysis, SC 2017, Denver, CO, USA, November 13, 2017, Proceedings
Brand: Springer
Price: 42.79 EUR
Availability: OnlineOnly

4th International Workshop, WACCPD 2017, Held in Conjunction with the International Conference for High Performance Computing, Networking, Storage and Analysis, SC 2017, Denver, CO, USA, November 13, 2017, Proceedings

Sunita Chandrasekaran Guido Juckeland(Herausgeber*in)

Springer (Verlag)

Erschienen am 9. Februar 2018

IX, 183 Seiten

E-Book

PDF mit Wasserzeichen-DRM

Systemvoraussetzungen

978-3-319-74896-2 (ISBN)

42,79 €inkl. 7% MwSt.

Systemvoraussetzungen

für PDF mit Wasserzeichen-DRM

E-Book Einzellizenz

Als Download verfügbar

Beschreibung

Weitere Details

Weitere Ausgaben

Inhalt

Intro
Preface
Organization
Contents
Applications
An Example of Porting PETSc Applications to Heterogeneous Platforms with OpenACC
Abstract
1 Introduction
2 Workflow and System Description
2.1 Workflow
2.2 System
3 Results and Discussion
3.1 Profiling with Score-P
3.2 The Most Expensive Kernel: MatMult_SeqAIJ
3.3 Four Steps Toward the Final Version of OpenACC Kernel
4 Speedups and Strong Scaling
5 Conclusion
Acknowledgement
References
Hybrid Fortran: High Productivity GPU Porting Framework Applied to Japanese Weather Prediction Model
1 Introduction
1.1 ASUCA on GPU
1.2 Parallelization Granularity
1.3 Memory Layout
1.4 Related Work
1.5 Problem Summary
2 Hybrid Fortran Language Extension and Code Transformation
2.1 Parallel Loop Abstraction
2.2 Compile-Time Defined Memory Layout and Device Data Region
2.3 Transformed Code
3 Code Transformation Method
4 Productivity- and Performance Results
5 Conclusion and Future Work
References
Implicit Low-Order Unstructured Finite-Element Multiple Simulation Enhanced by Dense Computation Using OpenACC
1 Introduction
2 Finite-Element Earthquake Simulation Designed for the K Computer
3 Proposed Solver for GPUs Using OpenACC
3.1 Modification of Algorithm for GPUs
3.2 Introduction of OpenACC
4 Performance Measurements
5 Application Example
6 Concluding Remarks
References
Runtime Environments
The Design and Implementation of OpenMP 4.5 and OpenACC Backends for the RAJA C++ Performance Portability Layer
1 Introduction
2 RAJA
2.1 Basic Execution Policies
2.2 RAJA::NestedPolicy and Loop Transformations
3 Embedding Directives in the C++ Type System
3.1 Defining Policy Tags for a Backend
3.2 Constructing Explicit Execution Policy Types
3.3 Implement forall Specializations
4 Case Study: OpenMP 4.5
5 Case Study: OpenACC
6 Evaluation
6.1 Test Set
6.2 Goals and Non-Goals
6.3 Compilation Overhead
6.4 Runtime Overhead
7 Future Work and Conclusion
References
Enabling GPU Support for the COMPSs-Mobile Framework
1 Introduction
2 Related Work
3 Programming Model
3.1 Extension for GPU Support
4 Runtime Support Implementation
4.1 COMPSs-Mobile Runtime Architecture
4.2 OpenCL Platform
5 Performance Evaluation
5.1 OpenCL Platform Performance
5.2 Load Balancing Policies
6 Conclusions and Future Work
References
Concurrent Parallel Processing on Graphics and Multicore Processors with OpenACC and OpenMP
Abstract
1 Introduction
2 MBFLO3 Application
2.1 Mathematical Formulation
2.2 Numerical Method
3 Heterogeneous Multiblock Computing Strategy
3.1 Multicore Host Parallelism
3.2 Manycore Accelerator Parallelism
3.3 Heterogeneous Host-Device Parallelism
4 Performance Results and Analysis
5 Conclusions
Acknowledgements
References
Program Evaluation
Exploration of Supervised Machine Learning Techniques for Runtime Selection of CPU vs. GPU Execution in Java Programs
1 Introduction
2 Motivation
3 Compiling Java to GPUs
3.1 Java Parallel Stream API
3.2 JIT Compilation for GPUs
4 Exploring Supervised Machine Learning Algorithms
4.1 Supervised Machine Learning
4.2 Generating Subsets of Features
4.3 Constructing Prediction Models
4.4 Integrating Prediction Models
5 Experimental Results
5.1 Experimental Protocol
5.2 Overall Summary
5.3 Accuracies on the Full Set of Features
5.4 Exploring ML Algorithms by Feature Subsetting
5.5 Lessons Learned
6 Related Work
6.1 GPU Code Generation from High-Level Languages
6.2 Offline Model Construction
7 Conclusions
A Appendix
References
Automatic Testing of OpenACC Applications
1 Introduction
2 Testing a GPU Port of a Numerical Application
3 Autocompare with OpenACC
4 Autocompare Implementation
5 Experiments
6 Related Work
7 Future Work
8 Conclusion
References
Evaluation of Asynchronous Offloading Capabilities of Accelerator Programming Models for Multiple Devices
1 Introduction
2 Related Work
3 Accelerator Programming Models
3.1 CUDA
3.2 OpenCL
3.3 OpenACC
3.4 OpenMP
4 Implementing the Conjugate Gradient Method
5 Performance Results on NVIDIA GPUs
5.1 Data Transfers with the Host
5.2 Single Device
5.3 Two Devices
6 Performance Results on Intel Xeon Phi Coprocessors
6.1 Single Device
6.2 Two Devices
7 Summary
References
Author Index

Systemvoraussetzungen

Als PDF speichern Als Link merken