The art of high performance computing for computational science. Advanced Techniques and Examples for Materials Science / Vol. 2, (2019)

• Record Type: Book
• Other Titles: Advanced Techniques and Examples for Materials Science
• Further Information: Note: Masaki Geshi, editor.
• Other Names: Geshi, Masaaki
• Contents: Intro; Preface; References; Contents; 1 Supercomputers and Application Performance; 1.1 What Is a Supercomputer?; 1.2 What Is a High-Performance Application?; 1.2.1 Important Points in Optimizing High Parallelism; 1.2.2 Important Points for Single-CPU Performance Optimization; 2 Performance Optimization of Applications; 2.1 Performance Evaluation Method; 2.2 Current State Recognition: Source Code Investigation; 2.3 Current State Recognition: Measurement Methods; 2.4 Current State Recognition: Determination of Computation and Communication Kernels 2.5 Understanding the Problems: Evaluation of High-Parallelism Problems2.5.1 Classification of Problems Related to High Parallelism; 2.6 Understanding the Problems: Evaluation Methods for Problems in Single-CPU Performance; 2.6.1 Application Classification for Single-CPU Performance; 2.6.2 Evaluation by Cutting Out the Computation Kernel; 2.7 Performance Optimization Techniques Using Problem Patterns for High Parallelism; 2.7.1 Mismatch of Parallelism Between Application and Hardware; 2.7.2 Remaining Nonparallel Part; 2.7.3 Large Communication Size and Frequent Global Communication 2.7.4 All-Node Global Communication2.7.5 Large Communication Size in Adjacent Communication, Communication Frequency; 2.7.6 Load Imbalances; 2.8 Performance Optimization of Single-CPU Performance; 2.8.1 Elements for Obtaining High Single-CPU Performance; 2.8.2 Relationship Between Factors for High Performance and Required B/F Values; 2.8.3 Thread Parallelization Common to Large and Small Required B/F Values; 2.8.4 When the Required B/F Value Is Small: DGEMM Conversion; 2.8.5 When the Required B/F Value Is Small: Cache Blocking 2.8.6 When the Required B/F Value Is Small and the Loop Body Is Complex2.8.7 Optimizing the Single-CPU Performance: When the Required B/F Value Is Large; 2.8.8 Performance Estimation: When the Required B/F Value Is Large; 2.8.9 Performance Optimization of the Sparse Matrix-Vector Product: When the Required B/F Value Is Large; 2.8.10 Performance Optimization of the Sparse Matrix-Vector Product: Required B/F Value Is Large and List Vectors Used; References; 3 Case Studies of Performance Optimization of Applications; 3.1 Applications for Demonstration of the Performance of the K Computer 3.1.1 Outline of Application Groups3.2 System Outline of the K Computer; 3.2.1 Processor Features; 3.2.2 Outline of Tofu Interconnect; 3.2.3 6D Network; 3.3 Computer Environment for Performance Evaluation; 3.4 Outline of the Examples of High-Parallelization Performance Optimization; 3.4.1 Enhancement of the Parallelization Axis; 3.4.2 Improvement of Communication and Load Imbalances; 3.5 Performance Optimization of RSDFT; 3.5.1 Overview of RSDFT; 3.5.2 Understanding the Software Application Characteristics by Investigating the RSDFT Source Code
• Publisher Details: Singapore : Springer
• Publication Date: 2019
• Extent: 1 online resource (209 pages)
• Subjects: 004.1/1; High performance computing; High performance computing; Electronic books
• Languages: English
• ISBNs: 9789811398025
• Related ISBNs: 981139802X; 9789811398018
• Notes: Note: Print version record.
• Access Rights: Legal Deposit; Only available on premises controlled by the deposit library and to one user at any one time; The Legal Deposit Libraries (Non-Print Works) Regulations (UK).
• Access Usage: Restricted: Printing from this resource is governed by The Legal Deposit Libraries (Non-Print Works) Regulations (UK) and UK copyright law currently in force.