OpenMP: Heterogenous Execution and Data Movements

Name: OpenMP: Heterogenous Execution and Data Movements | 11th International Workshop on OpenMP, IWOMP 2015, Aachen, Germany, October 1-2, 2015, Proceedings
Brand: Springer
Price: 50.28 EUR
Availability: OnlineOnly

11th International Workshop on OpenMP, IWOMP 2015, Aachen, Germany, October 1-2, 2015, Proceedings

Christian Terboven Bronis R. de Supinski Pablo Reble Barbara M. Chapman Matthias S. Müller(Editor)

Springer (Publisher)

Published on 30. September 2015

XI, 274 pages

E-Book

PDF with digital watermarking

System requirements

978-3-319-24595-9 (ISBN)

€50.28incl. 7% vat

System requirements

for PDF with digital watermarking

E-Book Single Licence

Available for download

Description

More details

Other editions

Content

Intro
Preface
Organization
Contents
Applications
PAGANtec: OpenMP Parallel Error Correction for Next-Generation Sequencing Data
1 Introduction
2 Related Work
2.1 k-mer Graph and Error Correction
2.2 Parallelization Options
3 PAGANtec Architecture
3.1 Graph Structure
3.2 Correction Strategies
3.3 Correcting Errors
4 Parallelization
4.1 Performance Analysis
5 Conclusion
References
Composing Low-Overhead Scheduling Strategies for Improving Performance of Scientific Applications
1 Introduction
2 Scheduling Strategies
3 Techniques for Composing Scheduling Strategies
3.1 uSched
3.2 slackSched
3.3 vSched
3.4 ComboSched
4 Code Transformation
5 Results
6 Related Work
7 Conclusions
References
Exploiting Fine- and Coarse-Grained Parallelism Using a Directive Based Approach
1 Introduction
2 Related Work
3 Background: OpenMP Accelerator Model
4 An Offloading Model for a Cluster
4.1 Definitions
4.2 Execution Model
4.3 Memory Model
5 Implementation
5.1 Runtime Support
6 Preliminary Results
7 Discussion
8 Conclusions
References
Accelerator Applications
Experiences of Using the OpenMP Accelerator Model to Port DOE Stencil Applications
1 Introduction
2 OpenMP 4.0's Accelerator Support
3 Applications
4 Porting to GPUs
4.1 Baseline Performance on CPU and GPU
4.2 Increasing Parallelism
4.3 Loop Scheduling
4.4 Exploiting Memory Hierarchy
4.5 Reducing Memory Movement Between Host and Device
4.6 Manual Tuning for GPU Performance
4.7 Productivity
5 Related Work
6 Discussion and Future Work
References
Evaluating the Impact of OpenMP 4.0 Extensions on Relevant Parallel Workloads
1 Introduction and Motivation
2 Application Parallelization
2.1 Facesim
2.2 Fluidanimate
2.3 Streamcluster
3 Evaluation
3.1 Performance Evaluation
3.2 Programmability
4 Related Work
5 Conclusions
References
First Experiences Porting a Parallel Application to a Hybrid Supercomputer with OpenMP 4.0 Device Constructs
1 Introduction
2 OpenMP Device Constructs
2.1 Data Regions
3 A High Level View of the Porting Method
3.1 Fusing Local Data Regions
4 Porting NekBone
5 Conclusions
References
Tools
Lessons Learned from Implementing OMPD: A Debugging Interface for OpenMP
1 Introduction
2 Prior Work
3 The OpenMP Debugging Interface
3.1 OMPT: A Runtime Interface for OpenMP Tools
3.2 Why Distinguish OMPD from OMPT?
3.3 The OMPD Architecture
4 Use Cases of OMPD
4.1 OpenMP-Aware Stack Trace
4.2 Stepping in and Out of a Parallel Region
5 OMPD Callback Interface
5.1 Functions for Operating System Interaction
5.2 Resolving Structures for Target Architecture
5.3 Access Application Memory
5.4 Debugger's Context Argument
6 OMPD API Function Specifications
6.1 Providing Information on Compatible Runtime Library
6.2 API Specification for Breakpoints
6.3 Missing Function to Identify Master
7 Future Challenges
7.1 Context Pointer for Accelerators
7.2 Addressing Accelerator Threads
7.3 Return Codes
8 Conclusions
References
False Sharing Detection in OpenMP Applications Using OMPT API
1 Introduction
2 Motivation
3 Related Work
4 OMPT- Application Programming Interface for Tools
5 Our Approach
5.1 OMPT for Capturing Unique Patterns
5.2 Hardware Performance Information
5.3 Binary Classifier for False Sharing Detection
5.4 Feature Selection
6 Experimentation and Results
6.1 Training Phase
6.2 Validation of the Approach
7 Conclusion and Future Work
References
Exception Handling with OpenMP in Object-Oriented Languages
1 Introduction
2 Related Work
3 Problem Overview
3.1 Current Situation
3.2 Problem Definition
4 Cancellations
5 Exception Handling
5.1 Overview of Categorization
5.2 Local Exception Handling
5.3 Global Exception Handling
6 Implementation
6.1 Adaptable Synchronization Barrier
6.2 Dynamic Work Redistribution
6.3 Exception from Synchronization Regions
6.4 Global Exception Throwing
7 Evaluation
7.1 Usability
7.2 Performance
8 Conclusion
References
Extensions
On the Algorithmic Aspects of Using OpenMP Synchronization Mechanisms II: User-Guided Speculative Locks
1 Introduction
2 Related Work
3 User-Guided Locking API with TSX
3.1 Intel Transactional Synchronization Extensions
3.2 Using the User-Guided Locking API
4 Applying Intel TSX to the Test Code
4.1 A Brief Review of the Algorithm
4.2 The Role of TSX
5 Experimental Results
5.1 Convergence
5.2 Transactional Memory Statistics
5.3 Performance Measurement
6 Conclusions and Future Work
References
Using Transactional Memory to Avoid Blocking in OpenMP Synchronization Directives
1 Introduction
2 Avoiding Blocking in OpenMP
2.1 Critical Sections
2.2 Barrier/Taskwait
3 Evaluation
3.1 Experimental Setup
3.2 Results
4 Limitations and Related Work
5 Conclusion
References
A Case Study of OpenMP Applied to Map/Reduce-Style Computations
1 Introduction
2 Related Work
3 Map-Reduce Programming Model
3.1 Phoenix++ Implementation
3.2 OpenMP Facilities for Map/Reduce-Style Computations
4 OpenMP Implementations
4.1 Histogram
4.2 Linear Regression
4.3 K-Means Clustering
4.4 Word Count
4.5 String Match
4.6 Matrix Multiply
4.7 Principal Component Analysis
5 Evaluation
5.1 Analysis
5.2 Coding Style Comparison
5.3 Implications to OpenMP
6 Conclusion
References
Compiler and Runtime
Enabling Region Merging Optimizations in OpenMP
1 Introduction
2 Region Merging and Control
2.1 Region Merging Validity in OpenMP
2.2 Syntax Extensions to Support Merging
3 Results and Evaluation
3.1 Back to Back Regions
3.2 Parallel Regions with Intervening Serial Regions
3.3 Lulesh
4 Related Work
5 Conclusion
References
Towards Task-Parallel Reductions in OpenMP
1 Introduction
2 Related Work
3 Discussion
3.1 Updates of a Reduction Variable Outside a Reduction Context
3.2 Over-Specifying the Reduction Identifier
3.3 Supporting Untied Tasks
3.4 Supporting Nested Taskgroups
3.5 Cancellation, Dependencies and Merged Tasks
4 Syntax Additions
5 Evaluation
5.1 System Environment
5.2 Benchmark Descriptions
5.3 Performance Results on Intel Xeon Processors
5.4 Performance Results on Intel Xeon Phi Coprocessors
6 Conclusions and Future Work
References
OpenMP 4.0 Device Support in the OMPi Compiler
1 Introduction
2 Background
3 Compiler Transformations
3.1 Target Data
3.2 Target
3.3 Declare Target
4 Runtime Support
4.1 Data Environment Handling
5 The Epiphany Accelerator as a Device
5.1 Runtime Organization
5.2 Experiments
6 Discussion and Current Status
References
Energy
Application-Level Energy Awareness for OpenMP
1 Introduction
2 Motivation
3 OpenMPE
4 Compilation and Runtime System
5 Evaluation
6 Related Work
7 Conclusion and Future Work
References
Evaluating the Energy Consumption of OpenMP Applications on Haswell Processors
1 Introduction
2 Related Works
3 Basic Characteristics
3.1 Energy-Saving Features of Haswell
3.2 Load-Dependent Behavior
4 Optimization Steps
4.1 Wait Strategies
4.2 Iterative Clock Adjustment
4.3 Evaluation
5 Conclusion
References
Parallelization Methods for Hierarchical SMP Systems
1 Introduction
2 The Test Code
3 SIMD Building Blocks
4 Nested Threading
5 Code Variants
5.1 Baseline
5.2 Hand Decomposed
5.3 Nested Parallelism
5.4 Hand Nested
5.5 Crew and Teams
5.6 SBB
6 Performance Experiments
7 Conclusions and Future Work
References
Supporting Indirect Data Mapping in OpenMP ????? ? ?????? ???
1 Introduction
2 The OpenMP 4.0 Data Environment
2.1 Mapping Syntax
2.2 Presence
3 Map Refinements
3.1 Data only Array Sections
3.2 Type-Based Implicit Mappings
4 Clause Grouping and Binding
5 Conclusion
References
Author Index

System requirements

Save as PDF Copy link into clipboard

Schweitzer Fachinformationen

OpenMP: Heterogenous Execution and Data Movements

Description

More details

Other editions

Additional editions

Content

System requirements