Scheduling Large Programs

For more info about our services contact : help@bestpfe.com

Table of contents

1 Introduction and Background 
1.1 Introduction
1.2 Background and Notations
1.2.1 Maps and Sets
1.2.2 Additional Definitions
1.2.3 Polyhedral Representation of Programs
1.2.4 Data Locality
1.2.5 Loop Transformations
1.3 Outline
2 Tiling Code With Memory-Based Dependences 
2.1 Introduction and related work
2.2 Sources of False Dependences
2.3 Motivating Example
2.4 Live Range Non-Interference
2.4.1 Live ranges
2.4.2 Construction of live ranges
2.4.3 Live range non-interference
2.4.4 Live range non-interference and tiling: a relaxed permutability criterion
2.4.5 Illustrative examples
2.4.6 Parallelism
2.5 Effect of 3AC on the Tilability of a Loop
2.6 Experimental Evaluation
2.7 Experimental Results
2.7.1 Evaluating the Effect of Ignoring False Dependences on Tiling
2.7.2 Performance evaluation for Polybench-3AC
2.7.3 Reasons for slowdowns
2.7.4 Performance evaluation for Polybench-PRE
2.8 Conclusion and Future Work
3 Scheduling Large Programs
3.1 Introduction
3.2 Definitions
3.3 Example of Statement Clustering
3.4 Clustering Algorithm
3.5 Using Offline Clustering with the Pluto Affine Scheduling Algorithm
3.5.1 Correctness of Loop Transformations after Clustering
3.6 Clustering Heuristics
3.6.1 Clustering Decision Validity
3.6.2 SCC Clustering
3.6.3 Basic-block Clustering
3.7 Experiments
3.8 Related Work
3.9 Conclusion and Future Work
4 The PENCIL Language 
4.1 Introduction
4.2 PENCIL Language
4.2.1 Overview of PENCIL
4.2.2 PENCIL Definition as a Subset of C99
4.2.3 PENCIL Extensions to C99
4.3 Detailed Description
4.3.1 stationst and restri
4.3.2 Description of the Memory Accesses of Functions
4.3.3 onst Function Attribute
4.3.4 Pragma Directives
4.3.5 Builtin Functions
4.4 Examples of PENCIL and non-PENCIL Code
4.4.1 BFS Example
4.4.2 Image Resizing Example
4.4.3 Recursive Data Structures and Recursive Function Calls
4.5 Polyhedral Compilation of PENCIL code
4.6 Related Work
5 Evaluating PENCIL 
5.1 Introduction
5.2 Benchmarks
5.2.1 Image Processing Benchmark Suite
5.2.2 Rodinia and SHOC Benchmark Suites
5.2.3 VOBLA DSL for Linear Algebra
5.2.4 SpearDE DSL for Data-Streaming Applications
5.3 Discussion of the Results
5.4 Conclusion and Future Work
6 Conclusions and Perspectives 
6.1 Contributions
6.2 Future Directions