high performance computing on graphics processing units: hgpu.org

Since 2004, the clock frequency of CPUs has not increased significantly. Computer Vision applications have an increasing demand for more processing power and are limited by the performance capabilities of sequential processor architectures. The only way to get better performance using commodity hardware is to adopt parallel programming. Many other related research projects have considered using one domain specific algorithm to compare the best sequential implementation with the best parallel implementation on a specific hardware platform. This project is distinctive because it investigated how to speed up a whole library by parallelizing the algorithms in an economical way and execute them on multiple platforms. In this work the author has: – Examined, compared and evaluated 22 programming languages and environments for parallel computing on multi-core CPUs and GPUs. – Chosen to use OpenMP as the standard for multi-core CPU programming and OpenCL for GPU programming. – Re-implemented a number of standard and well-known algorithms in Computer Vision using both standards. – Tested the performance of the implemented parallel algorithms and compared the performance to the sequential implementations of the commercially available software package VisionLab. – Evaluated the test results with a view to assessing: – Appropriateness of multi-core CPU and GPU architectures in Computer Vision. – Benefits and costs of parallel approaches to implementation of Computer Vision algorithms. Both the literature review and the results of the benchmarks in this work have confirmed that both multi-core CPU and GPU architectures are appropriate for accelerating sequential Computer Vision algorithms. Using OpenMP it was demonstrated that many algorithms of a library could be parallelized in an economical way and that adequate speedups were achieved on two multi-core CPU platforms. With a considerable amount of extra effort, OpenCL was used to achieve much higher speedups for specific algorithms on dedicated GPUs. At the end of the project, the choice of standards was re-evaluated including newly emerged ones. Recommendations are given for using standards in the future, and for future research and development. The following algorithmic improvements appear to be novel. The literature search has not found any previous use of them: – Vectorization of Convolution on grayscale images with variable sized mask utilizing padding width of vector with zeros. – Few-core Connect Component Labelling. – Optimization of a recent many-core Connect Component Labelling approach.