Thesis Details
Comparison of Parallel Programming APIs
This BSc Thesis was performed during a study stay at the Faculty of Information Technology of the Brno University of Technology. Parallel computing is a relevant issue nowadays since virtually every CPU manufactured is built with multi-core features. The way to introduce parallel code in our applications is still a long road ahead, but some tools have already proved a high level of maturity. This thesis aims to be a limited and practical approach to them: it begins briefly describing the main application programming models in use for parallelism and laying the theoretical foundations for this topic. In order to illustrate the comparison, the implementation of a video stabilization algorithm based in plane matching is presented and ported to different modern APIs, such as OpenMP, CUDA and OpenCL, along with a methodology to compare efficiency and speedup. Also, we introduce some opinionated metrics to evaluate the developer experience while using these tools.
Parallel Computing, GPU, GP-GPU, OpenMP, CUDA, OpenCL, Video Stabilization, Gray-Code Bit Plane Matching
@bachelorsthesis{FITBT19037, author = "Samuel Sainz Alfageme", type = "Bachelor's thesis", title = "Comparison of Parallel Programming APIs", school = "Brno University of Technology, Faculty of Information Technology", year = 2016, location = "Brno, CZ", language = "english", url = "https://www.fit.vut.cz/study/thesis/19037/" }