English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

. Data parallel programming provides an accessible model for exploiting thepower of parallel computing elements without resorting to the explicit use of low levelprogramming techniques based on locks, threads and monitors. The emergence ofGraphics Processing Units (GPUs) with hundreds or thousands of processing cores hasmade data parallel computing available to a wider class of programmers. GPUs can beused not only for accelerating the processing of computer graphics but also for generalpurpose data-parallel programming. Low level data-parallel programming languagesbased on the Compute Unied Device Architecture (CUDA) provide an approach fordeveloping programs for GPUs but these languages require explicit creation and coor-dination of threads and careful data layout and movement. This has created a demandfor higher level programming languages and libraries which raise the abstraction levelof data-parallel programming and increase programmer productivity. The Acceleratorsystem was developed by Microsoft for writing data parallel code in a high level man-ner which can execute on GPUs, multicore processors using SSE3 vector instructionsand FPGA chips. This paper compares the performance and development effort of thehigh level Accelerator system against lower level systems which are more difcult touse but may yield better results. Specically, we compare against the NVIDIA CUDAcompiler and sequential C++ code considering both the level of abstraction in the im-plementation code and the execution models. We compare the performance of thesesystems using several case studies. For some classes of problems, Accelerator has aperformance comparable to CUDA, but for others its performance is signicantly re-duced; however in all cases it provides a model which is easier to use and enablesgreater programmer productivity.Keywords. GPGPU, Accelerator, CUDA, comparisons.

An analysis of programmer productivity versus performance for high level data parallel programming