REDUCING THE DATA TRANSMISSION TIME FROM GPU TO CPU BY GENERATING BIT MEMORY CARD
Abstract
This article describes the research of algorithms of decreasing data transmission time from GPU to GPU by forming the bit memory card based on the results of calculations. Practical experiment showing the test results of the algorithm on different data quantities was performed.
References
1. Fujii Y., Azumi T., Nishio N., и др. Data transfer matters for GPU computing / Proc. Int. Conf. Parallel Distrib. Syst. ICPADS, pp. 275-282, 2013.
2. NVIDIA CORPORATION. CUDA C PROGRAMMING GUIDE» 2014. [Электронныйресурс]. URL: http://docs. nvidia.com/cuda/ pdf/CUDA_C_Programming_Guide.pdf.
3. Габдуллин В. В., Капустин А. И., Королев А. И. Применение технологии CUDA для задач голосовой биометрии на примере построения универсальной фоновой модели диктора. Санкт-Петербург, 2011.
4. Kardos J. Efficient CPU <~> GPU data transfers» Lugano, 2014.
5. Van Werkhoven B., Maassen J., Seinstra F. J., Bal H. E. Performance models for CPU-GPU data transfers» Proc. 14th IEEE /ACM Int. Symp. Clust. Cloud, Grid Comput. CCGrid 2014, pp. 11-20, 2014.
6. Boyer M., Meng J., Kumaran K. Improving GPU performance prediction with data transfer modeling / Proc. IEEE 27th Int. ParallelDistrib. Process. Symp. Work. PhDForum, IPDPSW 2013, pp. 1097-1106, 2013.
Review
For citations:
Voronkin V.S. REDUCING THE DATA TRANSMISSION TIME FROM GPU TO CPU BY GENERATING BIT MEMORY CARD. Modern Science and Innovations. 2016;(1):30-34. (In Russ.)
Views:
79
Email this article 