A test suite for analyzing the performance features of graphics processing units
Keywords:
graphics processing units
heterogeneous computing systems
atomic operations
Abstract
A test suite is proposed for graphics processors. These tests allow measuring various performance features, such as the delay and bandwidth of various types of memory, the efficiency of atomic operations, and the cache line size. With the aid of a specially developed memory test, we show that the coherency between threads has a much greater effect on the memory bandwidth than locality.
Section
Section 2. Programming
References
- https://www-01.ibm.com/chips/techlib/techlib.nsf/products/Cell_Broadband_Engine
- http://www.clearspeed.com/
- http://parallel.ru/
- http://www.khronos.org/opencl/
- http://sourceforge.net/projects/nuda/
- Wong H., Papadopoulou M., Sadooghi-Alvandi M., Moshovos A. Demystifying GPU microarchitecture through microbenchmarking // Trans. of IEEE Int. Symp. on Peformance Analysis &; Software. Toronto: IEEE, 2010. 235-246.
- Taylor R., Li X. A micro-benchmark suite for AMD GPUs // Proc. of the 39th Int. Conf. on Parallel Processing Workshops. San Diego: ACM Press, 2010. 387-396.
- Danalis A., Marin G., McCurdy C., Meredith J.S., Roth P.C., Spafford K., Tipparaju V., Vetter J. The Scalable Heterogeneous Computing (SHOC) benchmark suite // Proc. of the 3rd Workshop on General-Purpose Computation on Graphics Processing Units. Oak Ridge: ACM Press, 2010. 63-74.
- http://www.prace-project.eu/documents/17_apexmap_vw.pdf
- http://developer.amd.com/gpu/AMDAPPSDK/Pages/default.aspx
- http://sourceforge.net/projects/gpuperfomance/