Performance Comparison Between Cg-based and CUDA-based Matrix Multiplications
Dept. of Computer Science, Bowling Green State Univ., Bowling Green, OH 43403, U.S.A.
International Conference on Parallel and Distributed Processing Techniques and Applications (PDPTA’12), 2012
In this paper, we compare the performances of Cg-based and CUDA-based GPU programming APIs. In particular, their performances on squared matrix multiplications are considered. We also discuss other aspects of these widely-used GPU programming APIs. This work can help gain insight on various applications that involve matrix multiplication that are better suited for a specific GPU programming API.
August 28, 2012 by hgpu
