SC Conference - Activity Details

Comparing Cg and CUDA Implementations of Selected Transform Algorithms

Robin Eidissen  (Norwegian University of Science & Technology)
Anne C. Elster  (Norwegian University of Science & Technology)
Posters Session
Tuesday,  05:15PM - 07:00PM
Room Rotunda Lobby
For many years the field of GPGPU was forced to make do with graphics APIs like Cg, but in the recent past we have had the privilege of being able to program for GPUs using general purpose development toolkits like CUDA from Nvidia. This report examines three transforms: the Discrete Cosine Transform, the Lapped Orthogonal Transformand the Discrete Wavelet Transform. Of these the DCT and the DWT are widelyused in computer science, and the LOT has important applications as well. Existing Cg implementations of these are looked at, and CUDAreimplementations are described.Important differences between the corresponding implementations are examined, as well as the impact of choosing either Cg or CUDA in each case.
   IEEE Computer Society  /  ACM     2 0   Y E A R S   -   U N L E A S H I N G   T H E   P O W E R   O F   H P C