Exploring scalability of FIR filter realizations on Graphics Processing Units

摘要

General-Purpose Computing on Graphics Processing Units (GPGPU) has lately been of great interest due to the release of architectures and software that simplifies programming graphics cards. This study explores how performance scales with FIR digital filters by varying the number of taps and the samples. We also discuss the trade-offs with various techniques for GPGPU programming in CUDA.