Application of CUDA in DSP: Implementation of FIR filter and Cross Ambiguity Function
Description
This software package contains source code and documentation of DSP applications implemented on NVidia CUDA hardware computing platform. The first DSP application , we implemented on CUDA, was digital filter with finite impulse response (FIR). The goal of implementation of such a basic DSP task was to evaluate computational performance of CUDA platform in single-precision number representation. In summary, the FIR filter implementation enumerates output signal of the filter defined by its impulse response for a given input signal.
Other DSP application implemented on CUDA was the Cross Ambiguity Function (CAF) which is base task for Passive Coherent Locator systems.
The goal of the CAF implementation was to develop hardware accelerator with higher performance than other CAF accelerators we implemented in the past. The accelerator computes Cross Ambiguity Function between direct signal and signal reflected from a moving target. The accelerator is based on the NVidia CUDA platform and it is supplemented with an appropriate communication function in Matlab.
Package Summary
| Title | Application of CUDA in DSP |
| Filename | cudadsp.zip |
| License | Freeware (See the 'license.txt' file in the package.) |
| Package content | ZIP archive containing implementation source codes, supplementary Matlab scripts, testing data-files and documentation |
| Size | 6 858 526 Bytes |
| Required tools & platform |
Matlab 2008b, Signal Processing Toolbox, NVidia CUDA SDK v2.2, NVidia CUDA toolkit v2.2 and Graphics card NVidia GeForce GTX260 |
| Installation notes | See the Appendix B of 'cudadsp.pdf' file, contained in the package. |
Result Category
| Project number | Contract section | Year | RIV category | Comment |
| 7H09005 | VS1 | 1 | 2009 | Functional sample |