https://github.com/moznion/fftw-vs-cufft
Benchmark scripts to compare processing speed between FFTW and cuFFT
https://github.com/moznion/fftw-vs-cufft
Last synced: 8 months ago
JSON representation
Benchmark scripts to compare processing speed between FFTW and cuFFT
- Host: GitHub
- URL: https://github.com/moznion/fftw-vs-cufft
- Owner: moznion
- Created: 2015-03-25T07:45:08.000Z (about 11 years ago)
- Default Branch: master
- Last Pushed: 2015-03-25T07:47:33.000Z (about 11 years ago)
- Last Synced: 2025-01-31T07:34:20.439Z (over 1 year ago)
- Language: Cuda
- Size: 113 KB
- Stars: 8
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
FFTW VS cuFFT
==
Benchmark scripts to compare processing speed between [FFTW](http://www.fftw.org/) and [cuFFT](https://developer.nvidia.com/cuFFT).
Usage
--
```
make
./bench_fftw
./bench_cufft
```
Description
--
- `bench_fftw`: Run benchmark with FFTW
- `bench_cufft`: Run benchmark with cuFFT
Both of the binary have the same interfaces.
```
./bench_XXX [Number of Trials to Execute FFT] [Number of Trials to Execute Benchmark]
```
- Number of Trials to Execute FFT (int)
You omit this when it will use default value (default value: `10000`).
- Number of Trials to Execute Benchmark (int)
You omit this when it will use default value (default value: `10`).
And these binary accept `SIGNAL_LENGTH` environment variable.
It determines the signal length for signal to apply to FFT.
It must be int value. Default value is `4096`.
Author
--
moznion ()
License
--
MIT