[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Benchmarks and code for 1D/2D convolution of real signals
From: |
jeremyfix |
Subject: |
Benchmarks and code for 1D/2D convolution of real signals |
Date: |
Fri, 3 May 2013 06:51:27 -0700 (PDT) |
Hi,
If interested, I did some benchmarks for 1D/2D convolutions of real signals
using various implementations:
- FFT convolution using the GSL or FFTW
- nested for loops
- octave convn, fftconv/fftconv2
All the results and codes are available on github :
https://github.com/jeremyfix/FFTConvolution
Actually the implementation using the FFTW is faster than any of the other
implementations for circular and linear convolutions. For linear
convolutions, I only did benchmarks for the 'same' mode and octave is faster
if the kernel size is smaller than , say 10 x 10 ;
Best;
Jeremy.
--
View this message in context:
http://octave.1599824.n4.nabble.com/Benchmarks-and-code-for-1D-2D-convolution-of-real-signals-tp4652598.html
Sent from the Octave - General mailing list archive at Nabble.com.
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- Benchmarks and code for 1D/2D convolution of real signals,
jeremyfix <=