A Faster Convolutional 1D Operator

A faster conv for very large arrays.

Description

We have implemented a faster 1D convolutional operator. It is extremely fast for VERY large input arrays with comparable dimensions.

If the dimensions of the two vectors are quite different the best way to work is to compare execution times (in general the speed improvement depends by their sizes).

For smaller input vectors MATLAB implementation results more efficient.

Example
a=floor(10000*rand(200000,1));
b=floor(10000*rand(200000,1));
Matlab conv(a,b) -----> requires 1616 secs.
Our mcgo_conv(a,b) -->requires 639 secs.

• MATLAB release: MATLAB 6.5 (R13)