Because matrix x matrix mult is so highly optimized as discussed in class, this is a significantly more efficient operation than performing each of the matrix x vector mults in parallel.

jellybean

One downside of this method is that it requires duplication of the image's data, and thus has a very large memory footprint.

Because matrix x matrix mult is so highly optimized as discussed in class, this is a significantly more efficient operation than performing each of the matrix x vector mults in parallel.

One downside of this method is that it requires duplication of the image's data, and thus has a very large memory footprint.