How do I multiply two 'int32' data type matrices in MATLAB?

I would like to multiply a variable of size of (100,100) with another variable of size (100,100). Both of the variables are in 'int32' format.

Answer 1

[MathWorks Support Team]

The ability to do a direct matrix multiplication operation on two ‘int32’ variables is not available in MATLAB.

You can use the following code to work around this issue:

function z = mtimes(x,y)

if (isscalar(x) || isscalar(y))

      z = x .* y;

      return;

end

m = size(x,1);

n = size(x,2);

if (n ~= size(y,1))

      error('mmultmanual:size', 'matrix inner dimensions do not agree');

end

p = size(y,2);

z = zeros(m,p,class(x));

for i = 1:m

          z(i,1:p) = sum(bsxfun(@times, reshape(x(i,:),n,1), y), 1);

end

Place the above code in a file called ‘mtimes.m’ and save this file in a directory with name ‘@int32’. This directory must be placed in a directory which is on the MATLAB path but do not add ‘@int32’ to the MATLAB path. This will give you the feature of multiplying two ‘int32’ matrices.

This code has been written only as an example for your use and is not supported by the MathWorks.

Answer 2

[Richard Zapor]

function z = mtimes(x,y)
  if (isscalar(x) || isscalar(y))
      z = x .* y;
      return;
  end
  m = size(x,1);
  n = size(x,2);
  if (n ~= size(y,1))
      error('mmultmanual:size', 'matrix inner dimensions do not agree');
  end
  p = size(y,2);
  z = zeros(m,p,class(x));

%May be faster than bsxfun
  for rx=1:m
   for cy=1:p
     z(rx,cy)=sum(x(rx,:)'.*y(:,cy));
   end
  end
end

Answer 3

[Richard Zapor]

A special case of uint32/64 matrix multiply is where x is a binary matrix and y is a uint32/64 vector.

This can be calculated very quickly and outputs a uint32/64 vector.

function z=bintimes64(x,y)
% z=x*y  [m,n]*[n,1], where x is a binary matrix(real/int) and y is a vector uint64/uint32
% no error checking included
  [m,n]=size(x);
  z = zeros(m,1,class(y));
  for j=1:n
   v=x(:,j)==1;
   z(v)=z(v)+y(j);
  end
end