For some reason when optimization was on with the Intel compiler it
was removing the loop in some instances of the templated MatrixTranspose
function. I inserted an empty assembly statement that prevents the
compiler from removing the loop but does not add any actual code. That
seems to fix the problem.