For multiply (*), one operand can be a floating-point vector and the other a matrix, treating a right vector operand as a column vector and a left vector operand as a row vector, or the two operands can both be matrices. These cases require that the number of columns of the left operand is equal to the number of rows of the right operand .