@p @qint @pernia he multiplies 32 bit numbers in 16.16 format. 16x16 comes from the 8086 limitation of only having 16 bit regs to work with, but in this case each product will have 0, 16, or 32 fractional bits.