i can't believe i completely missed out 2 critical stages: normalisation of the mantissa prior to the multiplication. it's supposed to be for when the exponent is (in fp32) -126, very close to zero. i'll fix it tomorrow. l.