If you cut 50% of the sine wave you still need to include the samples where the value is 0 in the calculation - you're trying to compute energy over time, if you ignore some of the time, your result will be incorrect. Remember calcuating RMS is to give you the equivalent DC voltage that will cause the same heating effect in a resistor. You can optimise your code to skip the squaring function if the input value is zero to speed things up a little, but you're just saving a multiply (4 cycles) and a 24bit add (~18 cycles). Doing the square root in assembly is not too difficult - it's been done before! Personally I'd do the whole thing in C.