Primera página Regresar Continuar Última página Imagen
Floating point numbers are already normalized. We can skip the normalization step and use the exponent as j. But we must normalize the result. Similar complexity.
Q: FP values are normalized using 1.x and this core uses 0.x. Is that the same?
- A: That's why we used 0.24 instead of 0.23. The only adjust is in j where we already add a constant to balance the error.
Q: FP values have 23 bits of mantissa. Why the output is 0.26?
- A: To get 23 bits without error.