Primera página Regresar Continuar Última página Imagen
Normalization
For m=1 and n>=3 the 0.5<=x<1 is ok
In binary: 0.1<=x<1.0
How to find j? E integer bits, D decimal bits
- j=E
- Shift to the left until the MSB=1
- Decrease j in each iteration
Example: E=5, D=3, x=00100.101 (4.625)
- x=00100.101 j=5
- x=01001.010 j=4
- x=10010.100 j=3 => 0.10010100 (0.578125)
Up to E+D-1 clocks, but half takes 0 clocks and only one in 2E+D-1 takes the max.