Exponential
The inverse of logarithm. Takes a 12-bit floating-point number and produces a 35-bit integer. Reverses a logarithmic compression or calculation. Accurate to 10 bits.
*.
16-bit fraction multiply. Leaves multiplicand on stack
2*d
Standard 36-bit left shift. Carry should be 0
split
Extract the exponent from the floating number
ex2
- Save the exponent
- Mask the fraction
- Shift fraction to be 16 bits
- Compute cubic polynomial of fraction. Coefficients from Hart
- Shift fraction to 12 bits. Coefficients could be scaled to accomplish this
- Replace multiplicand with high-order 0
shift
Shift is only defined as a comment.
- If exponent negative, shift right
- Otherwise shift left. If carry is not 0, it adds negligibly to noise