Readme
bits
Converts input into bits (with explanations, where applicable).
Example
When run with e.g. bits - 0. 3 , the output is
f16: 1011010011001101
SEEEEEMMMMMMMMMM
S: Sign ( 1 bit)
E: Exponent ( 5 bits)
M: Fraction / Mantissa ( 10 bits)
bf16: 1011111010011010
SEEEEEEEEMMMMMMM
S: Sign ( 1 bit)
E: Exponent ( 8 bits)
M: Fraction / Mantissa ( 7 bits)
f32 : 10111110100110011001100110011010
SEEEEEEEEMMMMMMMMMMMMMMMMMMMMMMM
S: Sign ( 1 bit)
E: Exponent ( 8 bits)
M: Fraction / Mantissa ( 23 bits)
f64 : 1011111111010011001100110011001100110011001100110011001100110011
SEEEEEEEEEEEMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMM
S: Sign ( 1 bit)
E: Exponent ( 11 bits)
M: Fraction / Mantissa ( 52 bits)
If only some types are required, the --type =... option can be used, e.g. bin - - type = f16, f32 - 0. 3 which gives
f16: 1011010011001101
SEEEEEMMMMMMMMMM
S: Sign ( 1 bit)
E: Exponent ( 5 bits)
M: Fraction / Mantissa ( 10 bits)
f32 : 10111110100110011001100110011010
SEEEEEEEEMMMMMMMMMMMMMMMMMMMMMMM
S: Sign ( 1 bit)
E: Exponent ( 8 bits)
M: Fraction / Mantissa ( 23 bits)
By default, --display=full is implied, showing all information. If only type and bit information are wished,
--display=short can be used:
f16: 1011010011001101
f32 : 10111110100110011001100110011010
If only the bits are wished for, the --display=very-short format exists, removing even the type information:
1011010011001101
10111110100110011001100110011010
The very short format makes most sense in combination with specific types; in the example above, --type =f16,f32
was used.