Skip to content
Convert to/from IEEE-754 HP/SP/DP formats
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

CrackNum: Decode/Encode IEE754 Numbers

Hackage version Build Status

CrackNum now comes with vim bindings, see

Command line options:

crackNum v2.3, (c) Levent Erkok. Released with a BSD3 license.
Usage: crackNum precision bit/hex-pattern
          --hp        16 bit half     precision
          --sp        32 bit single   precision
          --dp        64 bit double   precision
          --sb         8 bit signed   byte
          --sw        16 bit signed   word
          --sd        32 bit signed   double
          --sq        64 bit signed   quad
          --ub         8 bit unsigned byte
          --uw        16 bit unsigned word
          --ud        32 bit unsigned double
          --uq        64 bit unsigned quad
          --toIEEE=n  Convert from decimal to IEEE SP/DP formats.
  -l n    --lanes=n   number of lanes
          --vim       output in vim friendly format
  -h, -?  --help      print help, with examples
  -v      --version   print version info


   crackNum --hp fc00
   crackNum --sp fc00 abcd
   crackNum --dp fc00 abc1 2345 6789
   crackNum --sp 01111111110000000000000000000000
   crackNum -l2 --hp 01111111110000000000000000000000
   crackNum --sb 7f
   crackNum --sp --toIEEE=-2.3e6
   crackNum --dp --toIEEE=max
   crackNum --dp --toIEEE=ulp

  - You can use hexadecimal or binary as input.
  - You can use _,- or space as a digit to improve readability.
  - You can give input for multiple lanes, we will guess the #of lanes for you.
    Or, you can specify number of lanes with the -l option.
  - For "toIEEE" option (case doesn't matter):
        - You can enter a number in decimal notation (like 2.3)
        - You can enter a number in hexadecimal notation (like 0x1.abcp+3)
        - OR, enter one of the following:
               * infinity, -infinity: Positive/Negative infinities
               * nan, snan, qnan: Not-A-Number; signaling/quiet
               * 0, -0: Both kinds of zeros
               * max : The maximum finite positive value
               * -max: The minimum finite negative value
               * min : The minimum normal positive value
               * -min: The maximum normal negative value
               * epsilon: The smallest possible value x s.t. 1+x /= 1.
               * ulp: The minimum subnormal value

Example: Decoding single-precision numbers on two lanes

$ crackNum --sp fc00 abc1 7F80 0001
== Lane: 1 ==========================================
                  3  2          1         0
                  1 09876543 21098765432109876543210
                  S ---E8--- ----------F23----------
          Binary: 1 11111000 00000001010101111000001
             Hex: FC00 ABC1
       Precision: SP
            Sign: Negative
        Exponent: 121 (Stored: 248, Bias: 127)
       Hex-float: -0x1.015782p121
           Value: -2.6723903e36 (NORMAL)
== Lane: 0 ==========================================
                  3  2          1         0
                  1 09876543 21098765432109876543210
                  S ---E8--- ----------F23----------
          Binary: 0 11111111 00000000000000000000001
             Hex: 7F80 0001
       Precision: SP
            Sign: Positive
        Exponent: 128 (Stored: 255, Bias: 127)
       Hex-float: NaN (Signaling)
           Value: NaN (Signaling)
            Note: Representation for NaN's is not unique.

Example: Encoding a float as a IEEE754 single-precision bit-pattern

$ crackNum --sp --toIEEE=-2.3e6
              3  2          1         0
              1 09876543 21098765432109876543210
              S ---E8--- ----------F23----------
      Binary: 1 10010100 00011000110000110000000
         Hex: CA0C 6180
   Precision: SP
        Sign: Negative
    Exponent: 21 (Stored: 148, Bias: 127)
   Hex-float: -0x1.18c3p21
       Value: -2300000.0 (NORMAL)
You can’t perform that action at this time.