Floating-point

The C3D format supports a single-precision floating-point format stored in 32 bits (two words) in the C3D file – this is called REAL*4 in FORTRAN documentation and REAL the original C3D documentation.  Each C3D file processor type (DEC, SGI/MIPS and Intel) defines a slightly different internal floating-point format. Intel and SGI/MPIS use the IEEE-754 format, stored in little endian for Intel and big endian format for SGI/MIPS processors.

The DEC floating-point format has the same mantissa with "hidden 1 bit", an offset binary exponent to the left of the mantissa, but when the numbers are negative, DEC stores the value as the 2's complement of the positive value. This means that the DEC format can only store a zero with no sign associated because, unlike the Intel format, there is no ability to store both positive and negative zero representations.