todd@hpcuhd.HP.COM (Todd Poynor) (02/21/91)
The HP 1000 single-precision floating point format, a 4-byte representation which corresponds to the FORTRAN "REAL*4" data format, is: +--- implied binary point 1|1 5v4 0 +-+-----------------------------+ | | fraction bits | word 1 +-+-----------------------------+ ^ | +---- sign of fraction 1 5 8 7 1 0 +---------------+-------------+-+ | fraction bits | exponent | | word 2 +---------------+-------------+-+ ^ | sign of exponent ------+ -39 Single-precision format has an approximate range of 1.47 x 10 to 38 1.70 x 10 , with a 23-bit mantissa and a 7-bit exponent. Significance is 6.6 to 6.9 decimal digits, depending on the magnitude of the leading bits/ 6.6 6.9 digits in the fraction (that is, one part in 10 to 10 ). If either the mantissa or the exponent is negative, that part is stored in two's complement form. Lotsa' luck! Todd "still an RTEsian after all these years"