Floating point numbers for PDP computers are stored in either
2 words (totaling 32 bits) or 4 words (totaling 64 bits).

The normalized bit is assumed to be 1 unless the exponent is 0 (corresponding to 2 to the 128) in which case that bit is assumed to be 0. The value 0 is represented by two (or four) words of all zeros.

The normalized bit is assumed to be 1 unless the exponent is 0 (corresponding to 2 to the 128) in which case that bit is assumed to be 0. The value 0 is represented by two (or four) words of all zeros.

The following is the representation of a *2-word floating point number:*

PDP-11 Floating Point Representations:
(Word 1 of 2) |
|||||||||||||||

15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |

Sign Bit | Exponent | High-order fraction |

PDP-11 Floating Point Representations:
(Word 2 of 2) |
|||||||||||||||

15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 |

Low-order fraction |

NOTE: The exponent is biased by 128.

For instance, 2.0 is 40200 in octal or 0100000010000000 in binary (remember the 128 bias and the hidden bit), plus a following 16-bit word of all zeros.

** Four word floating point numbers** add two more words for the
fraction of decreasing significance and increasing precision.