The Power of the NORC 3756
in part:
On the Motorola computers, they put 80-bit extended in a 128-bit storage cell, leaving the rest empty.
I tend to trace the lineage of IEEE-754 this way:
The IBM 704 begat...
a) The PDP-10, and
b) The Univac 1107, which, instead of having a duplicate of the exponent in the second word, put every bit to good use - extending the precision of the mantissa a bit, but also using some of the extra bits to make the exponent larger.
The PDP-10 begat...
in part: In that case, you *will* know more about this sort of thing than I do... The 6...
The PDP-11, which used floating-point formats as similar as possible to the PDP-10. But since going from 36 bits down to 32 bits reduced floating-point precision, they clawed one bit back by suppressing the first bit of the mantissa.
So it seems to be based on something 32-bit from DEC, and something 36-bit from Univac, both of which trace their ancestry to something 36-bit from IBM.
If there *is* a "32-bit mistake" from IBM that contributed to IEEE-754 (it seems to be a *conscious rejection* of everything to do with the System-360 floating-point format) I would be interested in learning about it.
CONTROL's program was functionally equivalent to an empty program. It was simply the optimizer to work it out. Oh you had not seen a similar program...
John Savard Usenet Zone Free Binaries Usenet Server More than 140,000 groups Unlimited download
Alt Folklore Computers from Newsgroups