This math scares me

[John W. Baxter]

In article <3AAEAB92.DC8E984A at san.rr.com>, Darren New <dnew at san.rr.com> 
wrote:

> Grant Edwards wrote:
> > Of course we'd never imagined we'd ever see hardware FP and BCD
> > libraries were only marginally slower than binary ones. Now
> > that FP is usually in hardware, BCD FP (or fixed point) will be
> > pretty slow by comparison.
> 
> One mainframe I worked on (Xerox 560, AKA Sigma 9) had the "scientific
> module" which was hardware floating point, and the "business module" which
> was the hardware BCD. The latter added instructions like "Edit byte string"
> that would copy packed BCD into a string according to COBOL display-edit
> formatting. (Or whatever COBOL called the "print using" formatting. Too many
> years. Not enough COBOL. Or something like that.)

Sigmas were quite nice.

And the NCR C-315 line had a company-supplied Fortran with (obviously!) 
floating point routines, and a user group supplied set of floating point 
libraries which actually worked, which most sites used.  The 315 was a 
decimal machine (BCD)...a word was 12 bits...either 2 characters or 3 
digits.  The accumulator was as long as it needed to be (up to 12 slabs) 
to contain the current value...there was a T@ register (Tally of 
Accumulator) which held its current length.

Words were called syllables (ergo  slabs) because someone in engineering 
had thought 12 bits too short to be dignified as a word.

  --John

-- 
John W. Baxter   Port Ludlow, WA USA  jwbnews at scandaroon.com