 |
|
| Computers Forum Index » Computer Architecture - Arithmetic » Markstein, Cocanougher... and Metropolis. |
|
Page 1 of 1 |
|
| Author |
Message |
| John Savard |
 Posted: Mon May 30, 2005 1:00 pm |
|
|
|
Guest
|
On 29 May 2005 20:56:17 -0500, hrubin@odds.stat.purdue.edu (Herman
Rubin) wrote, in part:
Quote: The reason for this was that shifting was slow, so a shift
of 16 was put in the hardware. At that time, the cost of
a shift of n was about n times the cost of a transfer or a
test, so putting in a hardware shift of 16 would reduce the
cost of normalization greatly.
Thank you, but I did realize that each shift would take time, as a step
in doing any arithmetic procedure. Even with improved techniques, such
as barrel shifters, steps are still saved by using a larger basic shift.
If one starts with a 48-bit word, and takes 4 bits for the exponent and
its sign, even with this kind of normalization, one still has a minimum
precision of 28 bits, which is enough for single precision.
John Savard
http://www.quadibloc.com/index.html
_________________________________________
Usenet Zone Free Binaries Usenet Server
More than 120,000 groups
Unlimited download
http://www.usenetzone.com to open account |
|
|
| Back to top |
|
|
|
|
|
All times are GMT
The time now is Wed Mar 17, 2010 5:18 am
|
|