[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Single and Double Precision



Mr. Frank Morrison kindly offers the following definition regarding Single 
and Double precision:



<...I did not stop to think that the terms single/double precision are not
necessarily known to all computer users.

Here is my definition, for what it may be worth, and I hope it helps.

Numerical calculations on the IBM PC (and MACs) can be done using 
either single precision 32 bit words (4 bytes) or double precision 64 
bit words (8 bytes). 

Calculations can also be done in either integer arithmetic or in 
floating point arithmetic. Integer arithmetic can normally use only 
single precision 32 bit words. Floating point can use either precision.
Floating point numbers are represented in the same fashion that we use
in so-called scientific notation, i. e. 5000 can be written as
.5 time 10 raised to the fourth power, or .5E4. Here the 4 is an 
exponent, called the characteristic, and the .5 is the number shifted 
so that it lies to the right of the decimal point, and is called the 
mantissa. Single precision floating point uses an eight bit 
characteristic and a twenty-four bit mantissa, while double precision
uses a 16 bit characteristic and a 48 bit mantissa. The characteristic 
and mantissa are separated by a hypothetical binary point.

		Sincerely,

			Frank P. Morrison>

Thankyou Frank.

So if 4 bits = 1 byte then I guess that single point requires 2 bytes for 
the characteristic and 7 bytes for the mantissa? Nine bytes for the full 
number? Is this right? It seems like I've got something wrong here.

I went back to the article "FPU Precision" by Oliver Sharp, January 1995 
Byte, which I faxed Howard earlier, and the answer was staring me in the 
face. The definition is similar to what you have given.

PS.
Afterwards emailing you, while editing .newsrec, so I could start actually 
using the newsgroups, I found some comp.xxx listings that might have offered 
an answer, but being a newbie on internet, I'm sure it would have been 
awhile before I had found anything pertinent. I swear I'm going to really 
start using more than just email this week!

Thanks again.
Rick Gleason

Rick Gleason   rgleason@world.std.com
               The Gleason Partnership
               114 Commonwealth Ave.
               Boston, MA 02116
               617-267-6980, 617-267-6864 fax

[New Search][DataCAD Home]