home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Merciful 1
/
Merciful - Disc 1.iso
/
software
/
m
/
math_vision
/
mathvisionv2.1b.dms
/
mathvisionv2.1b.adf
/
README
/
MV_Precision.doc
< prev
next >
Wrap
Text File
|
1992-02-14
|
2KB
|
52 lines
There are two versions of MathVISION: single precision (FFP) and
double precision (IEEE). The IEEE version can represent a wider
range of numbers and a greater degree of precision than the FFP
version. Both versions of MathVISION are supplied, the FFP version
is on the MathVISION disk and the IEEE version is on the MathVISION2
disk. The drawers MathVISION.ffp and MathVISION.ieee contain the
particular versions.
SINGLE PRECISION
~~~~~~~~~~~~~~~~
MathVISION.ffp uses the Motorola Fast Floating Point library, with a
custom error-trapping interface. The limits of the numbers are:
9.22337377 x 10^18 > +value > 5.42101070 x 10^(-20)
-9.22337377 x 10^18 < -value < -2.71050535 x 10^(-20)
All arithmetic errors are patched up when they are encountered.
For example, -3/0 would produce negative infinity (-.922337377e-19), and
sqrt( - 25 ) would yield 5, although the error is recorded. The Analyze
feature will tell you where the function is in error.
When numbers are displayed, they are converted from scientific notation
to normal fixed point notation unless they are too large. The last digit
is rounded off, to avoid .99999998 problems.
Overflow errors may occur when converting from Fast Floating Point
numbers to integers, as when doing modular division, or as the final step
in getting a Contour value.
DOUBLE PRECISION
~~~~~~~~~~~~~~~~
The IEEE version of MathVISION uses the IEEE Floating Point library.
The approximate limits of the numbers are:
1.8 x 10^308 > +value > 2.2 x 10^(-308)
-2.2 x 10^(-308) < -value < -1.8 x 10^308
When numbers are displayed, they are converted from scientific notation
to normal fixed point notation unless they are too large. The last digit
is rounded off, to avoid .99999998 problems.
Overflow errors may occur when converting from IEEE numbers to integers,
as when doing modular division, or as the final step in getting
a Contour value.
In Contour mode the numbers are rounded.