Organization: DRD Corporation
Lines: 250
Archive-name: ieee_mr/part01

BEGIN_DOC ieee.doc
This is the documentation for a little directory which I find useful from
time to time when a mysterious bit pattern shows up unexpectedly in my work.
The user-rpl routines in this directory convert back and forth between HP48 
real numbers and HP48 binary number objects.  The binary number objects
contain the bit pattern for the IEEE format for both 4-byte and 8-byte 
floating point reals.

The IEEE directory contains a CST menu which accesses two routines, IEEE4
and IEEE8.  These routines take either an HP48 real number object or an
HP48 binary number object as the single argument.  They perform whichever
conversion is necessary depending on the type of the argument.

For documentation on the IEEE format, see the Microsoft C Bible or any
good PC-type computer language manual.

This is plain vanilla coding, no tricks, nothing impressive.  Very
little error checking is done.  Don't feed these routines bogus values.
The routines will most likely terminate prematurely with an ERROR
condition.  You won't like the condition of your user flags if this
happens.

The binary number object results are rounded up depending on the value
of an extra (lsb) bit of precision.  Binary number object results are
temporarily displayed in HEX format, regardless of HP48 binary mode.

The 4-byte binary representation suffers from a limited precision 
relative to the HP48.  A 4-byte real has at most 7 digits of precision, 
and therefore a temporary display of the real number object will be 
rounded to 7 significant digits.  Because the HP48 has greater precision,
you should eventually get consistent results during cyclic conversions.

The precision of the 8-byte reals (14 digits) exceeds that of the HP48 
(12 digits) and so a problem exists in both conversion directions.  In 
some cases, the HP48 will be unable to produce the correctly rounded
binary number object result (using the current algorithm) since the 
lsb represents a quantity two orders of magnitude smaller than the last 
digit the HP48 can produce at the user level.  Perhaps a future system 
rpl version can handle this deficiency with extended reals, if the 
demand exists.  Real number objects are temporarily displayed with 
11 significant digits.

These routines provide sufficient precision for their intended use,
which is diagnostic in nature.

========================================================================
END_DOC