Just to add my 2 cents to this discussion, DFP numbers retain zero digits to 
both the left and right of the radix point.   This is different from both 
binary floating point and hexadecimal floating point where for normalised 
numbers, there is a single representation of a value (this has the advantage of 
increasing the precision of the fraction).

DFP introduces the concept of  'cohort' - the set of different representations 
of the same value - and 'quantum' - the magnitude of the rightmost digit in the 
significand (fraction).   Each member of a cohort will have a different 
quantum.  
As a result of cohorts, it is important that any comparison of DFP numbers use 
the compare DFP instructions - even zeroes may have differing quantums so a 
comparison of the binary representations would not give the correct result

The Extract Biased component instruction allows you to determine the quantum of 
a value.  While the Quantize instruction adjusts the quantum -  this might be 
used where the value is say a balance in dollars and you just want to keep 2 
digits to the right of the decimal point. 

Kind regards 
John Dravnieks

Reply via email to