Floating point changes

System support for floating point (both binary floating point and decimal floating point) in IBM® i 7.2 is changed to correct defects and improve performance. Some floating point operations produce different results and some raise different exceptions than prior releases. Here are a few examples:

• Conversions between binary floating point and decimal forms (including decimal floating point) are changed to improve accuracy and correctly handle all rounding modes. Floating point results might differ from prior releases by as much as eight units in the last place.
• The modulus operation (used for the SQL MOD function) now produces correct results for all binary floating point input values.
• Several decimal floating point operations now produce an MCH1209 - floating point invalid operation exception instead of the (incorrect) MCH1212 - floating point conversion exception produced in prior releases (for example, a request to compute the logarithm of a negative number).