Improve Decimal Floating Point Implementation

[srorso]

As part of implementing the Floating Point Extension Facility, the following additions are needed to decimal floating point support:

• The rounding mode Round For Shorter Precision needs to be added to decNumber
• Eight new DFP instructions need to be added to dfp.c: CDPT, CDZT, CPDT, CPXT, CXPT, CXZT, CZDT, CZXT (Convert to/from Zoned/Packed)
• Twelve instructions need to be changed to accept an M4 rounding modifier field: ADTRA, AXTRA, CDGTRA, CGDTRA, CGXTRA, CXGTRA, DDTRA, DXTRA, MDTRA, MXTRA, SDTRA, SXTRA

[srorso]

In addition to the additions noted above, a review of results generated by Hercules versus hardware shows that the following additions are also required:

• Support for the return of scaled results on trappable overflow or underflow. While such results are not required by IEEE 754 2008, they are consistent with that standard's allowance for the return of an alternate result on underflow or overflow. And scaled DFP results make the DFP arithmetic suite consistent with the BFP suite.

• Support for the Quantum exception.

Both of these items are present in the Principles of Operation SA22-7832-10; the Quantum exception is not present in earlier PoO documents.