I have not done a thorough microscopic evaluation of IEEE Std 754-2008 in this regard, but what I have seen so far does not lead me to confidence that conversion from or to anything but decimal CHARACTER representations is included or described in the specifications. I am investigating the possibility of a "strict compliance" option for a future revision (not the one currently under development) or possibly a "sooner" amendment to that proposed standard. |
COBOL has historically had various forms of "packed-decimal" encodings -- typically representing 4-bit decimal digits in the range 0-9 with nondigit values used for the sign and typically otherwise invalid. Note also that this definition is decidedly "loose" in 2002 COBOL. For example, directly using "declets" in an implementor-specified format would qualify as "packed-decimal" under the current standard, for example.
1) For conversion of values to and from IEEE formats, is the specification, in the view of the IEEE community that the information in such field will be "treated as if" it had first been converted to decimal CHARACTER format, and THEN converted to a floating-point format "on the way in", and converted from a floating-point format to a decimal CHARACTER format and from there converted to the "packed-decimal" form "on the way out", an ADEQUATE specification for conformance with IEEE Std 754-2008?
2) Is there a specification in IEEE Std 754-2008 that would contradict such an approach?