Odd/Even packed numbers.

["Jon Paris" <Jon.Paris@xxxxxxxxxxxxxx>]

> In the past (at S/38 times) we have been told that odd packed numbers were
> not so efficient as even ones,

I think you have it the wrong way round - it is _odd_ numbers (i.e. those
that will "fill" the bytes allocated to them) that are most efficient.  This
should still be true for risc boxes since the compilers have always
generated additional instructions to ensure that the upper nyble (nibble?)
is set to zero.  The overhead is not that high since the code is only
generated when data is stored not when it is referenced.  Although it may be
cleared always when the RPG IV keyword Pack Even is used - I'm not sure.

You can demonstrate this by overlaying a 9 digit field with an 8 digit.  If
you store hival in the 9 digit field and then reference the data via the 8
digit field -adding/moving it to (say) a ten digit field you will find that
all 9 digits are used by the 8 digit operation.

Jon Paris
Partner400