Re: Odd/Even packed numbers.

> Leif,
> apart from rough testing . . . I don't have rights to use sst and not even
> enough knowledge about RISC instructions, but maybe it could be interesting
> to compare how both versions are compiled to see if there are some machine
> instruction added to fix the missing semi-byte.
>

no need for SST, a simple RPG program or MI program or COBOL
program will do. Since I said that it was trivial, I just wrote one
and added packed 123 up 100,000,000 times.
With PKD(14,0) it took 33.0 seconds (on my slow S20)
With PKD(15,0) it took 29.3 seconds,
so, yes, it is still true. there is a 10% difference. Or 3.7 seconds
for one hundred million additions. You must evaluate if that
difference is a show stopper for your application (it wouldn't be
it for mine).