On Wed, 8 Mar 2017 14:32:10 -0600, John McKown wrote:
>>
>> AP goes right to left because it would otherwise have to do more work to
>> propagate carry.
>
>​Right. But it could go to the left if the nybbles in the packed decimal
>number were in reverse order, with the sign nybble being the first
>(leftmost) nybble in the data stream. I.e. instead of 01234F be F43210 .
>But that was likely not acceptable because one reason that programmers love
>packed rather than binary is that they can read it directly in the hex
>dump. Said dump being far more prevalent tool for debugging in the far
>past. Some decisions are not really hardware dictated. They're cultural.
> 
DFP must have been a great disappointment to programmers who expected
it would facilitate reading floating point numbers in dumps.

And appearance of dumps is the only reason I can imagine that packed
decimal is sign-magnitude rather than 10's complement.

o 10's complement would allow a greater range in the same storage.  E.g.
  -5000 to +4999 rather than -999 to +999 in two bytes.

o 10's complement would obviate the need for a recomplement pass half
  the time when adding numbers with unlike signs.

Once I saw a VAX dump.  The ASCII read left-to-right.  The hex had the
higher addresses on the left, so it read as if big-endian.  I.e. where the
ASCII showed ABCD; the hex showed 44434241.

-- gil

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to