Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Motion 45



John Pryce wrote:
> ...must pay respect to the 754 standard.  I suppose CA, when adding a
> whole lot of zeros, doesn't check whether they were all +0, all -0, etc?

Actually, as described in VanSnyderP1788.pdf (in our list of position papers),
it does.

The one implementation I know of, in S/370, did not because IBM Hex FP did
not distinguish signed zeros, and always generated +0.  If the instructions
were to be resurrected for IEEE FP (supported in IBM mainframes in addition
to the original HFP since 1998 for BFP) in my lifetime at IBM, I would of
course insist that they should follow the IEEE 754 rules for +-0.  (I might
balk at supporting BFP128 or DFP128 however -- though perhaps 10KB-wide
accumulators wouldn't faze anybody by then!)

Michel.
---Sent: 2013-08-03 15:30:36 UTC