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

Re: Motion 4: revised wording



> Date: Fri, 10 Apr 2009 14:35:11 -0400
> To: stds-1788                           <stds-1788@xxxxxxxxxxxxxxxxx>
> From: Michel Hack                                 <hack@xxxxxxxxxxxxxx>
> Subject: Motion 4: revised wording
> 
> I agree with the rationale, but what bothers me is that the motion
> itself appears to exclude variants of 754-2008 explicitly designed
> to simplify support of IA.  Such a deviation would most likely be
> under some configuration control, e.g. a new FP control bit that
> says "in directed rounding modes, do this rather than that", so
> that most of ordinary FP arithmetic (with default rounding) would
> not be affected.  Nevertheless, setting this bit would technically
> render the system non-754-compliant, *unless* 754-2013 explicitly
> supports that mode...

	I'm not quite sure what you have in mind here, Michel, but
	it doesn't sound like anything that violates 754.  Indeed,
	if you have in mind special purpose hardware for support of
	intervals, you will recall the 'static' operation language
	in 754-2008 for things like operations with a specified
	rounding attribute to make this sort of thing easier than
	it was in 754-1985.

> 
> Now if the motion said "except for minor deviations specifically
> designed to help IA, or known not to affect IA", it would be ok.
> The second caveat would cover things like missing subnormals.
> 
> Michel.
> ---Sent: 2009-04-10 18:45:15 UTC

	As I mentioned to John, I am reluctant to do too much
	wordsmithing on something that is intended merely to guide
	us & not appear in the standard itself.

	I think you have a pretty clear idea of what is intended.
	Details of 'minor deviations' or subnormals are just that,
	details.  Once we can confine our discussion to 754 we
	will have a context within which to make those detailed
	decisions.

	But later.  This is not about that.


				Dan