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

Re: P1788/M0014.01: 6.1_and_6.2 (compatibility with multi-precision)



> Date: Tue, 4 May 2010 13:34:03 +0300
> Subject: Re: P1788/M0014.01: 6.1_and_6.2 (compatibility with multi-precision)
> From: "Hossam A. H. Fahmy" <hfahmy@xxxxxxxxxxxxxxxxxxxxxxx>
> To: stds-1788@xxxxxxxxxxxxxxxxx
> 
> 
> Vincent and P1788 group,
> 
> 2010/5/4 Vincent Lefevre <vincent@xxxxxxxxxx>
> 
> > On 2010-05-02 03:54:20 -0700, Dan Zuras Intervals wrote:
> > >       The 'dubious' nature of my concerns surround the
> > >       problem that mid-rad intervals represent a quite
> > >       different subset of the Real intervals than do the
> > >       inf-sup forms.  I believe that it will require us
> > >       to burden mid-rad forms further to represent these
> > >       intervals (like [1e-100,1e+100] & [3,+oo]) somehow.
> > >       Is it sufficient to represent them as say,
> > >       (5e+99,0,1e+100) & (something+3,-something,+oo)?
> >
> > I think that if the interval is large enough, one can still choose
> > mid = 0. Then the representation is equivalent to inf-sup, isn't it?
> >
> >
> My understanding of the form (mid, del1, del2) that represents  [mid-del1,
> mid+del2]  was that both del1 and del2 are always considered as zero or
> positive numbers (the distance of the radius). However, taking mid=0 for
> large intervals leads to negative or positive delimiters as in:
> 
> [-1e+10, 2e+20] = (0, 1e+10, 2e+20)
> 
> [+1e+10, 2e+20] = (0,-1e+10, 2e+20)
> 
> [-1e+10, -2e+20] = (0, 1e+10, -2e+20)
> 
> Vincent's proposal solves the issue of representation for large intervals.
> However, can the colleagues who use mid-rad representations in their work
> comment on whether they depend on the fact that rad is always non-negative?
> If yes, what is the impact of having two delimiters (not just one) and with
> both being either positive or negative?
> 
> --
> Hossam A. H. Fahmy
> Assistant Professor
> Electronics and Communications Engineering Department
> Cairo University
> Egypt
> 

	Vincent's proposal does seem to solve the problem of
	conversion from inf-sup to mid-rad.  So long as the
	rad components have sufficient precision to hold the
	inf & sup, that is.  And so long as the mid-rad people
	will tolerate having the mid strictly outside the
	resulting interval.

	But I don't know if that's true.  Nor do I know if
	there are any other more subtle problems.

	Remember the mid-rad form is chosen because of its
	utility in the case of very narrow intervals.  There
	may be algorithms that have the flavor of

		principleComponent + tinyAdjustment

	that would fail should the interval be too wide or
	the midpoint too far from the interval itself.

	Still, I trust you mid-rad guys can enlighten us.


				Dan