Re: Sun F95 NaN's variable to negative exponent
From: Dr. Richard E. Hawkins (hawk_at_slytherin.ds.psu.edu)
Date: 10/08/04
- Next message: Dr. Richard E. Hawkins: "Re: Sun F95 NaN's variable to negative exponent"
- Previous message: B52B: "Re: CVF bug ?"
- In reply to: Richard Maine: "Re: Sun F95 NaN's variable to negative exponent"
- Next in thread: Jugoslav Dujic: "Re: Sun F95 NaN's variable to negative exponent"
- Reply: Jugoslav Dujic: "Re: Sun F95 NaN's variable to negative exponent"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Fri, 8 Oct 2004 14:47:37 +0000 (UTC)
In article <87k6u48yv1.fsf@vega.site>,
Richard Maine <nospam@see.signature> wrote:
>hawk@slytherin.ds.psu.edu (Dr. Richard E. Hawkins) writes:
>
>> f=a*r*(b*r-one)**(alpha-one) - (b*r-one)**(alpha-one) +c*r**alpha
>...
>> It doesn't seem to like *any* fractional exponents to a*r*(b*r-one):
>...
>> Am I missing something obvious (again)?
>I don't see any "exponents to a*r*(b*r-one)" in any of the stuff
>you showed. Are you perhaps missing the fact that
> a*r*(b*r-one)**(alpha-one)
>parses as
> a*r*((b*r-one)**(alpha-one))
Yep :) After losing much of the day to a calculus error, I don't think
I was seeing well.
However, now I'm still scratching my head: that code never should have
run with some of my past values, either, as near as I can tell.
*sigh*
And the really ugly epilog: it was all for naught. I was trying to use
values matching what the authors had suggested in a suggestion. I
finally realized that to get the values they suggested, one had to
satisfy three equations in two unknowns. Ugh.
>The business abouit not liking any factional exponents sounds like
>what would be expected for negative numbers. It is illegal to raise
>a negative number to a real power...even if the real happens to
>have an exact integer value, though some compilers have been known
>to quietly "fix" things by turning, for example, x**2.0 into
>x**2. That "fix" sometimes only applies when the exponent is known at
>compile time.
It was a compile time constant. Is it conceivable that values of -.5
and -1.5 would have been somehow fixed?
I have discovered that I have to use different starting values for
Householder's iteration (above or below 1, depending upon where the
solution lies).
Thanks
hawk
-- Richard E. Hawkins, Asst. Prof. of Economics /"\ ASCII ribbon campaign dochawk@psu.edu 111 Hiller (814) 375-4846 \ / against HTML mail These opinions will not be those of X and postings. Penn State until it pays my retainer. / \
- Next message: Dr. Richard E. Hawkins: "Re: Sun F95 NaN's variable to negative exponent"
- Previous message: B52B: "Re: CVF bug ?"
- In reply to: Richard Maine: "Re: Sun F95 NaN's variable to negative exponent"
- Next in thread: Jugoslav Dujic: "Re: Sun F95 NaN's variable to negative exponent"
- Reply: Jugoslav Dujic: "Re: Sun F95 NaN's variable to negative exponent"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|