Re: Question about shuttle / ISS

On Fri, 15 Sep 2006 16:13:38 -0500, Dr John Stockton wrote
(in article <qpomWQECexCFFwoZ@xxxxxxxxxxxxxxxxxx>):

Rightly so. Both NASA and you should be able to multiply 25.4 by 12 and get
the *exact* value of 304.8 mm for an Imperial Foot (and say 300 mm when
appropriate).

The value of 304.7999 can be obtained by doing the calculation in IEEE
Doubles (25.4 cannot be represented exactly in a Double) and then truncating
instead of rounding to the desired number of digits.

More to the point (and at risk of puncturing an apparently-smug
SI-superiority complex), how often is a computation error of such
magnitude at all relevant in any project where measurements on the
scale of feet are used? Hint: not very many. The solution (besides
rigorous cross-checking) is a well-established set of dimensioning and
tolerancing standards for the drawings, skilled drafters and checkers,
and engineers who actually *think* about what they're doing.

--
Herb Schaltegger
"You can run on for a long time . . . sooner or later, God'll cut you
down." - Johnny Cash
<http://www.angryherb.net>

.

Relevant Pages

  • Re: Possreps and numeric types
    ... As you can verify easily with hand calculation, ... basically *all* languages that rely on the Intel ... Floating Point architecture). ... programmer demonstration that IEEE 754/854 floating point implementation ...
    (comp.databases.theory)
  • Re: Formatted IO of huge/tiny values
    ... radix-10 will still accumulate error ... it won't be so for the IEEE decimals. ... a long calculation just to end up keeping even. ...
    (comp.lang.fortran)
  • Re: Working with floating point values
    ... If precision is an issue in your calculation, then you have to use decimal. ... otherwise, keep using doubles is fine, and use the right output format to ... And the time it would take to convert to string, ...
    (microsoft.public.dotnet.languages.csharp)
  • Re: status of quadruple precision arithmetic in g95 and gfortran?
    ... Just because 1.5 is exactly representable in IEEE double does not ... mean that all the numbers involved in the calculation are. ... particular loop. ...
    (comp.lang.fortran)
  • Re: [9fans] fun and scary evil C code
    ... They're actually using that stuff to pick apart the (presumed IEEE 754) ... doubles into exponent, mantissa, ... ...
    (comp.os.plan9)