Mail Archives: djgpp/1993/08/26/04:13:52
Does anyone have a package for exact arithmetic, which computes a result and
*guarantees* its exactness within bounds that are computed by that package?
I'm thinking of something like ARITHMOS (that's a FORTRAN 77 subroutine
package for Siemens BS2000) or something like that. First, you have
interval arithmetic available and second you can demand a calculation (for
instance, matrix inversion) and as a result, you get a matrix of intervals
the entries of which are guaranteed to hold the according elements of the
*real* solution. Furthermore, it is assured that this result cannot be
computed more precisely on that machine with that algorithm... I read about
such packages as a base for pascal systems, but I'm searching one for gcc.
Any hints?
Martin (haltmayer AT uni-augsburg DOT de)
- Raw text -