exact dot product
Ulrich Kulisch wrote:
> Techniques for computing /close/ bounds for solutions should also be
> considered in IEEE P1788.
I agree, and this is a new strong argument for including exact dot product.
It is the same argument that I defended for explicit inclusion of the
requirement for scalar operations with directed rounding: these are
essential for implementing interval library functions -- by which I
mean application-specific libraries, not (just) the basic libraries
containing the functions required by the standard.
My earlier arguments were based on the fact that P1788 was at some point
supposed to be a standard for *dependable floating-point arithmetic*,
which would include complete arithmetic as well as interval arithmetic.
Michel.
---Sent: 2013-05-18 14:38:53 UTC