New results on floating-point
Speaker(s) : Siegfried M. Rump (TU Hamburg & université Waseda)
We show how to optimize a rounding with respect to the
first, or the second, or w.r.t. to both error models.
Moreover, proving techniques are discussed avoiding
digging in bits and pieces but restricting proofs to
the application of inequalities. The proofs become
more clear and are sometimes much shorter.
More details here
Marc.Mezzarobba (at) nulllip6.fr