Formal verification of a floating-point expansion renormalization algorithm

Abstract

Many numerical problems require a higher computing precision than the one offered by standard floating-point formats. A common way of extending the precision is to use floating-point expansions. As the problems may be critical and as the algorithms used have very complex proofs (many sub-cases), a formal guarantee of correctness is a wish that can now be fulfilled, using interactive theorem proving. In this article we give a formal proof in Coq for one of the algorithms used as a basic brick when computing with floating-point expansions, the renormalization, which is usually applied after each operation. It is a critical step needed to ensure that the resulted expansion has the same property as the input one, and is more “compressed”. The formal proof uncovered several gaps in the pen-and-paper proof and gives the algorithm a very high level of guarantee.