coq-club AT inria.fr
Subject: The Coq mailing list
List archive
- From: Laurent Théry <Laurent.Thery AT inria.fr>
- To: coq-club AT inria.fr
- Subject: Re: [Coq-Club] Machine floats in Coq
- Date: Sun, 05 Jul 2015 15:33:53 +0200
On 05/07/15 01:11, Eric Mullen wrote:
There's Flocq, which is used by CompCert to formalize floating point:
http://flocq.gforge.inria.fr/
Flocq floats are executable inside Coq but I am not sure it extracts
to something more efficient outside Coq.
As for relating it to rationals, that will be tough. Rational arithmetic isI think it is more associativity that is a problem.
quite fundamentally different than floating point: e.g. addition commutes
in the rationals, but doesn't in the floats.
It would be awesome if this interface also includes a logical
specification of the floating point operations, perhaps in terms of Coq
rationals (Coq.QArith.QArith_base.Q).
Proving anything about floats, you often need to know exactly which kind of floats you are manipulating (precision, format, round modes).
So it makes sense to stick to the IEEE754 standard. I am not sure Haskell is compliant with it.
--
Laurent
- [Coq-Club] Machine floats in Coq, Abhishek Anand, 07/04/2015
- Re: [Coq-Club] Machine floats in Coq, Eric Mullen, 07/05/2015
- Re: [Coq-Club] Machine floats in Coq, Laurent Théry, 07/05/2015
- RE: [Coq-Club] Machine floats in Coq, Soegtrop, Michael, 07/06/2015
- RE: [Coq-Club] Machine floats in Coq, Christophe Bal, 07/06/2015
- Re: [Coq-Club] Machine floats in Coq, Guillaume Melquiond, 07/06/2015
- RE: [Coq-Club] Machine floats in Coq, Soegtrop, Michael, 07/06/2015
- Re: [Coq-Club] Machine floats in Coq, Guillaume Melquiond, 07/06/2015
- RE: [Coq-Club] Machine floats in Coq, Soegtrop, Michael, 07/06/2015
- Re: [Coq-Club] Machine floats in Coq, Bas Spitters, 07/05/2015
- Re: [Coq-Club] Machine floats in Coq, Eric Mullen, 07/05/2015
Archive powered by MHonArc 2.6.18.