Yeah compensated is a great foundation. Bear in mind that is can’t do as extreme a dynamic range as a vanilla quad double but will be good for extra precision.

On Fri, Aug 30, 2019 at 1:53 AM Henning Thielemann <lemming@henning-thielemann.de> wrote:

On Fri, 30 Aug 2019, Dannyu NDos wrote:

> Could we implement IEEE quadruple-precision (binary128) format?

I think this is a pure Haskell implementation of the idea of representing
high precision floats as sums of low precision floats:
https://hackage.haskell.org/package/compensated
_______________________________________________
Libraries mailing list
Libraries@haskell.org
http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries