On 6/13/2016 3:33 AM, Ola Fosheim Grøstad wrote: > But would it really have an effect if I wrote a DIP on getting predictable > floating point behaviour? If there is a chance that it would, then I might > consider it :-). I encourage you to consider it.