Start of IEEE 754 support (Discussion)

[ridgeworks]

All good questions.

@jan

As you probably also want the other bound, what about bounds(Exp, Low, High)?

This doesn’t help. A new lower bound is calculated from the interval argument’s lower bounds, and similarly for the upper bound. So the two expressions aren’t the same.

@jan

there is typically more to do when evaluating entire expressions that consist of negation, division and other operations that swap the upper and lower bound. Where is that handled?

This is all handled at the application level which is doing the equivalent of breal arithmetic. Of course, I’m open to the suggestion of adding breals to SWIP I am sympathetic to your view that:

@jan

It feels as a bit of an overkill to add another atomic type to deal with this. ECLiPSe is primarily a constraint system where this often makes sense, SWI-Prolog is not.

But I think that, with the support of attributed variables, you can build effective constraint libraries on top of SWIP so you can take advantage of the rest of SWIP’s considerable eco-system. I believe I can demonstrate that with a CLP over reals implemented entirely in Prolog, but it’s spending upwards of 25% of the time just calculating interval bounds so I’m somewhat motivated to see if anything can be done about that without upsetting the apple cart too much.