IntervalArithmetic.jl

IntervalArithmetic.jl copied to clipboard

Currently in the `:accurate` directed rounding mode, we round the results of operations outward by adding or subtracting 1 ulp. However, different functions at different precisions may have different numbers...

dpsanders

`setrounding` is slow and unreliable

23

`setrounding` works by redefining a lot of methods. I suppose that can be good for performance, but has the serious costs of (1) `setrounding` itself is very slow (including file...

JeffBezanson

Bugs with tiny numbers

34

```jl julia> tiny = interval(0, floatmin()) [0, 2.22508e-308] julia> huge = interval(floatmax(), Inf) [1.79769e+308, ∞] julia> tiny * tiny [0, 0] julia> huge * huge [1.79769e+308, ∞] julia> huge /...

dpsanders

Use IntervalSets.jl for set operations

2

cf. discussion in https://github.com/JuliaMath/IntervalSets.jl/pull/55 This is related to redefining `..` to *not* do clever tricks and just take floating-point endpoints at face value.

dpsanders

Float interval + BigFloat interval should be Float interval?

5

`x::Interval{Float64}` + `y::Interval{BigFloat}` should give `Interval{Float64}`?

dpsanders

Print intervals as `3..4`?

7

Maybe it would be less confusing to print out intervals as `3..4` instead of `[3, 4]`?

dpsanders

API for choosing power function

13

We need to decide on an API for how to specify whether powers will use the slow but accurate version (via MPFR) or the fast but less accurate version (via...

dpsanders

See [this comment](https://github.com/JuliaIntervals/IntervalArithmetic.jl/issues/355#issuecomment-581478308).

mforets

Speed up pow by 4x or more

19

dpsanders

Rethink ..

9

I believe this has been raised before, but I think `a..b` should no longer do "magic" rounding. Apart from the performance aspect, this is to bring this usage into line...

dpsanders