We are currently at splitting std.math into sub modules [1]. There the question arose, how to handle the floating point stuff, that's in there: Functions like feqrel, nextDown, isClose but also IeeeFlags and FloatingPointControl.

Some ideas:

a) IeeeFlags and FloatingPointControl goes into std.math.hardware and all else into std.math.floats.

b) The same, but std.math.floats is split into two modules, e.g. std.math.comparison (feqrel, isClose, ...) and std.math.floatmanip (nextDown, ilogb, ...)

c) Moving everything out of std.math and into it's own module, e.g. std.floats.

Any opinions or ideas on this?

[1] https://github.com/dlang/phobos/pull/7924

On Tuesday, 30 March 2021 at 15:04:28 UTC, Berni44 wrote:
> We are currently at splitting std.math into sub modules [1]. There the question arose, how to handle the floating point stuff, that's in there: Functions like feqrel, nextDown, isClose but also IeeeFlags and FloatingPointControl.
>
> Some ideas:
>
> a) IeeeFlags and FloatingPointControl goes into std.math.hardware and all else into std.math.floats.
>
> b) The same, but std.math.floats is split into two modules, e.g. std.math.comparison (feqrel, isClose, ...) and std.math.floatmanip (nextDown, ilogb, ...)
>
> c) Moving everything out of std.math and into it's own module, e.g. std.floats.
>
> Any opinions or ideas on this?
>
> [1] https://github.com/dlang/phobos/pull/7924

You might find interesting mir-core's structure [1].It has mir.math.common, mir.math.constant, and mir.math.ieee. mir-algorithm [2] has some other math functions.

[1] http://mir-core.libmir.org/mir_math.html
[2] http://mir-algorithm.libmir.org/