std.math: added frexp for double and float

[9il]

For speed.

[dnadlinger]

Pinging @ibuclaw for review. Hopefully I'll be able to look into the intrinsics side of things soon.

[ibuclaw]

Looks fine to me.

[quickfur]

Nitpick: please indent the sig constraint by 4 spaces.

[9il]

Can I remove prototypes and leave only template?

[quickfur]

Prototypes?

[9il]

This stuff:

///ditto
double frexp(double value, out int exp) @trusted pure nothrow @nogc
{ return frexpImpl(value, exp); }

[9il]

For example isNaN is a template.

[quickfur]

You could try it, but I remember recently there was a problem where the compiler autopromoted everything to double (or was it real?), resulting in inefficient generated code. You'll have to check the disassembly to make sure this isn't happening if you remove the "prototypes".

[9il]

A have checked 32/64 bit, with and without -O. All works fine.

[quickfur]

LGTM.

[quickfur]

Auto-merge toggled on

[9il]

Thanks!