compiler-rt: add floatditf and floatunditf

[emaste]

Summary

These are long integer (di_int/du_int) to quad precision floating point conversions.

[ed]

contrib/compiler-rt/lib/builtins/floatditf.c

12 ↗

(On Diff #4509)

Out of curiosity, is this also what this function is supposed to do unconditionally? These functions are not supposed to integrate with fegetround() / FLT_ROUNDS?

47 ↗

(On Diff #4509)

These functions do not support subnormal floating point numbers. Should this be implemented, or at least documented?

[emaste]

contrib/compiler-rt/lib/builtins/floatditf.c

12 ↗

(On Diff #4509)

Good question; this was inherited from the floatsitf function I started from.

The thread I started on llvmdev is here:
http://lists.cs.uiuc.edu/pipermail/llvmdev/2015-March/084064.html

47 ↗

(On Diff #4509)

I'm missing something - how would you implement int to subnormal float conversion?

[ed]

[andrew]

Any plan to commit this? It will be needed for the arm64 buildworld.

[emaste]

Closed by commit rS281223 (authored by @emaste).