[libgfortran,patch] Use type-generic built-ins in float I/O routines

[FX]

This patch makes us use type-generic built-ins for isfinite(), isnan(), and signbit(), through their “safe” definitions in libgfortran.h, rather than manually handle the suffixed variants (isfinitef/isfinite/isfinitel/finiteq, and the like). This reverses the patch that originally introduced this suffix handling (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57496), because the built-ins have since been made type-generic in the C front-end (the last one last week, by me).

This is an improvement in code readability and size, and might also yield minor performance gains: the built-ins are expanded inline, rather than requiring a function call.

Bootstrapped and regtested on x86_64-apple-darwin15.
OK to commit to trunk?

FX



-------------- next part --------------
A non-text attachment was scrubbed...
Name: float.ChangeLog
Type: application/octet-stream
Size: 203 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20150824/cdc64bd2/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: float.diff
Type: application/octet-stream
Size: 1735 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20150824/cdc64bd2/attachment-0001.obj>