[Patch] Turn -fexcess-precision=fast on when in -ffast-math
commit9660cb79f2c10cac66a1a635ea6cc3c22f29f99b
authorjgreenhalgh <jgreenhalgh@138bc75d-0d04-0410-961f-82ee72b054a4>
Tue, 20 Dec 2016 16:17:27 +0000 (20 16:17 +0000)
committerjgreenhalgh <jgreenhalgh@138bc75d-0d04-0410-961f-82ee72b054a4>
Tue, 20 Dec 2016 16:17:27 +0000 (20 16:17 +0000)
treeb7b1211e081da6609886060d597693664c100be5
parent3e677d45627d30c9d031ab2b0ba3cb9e2c235111
[Patch] Turn -fexcess-precision=fast on when in -ffast-math

* common.opt (excess_precision): Tag as SetByCombined.
* opts.c (set_fast_math_flags): Also set
flag_excess_precision_cmdline.
(fast_math_flags_set_p): Also check flag_excess_precision_cmdline.
* doc/invoke.texi (-fexcess-precision): Drop text saying the
option has no effect under -ffast-math, make it clear that
-ffast-math will cause -fexcess-precision=fast by default even for
standards compliant modes.
(-ffast-math): Document that this sets -fexcess-precision=fast.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@243824 138bc75d-0d04-0410-961f-82ee72b054a4
gcc/ChangeLog
gcc/common.opt
gcc/doc/invoke.texi
gcc/opts.c