match.pd: Allow associating FLOAT_TYPE_P when flag_associative_math.
authorRichard Biener <rguenther@suse.de>
Fri, 26 Jun 2015 07:50:12 +0000 (07:50 +0000)
committerRichard Biener <rguenth@gcc.gnu.org>
Fri, 26 Jun 2015 07:50:12 +0000 (07:50 +0000)
commit7318e44f9f27debf0d97a03a465dc602fcd96443
treedbb5b991aae627683d1801a668edd10a4587bca7
parent6c1a2b3d11403f75927cd065580acf1678c32368
match.pd: Allow associating FLOAT_TYPE_P when flag_associative_math.

2015-06-26  Richard Biener  <rguenther@suse.de>

* match.pd: Allow associating FLOAT_TYPE_P when flag_associative_math.

* gfortran.dg/reassoc_3.f90: Adjust.

2015-06-26  Richard Biener  <rguenther@suse.de>

* match.pd: Allow (p +p off1) +p off2 to (p +p (off1 + off2))
irrespective on whether the inner operation has a single use
of both off are constant.

From-SVN: r224995
gcc/ChangeLog
gcc/match.pd
gcc/testsuite/ChangeLog
gcc/testsuite/gfortran.dg/reassoc_3.f90