re PR target/65787 (Miscompile due to bad vector swap optimization for little endian)
authorBill Schmidt <wschmidt@linux.vnet.ibm.com>
Fri, 17 Apr 2015 22:05:12 +0000 (22:05 +0000)
committerWilliam Schmidt <wschmidt@gcc.gnu.org>
Fri, 17 Apr 2015 22:05:12 +0000 (22:05 +0000)
commit67e6c405298dcb1539e7b1670018cf41bbdec6b6
tree7f4f848d7e27ca403568e9bd53c5e2b11dc77fc7
parent46f120ca7bc8b4a49c0d5a596b85aeca4a254698
re PR target/65787 (Miscompile due to bad vector swap optimization for little endian)

[gcc]

2015-04-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

PR target/65787
* config/rs6000/rs6000.c (rtx_is_swappable_p): Ensure that a
subsequent SH_NONE operand does not overwrite an existing *special
value.
(adjust_extract): Handle case where a vec_extract operation is
wrapped in a PARALLEL.

[gcc/testsuite]

2015-04-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

PR target/65787
* gcc.target/powerpc/pr65787.c: New.

From-SVN: r222205
gcc/ChangeLog
gcc/config/rs6000/rs6000.c
gcc/testsuite/ChangeLog
gcc/testsuite/gcc.target/powerpc/pr65787.c [new file with mode: 0644]