This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/70130] [6 Regression] h264ref fails with verification error starting with r231674
- From: "rguenth at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Wed, 13 Apr 2016 12:31:56 +0000
- Subject: [Bug tree-optimization/70130] [6 Regression] h264ref fails with verification error starting with r231674
- Auto-submitted: auto-generated
- References: <bug-70130-4 at http dot gcc dot gnu dot org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70130
--- Comment #13 from Richard Biener <rguenth at gcc dot gnu.org> ---
Does the following fix it? I'm not entirely happy with it because of the
comment.
Index: gcc/tree-vect-stmts.c
===================================================================
--- gcc/tree-vect-stmts.c (revision 234894)
+++ gcc/tree-vect-stmts.c (working copy)
@@ -6876,6 +6875,12 @@ vectorizable_load (gimple *stmt, gimple_
gcc_assert (!load_lanes_p
|| alignment_support_scheme == dr_aligned
|| alignment_support_scheme == dr_unaligned_supported);
+ /* If vectors are not adjacent we can't do dr_explicit_realign_optimized.
+ ??? This fails to account for the cost of this but it's messy to
+ compute all this from within vect_supportable_dr_alignment. */
+ if (group_gap_adj != 0
+ && alignment_support_scheme == dr_explicit_realign_optimized)
+ alignment_support_scheme = dr_explicit_realign;
/* In case the vectorization factor (VF) is bigger than the number
of elements that we can fit in a vectype (nunits), we have to generate