The following fixes an old vectorizer issue with realignment support
(thus only powerpc is affected) and BB vectorization. The realignment
token is set up from the wrong data-ref which causes an SSA verification
failure but in other circumstances might simply generate wrong code.
Bootstrap running on x86_64-unknown-linux-gnu, I'll install this
as obvious on trunk.
PPC folks - you know best how to appropriately test a target
where we use the re-alignment optimization. IIRC on later
powerpc hardware this isn't exercised at all since we can use
unaligned accesses.
The issue is at least present on the GCC 9 branch as well but I'd
appreciate testing where it exercises the path before considering
a backport.
Thanks,
Richard.
2019-09-17 Richard Biener <rguent...@suse.de>
PR tree-optimization/91790
* tree-vect-stmts.c (vectorizable_load): For BB vectorization
use the correct DR for setting up realignment.
Index: gcc/tree-vect-stmts.c
===================================================================
--- gcc/tree-vect-stmts.c (revision 275747)
+++ gcc/tree-vect-stmts.c (working copy)
@@ -9169,7 +9169,9 @@ vectorizable_load (stmt_vec_info stmt_in
|| alignment_support_scheme == dr_explicit_realign)
&& !compute_in_loop)
{
- msq = vect_setup_realignment (first_stmt_info, gsi, &realignment_token,
+ msq = vect_setup_realignment (first_stmt_info_for_drptr
+ ? first_stmt_info_for_drptr
+ : first_stmt_info, gsi, &realignment_token,
alignment_support_scheme, NULL_TREE,
&at_loop);
if (alignment_support_scheme == dr_explicit_realign_optimized)
