[Bug tree-optimization/68776] New: [6 regression] gcc.dg/vect/vect-widen-mult-const-[su]16.c failures since r226675

[wschmidt at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68776

            Bug ID: 68776
           Summary: [6 regression]
                    gcc.dg/vect/vect-widen-mult-const-[su]16.c failures
                    since r226675
           Product: gcc
           Version: 6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: wschmidt at gcc dot gnu.org
                CC: rguenth at gcc dot gnu.org
  Target Milestone: ---
              Host: powerpc*-*-linux*
            Target: powerpc*-*-linux*
             Build: powerpc*-*-linux*

Created attachment 36946
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=36946&action=edit
Vectorization dump for r226674

Since r226675, we have been seeing these failures:

FAIL: gcc.dg/vect/vect-widen-mult-const-s16.c -flto -ffat-lto-objects
scan-tree-dump-times vect "pattern recognized" 2
FAIL: gcc.dg/vect/vect-widen-mult-const-s16.c scan-tree-dump-times vect
"pattern recognized" 2
FAIL: gcc.dg/vect/vect-widen-mult-const-u16.c -flto -ffat-lto-objects
scan-tree-dump-times vect "pattern recognized" 2
FAIL: gcc.dg/vect/vect-widen-mult-const-u16.c scan-tree-dump-times vect
"pattern recognized" 2

Comparing the vect-details dumps from r226674 to r226675, I see these as
the reason:

63a64,66
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:16:3: note: vect_recog_mult_pattern: detected:
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:16:3: note: patt_47 = _6 << 2;
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:16:3: note: pattern recognized: patt_47 = _6 << 2;
70a74,76
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:16:3: note: vect_recog_mult_pattern: detected:
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:16:3: note: patt_40 = _6 << 1;
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:16:3: note: pattern recognized: patt_40 = _6 << 1;

747a754,756
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:31:3: note: vect_recog_mult_pattern: detected:
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:31:3: note: patt_47 = _6 << 2;
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:31:3: note: pattern recognized: patt_47 = _6 << 2;
754a764,766
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:31:3: note: vect_recog_mult_pattern: detected:
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:31:3: note: patt_40 = _6 << 1;
> /home/wschmidt/gcc/gcc-mainline-base/gcc/testsuite/gcc.dg/vect/vect-widen-mult-const-s16.c:31:3: note: pattern recognized: patt_40 = _6 << 1;

These do not show up for x86_64, so the DejaGnu patterns:

/* { dg-final { scan-tree-dump-times "pattern recognized" 2 "vect" { target
vect_widen_mult_hi_to_si_pattern } } } */

have different results for these different targets, even though both support
vect_widen_mult_hi_to_si_pattern.

Detailed vectorization dumps from powerpc64le-unknown-linux-gnu before and
after this revision are attached.