[Bug tree-optimization/97334] New: inefficient vectorization of gcc.dg/vect/bb-slp-pr65935.c

[rguenth at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97334

            Bug ID: 97334
           Summary: inefficient vectorization of
                    gcc.dg/vect/bb-slp-pr65935.c
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: rguenth at gcc dot gnu.org
  Target Milestone: ---

For

void rephase (void)
{
  int i,j,k,dir;
  struct site *s;
  for(i=0,s=lattice;i<sites_on_node;i++,s++)
    for(dir=0;dir<32;dir++)
      for(j=0;j<3;j++)for(k=0;k<3;k++)
        {
          s->link[dir].e[j][k].real *= s->phase[dir];
          s->link[dir].e[j][k].imag *= s->phase[dir];
        }

where SLP faces unrolled j and k loops we fail to vectorize the
loads of { s->link[dir].e[j][k].real, s->link[dir].e[j][k].imag }
because SLP discovery rejects the s->phase[dir] load:

/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
note:   Build SLP for _59 = s_8->phase[dir_80];
/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
missed:   Build SLP failed: unvectorizable statement _59 = s_8->phase[dir_80];
/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
note:   Building vector operands from scalars
/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
note:   Build SLP for _59 = s_8->phase[dir_80];
/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
missed:   Build SLP failed: unvectorizable statement _59 = s_8->phase[dir_80];
/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
note:   Building vector operands from scalars

and then ends up building the multiplication operands from scalars:

/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
note:   node 0x4a36cb0 (max_nunits=2, refcnt=2)
/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
note:        stmt 0 _60 = _58 * _59;
..
/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
note:        children 0x4931310 0x4a6e690
/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
note:   node (external) 0x4931310 (max_nunits=1, refcnt=1)
/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
note:        { _58, _59, _59, _59, _59, _59, _59, _59, _59, _59, _59, _59, _59,
_59, _59, _59, _59, _59 }
/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
note:   node (external) 0x4a6e690 (max_nunits=1, refcnt=1)
/home/rguenther/src/gcc3/gcc/testsuite/gcc.dg/vect/bb-slp-pr65935.c:28:30:
note:        { _59, _62, _65, _68, _71, _74, _82, _86, _89, _92, _95, _98,
_106, _110, _113, _116, _119, _122 }