This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Bug tree-optimization/81245] New: [8 Regression] ICE building calculix with -Ofast


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81245

            Bug ID: 81245
           Summary: [8 Regression] ICE building calculix with -Ofast
           Product: gcc
           Version: 7.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---
            Target: aarch64-linux-gnu

This started with my copysign patch.
Here is a reduced testcase:
double sg[18];
void f(void)
{
  for (int i = 0 ;i < 18;i++)
  {
    if (sg[i] < 0.0)
      sg[i] = -1.0;
    else
      sg[i] = 1.0;
  }
}
---- CUT ---
Compile with -Ofast.
0x6f53b4 flow_bb_inside_loop_p(loop const*, basic_block_def const*)
        ../../gcc/gcc/cfgloop.c:794
0xd82b5b vect_stmt_relevant_p
        ../../gcc/gcc/tree-vect-stmts.c:326
0xd8366f vect_mark_stmts_to_be_vectorized(_loop_vec_info*)
        ../../gcc/gcc/tree-vect-stmts.c:664
0xdae127 vect_analyze_loop_2
        ../../gcc/gcc/tree-vect-loop.c:1944
0xdae127 vect_analyze_loop(loop*, _loop_vec_info*)
        ../../gcc/gcc/tree-vect-loop.c:2403
0xdc4e4f vectorize_loops()
        ../../gcc/gcc/tree-vectorizer.c:614
Please submit a full bug report,


>From what I can tell, is ifcvt clones the loop and ifconverts that loop only
and enables it only for vectorization and that cloned loop has the
a>0.0?1.0:-1.0 optimized to copysign.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]