https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81118
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Priority|P3 |P1
Status|UNCONFIRMED |ASSIGNED
Last reconfirmed| |2017-06-19
Component|c |tree-optimization
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
Ever confirmed|0 |1
--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
It's another incarnation of PR81090 (as I said in the patch for that PR the fix
is incomplete). We again have:
#0 0x00000000010b4112 in remove_redundant_iv_tests (loop=0x7ffff69d6108)
at /space/rguenther/src/svn/early-lto-debug/gcc/tree-ssa-loop-ivcanon.c:566
566 edge exit_edge = EDGE_SUCC (bb, 0);
(gdb) l
561 upper bound. */
562 if (elt->is_exit && loop->any_upper_bound
563 && wi::ltu_p (loop->nb_iterations_upper_bound, elt->bound))
564 {
565 basic_block bb = gimple_bb (elt->stmt);
566 edge exit_edge = EDGE_SUCC (bb, 0);
567 struct tree_niter_desc niter;
568
569 if (!loop_exit_edge_p (loop, exit_edge))
570 exit_edge = EDGE_SUCC (bb, 1);
(gdb) p elt->stmt
$1 = <gimple_cond 0x7ffff69bbe60>
(gdb) p gimple_bb (elt->stmt)
$2 = <basic_block 0x0>
and thus loop->bounds is not up-to-date (as I said in the audit trail of
PR81090
the situation in cunroll is quite "interesting").