https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113385
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Last reconfirmed| |2024-01-15
Priority|P3 |P1
Ever confirmed|0 |1
Status|UNCONFIRMED |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed. The ICE means that loop->num_nodes is out-of-date.
I see an early exit created BB which is in the wrong loop. This:
for (auto exit : loop_exits)
{
basic_block dest = main_loop_exit_block;
if (exit != loop_exit)
{
if (!alt_loop_exit_block)
{
alt_loop_exit_block = split_edge (exit);
splits an edge to __assert_fail but is then re-wried
edge res = redirect_edge_and_branch (
single_succ_edge (alt_loop_exit_block),
new_preheader);
flush_pending_stmts (res);
continue;
not sure why we split the edge before redirecting, swapping fixes the
testcase. remove_bb_from_loops/add_bb_to_loop could also fixup with
the original order.
I'm testing swapping.
