https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87124

            Bug ID: 87124
           Summary: [9 Regression] ICE have integer_cst in SSA_VAL, at
                    tree-ssa-sccvn.c:462 while building 541.leela_r from
                    SPEC2017
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ktkachov at gcc dot gnu.org
  Target Milestone: ---

class A {
  int m_fn1();
};

int A::m_fn1() {
  A *a = this;
  for (int i; i && a;)
    a = 0;
}


ICEs on aarch64 with -O2

SGFTree.ii:8:1: internal compiler error: tree check: expected ssa_name, have
integer_cst in SSA_VAL, at tree-ssa-sccvn.c:462
0x62d6e5 tree_check_failed(tree_node const*, char const*, int, char const*,
...)
        $SRC/gcc/tree.c:9371
0x1018d87 tree_check(tree_node*, char const*, int, char const*, tree_code)
        $SRC/gcc/tree.h:3106
0x1018d87 SSA_VAL(tree_node*)
        $SRC/gcc/tree-ssa-sccvn.c:462
0x100592b rpo_elim::eliminate_avail(basic_block_def*, tree_node*)
        $SRC/gcc/tree-ssa-sccvn.c:5690
0x1009db5 vn_lookup_simplify_result
        $SRC/gcc/tree-ssa-sccvn.c:5673
0xb5ec75 gimple_fold_stmt_to_constant_1(gimple*, tree_node* (*)(tree_node*),
tree_node* (*)(tree_node*))
        $SRC/gcc/gimple-fold.c:6098
0x1013c83 try_to_simplify
        $SRC/gcc/tree-ssa-sccvn.c:4225
0x1013c83 visit_stmt
        $SRC/gcc/tree-ssa-sccvn.c:4268
0x10157fb process_bb
        $SRC/gcc/tree-ssa-sccvn.c:5984
0x1017466 do_rpo_vn
        $SRC/gcc/tree-ssa-sccvn.c:6408
0x101833c execute
        $SRC/gcc/tree-ssa-sccvn.c:6597
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.


This is reduced from 541.leela_r from SPEC2017.
Apologies is this is a duplicate of recently fixed sscvn bugs, I couldn't find
an exact duplicate of the backtrace

Reply via email to