fix backend Cond/Cmp flag optimization failing for unoptimized code

diff --git a/ir/be/ia32/ia32_transform.c b/ir/be/ia32/ia32_transform.c
index 0b15ad7..e3d86b5 100644 (file)
--- a/ir/be/ia32/ia32_transform.c
+++ b/ir/be/ia32/ia32_transform.c
@@ -2110,13 +2110,16 @@ static ir_node *get_flags_node(ir_node *cmp, ia32_condition_code_t *cc_out)
                }
        }
 
-       /* the middle-end tries to eliminate impossible relations, so a ptr != 0
+       /* the middle-end tries to eliminate impossible relations, so a ptr <> 0
         * test becomes ptr > 0. But for x86 an equal comparison is preferable to
         * a >0 (we can sometimes eliminate the cmp in favor of flags produced by
-        * a predecessor node). So add the < bit */
+        * a predecessor node). So add the < bit.
+        * (Note that we do not want to produce <=> (which can happen for
+        * unoptimized code), because no x86 flag can represent that */
        possible = ir_get_possible_cmp_relations(l, r);
-       if (((relation & ir_relation_less) && !(possible & ir_relation_greater))
-           || ((relation & ir_relation_greater) && !(possible & ir_relation_less)))
+       if (!(relation & ir_relation_equal) &&
+               ( ((relation & ir_relation_less) && !(possible & ir_relation_greater))
+          || ((relation & ir_relation_greater) && !(possible & ir_relation_less))))
                relation |= ir_relation_less_greater;
 
        overflow_possible = true;