https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80104
Martin Liška <marxin at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |ASSIGNED
CC| |hubicka at ucw dot cz
Ever confirmed|0 |1
--- Comment #1 from Martin Liška <marxin at gcc dot gnu.org> ---
Problem is that we produce a thunk which has an invisible reference type of an
argument. Optimized tree dumps looks as follows:
;; Function c (c, funcdef_no=3, decl_uid=2001, cgraph_uid=1, symbol_order=1)
c (complex float b)
{
complex float arg.1;
float retval.0;
<bb 2> [100.00%]:
arg.1_2 = b;
retval.0_4 = a (arg.1_2); [tail call]
return retval.0_4;
}
Problem is that without any optimization level, arg.1_2 isn't a default def.
Following patch can fix that, but I'm not sure whether it's OK, or it's more a
workaround?
diff --git a/gcc/calls.c b/gcc/calls.c
index 61caf4ca752..404f9e303b7 100644
--- a/gcc/calls.c
+++ b/gcc/calls.c
@@ -1745,8 +1745,15 @@ initialize_argument_information (int num_actuals
ATTRIBUTE_UNUSED,
address. */
if (TREE_CODE (args[i].tree_value) == SSA_NAME)
{
- gcc_assert (SSA_NAME_IS_DEFAULT_DEF (args[i].tree_value));
- args[i].tree_value = SSA_NAME_VAR (args[i].tree_value);
+ if (SSA_NAME_IS_DEFAULT_DEF (args[i].tree_value))
+ args[i].tree_value = SSA_NAME_VAR (args[i].tree_value);
+ else
+ {
+ gimple *g = SSA_NAME_DEF_STMT (args[i].tree_value);
+ if (gimple_assign_single_p (g))
+ args[i].tree_value = gimple_assign_rhs1 (g);
+ }
+
gcc_assert (TREE_CODE (args[i].tree_value) == PARM_DECL);
}
/* Argument setup code may have copied the value to register. We
