Hi,
this patch adds checking virtual_p flag of refernces (to ensure proper answers
of polymorphic-call-context analysis), of resrict flags on parameters, and of
type attributes on functions (because these do matter for codegen if we decide
to redirect).
I am leavin the PR open, there are remaining issues to track.
ICF specific:
- ipa-icf-gimple.c needs to match dependence analysis
Richard has propsed a patch for it, so I hope he will commit it tomorrow.
- restrict flag may need to be matched when considering two references
to variables being equal.
Here I am waiting for Richards comment
non-ICF specific
- tree-vectorizer is picking up wrong alignment
- fold-const.c's operands_equal_p probably needs same treatment for
comparing mem-ref as ipa-icf-gimple has. I think in all cases one can
construct testcase where tree-tail-merge would produce same incorrect
merging as ipa-icf does.
Bootstrapped/regtested x86_64-linux, comitted.
Honza
PR ipa/65270
* ipa-icf.c (sem_item::compare_cgraph_references): Compare
vtable references for their containing type.
(sem_function::equals_wpa): Compare TYPE_RESTRICT
and type attributes.
Index: ipa-icf.c
===================================================================
--- ipa-icf.c (revision 221170)
+++ ipa-icf.c (working copy)
@@ -345,6 +345,20 @@ sem_item::compare_cgraph_references (
{
enum availability avail1, avail2;
+ if (n1 == n2)
+ return true;
+
+ /* Merging two definitions with a reference to equivalent vtables, but
+ belonging to a different type may result in ipa-polymorphic-call analysis
+ giving a wrong answer about the dynamic type of instance. */
+ if (is_a <varpool_node *> (n1)
+ && (DECL_VIRTUAL_P (n1->decl) || DECL_VIRTUAL_P (n2->decl))
+ && (DECL_VIRTUAL_P (n1->decl) != DECL_VIRTUAL_P (n2->decl)
+ || !types_must_be_same_for_odr (DECL_CONTEXT (n1->decl),
+ DECL_CONTEXT (n2->decl))))
+ return return_false_with_msg
+ ("references to virtual tables can not be merged");
+
if (address && n1->equal_address_to (n2) == 1)
return true;
if (!address && n1->semantically_equivalent_p (n2))
@@ -407,6 +421,10 @@ sem_function::equals_wpa (sem_item *item
m_compared_func->arg_types[i],
is_not_leaf, i == 0))
return return_false_with_msg ("argument type is different");
+ if (POINTER_TYPE_P (arg_types[i])
+ && (TYPE_RESTRICT (arg_types[i])
+ != TYPE_RESTRICT (m_compared_func->arg_types[i])))
+ return return_false_with_msg ("argument restrict flag mismatch");
}
/* Result type checking. */
@@ -417,6 +435,10 @@ sem_function::equals_wpa (sem_item *item
if (node->num_references () != item->node->num_references ())
return return_false_with_msg ("different number of references");
+ if (comp_type_attributes (TREE_TYPE (decl),
+ TREE_TYPE (item->decl)) != 1)
+ return return_false_with_msg ("different type attributes");
+
ipa_ref *ref = NULL, *ref2 = NULL;
for (unsigned i = 0; node->iterate_reference (i, ref); i++)
{
