This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Improve inliner's handling of aliases
- From: Jan Hubicka <hubicka at ucw dot cz>
- To: gcc-patches at gcc dot gnu dot org
- Date: Mon, 24 Nov 2014 17:18:04 +0100
- Subject: Improve inliner's handling of aliases
- Authentication-results: sourceware.org; auth=none
Hi,
the original update of inliner for aliases contined some lazyness. This patch fixes I think
last remainder of it where inliner did not remove symbol with aliases on it.
This is more important now when ICF uses aliases quite often.
Bootstrapped/regtested x86_64-linux, comitted.
PR ipa/63671
* ipa-inline-transform.c (can_remove_node_now_p_1): Handle alises
and -fno-devirtualize more carefully.
(can_remove_node_now_p): Update.
Index: ipa-inline-transform.c
===================================================================
--- ipa-inline-transform.c (revision 217980)
+++ ipa-inline-transform.c (working copy)
@@ -93,19 +93,28 @@ update_noncloned_frequencies (struct cgr
copy of function was removed. */
static bool
-can_remove_node_now_p_1 (struct cgraph_node *node)
+can_remove_node_now_p_1 (struct cgraph_node *node, struct cgraph_edge *e)
{
+ ipa_ref *ref;
+
+ FOR_EACH_ALIAS (node, ref)
+ {
+ cgraph_node *alias = dyn_cast <cgraph_node *> (ref->referring);
+ if ((alias->callers && alias->callers != e)
+ || !can_remove_node_now_p_1 (alias, e))
+ return false;
+ }
/* FIXME: When address is taken of DECL_EXTERNAL function we still
can remove its offline copy, but we would need to keep unanalyzed node in
the callgraph so references can point to it. */
return (!node->address_taken
- && !node->has_aliases_p ()
&& node->can_remove_if_no_direct_calls_p ()
/* Inlining might enable more devirtualizing, so we want to remove
those only after all devirtualizable virtual calls are processed.
Lacking may edges in callgraph we just preserve them post
inlining. */
- && !DECL_VIRTUAL_P (node->decl)
+ && (!DECL_VIRTUAL_P (node->decl)
+ || !opt_for_fn (node->decl, flag_devirtualize))
/* During early inlining some unanalyzed cgraph nodes might be in the
callgraph and they might reffer the function in question. */
&& !cgraph_new_nodes.exists ());
@@ -119,7 +128,7 @@ static bool
can_remove_node_now_p (struct cgraph_node *node, struct cgraph_edge *e)
{
struct cgraph_node *next;
- if (!can_remove_node_now_p_1 (node))
+ if (!can_remove_node_now_p_1 (node, e))
return false;
/* When we see same comdat group, we need to be sure that all
@@ -128,9 +137,13 @@ can_remove_node_now_p (struct cgraph_nod
return true;
for (next = dyn_cast<cgraph_node *> (node->same_comdat_group);
next != node; next = dyn_cast<cgraph_node *> (next->same_comdat_group))
- if ((next->callers && next->callers != e)
- || !can_remove_node_now_p_1 (next))
- return false;
+ {
+ if (next->alias)
+ continue;
+ if ((next->callers && next->callers != e)
+ || !can_remove_node_now_p_1 (next, e))
+ return false;
+ }
return true;
}