lto: Compute partition boundary with asm_nodes

Previous patch added asm_node streaming, so we need to add referenced
symbols to partition.

asm_nodes must be added to partition before computing the boundary.

gcc/ChangeLog:

	* lto-cgraph.cc (compute_ltrans_boundary): Add symbols
	referenced from asm_nodes.
	* lto-streamer-out.cc (lto_output): Move adding asm_nodes
	to...
	* passes.cc (ipa_write_summaries): ...here.

gcc/testsuite/ChangeLog:

	* gcc.dg/lto/toplevel-extended-asm-1_0.c: New test.
	* gcc.dg/lto/toplevel-extended-asm-1_1.c: New test.

gcc/lto-cgraph.cc

@@ -899,6 +899,24 @@ compute_ltrans_boundary (lto_symtab_encoder_t in_encoder)
       lto_set_symtab_encoder_encode_initializer (encoder, vnode);
       create_references (encoder, vnode);
     }
+  for (lsei = lsei_start (in_encoder); !lsei_end_p (lsei); lsei_next (&lsei))
+    {
+      toplevel_node *tnode = lsei_node (lsei);
+      if (asm_node* node = dyn_cast <asm_node*> (tnode))
+	{
+	  symtab_node* ref;
+	  for (unsigned i = 0; node->symbols_referenced.iterate (i, &ref); i++)
+	    {
+	      if (!lto_symtab_encoder_in_partition_p (encoder, ref))
+		{
+		  if (cgraph_node* cref = dyn_cast <cgraph_node*> (ref))
+		    add_node_to (encoder, cref, false);
+		  else if (varpool_node *vref = dyn_cast <varpool_node *> (ref))
+		    lto_symtab_encoder_encode (encoder, vref);
+		}
+	    }
+	}
+    }
   /* Pickle in also the initializer of all referenced readonly variables
      to help folding.  Constant pool variables are not shared, so we must
      pickle those too.  */

gcc/lto-streamer-out.cc

@@ -2817,15 +2817,6 @@ lto_output (void)
   lto_symtab_encoder_t encoder = lto_get_out_decl_state ()->symtab_node_encoder;
   auto_vec<symtab_node *> symbols_to_copy;
 
-  if (!flag_wpa)
-    {
-      asm_node *anode;
-      for (anode = symtab->first_asm_symbol ();
-	   anode;
-	   anode = safe_as_a<asm_node*>(anode->next))
-	lto_set_symtab_encoder_in_partition (encoder, anode);
-    }
-
   create_order_remap (encoder);
 
   prune_offload_funcs ();

gcc/passes.cc

@@ -2911,6 +2911,11 @@ ipa_write_summaries (void)
     if (vnode->need_lto_streaming)
       lto_set_symtab_encoder_in_partition (encoder, vnode);
 
+  asm_node *anode;
+  for (anode = symtab->first_asm_symbol (); anode;
+       anode = safe_as_a<asm_node*> (anode->next))
+    lto_set_symtab_encoder_in_partition (encoder, anode);
+
   ipa_write_summaries_1 (compute_ltrans_boundary (encoder),
 			 flag_generate_offload);
 

gcc/testsuite/gcc.dg/lto/toplevel-extended-asm-1_0.c

@@ -0,0 +1,19 @@
+/* { dg-lto-do link } */
+/* { dg-lto-options {{-O2 -flto -flto-partition=1to1} } } */
+
+void asm_fn();
+void asm_fn_used();
+
+asm(".global %cc0\n%cc0:" :: ":" (asm_fn));
+asm(".global %cc0\n%cc0:" :: ":" (asm_fn_used));
+
+
+__attribute__((noinline))
+int privatized_fn(int v) { return v + v;}
+
+extern void call_privatized_fn();
+
+int main() {
+  privatized_fn (0);
+  call_privatized_fn ();
+}

gcc/testsuite/gcc.dg/lto/toplevel-extended-asm-1_1.c

@@ -0,0 +1,12 @@
+extern void asm_fn_used();
+
+__attribute__((used))
+void local_caller() {
+  asm_fn_used();
+}
+
+__attribute__((noipa))
+static void privatized_fn() { asm volatile ("");}
+asm(".long %cc0" :: "s"(privatized_fn));
+
+void call_privatized_fn() { privatized_fn();}