cgraph.c: Add overall comment.

	* cgraph.c: Add overall comment.
	(cgraph_inline_hash): New global variable.
	(cgraph_create_node): Break out from ...
	(cgraph_node): ... here.
	(cgraph_edge): New function.
	(cgraph_create_edge): New CALL_EXPR argument; some sanity checking.
	(cgraph_remove_edge): Accept edge, intead of source and destination.
	(cgraph_redirect_edge_callee): New.
	(cgraph_remove_node): Update all new datastructures.
	(cgraph_record_call, cgraph_remove_call): Kill.
	(dump_cgraph_node): Break out from ... ; dump new datastructures.
	(dump_cgraph): ... here.
	(cgraph_function_possibly_inlined_p): Use new hashtable.
	(cgraph_clone_edge, cgraph_clone_node): New.
	* cgraph.h: Include hashtab.h
	(struct cgraph_global_info): Kill cloned_times, inline_once, will_be_output
	fields, add inlined_to pointer.
	(cgraph_node): Add pointer to next_clone.
	(cgraph_remove_edge, cgraph_create_edge): Update prototype.
	(cgraph_remove_call, cgraph_record_call): Kill.
	(cgraph_inline_hash): Declare.
	(dump_cgraph_node, cgraph_edge, cg4raph_clone_edge, cgraph_clone_node,
	cgraph_redirect_edge_callee): Declare.
	(cgraph_create_edges, cgraph_inline_p): Update prorotype.
	(cgraph_preserve_function_body_p, verify_cgraph, verify_cgraph_node,
	cgraph_mark_inline_edge, cgraph_clone_inlined_nodes): Declare.
	* cgraphunit.c: Add overall comment.
	(cgraph_optimize_function): Kill.
	(cgraph_assemble_pending_functions): Do not assemble inline clones.
	(cgraph_finalize_function): Update call of cgraph_remove_node
	(record_call_1): Record call sites.
	(cgraph_create_edges): Accept node instead of decl argument.
	(error_found): New static variable.
	(verify_cgraph_node_1, verify_cgraph_node, verify_cgraph): New functions.
	(cgraph_analyze_function): Update for new datastructures.
	(cgraph_finalize_compilation_unit): Plug memory leak.
	(cgraph_optimize_function): Kill.
	(cgraph_expand_function): Do not use cgraph_optimize_function.
	(INLINED_TIMES, SET_INLINED_TIMES, cgraph_inlined_into,
	cgraph_inlined_callees): Kill.
	(cgraph_remove_unreachable_nodes): Verify cgraph; update handling of
	clones.
	(estimate_growth): Simplify.
	(cgraph_clone_inlined_nodes): New function.
	(cgraph_mark_inline_edge): Re-implement.
	(cgraph_mark_inline): Likewise.
	(cgraph_check_inline_limits): Simplify.
	(cgraph_recursive_inlining_p): New.
	(update_callee_keys): Break out from ...
	(cgraph_decide_inlining_of_small_functions): ... here; simplify.
	(cgraph_decide_inlining, cgraph_decide_inlining_incrementally):
	Likewise.
	(cgraph_expand_all_functions):  Remove inline clones from the ordered
	list.
	(cgraph_preserve_function_body_p): New predicate.
	(cgraph_optimize): Verify cgraph.
	* function.h (struct function): Add fields saved_tree/saved_args.
	* timevar.def (TV_CGRAPH_VERIFY): Use verifier.
	* toplev.c (rest_of_compilation): Do not free cfun.
	* tree-inline.c: Include function.h
	(struct inline_data): Add saving_p field; replace decl/current_decl by
	node/current_node.
	(insert_decl_map): New function.
	(copy_body_r): Handle saving; update cgraph datastructure.
	(copy_body): Handle recursive inlining.
	(initialize_inlined_parameters): Likewise.
	(expand_call_inline): Propagate node attributes; update cgraph.
	(optimize_inline_calls): Verify that datastructure still match.
	(save_body): New function.
	* tree-inline.h (save_body): New.
	* tree-optimize.c (tree_rest_of_compilation): preserve function body; do inlining.
	* langhooks-def.c (LANG_HOOKS_UPDATE_DECL_AFTER_SAVING): New.
	* langhooks.c (lang_hooks): Add update_decl_after_saving.

	* cp-lang. (LANG_HOOKS_UPDATE_DECL_AFTER_SAVING): Define.
	* cp-tree.h (cp_update_decl_after_saving): Declare.
	* tree.c (cp_update_decl_after_saving): Define.

	* Make-lang.in (com.o): Add dependnecy on function.h
	* com.c: Include function.h
	(finish_function): Clear DECL_STRUCT_FUNCTION.

	* utils.c: Include function.h
	(end_subprog_body): Clear DECL_STRUCT_FUNCTION.

From-SVN: r80334

1 files changed, 624 insertions, 560 deletions

diff --git a/gcc/cgraphunit.c b/gcc/cgraphunit.c
index f871ef4..7495a75 100644
--- a/gcc/cgraphunit.c
+++ b/gcc/cgraphunit.c
@@ -19,6 +19,150 @@ along with GCC; see the file COPYING.  If not, write to the Free
 Software Foundation, 59 Temple Place - Suite 330, Boston, MA
 02111-1307, USA.  */
 
+/* This module implements main driver of compilation process as well as
+   few basic intraprocedural optimizers.
+
+   The main scope of this file is to act as an interface in between
+   tree based frontends and the backend (and middle end)
+
+   The front-end is supposed to use following functionality:
+
+    - cgraph_finalize_function
+
+      This function is called once front-end has parsed whole body of function
+      and it is certain that the function body nor the declaration will change.
+
+      (There is one exception needed for implementing GCC extern inline function.)
+
+    - cgraph_varpool_finalize_variable
+
+      This function has same behaviour as the above but is used for static
+      variables.
+
+    - cgraph_finalize_compilation_unit
+
+      This function is called once compilation unit is finalized and it will
+      no longer change.
+
+      In the unit-at-a-time the call-graph construction and local function
+      analysis takes place here.  Bodies of unreachable functions are released
+      to conserve memory usage.
+
+      ???  The compilation unit in this point of view should be compilation
+      unit as defined by the language - for instance C frontend allows multiple
+      compilation units to be parsed at once and it should call function each
+      time parsing is done so we save memory.
+
+    - cgraph_optimize
+
+      In this unit-at-a-time compilation the intra procedural analysis takes
+      place here.  In particular the static functions whose address is never
+      taken are marked as local.  Backend can then use this information to
+      modify calling conventions, do better inlining or similar optimizations.
+
+    - cgraph_assemble_pending_functions
+    - cgraph_varpool_assemble_pending_variables
+
+      In non-unit-at-a-time mode these functions can be used to force compilation
+      of functions or variables that are known to be needed at given stage
+      of compilation
+
+    - cgraph_mark_needed_node
+    - cgraph_varpool_mark_needed_node
+
+      When function or variable is referenced by some hidden way (for instance
+      via assembly code and marked by attribute "used"), the call-graph data structure
+      must be updated accordingly by this function.
+
+    - analyze_expr callback
+
+      This function is responsible for lowering tree nodes not understood by
+      generic code into understandable ones or alternatively marking
+      callgraph and varpool nodes referenced by the as needed.
+
+      ??? On the tree-ssa genericizing should take place here and we will avoid
+      need for these hooks (replacing them by genericizing hook)
+
+    - expand_function callback
+
+      This function is used to expand function and pass it into RTL back-end.
+      Front-end should not make any assumptions about when this function can be
+      called.  In particular cgraph_assemble_pending_functions,
+      cgraph_varpool_assemble_pending_variables, cgraph_finalize_function,
+      cgraph_varpool_finalize_function, cgraph_optimize can cause arbitrarily
+      previously finalized functions to be expanded.
+
+    We implement two compilation modes.
+
+      - unit-at-a-time:  In this mode analyzing of all functions is deferred
+	to cgraph_finalize_compilation_unit and expansion into cgraph_optimize.
+
+	In cgraph_finalize_compilation_unit the reachable functions are
+	analyzed.  During analysis the call-graph edges from reachable
+	functions are constructed and their destinations are marked as
+	reachable.  References to functions and variables are discovered too
+	and variables found to be needed output to the assembly file.  Via
+	mark_referenced call in assemble_variable functions referenced by
+	static variables are noticed too.
+
+	The intra-procedural information is produced and it's existence
+	indicated by global_info_ready.  Once this flag is set it is impossible
+	to change function from !reachable to reachable and thus
+	assemble_variable no longer call mark_referenced.
+
+	Finally the call-graph is topologically sorted and all reachable functions
+	that has not been completely inlined or are not external are output.
+
+	??? It is possible that reference to function or variable is optimized
+	out.  We can not deal with this nicely because topological order is not
+	suitable for it.  For tree-ssa we may consider another pass doing
+	optimization and re-discovering reachable functions.
+
+	??? Reorganize code so variables are output very last and only if they
+	really has been referenced by produced code, so we catch more cases
+	where reference has been optimized out.
+
+      - non-unit-at-a-time
+
+	All functions are variables are output as early as possible to conserve
+	memory consumption.  This may or may not result in less memory used but
+	it is still needed for some legacy code that rely on particular ordering
+	of things output from the compiler.
+
+	Varpool data structures are not used and variables are output directly.
+
+	Functions are output early using call of
+	cgraph_assemble_pending_function from cgraph_finalize_function.  The
+	decision on whether function is needed is made more conservative so
+	uninlininable static functions are needed too.  During the call-graph
+	construction the edge destinations are not marked as reachable and it
+	is completely relied upn assemble_variable to mark them.
+	
+     Inlining decision heuristics
+        ??? Move this to separate file after tree-ssa merge.
+
+	We separate inlining decisions from the inliner itself and store it
+	inside callgraph as so called inline plan.  Reffer to cgraph.c
+	documentation about particular representation of inline plans in the
+	callgraph
+
+	The implementation of particular heuristics is separated from
+	the rest of code to make it easier to replace it with more complicated
+	implementation in the future.  The rest of inlining code acts as a
+	library aimed to modify the callgraph and verify that the parameters
+	on code size growth fits.
+
+	To mark given call inline, use cgraph_mark_inline function, the
+	verification is performed by cgraph_default_inline_p and
+	cgraph_check_inline_limits.
+
+	The heuristics implements simple knapsack style algorithm ordering
+	all functions by their "profitability" (estimated by code size growth)
+	and inlining them in priority order.
+
+	cgraph_decide_inlining implements heuristics taking whole callgraph
+	into account, while cgraph_decide_inlining_incrementally considers
+	only one function at a time and is used in non-unit-at-a-time mode.  */
 #include "config.h"
 #include "system.h"
 #include "coretypes.h"
@@ -47,7 +191,6 @@ static void cgraph_mark_functions_to_output (void);
 static void cgraph_expand_function (struct cgraph_node *);
 static tree record_call_1 (tree *, int *, void *);
 static void cgraph_mark_local_functions (void);
-static void cgraph_optimize_function (struct cgraph_node *);
 static bool cgraph_default_inline_p (struct cgraph_node *n);
 static void cgraph_analyze_function (struct cgraph_node *node);
 static void cgraph_decide_inlining_incrementally (struct cgraph_node *);
@@ -138,7 +281,8 @@ cgraph_assemble_pending_functions (void)
       struct cgraph_node *n = cgraph_nodes_queue;
 
       cgraph_nodes_queue = cgraph_nodes_queue->next_needed;
-      if (!n->origin && !DECL_EXTERNAL (n->decl))
+      n->next_needed = NULL;
+      if (!n->origin && !n->global.inlined_to && !DECL_EXTERNAL (n->decl))
 	{
 	  cgraph_expand_function (n);
 	  output = true;
@@ -186,7 +330,7 @@ cgraph_finalize_function (tree decl, bool nested)
       node->analyzed = false;
       node->local.redefined_extern_inline = true;
       while (node->callees)
-	cgraph_remove_edge (node, node->callees->callee);
+	cgraph_remove_edge (node->callees);
 
       /* We may need to re-queue the node for assembling in case
          we already proceeded it and ignored as not needed.  */
@@ -229,7 +373,7 @@ cgraph_finalize_function (tree decl, bool nested)
   if (!TREE_ASM_WRITTEN (decl))
     (*debug_hooks->deferred_inline_function) (decl);
 
-  /* We will never really output the function body, clear the SAVED_INSNS array
+  /* We will never really output the function body, clear the STRUCT_FUNCTION array
      early then.  */
   if (DECL_EXTERNAL (decl))
     DECL_STRUCT_FUNCTION (decl) = NULL;
@@ -267,7 +411,7 @@ record_call_1 (tree *tp, int *walk_subtrees, void *data)
 	tree decl = get_callee_fndecl (*tp);
 	if (decl && TREE_CODE (decl) == FUNCTION_DECL)
 	  {
-	    cgraph_record_call (data, decl);
+	    cgraph_create_edge (data, cgraph_node (decl), *tp);
 
 	    /* When we see a function call, we don't want to look at the
 	       function reference in the ADDR_EXPR that is hanging from
@@ -300,20 +444,162 @@ record_call_1 (tree *tp, int *walk_subtrees, void *data)
   return NULL;
 }
 
-/* Create cgraph edges for function calls inside BODY from DECL.  */
+/* Create cgraph edges for function calls inside BODY from NODE.  */
 
 void
-cgraph_create_edges (tree decl, tree body)
+cgraph_create_edges (struct cgraph_node *node, tree body)
 {
   /* The nodes we're interested in are never shared, so walk
      the tree ignoring duplicates.  */
   visited_nodes = htab_create (37, htab_hash_pointer,
 				    htab_eq_pointer, NULL);
-  walk_tree (&body, record_call_1, decl, visited_nodes);
+  walk_tree (&body, record_call_1, node, visited_nodes);
   htab_delete (visited_nodes);
   visited_nodes = NULL;
 }
 
+static bool error_found;
+
+/* Callbrack of verify_cgraph_node.  Check that all call_exprs have cgraph nodes.  */
+static tree
+verify_cgraph_node_1 (tree *tp, int *walk_subtrees, void *data)
+{
+  tree t = *tp;
+  tree decl;
+
+  if (TREE_CODE (t) == CALL_EXPR && (decl = get_callee_fndecl (t)))
+    {
+      struct cgraph_edge *e = cgraph_edge (data, t);
+      if (e)
+	{
+	  if (e->aux)
+	    {
+	      error ("Shared call_expr:");
+	      debug_tree (t);
+	      error_found = true;
+	    }
+	  if (e->callee->decl != cgraph_node (decl)->decl)
+	    {
+	      error ("Edge points to wrong declaration:");
+	      debug_tree (e->callee->decl);
+	      fprintf (stderr," Instead of:");
+	      debug_tree (decl);
+	    }
+	  e->aux = (void *)1;
+	}
+      else
+	{
+	  error ("Missing callgraph edge for call expr:");
+	  debug_tree (t);
+	  error_found = true;
+	}
+    }
+  /* Save some cycles by not walking types and declaration as we
+     won't find anything useful there anyway.  */
+  if (DECL_P (*tp) || TYPE_P (*tp))
+    {
+      *walk_subtrees = 0;
+    }
+  return NULL_TREE;
+}
+
+/* Verify cgraph nodes of given cgraph node.  */
+void
+verify_cgraph_node (struct cgraph_node *node)
+{
+  struct cgraph_edge *e;
+  struct cgraph_node *main_clone;
+
+  timevar_push (TV_CGRAPH_VERIFY);
+  error_found = false;
+  for (e = node->callees; e; e = e->next_callee)
+    if (e->aux)
+      {
+	error ("Aux field set for edge %s->%s",
+	       cgraph_node_name (e->caller), cgraph_node_name (e->callee));
+	error_found = true;
+      }
+  for (e = node->callers; e; e = e->next_caller)
+    {
+      if (!e->inline_failed)
+	{
+	  if (node->global.inlined_to
+	      != (e->caller->global.inlined_to
+		  ? e->caller->global.inlined_to : e->caller))
+	    {
+	      error ("Inlined_to pointer is wrong");
+	      error_found = true;
+	    }
+	  if (node->callers->next_caller)
+	    {
+	      error ("Multiple inline callers");
+	      error_found = true;
+	    }
+	}
+      else
+	if (node->global.inlined_to)
+	  {
+	    error ("Inlined_to pointer set for noninline callers");
+	    error_found = true;
+	  }
+    }
+  if (!node->callers && node->global.inlined_to)
+    {
+      error ("Inlined_to pointer is set but no predecesors found");
+      error_found = true;
+    }
+  if (node->global.inlined_to == node)
+    {
+      error ("Inlined_to pointer reffers to itself");
+      error_found = true;
+    }
+
+  for (main_clone = cgraph_node (node->decl); main_clone;
+       main_clone = main_clone->next_clone)
+    if (main_clone == node)
+      break;
+  if (!node)
+    {
+      error ("Node not found in DECL_ASSEMBLER_NAME hash");
+      error_found = true;
+    }
+  
+  if (node->analyzed
+      && DECL_SAVED_TREE (node->decl) && !TREE_ASM_WRITTEN (node->decl)
+      && (!DECL_EXTERNAL (node->decl) || node->global.inlined_to))
+    {
+      walk_tree_without_duplicates (&DECL_SAVED_TREE (node->decl),
+				    verify_cgraph_node_1, node);
+      for (e = node->callees; e; e = e->next_callee)
+	{
+	  if (!e->aux)
+	    {
+	      error ("Edge %s->%s has no corresponding call_expr",
+		     cgraph_node_name (e->caller),
+		     cgraph_node_name (e->callee));
+	      error_found = true;
+	    }
+	  e->aux = 0;
+	}
+    }
+  if (error_found)
+    {
+      dump_cgraph_node (stderr, node);
+      internal_error ("verify_cgraph_node failed.");
+    }
+  timevar_pop (TV_CGRAPH_VERIFY);
+}
+
+/* Verify whole cgraph structure.  */
+void
+verify_cgraph (void)
+{
+  struct cgraph_node *node;
+
+  for (node = cgraph_nodes; node; node = node->next)
+    verify_cgraph_node (node);
+}
+
 /* Analyze the function scheduled to be output.  */
 static void
 cgraph_analyze_function (struct cgraph_node *node)
@@ -324,7 +610,7 @@ cgraph_analyze_function (struct cgraph_node *node)
   current_function_decl = decl;
 
   /* First kill forward declaration so reverse inlining works properly.  */
-  cgraph_create_edges (decl, DECL_SAVED_TREE (decl));
+  cgraph_create_edges (node, DECL_SAVED_TREE (decl));
 
   node->local.inlinable = tree_inlinable_function_p (decl);
   if (!node->local.self_insns)
@@ -334,25 +620,19 @@ cgraph_analyze_function (struct cgraph_node *node)
     node->local.disregard_inline_limits
       = lang_hooks.tree_inlining.disregard_inline_limits (decl);
   for (e = node->callers; e; e = e->next_caller)
-    if (e->inline_failed)
-      {
-	if (node->local.redefined_extern_inline)
-	  e->inline_failed = N_("redefined extern inline functions are not "
-				"considered for inlining");
-	else if (!node->local.inlinable)
-	  e->inline_failed = N_("function not inlinable");
-	else
-	  e->inline_failed = N_("function not considered for inlining");
-      }
+    {
+      if (node->local.redefined_extern_inline)
+	e->inline_failed = N_("redefined extern inline functions are not "
+			   "considered for inlining");
+      else if (!node->local.inlinable)
+	e->inline_failed = N_("function not inlinable");
+      else
+	e->inline_failed = N_("function not considered for inlining");
+    }
   if (flag_really_no_inline && !node->local.disregard_inline_limits)
     node->local.inlinable = 0;
   /* Inlining characteristics are maintained by the cgraph_mark_inline.  */
   node->global.insns = node->local.self_insns;
-  if (!DECL_EXTERNAL (decl))
-    {
-      node->global.cloned_times = 1;
-      node->global.will_be_output = true;
-    }
 
   node->analyzed = true;
   current_function_decl = NULL;
@@ -396,6 +676,7 @@ cgraph_finalize_compilation_unit (void)
 
       node = cgraph_nodes_queue;
       cgraph_nodes_queue = cgraph_nodes_queue->next_needed;
+      node->next_needed = NULL;
 
       /* ??? It is possible to create extern inline function and later using
 	 weak alas attribute to kill its body. See
@@ -436,9 +717,9 @@ cgraph_finalize_compilation_unit (void)
 
       if (!node->reachable && DECL_SAVED_TREE (decl))
 	{
-	  cgraph_remove_node (node);
 	  if (cgraph_dump_file)
 	    fprintf (cgraph_dump_file, " %s", cgraph_node_name (node));
+	  cgraph_remove_node (node);
 	}
       else
 	node->next_needed = NULL;
@@ -475,45 +756,20 @@ cgraph_mark_functions_to_output (void)
 	 always inlined, as well as those that are reachable from
 	 outside the current compilation unit.  */
       if (DECL_SAVED_TREE (decl)
+	  && !node->global.inlined_to
 	  && (node->needed
 	      || (e && node->reachable))
 	  && !TREE_ASM_WRITTEN (decl) && !node->origin
 	  && !DECL_EXTERNAL (decl))
 	node->output = 1;
-      else
-        DECL_STRUCT_FUNCTION (decl) = NULL;
-    }
-}
-
-/* Optimize the function before expansion.  */
-
-static void
-cgraph_optimize_function (struct cgraph_node *node)
-{
-  tree decl = node->decl;
-
-  timevar_push (TV_INTEGRATION);
-  /* optimize_inline_calls avoids inlining of current_function_decl.  */
-  current_function_decl = decl;
-  if (flag_inline_trees)
-    {
-      struct cgraph_edge *e;
-
-      for (e = node->callees; e; e = e->next_callee)
-	if (!e->inline_failed || warn_inline
-	    || (DECL_DECLARED_INLINE_P (e->callee->decl)
-		&& lookup_attribute ("always_inline",
-				     DECL_ATTRIBUTES (e->callee->decl))))
-	  break;
-      if (e)
-        optimize_inline_calls (decl);
-    }
-  if (node->nested)
-    {
-      for (node = node->nested; node; node = node->next_nested)
-	cgraph_optimize_function (node);
+      /* We should've reclaimed all functions that are not needed.  */
+      else if (!node->global.inlined_to && DECL_SAVED_TREE (decl)
+	       && !node->origin && !DECL_EXTERNAL (decl))
+	{
+	  dump_cgraph_node (stderr, node);
+	  abort ();
+	}
     }
-  timevar_pop (TV_INTEGRATION);
 }
 
 /* Expand function specified by NODE.  */
@@ -523,17 +779,24 @@ cgraph_expand_function (struct cgraph_node *node)
 {
   tree decl = node->decl;
 
+  /* We ought to not compile any inline clones.  */
+  if (node->global.inlined_to)
+    abort ();
+
   if (flag_unit_at_a_time)
     announce_function (decl);
 
-  cgraph_optimize_function (node);
-
   /* Generate RTL for the body of DECL.  Nested functions are expanded
      via lang_expand_decl_stmt.  */
   lang_hooks.callgraph.expand_function (decl);
   if (DECL_DEFER_OUTPUT (decl))
     abort ();
 
+  /* Make sure that BE didn't gave up on compiling.  */
+  if (!TREE_ASM_WRITTEN (node->decl)
+      && !(sorrycount || errorcount))
+    abort ();
+
   current_function_decl = NULL;
 }
 
@@ -599,207 +862,6 @@ cgraph_postorder (struct cgraph_node **order)
   return order_pos;
 }
 
-#define INLINED_TIMES(node) ((size_t)(node)->aux)
-#define SET_INLINED_TIMES(node,times) ((node)->aux = (void *)(times))
-
-/* Return list of nodes we decided to inline NODE into, set their output
-   flag and compute INLINED_TIMES.
-
-   We do simple backtracing to get INLINED_TIMES right.  This should not be
-   expensive as we limit the amount of inlining.  Alternatively we may first
-   discover set of nodes, topologically sort these and propagate
-   INLINED_TIMES  */
-
-static int
-cgraph_inlined_into (struct cgraph_node *node, struct cgraph_node **array)
-{
-  int nfound = 0;
-  struct cgraph_edge **stack;
-  struct cgraph_edge *e, *e1;
-  int sp;
-  int i;
-
-  /* Fast path: since we traverse in mostly topological order, we will likely
-     find no edges.  */
-  for (e = node->callers; e; e = e->next_caller)
-    if (!e->inline_failed)
-      break;
-
-  if (!e)
-    return 0;
-
-  /* Allocate stack for back-tracking up callgraph.  */
-  stack = xmalloc ((cgraph_n_nodes + 1) * sizeof (struct cgraph_edge));
-  sp = 0;
-
-  /* Push the first edge on to the stack.  */
-  stack[sp++] = e;
-
-  while (sp)
-    {
-      struct cgraph_node *caller;
-
-      /* Look at the edge on the top of the stack.  */
-      e = stack[sp - 1];
-      caller = e->caller;
-
-      /* Check if the caller destination has been visited yet.  */
-      if (!caller->output)
-	{
-	  array[nfound++] = e->caller;
-	  /* Mark that we have visited the destination.  */
-	  caller->output = true;
-	  SET_INLINED_TIMES (caller, 0);
-	}
-      SET_INLINED_TIMES (caller, INLINED_TIMES (caller) + 1);
-
-      for (e1 = caller->callers; e1; e1 = e1->next_caller)
-	if (!e1->inline_failed)
-	  break;
-
-      if (e1)
-	stack[sp++] = e1;
-      else
-	{
-	  while (true)
-	    {
-	      for (e1 = e->next_caller; e1; e1 = e1->next_caller)
-		if (!e1->inline_failed)
-		  break;
-
-	      if (e1)
-		{
-		  stack[sp - 1] = e1;
-		  break;
-		}
-	      else
-		{
-		  sp--;
-		  if (!sp)
-		    break;
-		  e = stack[sp - 1];
-		}
-	    }
-	}
-    }
-
-  free (stack);
-
-
-  if (cgraph_dump_file)
-    {
-      fprintf (cgraph_dump_file, " Found inline predecesors of %s:",
-	       cgraph_node_name (node));
-      for (i = 0; i < nfound; i++)
-	{
-	  fprintf (cgraph_dump_file, " %s", cgraph_node_name (array[i]));
-	  if (INLINED_TIMES (array[i]) != 1)
-	    fprintf (cgraph_dump_file, " (%i times)",
-		     (int)INLINED_TIMES (array[i]));
-	}
-      fprintf (cgraph_dump_file, "\n");
-    }
-
-  return nfound;
-}
-
-/* Return list of nodes we decided to inline into NODE, set their output
-   flag and compute INLINED_TIMES.
-
-   This function is identical to cgraph_inlined_into with callers and callees
-   nodes swapped.  */
-
-static int
-cgraph_inlined_callees (struct cgraph_node *node, struct cgraph_node **array)
-{
-  int nfound = 0;
-  struct cgraph_edge **stack;
-  struct cgraph_edge *e, *e1;
-  int sp;
-  int i;
-
-  /* Fast path: since we traverse in mostly topological order, we will likely
-     find no edges.  */
-  for (e = node->callees; e; e = e->next_callee)
-    if (!e->inline_failed)
-      break;
-
-  if (!e)
-    return 0;
-
-  /* Allocate stack for back-tracking up callgraph.  */
-  stack = xmalloc ((cgraph_n_nodes + 1) * sizeof (struct cgraph_edge));
-  sp = 0;
-
-  /* Push the first edge on to the stack.  */
-  stack[sp++] = e;
-
-  while (sp)
-    {
-      struct cgraph_node *callee;
-
-      /* Look at the edge on the top of the stack.  */
-      e = stack[sp - 1];
-      callee = e->callee;
-
-      /* Check if the callee destination has been visited yet.  */
-      if (!callee->output)
-	{
-	  array[nfound++] = e->callee;
-	  /* Mark that we have visited the destination.  */
-	  callee->output = true;
-	  SET_INLINED_TIMES (callee, 0);
-	}
-      SET_INLINED_TIMES (callee, INLINED_TIMES (callee) + 1);
-
-      for (e1 = callee->callees; e1; e1 = e1->next_callee)
-	if (!e1->inline_failed)
-	  break;
-      if (e1)
-	stack[sp++] = e1;
-      else
-	{
-	  while (true)
-	    {
-	      for (e1 = e->next_callee; e1; e1 = e1->next_callee)
-		if (!e1->inline_failed)
-		  break;
-
-	      if (e1)
-		{
-		  stack[sp - 1] = e1;
-		  break;
-		}
-	      else
-		{
-		  sp--;
-		  if (!sp)
-		    break;
-		  e = stack[sp - 1];
-		}
-	    }
-	}
-    }
-
-  free (stack);
-
-  if (cgraph_dump_file)
-    {
-      fprintf (cgraph_dump_file, " Found inline successors of %s:",
-	       cgraph_node_name (node));
-      for (i = 0; i < nfound; i++)
-	{
-	  fprintf (cgraph_dump_file, " %s", cgraph_node_name (array[i]));
-	  if (INLINED_TIMES (array[i]) != 1)
-	    fprintf (cgraph_dump_file, " (%i times)",
-		     (int)INLINED_TIMES (array[i]));
-	}
-      fprintf (cgraph_dump_file, "\n");
-    }
-
-  return nfound;
-}
-
 /* Perform reachability analysis and reclaim all unreachable nodes.
    This function also remove unneeded bodies of extern inline functions
    and thus needs to be done only after inlining decisions has been made.  */
@@ -811,6 +873,9 @@ cgraph_remove_unreachable_nodes (void)
   bool changed = false;
   int insns = 0;
 
+#ifdef ENABLE_CHECKING
+  verify_cgraph ();
+#endif
   if (cgraph_dump_file)
     fprintf (cgraph_dump_file, "\nReclaiming functions:");
 #ifdef ENABLE_CHECKING
@@ -850,10 +915,11 @@ cgraph_remove_unreachable_nodes (void)
   /* Remove unreachable nodes.  Extern inline functions need special care;
      Unreachable extern inline functions shall be removed.
      Reachable extern inline functions we never inlined shall get their bodies
-     eliminated.
+     eliminated
      Reachable extern inline functions we sometimes inlined will be turned into
      unanalyzed nodes so they look like for true extern functions to the rest
-     of code.  */
+     of code.  Body of such functions is relased via remove_node once the
+     inline clones are eliminated.  */
   for (node = cgraph_nodes; node; node = node->next)
     {
       if (!node->aux)
@@ -878,9 +944,21 @@ cgraph_remove_unreachable_nodes (void)
 		  break;
 	      if (e || node->needed)
 		{
-		  DECL_SAVED_TREE (node->decl) = NULL_TREE;
+		  struct cgraph_node *clone;
+
+		  for (clone = node->next_clone; clone;
+		       clone = clone->next_clone)
+		    if (clone->aux)
+		      break;
+		  if (!clone)
+		    {
+		      DECL_SAVED_TREE (node->decl) = NULL;
+		      DECL_STRUCT_FUNCTION (node->decl) = NULL;
+		      DECL_ARGUMENTS (node->decl) = NULL;
+		      DECL_INITIAL (node->decl) = error_mark_node;
+		    }
 		  while (node->callees)
-		    cgraph_remove_edge (node, node->callees->callee);
+		    cgraph_remove_edge (node->callees);
 		  node->analyzed = false;
 		}
 	      else
@@ -898,7 +976,6 @@ cgraph_remove_unreachable_nodes (void)
   return changed;
 }
 
-
 /* Estimate size of the function after inlining WHAT into TO.  */
 
 static int
@@ -914,113 +991,153 @@ static int
 cgraph_estimate_growth (struct cgraph_node *node)
 {
   int growth = 0;
-  int calls_saved = 0;
-  int clones_added = 0;
   struct cgraph_edge *e;
 
   for (e = node->callers; e; e = e->next_caller)
     if (e->inline_failed)
-      {
-	growth += ((cgraph_estimate_size_after_inlining (1, e->caller, node)
-		    -
-		    e->caller->global.insns) *e->caller->global.cloned_times);
-	calls_saved += e->caller->global.cloned_times;
-	clones_added += e->caller->global.cloned_times;
-      }
+      growth += (cgraph_estimate_size_after_inlining (1, e->caller, node)
+		 - e->caller->global.insns);
 
   /* ??? Wrong for self recursive functions or cases where we decide to not
      inline for different reasons, but it is not big deal as in that case
      we will keep the body around, but we will also avoid some inlining.  */
   if (!node->needed && !node->origin && !DECL_EXTERNAL (node->decl))
-    growth -= node->global.insns, clones_added--;
-
-  if (!calls_saved)
-    calls_saved = 1;
+    growth -= node->global.insns;
 
   return growth;
 }
 
-/* Update insn sizes after inlining WHAT into TO that is already inlined into
-   all nodes in INLINED array.  */
+/* E is expected to be an edge being inlined.  Clone destination node of
+   the edge and redirect it to the new clone.
+   DUPLICATE is used for bookeeping on whether we are actually creating new
+   clones or re-using node originally representing out-of-line function call.
+   */
+void
+cgraph_clone_inlined_nodes (struct cgraph_edge *e, bool duplicate)
+{
+  struct cgraph_node *n;
+
+  /* We may elliminate the need for out-of-line copy to be output.  In that
+     case just go ahead and re-use it.  */
+  if (!e->callee->callers->next_caller
+      && (!e->callee->needed || DECL_EXTERNAL (e->callee->decl))
+      && !e->callee->origin
+      && duplicate
+      && flag_unit_at_a_time)
+    {
+      if (e->callee->global.inlined_to)
+	abort ();
+      if (!DECL_EXTERNAL (e->callee->decl))
+        overall_insns -= e->callee->global.insns, nfunctions_inlined++;
+      duplicate = 0;
+    }
+   else if (duplicate)
+    {
+      n = cgraph_clone_node (e->callee);
+      cgraph_redirect_edge_callee (e, n);
+    }
 
-static void
-cgraph_mark_inline (struct cgraph_node *to, struct cgraph_node *what,
-		    struct cgraph_node **inlined, int ninlined,
-		    struct cgraph_node **inlined_callees,
-		    int ninlined_callees)
+  if (e->caller->global.inlined_to)
+    e->callee->global.inlined_to = e->caller->global.inlined_to;
+  else
+    e->callee->global.inlined_to = e->caller;
+
+  /* Recursivly clone all bodies.  */
+  for (e = e->callee->callees; e; e = e->next_callee)
+    if (!e->inline_failed)
+      cgraph_clone_inlined_nodes (e, duplicate);
+}
+
+/* Mark edge E as inlined and update callgraph accordingly.  */
+
+void
+cgraph_mark_inline_edge (struct cgraph_edge *e)
 {
-  int i;
-  int times = 0;
-  int clones = 0;
-  struct cgraph_edge *e;
-  bool called = false;
-  int new_insns;
+  int old_insns = 0, new_insns = 0;
+  struct cgraph_node *to = NULL, *what;
+
+  if (!e->inline_failed)
+    abort ();
+  e->inline_failed = NULL;
 
-  what->global.inlined = 1;
-  for (e = what->callers; e; e = e->next_caller)
+  if (!e->callee->global.inlined && flag_unit_at_a_time)
     {
-      if (e->caller == to)
-	{
-	  if (!e->inline_failed)
-	    continue;
-	  e->inline_failed = NULL;
-	  times++;
-	  clones += e->caller->global.cloned_times;
-	}
-      else if (e->inline_failed)
-	called = true;
+      void **slot;
+      if (!cgraph_inline_hash)
+        cgraph_inline_hash = htab_create_ggc (42, htab_hash_pointer,
+					      htab_eq_pointer, NULL);
+      slot = htab_find_slot (cgraph_inline_hash,
+			     DECL_ASSEMBLER_NAME (e->callee->decl), INSERT);
+      *slot = DECL_ASSEMBLER_NAME (e->callee->decl);
     }
-  if (!times)
-    abort ();
-  ncalls_inlined += times;
+  e->callee->global.inlined = true;
+
+  cgraph_clone_inlined_nodes (e, true);
 
-  new_insns = cgraph_estimate_size_after_inlining (times, to, what);
-  if (to->global.will_be_output)
-    overall_insns += new_insns - to->global.insns;
-  to->global.insns = new_insns;
+  what = e->callee;
 
-  if (!called && !what->needed && !what->origin
-      && flag_unit_at_a_time
-      && !DECL_EXTERNAL (what->decl))
+  /* Now update size of caller and all functions caller is inlined into. */
+  for (;e && !e->inline_failed; e = e->caller->callers)
     {
-      if (!what->global.will_be_output)
+      old_insns = e->caller->global.insns;
+      new_insns = cgraph_estimate_size_after_inlining (1, e->caller,
+						       what);
+      if (new_insns < 0)
 	abort ();
-      clones--;
-      nfunctions_inlined++;
-      what->global.will_be_output = 0;
-      overall_insns -= what->global.insns;
+      to = e->caller;
+      to->global.insns = new_insns;
     }
-  what->global.cloned_times += clones;
-  for (i = 0; i < ninlined; i++)
-    {
-      new_insns =
-	cgraph_estimate_size_after_inlining (INLINED_TIMES (inlined[i]) *
-					     times, inlined[i], what);
-      if (inlined[i]->global.will_be_output)
-	overall_insns += new_insns - inlined[i]->global.insns;
-      inlined[i]->global.insns = new_insns;
-    }
-  for (i = 0; i < ninlined_callees; i++)
+  if (what->global.inlined_to != to)
+    abort ();
+  overall_insns += new_insns - old_insns;
+  ncalls_inlined++;
+}
+
+/* Mark all calls of EDGE->CALLEE inlined into EDGE->CALLER.
+   Return following unredirected edge in the list of callers
+   of EDGE->CALLEE  */
+
+static struct cgraph_edge *
+cgraph_mark_inline (struct cgraph_edge *edge)
+{
+  struct cgraph_node *to = edge->caller;
+  struct cgraph_node *what = edge->callee;
+  struct cgraph_edge *e, *next;
+  int times = 0;
+
+  /* Look for all calls, mark them inline and clone recursivly
+     all inlined functions.  */
+  for (e = what->callers; e; e = next)
     {
-      inlined_callees[i]->global.cloned_times +=
-	INLINED_TIMES (inlined_callees[i]) * clones;
+      next = e->next_caller;
+      if (e->caller == to && e->inline_failed)
+	{
+          cgraph_mark_inline_edge (e);
+	  if (e == edge)
+	    edge = next;
+	  times ++;
+	}
     }
+  if (!times)
+    abort ();
+  return edge;
 }
 
-/* Return false when inlining WHAT into TO is not good idea as it would cause
-   too large growth of function bodies.  */
+/* Return false when inlining WHAT into TO is not good idea
+   as it would cause too large growth of function bodies.  */
 
 static bool
 cgraph_check_inline_limits (struct cgraph_node *to, struct cgraph_node *what,
-			    struct cgraph_node **inlined, int ninlined,
 			    const char **reason)
 {
-  int i;
   int times = 0;
   struct cgraph_edge *e;
   int newsize;
   int limit;
 
+  if (to->global.inlined_to)
+    to = to->global.inlined_to;
+
   for (e = to->callees; e; e = e->next_callee)
     if (e->callee == what)
       times++;
@@ -1038,23 +1155,10 @@ cgraph_check_inline_limits (struct cgraph_node *to, struct cgraph_node *what,
   if (newsize > PARAM_VALUE (PARAM_LARGE_FUNCTION_INSNS)
       && newsize > limit)
     {
-      *reason = N_("--param large-function-growth limit reached");
+      if (reason)
+        *reason = N_("--param large-function-growth limit reached");
       return false;
     }
-  for (i = 0; i < ninlined; i++)
-    {
-      newsize =
-	cgraph_estimate_size_after_inlining (INLINED_TIMES (inlined[i]) *
-					     times, inlined[i], what);
-      if (newsize > PARAM_VALUE (PARAM_LARGE_FUNCTION_INSNS)
-	  && newsize >
-	  inlined[i]->local.self_insns *
-	  (100 + PARAM_VALUE (PARAM_LARGE_FUNCTION_GROWTH)) / 100)
-	{
-	  *reason = N_("--param large-function-growth limit reached while inlining the caller");
-	  return false;
-	}
-    }
   return true;
 }
 
@@ -1071,6 +1175,54 @@ cgraph_default_inline_p (struct cgraph_node *n)
     return n->global.insns < MAX_INLINE_INSNS_AUTO;
 }
 
+/* Return true when inlining WHAT would create recursive inlining.
+   We call recursive inlining all cases where same function appears more than
+   once in the single recusion nest path in the inline graph.  */
+
+static bool
+cgraph_recursive_inlining_p (struct cgraph_node *to,
+			     struct cgraph_node *what,
+			     const char **reason)
+{
+  struct cgraph_node *node;
+
+  /* Walk TO and all functions TO is inlined in.  */
+  while (1)
+    {
+      /* We create recursive inlining either by inlining WHAT into something
+	 already inlined in possibly different clone of WHAT.  */
+      if (what->decl == to->decl)
+	goto recursive;
+      /* Or by inlining WHAT into something that is already inlined in WHAT.  */
+      for (node = cgraph_node (to->decl); node; node = node->next_clone)
+	if (node->global.inlined_to == what)
+	  goto recursive;
+      if (!to->callers || to->callers->inline_failed)
+	return false;
+      to = to->callers->caller;
+    }
+recursive:
+  if (reason)
+    *reason = (what->local.disregard_inline_limits
+	       ? N_("recursive inlining") : "");
+  return true;
+}
+
+/* Recompute heap nodes for each of callees.  */
+static void
+update_callee_keys (fibheap_t heap, struct fibnode **heap_node,
+		    struct cgraph_node *node)
+{
+  struct cgraph_edge *e;
+
+  for (e = node->callees; e; e = e->next_callee)
+    if (e->inline_failed && heap_node[e->callee->uid])
+      fibheap_replace_key (heap, heap_node[e->callee->uid],
+			   cgraph_estimate_growth (e->callee));
+    else if (!e->inline_failed)
+      update_callee_keys (heap, heap_node, e->callee);
+}
+
 /* Set inline_failed for all callers of given function to REASON.  */
 
 static void
@@ -1093,15 +1245,12 @@ cgraph_set_inline_failed (struct cgraph_node *node, const char *reason)
    to be passed to cgraph_inlined_into and cgraph_inlined_callees.  */
 
 static void
-cgraph_decide_inlining_of_small_functions (struct cgraph_node **inlined,
-					   struct cgraph_node **inlined_callees)
+cgraph_decide_inlining_of_small_functions (void)
 {
-  int i;
   struct cgraph_node *node;
   fibheap_t heap = fibheap_new ();
   struct fibnode **heap_node =
     xcalloc (cgraph_max_uid, sizeof (struct fibnode *));
-  int ninlined, ninlined_callees;
   int max_insns = ((HOST_WIDEST_INT) initial_insns
 		   * (100 + PARAM_VALUE (PARAM_INLINE_UNIT_GROWTH)) / 100);
 
@@ -1127,7 +1276,7 @@ cgraph_decide_inlining_of_small_functions (struct cgraph_node **inlined,
     fprintf (cgraph_dump_file, "\nDeciding on smaller functions:\n");
   while (overall_insns <= max_insns && (node = fibheap_extract_min (heap)))
     {
-      struct cgraph_edge *e;
+      struct cgraph_edge *e, *next;
       int old_insns = overall_insns;
 
       heap_node[node->uid] = NULL;
@@ -1143,77 +1292,48 @@ cgraph_decide_inlining_of_small_functions (struct cgraph_node **inlined,
 	    N_("--param max-inline-insns-single limit reached after inlining into the callee"));
 	  continue;
 	}
-      ninlined_callees = cgraph_inlined_callees (node, inlined_callees);
-      for (e = node->callers; e; e = e->next_caller)
-	if (e->inline_failed)
-	  {
-	    /* Marking recursive function inlinine has sane semantic and
-	       thus we should not warn on it.  */
- 	    if (e->caller == node)
- 	      {
- 	        e->inline_failed = "";
- 		continue;
- 	      }
-	    ninlined = cgraph_inlined_into (e->caller, inlined);
-	    if (e->callee->output)
-	      e->inline_failed = "";
-	    if (e->callee->output
-		|| !cgraph_check_inline_limits (e->caller, node, inlined,
-						ninlined, &e->inline_failed))
-	      {
-		for (i = 0; i < ninlined; i++)
-		  inlined[i]->output = 0, inlined[i]->aux = 0;
-		if (cgraph_dump_file)
-		  fprintf (cgraph_dump_file, " Not inlining into %s.\n",
-			   cgraph_node_name (e->caller));
-		continue;
-	      }
-	    cgraph_mark_inline (e->caller, node, inlined, ninlined,
-				inlined_callees, ninlined_callees);
-	    if (heap_node[e->caller->uid])
-	      fibheap_replace_key (heap, heap_node[e->caller->uid],
-				   cgraph_estimate_growth (e->caller));
-
-	    /* Size of the functions we updated into has changed, so update
-	       the keys.  */
-	    for (i = 0; i < ninlined; i++)
-	      {
-		inlined[i]->output = 0, inlined[i]->aux = 0;
-		if (heap_node[inlined[i]->uid])
-		  fibheap_replace_key (heap, heap_node[inlined[i]->uid],
-				       cgraph_estimate_growth (inlined[i]));
-	      }
-	    if (cgraph_dump_file)
-	      fprintf (cgraph_dump_file, 
-		       " Inlined into %s which now has %i insns.\n",
-		       cgraph_node_name (e->caller),
-		       e->caller->global.insns);
-	  }
+      for (e = node->callers; e; e = next)
+	{
+	  next = e->next_caller;
+	  if (e->inline_failed)
+	    {
+	      struct cgraph_node *where;
+
+	      if (cgraph_recursive_inlining_p (e->caller, e->callee,
+				      	       &e->inline_failed)
+		  || !cgraph_check_inline_limits (e->caller, e->callee,
+			  			  &e->inline_failed))
+		{
+		  if (cgraph_dump_file)
+		    fprintf (cgraph_dump_file, " Not inlining into %s:%s.\n",
+			     cgraph_node_name (e->caller), e->inline_failed);
+		  continue;
+		}
+	      next = cgraph_mark_inline (e);
+	      where = e->caller;
+	      if (where->global.inlined_to)
+		where = where->global.inlined_to;
+
+	      if (heap_node[where->uid])
+		fibheap_replace_key (heap, heap_node[where->uid],
+				     cgraph_estimate_growth (where));
+
+	      if (cgraph_dump_file)
+		fprintf (cgraph_dump_file, 
+			 " Inlined into %s which now has %i insns.\n",
+			 cgraph_node_name (e->caller),
+			 e->caller->global.insns);
+	    }
+	}
 
       /* Similarly all functions called by the function we just inlined
          are now called more times; update keys.  */
+      update_callee_keys (heap, heap_node, node);
 
-      for (e = node->callees; e; e = e->next_callee)
-	if (e->inline_failed && heap_node[e->callee->uid])
-	  fibheap_replace_key (heap, heap_node[e->callee->uid],
-			       cgraph_estimate_growth (e->callee));
-
-      for (i = 0; i < ninlined_callees; i++)
-	{
-	  struct cgraph_edge *e;
-
-	  for (e = inlined_callees[i]->callees; e; e = e->next_callee)
-	    if (e->inline_failed && heap_node[e->callee->uid])
-	      fibheap_replace_key (heap, heap_node[e->callee->uid],
-				   cgraph_estimate_growth (e->callee));
-
-	  inlined_callees[i]->output = 0;
-	  inlined_callees[i]->aux = 0;
-	}
       if (cgraph_dump_file)
 	fprintf (cgraph_dump_file, 
-		 " Inlined %i times for a net change of %+i insns.\n",
-		 node->global.cloned_times, overall_insns - old_insns);
+		 " Inlined for a net change of %+i insns.\n",
+		 overall_insns - old_insns);
     }
   while ((node = fibheap_extract_min (heap)) != NULL)
     if (!node->local.disregard_inline_limits)
@@ -1232,14 +1352,8 @@ cgraph_decide_inlining (void)
   int nnodes;
   struct cgraph_node **order =
     xcalloc (cgraph_n_nodes, sizeof (struct cgraph_node *));
-  struct cgraph_node **inlined =
-    xcalloc (cgraph_n_nodes, sizeof (struct cgraph_node *));
-  struct cgraph_node **inlined_callees =
-    xcalloc (cgraph_n_nodes, sizeof (struct cgraph_node *));
-  int ninlined;
-  int ninlined_callees;
   int old_insns = 0;
-  int i, y;
+  int i;
 
   for (node = cgraph_nodes; node; node = node->next)
     initial_insns += node->local.self_insns;
@@ -1257,11 +1371,6 @@ cgraph_decide_inlining (void)
 
   if (cgraph_dump_file)
     fprintf (cgraph_dump_file, "\nInlining always_inline functions:\n");
-#ifdef ENABLE_CHECKING
-  for (node = cgraph_nodes; node; node = node->next)
-    if (node->aux || node->output)
-      abort ();
-#endif
 
   /* In the first pass mark all always_inline edges.  Do this with a priority
      so none of our later choices will make this impossible.  */
@@ -1280,51 +1389,30 @@ cgraph_decide_inlining (void)
 	fprintf (cgraph_dump_file,
 		 "\nConsidering %s %i insns (always inline)\n",
 		 cgraph_node_name (e->callee), e->callee->global.insns);
-      ninlined = cgraph_inlined_into (order[i], inlined);
       for (; e; e = e->next_callee)
 	{
 	  old_insns = overall_insns;
-	  if (!e->inline_failed || !e->callee->local.inlinable
-	      || !e->callee->local.disregard_inline_limits)
-  	    continue;
-  	  if (e->callee->output || e->callee == node)
-	    {
-	      e->inline_failed = N_("recursive inlining");
-	      continue;
-	    }
-	  ninlined_callees =
-	    cgraph_inlined_callees (e->callee, inlined_callees);
-	  cgraph_mark_inline (node, e->callee, inlined, ninlined,
-			      inlined_callees, ninlined_callees);
-	  for (y = 0; y < ninlined_callees; y++)
-	    inlined_callees[y]->output = 0, inlined_callees[y]->aux = 0;
+	  if (!e->inline_failed || !e->callee->local.disregard_inline_limits)
+	    continue;
+	  if (cgraph_recursive_inlining_p (order[i], e->callee,
+				  	   &e->inline_failed))
+	    continue;
+	  cgraph_mark_inline (e);
 	  if (cgraph_dump_file)
 	    fprintf (cgraph_dump_file, 
 		     " Inlined into %s which now has %i insns.\n",
 		     cgraph_node_name (node->callees->caller),
 	             node->callees->caller->global.insns);
 	}
-      if (cgraph_dump_file && node->global.cloned_times > 0)
-	fprintf (cgraph_dump_file, 
-		 " Inlined %i times for a net change of %+i insns.\n",
-		 node->global.cloned_times, overall_insns - old_insns);
-      for (y = 0; y < ninlined; y++)
-	inlined[y]->output = 0, inlined[y]->aux = 0;
+	if (cgraph_dump_file)
+	  fprintf (cgraph_dump_file, 
+		   " Inlined for a net change of %+i insns.\n",
+		   overall_insns - old_insns);
     }
-#ifdef ENABLE_CHECKING
-  for (node = cgraph_nodes; node; node = node->next)
-    if (node->aux || node->output)
-      abort ();
-#endif
 
   if (!flag_really_no_inline)
     {
-      cgraph_decide_inlining_of_small_functions (inlined, inlined_callees);
-#ifdef ENABLE_CHECKING
-      for (node = cgraph_nodes; node; node = node->next)
-	if (node->aux || node->output)
-	  abort ();
-#endif
+      cgraph_decide_inlining_of_small_functions ();
 
       if (cgraph_dump_file)
 	fprintf (cgraph_dump_file, "\nDeciding on functions called once:\n");
@@ -1350,7 +1438,6 @@ cgraph_decide_inlining (void)
 		  ok = false;
 	      if (ok)
 		{
-		  const char *dummy_reason;
 		  if (cgraph_dump_file)
 		    fprintf (cgraph_dump_file,
 			     "\nConsidering %s %i insns.\n"
@@ -1358,22 +1445,13 @@ cgraph_decide_inlining (void)
 			     cgraph_node_name (node), node->global.insns,
 			     cgraph_node_name (node->callers->caller),
 			     node->callers->caller->global.insns);
-		  ninlined = cgraph_inlined_into (node->callers->caller,
-		      				  inlined);
+
 		  old_insns = overall_insns;
 
-		  /* Inlining functions once would never cause inlining warnings.  */
-		  if (cgraph_check_inline_limits
-		      (node->callers->caller, node, inlined, ninlined,
-		       &dummy_reason))
+		  if (cgraph_check_inline_limits (node->callers->caller, node,
+					  	  NULL))
 		    {
-		      ninlined_callees =
-			cgraph_inlined_callees (node, inlined_callees);
-		      cgraph_mark_inline (node->callers->caller, node, inlined,
-					  ninlined, inlined_callees,
-					  ninlined_callees);
-		      for (y = 0; y < ninlined_callees; y++)
-			inlined_callees[y]->output = 0, inlined_callees[y]->aux = 0;
+		      cgraph_mark_inline (node->callers);
 		      if (cgraph_dump_file)
 			fprintf (cgraph_dump_file,
 				 " Inlined into %s which now has %i insns"
@@ -1388,12 +1466,14 @@ cgraph_decide_inlining (void)
 			fprintf (cgraph_dump_file,
 				 " Inline limit reached, not inlined.\n");
 		    }
-		  for (y = 0; y < ninlined; y++)
-		    inlined[y]->output = 0, inlined[y]->aux = 0;
 		}
 	    }
 	}
     }
+
+  /* We will never output extern functions we didn't inline. 
+     ??? Perhaps we can prevent accounting of growth of external
+     inline functions.  */
   cgraph_remove_unreachable_nodes ();
 
   if (cgraph_dump_file)
@@ -1403,8 +1483,6 @@ cgraph_decide_inlining (void)
 	     ncalls_inlined, nfunctions_inlined, initial_insns,
 	     overall_insns);
   free (order);
-  free (inlined);
-  free (inlined_callees);
 }
 
 /* Decide on the inlining.  We do so in the topological order to avoid
@@ -1414,95 +1492,45 @@ static void
 cgraph_decide_inlining_incrementally (struct cgraph_node *node)
 {
   struct cgraph_edge *e;
-  struct cgraph_node **inlined =
-    xmalloc (sizeof (struct cgraph_node *) * cgraph_n_nodes);
-  struct cgraph_node **inlined_callees =
-    xmalloc (sizeof (struct cgraph_node *) * cgraph_n_nodes);
-  int ninlined;
-  int ninlined_callees;
-  int y;
-
-  ninlined = cgraph_inlined_into (node, inlined);
 
   /* First of all look for always inline functions.  */
   for (e = node->callees; e; e = e->next_callee)
-    if (e->callee->local.disregard_inline_limits && e->inline_failed
+    if (e->callee->local.disregard_inline_limits
+	&& e->inline_failed
+        && !cgraph_recursive_inlining_p (node, e->callee, &e->inline_failed)
 	/* ??? It is possible that renaming variable removed the function body
 	   in duplicate_decls. See gcc.c-torture/compile/20011119-2.c  */
 	&& DECL_SAVED_TREE (e->callee->decl))
-      {
-	if (e->callee->output || e->callee == node)
-	  {
- 	    e->inline_failed = N_("recursive inlining");
-	    continue;
-	  }
-	ninlined_callees = cgraph_inlined_callees (e->callee, inlined_callees);
-	cgraph_mark_inline (node, e->callee, inlined, ninlined,
-			    inlined_callees, ninlined_callees);
-	for (y = 0; y < ninlined_callees; y++)
-	  inlined_callees[y]->output = 0, inlined_callees[y]->aux = 0;
-      }
+      cgraph_mark_inline (e);
 
+  /* Now do the automatic inlining.  */
   if (!flag_really_no_inline)
-    {
-      /* Now do the automatic inlining.  */
-      for (e = node->callees; e; e = e->next_callee)
-	if (e->callee->local.inlinable && e->inline_failed
-	    && cgraph_default_inline_p (e->callee)
-	    && cgraph_check_inline_limits (node, e->callee, inlined,
-					   ninlined, &e->inline_failed)
-	    && DECL_SAVED_TREE (e->callee->decl))
-	  {
-	    /* Marking recursive function inlinine has sane semantic and thus
-	       we should not warn on it.  */
-	    if (e->callee->output || e->callee == node)
-	      {
-		e->inline_failed = "";
-		continue;
-	      }
-	    ninlined_callees = cgraph_inlined_callees (e->callee,
-						       inlined_callees);
-	    cgraph_mark_inline (node, e->callee, inlined, ninlined,
-				inlined_callees, ninlined_callees);
-	    for (y = 0; y < ninlined_callees; y++)
-	      inlined_callees[y]->output = 0, inlined_callees[y]->aux = 0;
-	  }
-    }
-
-  /* Clear the flags set by cgraph_inlined_into.  */
-  for (y = 0; y < ninlined; y++)
-    inlined[y]->output = 0, inlined[y]->aux = 0;
-
-  free (inlined);
-  free (inlined_callees);
+    for (e = node->callees; e; e = e->next_callee)
+      if (e->callee->local.inlinable
+	  && e->inline_failed
+	  && !e->callee->local.disregard_inline_limits
+	  && !cgraph_recursive_inlining_p (node, e->callee, &e->inline_failed)
+	  && cgraph_check_inline_limits (node, e->callee, &e->inline_failed)
+	  && DECL_SAVED_TREE (e->callee->decl))
+	{
+	  if (cgraph_default_inline_p (e->callee))
+	    cgraph_mark_inline (e);
+	  else
+	    e->inline_failed
+	      = N_("--param max-inline-insns-single limit reached");
+	}
 }
 
 
-/* Return true when CALLER_DECL should be inlined into CALLEE_DECL.
-   When returned false and reason is non-NULL, set it to the reason
-   why the call was not inlined.  */
+/* Return true when CALLER_DECL should be inlined into CALLEE_DECL.  */
 
 bool
-cgraph_inline_p (tree caller_decl, tree callee_decl, const char **reason)
+cgraph_inline_p (struct cgraph_edge *e, const char **reason)
 {
-  struct cgraph_node *caller = cgraph_node (caller_decl);
-  struct cgraph_node *callee = cgraph_node (callee_decl);
-  struct cgraph_edge *e;
-
-  for (e = caller->callees; e; e = e->next_callee)
-    if (e->callee == callee)
-      {
-	if (e->inline_failed && reason)
-	  *reason = e->inline_failed;
-        return !e->inline_failed;
-      }
-  /* We do not record builtins in the callgraph.  Perhaps it would make more
-     sense to do so and then prune out those not overwritten by explicit
-     function body.  */
-  if (reason)
-    *reason = "originally indirect function calls never inlined";
-  return false;
+  *reason = e->inline_failed;
+  return !e->inline_failed;
 }
+
 /* Expand all functions that must be output.
 
    Attempt to topologically sort the nodes so function is output when
@@ -1519,14 +1547,22 @@ cgraph_expand_all_functions (void)
   struct cgraph_node *node;
   struct cgraph_node **order =
     xcalloc (cgraph_n_nodes, sizeof (struct cgraph_node *));
-  int order_pos = 0;
+  int order_pos = 0, new_order_pos = 0;
   int i;
 
   cgraph_mark_functions_to_output ();
 
   order_pos = cgraph_postorder (order);
+  if (order_pos != cgraph_n_nodes)
+    abort ();
 
-  for (i = order_pos - 1; i >= 0; i--)
+  /* Garbage collector may remove inline clones we elliminate during
+     optimization.  So we must be sure to not reference them.  */
+  for (i = 0; i < order_pos; i++)
+    if (order[i]->output)
+      order[new_order_pos++] = order[i];
+
+  for (i = new_order_pos - 1; i >= 0; i--)
     {
       node = order[i];
       if (node->output)
@@ -1569,11 +1605,32 @@ cgraph_mark_local_functions (void)
     fprintf (cgraph_dump_file, "\n\n");
 }
 
+/* Return true when function body of DECL still needs to be kept around
+   for later re-use.  */
+bool
+cgraph_preserve_function_body_p (tree decl)
+{
+  struct cgraph_node *node;
+  /* Keep the body; we're going to dump it.  */
+  if (dump_enabled_p (TDI_all))
+    return true;
+  if (!cgraph_global_info_ready)
+    return (DECL_INLINE (decl) && !flag_really_no_inline);
+  /* Look if there is any clone around.  */
+  for (node = cgraph_node (decl); node; node = node->next_clone)
+    if (node->global.inlined_to)
+      return true;
+  return false;
+}
+
 /* Perform simple optimizations based on callgraph.  */
 
 void
 cgraph_optimize (void)
 {
+#ifdef ENABLE_CHECKING
+  verify_cgraph ();
+#endif
   if (!flag_unit_at_a_time)
     return;
   timevar_push (TV_CGRAPHOPT);
@@ -1587,7 +1644,8 @@ cgraph_optimize (void)
       dump_cgraph (cgraph_dump_file);
     }
 
-  cgraph_decide_inlining ();
+  if (flag_inline_trees)
+    cgraph_decide_inlining ();
   cgraph_global_info_ready = true;
   if (cgraph_dump_file)
     {
@@ -1599,10 +1657,16 @@ cgraph_optimize (void)
   /* Output everything.  */
   if (!quiet_flag)
     fprintf (stderr, "Assembling functions:\n");
+#ifdef ENABLE_CHECKING
+  verify_cgraph ();
+#endif
   cgraph_expand_all_functions ();
   if (cgraph_dump_file)
     {
       fprintf (cgraph_dump_file, "\nFinal ");
       dump_cgraph (cgraph_dump_file);
     }
+#ifdef ENABLE_CHECKING
+  verify_cgraph ();
+#endif
 }