public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [RFC PATCH 0/3] Call summary class
@ 2017-02-27 16:39 Martin Jambor
  2017-02-27 16:39 ` [RFC PATCH 1/3] call_summary to keep info about cgraph_edges Martin Jambor
                   ` (3 more replies)
  0 siblings, 4 replies; 9+ messages in thread
From: Martin Jambor @ 2017-02-27 16:39 UTC (permalink / raw)
  To: GCC Patches; +Cc: Jan Hubicka, Martin Liska

Hello,

the patch sequence in this thread adds a call_summary class, which is
analogous to function_summary we already have but which gathers
information about call graph edges, rather than nodes.

The first patch implements the class itself, the second modifies
ipa-prop.[ch] and ipa-cp.c to use it instead of a vector indexed by
edge->uid and the third patch is a semi-related cleanup I spotted
along the way.

I have LTO-bootstrapped and tested the patches on x86_64-linux and
successfully LTO-built Mozilla Firefox with it.  I'll be grateful for
any comments, otherwise I'll queue the series for the next stage1.

Thanks,

Martin



Martin Jambor (3):
  call_summary to keep info about cgraph_edges
  Use call_summary in ipa-prop and ipa-cp
  Remove ipa_update_after_lto_read

 gcc/ipa-cp.c              |   4 -
 gcc/ipa-inline-analysis.c |   2 +-
 gcc/ipa-inline.c          |   3 -
 gcc/ipa-profile.c         |   2 +-
 gcc/ipa-prop.c            |  82 ++++--------------
 gcc/ipa-prop.h            |  60 +++++++++----
 gcc/symbol-summary.h      | 214 +++++++++++++++++++++++++++++++++++++++++++++-
 7 files changed, 273 insertions(+), 94 deletions(-)

-- 
2.11.1

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [RFC PATCH 3/3] Remove ipa_update_after_lto_read
  2017-02-27 16:39 [RFC PATCH 0/3] Call summary class Martin Jambor
  2017-02-27 16:39 ` [RFC PATCH 1/3] call_summary to keep info about cgraph_edges Martin Jambor
@ 2017-02-27 16:39 ` Martin Jambor
  2017-02-28 10:50   ` jh
  2017-02-27 16:56 ` [RFC PATCH 2/3] Use call_summary in ipa-prop and ipa-cp Martin Jambor
  2017-05-03 16:54 ` [RFC PATCH 0/3] Call summary class Martin Jambor
  3 siblings, 1 reply; 9+ messages in thread
From: Martin Jambor @ 2017-02-27 16:39 UTC (permalink / raw)
  To: GCC Patches; +Cc: Jan Hubicka, Martin Liska

Hello,

when working on call summaries, I have found a weird function
ipa_update_after_lto_read which currently only makes sure that IPA-CP
function and edge summaries exist, at times when they already have to
exist.

So I did some digging in history and found out that originally the
function was meant for setting a called_with_var_arguments flags where
appropriate after LTO stream-in.  We have however removed that flag in
2011 and the function is basically useless since then (although only
with call summaries it is definitely a no-OP).  Thus, I'll propose to
remove it next stage1.

Thanks,

Martin

2017-02-27  Martin Jambor  <mjambor@suse.cz>

	* ipa-prop.c (ipa_update_after_lto_read): Removed.
	* ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
	* ipa-cp.c (ipcp_propagate_stage): Do not call
	ipa_update_after_lto_read.
	* ipa-inline.c (ipa_inline): Likewise.
---
 gcc/ipa-cp.c     |  4 ----
 gcc/ipa-inline.c |  3 ---
 gcc/ipa-prop.c   | 11 -----------
 gcc/ipa-prop.h   |  1 -
 4 files changed, 19 deletions(-)

diff --git a/gcc/ipa-cp.c b/gcc/ipa-cp.c
index aa3c9973a66..6a6467ce8ba 100644
--- a/gcc/ipa-cp.c
+++ b/gcc/ipa-cp.c
@@ -3234,10 +3234,6 @@ ipcp_propagate_stage (struct ipa_topo_info *topo)
   if (dump_file)
     fprintf (dump_file, "\n Propagating constants:\n\n");
 
-  if (in_lto_p)
-    ipa_update_after_lto_read ();
-
-
   FOR_EACH_DEFINED_FUNCTION (node)
   {
     struct ipa_node_params *info = IPA_NODE_REF (node);
diff --git a/gcc/ipa-inline.c b/gcc/ipa-inline.c
index 4843f8a2f8e..b681c568c89 100644
--- a/gcc/ipa-inline.c
+++ b/gcc/ipa-inline.c
@@ -2379,9 +2379,6 @@ ipa_inline (void)
 
   order = XCNEWVEC (struct cgraph_node *, symtab->cgraph_count);
 
-  if (in_lto_p && optimize)
-    ipa_update_after_lto_read ();
-
   if (dump_file)
     dump_inline_summaries (dump_file);
 
diff --git a/gcc/ipa-prop.c b/gcc/ipa-prop.c
index 6d598763d02..47936a92adb 100644
--- a/gcc/ipa-prop.c
+++ b/gcc/ipa-prop.c
@@ -5096,17 +5096,6 @@ ipa_prop_read_jump_functions (void)
     }
 }
 
-/* After merging units, we can get mismatch in argument counts.
-   Also decl merging might've rendered parameter lists obsolete.
-   Also compute called_with_variable_arg info.  */
-
-void
-ipa_update_after_lto_read (void)
-{
-  ipa_check_create_node_params ();
-  ipa_check_create_edge_args ();
-}
-
 void
 write_ipcp_transformation_info (output_block *ob, cgraph_node *node)
 {
diff --git a/gcc/ipa-prop.h b/gcc/ipa-prop.h
index 4d8ee0d351d..bf5d02f8e3a 100644
--- a/gcc/ipa-prop.h
+++ b/gcc/ipa-prop.h
@@ -867,7 +867,6 @@ void ipa_prop_write_jump_functions (void);
 void ipa_prop_read_jump_functions (void);
 void ipcp_write_transformation_summaries (void);
 void ipcp_read_transformation_summaries (void);
-void ipa_update_after_lto_read (void);
 int ipa_get_param_decl_index (struct ipa_node_params *, tree);
 tree ipa_value_from_jfunc (struct ipa_node_params *info,
 			   struct ipa_jump_func *jfunc);
-- 
2.11.1

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [RFC PATCH 1/3] call_summary to keep info about cgraph_edges
  2017-02-27 16:39 [RFC PATCH 0/3] Call summary class Martin Jambor
@ 2017-02-27 16:39 ` Martin Jambor
  2017-02-28 10:48   ` jh
  2017-02-27 16:39 ` [RFC PATCH 3/3] Remove ipa_update_after_lto_read Martin Jambor
                   ` (2 subsequent siblings)
  3 siblings, 1 reply; 9+ messages in thread
From: Martin Jambor @ 2017-02-27 16:39 UTC (permalink / raw)
  To: GCC Patches; +Cc: Jan Hubicka, Martin Liska

Hello,

this patch is an actual implementation of the call_summary class (I
hope the name is a good analogy to the function_summary we have, I am
opened to other suggestions).  I have kept the implementation close to
the existing one of function-summary, there are I think only two
notable differences:

  - there is no call_graph hook for new edge creation, which is in
    fact quite natural, so there is no virtual function for that
    event, and

  - edge UIDs start with zero, and the inliner does use it to index a
    vector, so not only did I need to remove asserts that UIDs are
    non-zero, I also had to actually increment them before using them
    because apparently our hash-map does not like zero hashes.

Apart from that, I have introduced a predicate method exists to both
summaries, in order to figure out whether a summary for a given
node/edge already exists.  This is useful for example for dumping,
there is no need for dumping to allocate new stuff by querying.

Moreover, I have also spotted what I think is a bug in
function_summary::symtab_removal which does not call the destructor of
the item it holds when a function is removed and the item is
garbage-collecotr allocated.  Fixed in a simple hunk below.

I'll be grateful for any comments and/or suggestions,

Martin

2017-02-24  Martin Jambor  <mjambor@suse.cz>

	* symbol-summary.h (function_summary): New method exists.
	(function_summary::symtab_removal): Deallocate through release.
	(call_summary): New class.
	(gt_ggc_mx): New overload.
	(gt_pch_nx): Likewise.
	(gt_pch_nx): Likewise.
---
 gcc/symbol-summary.h | 214 ++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 210 insertions(+), 4 deletions(-)

diff --git a/gcc/symbol-summary.h b/gcc/symbol-summary.h
index 3bcd14522c8..313082d7d5d 100644
--- a/gcc/symbol-summary.h
+++ b/gcc/symbol-summary.h
@@ -126,6 +126,12 @@ public:
     return m_map.elements ();
   }
 
+  /* Return true if a summary for the given NODE already exists.  */
+  bool exists (cgraph_node *node)
+  {
+    return m_map.get (node->summary_uid) != NULL;
+  }
+
   /* Enable insertion hook invocation.  */
   void enable_insertion_hook ()
   {
@@ -160,10 +166,7 @@ public:
     if (v)
       {
 	summary->remove (node, *v);
-
-	if (!summary->m_ggc)
-	  delete (*v);
-
+	summary->release (*v);
 	summary->m_map.remove (summary_uid);
       }
   }
@@ -251,4 +254,207 @@ gt_pch_nx(function_summary<T *>* const& summary, gt_pointer_operator op,
   gt_pch_nx (&summary->m_map, op, cookie);
 }
 
+/* An impossible class templated by non-pointers so, which makes sure that only
+   summaries gathering pointers can be created.  */
+
+template <class T>
+class call_summary
+{
+private:
+  call_summary();
+};
+
+/* Class to store auxiliary information about call graph edges.  */
+
+template <class T>
+class GTY((user)) call_summary <T *>
+{
+public:
+  /* Default construction takes SYMTAB as an argument.  */
+  call_summary (symbol_table *symtab, bool ggc = false): m_ggc (ggc),
+    m_map (13, ggc), m_released (false), m_symtab (symtab)
+  {
+    m_symtab_removal_hook =
+      symtab->add_edge_removal_hook
+      (call_summary::symtab_removal, this);
+    m_symtab_duplication_hook =
+      symtab->add_edge_duplication_hook
+      (call_summary::symtab_duplication, this);
+  }
+
+  /* Destructor.  */
+  virtual ~call_summary ()
+  {
+    release ();
+  }
+
+  /* Destruction method that can be called for GGT purpose.  */
+  void release ()
+  {
+    if (m_released)
+      return;
+
+    m_symtab->remove_edge_removal_hook (m_symtab_removal_hook);
+    m_symtab->remove_edge_duplication_hook (m_symtab_duplication_hook);
+
+    /* Release all summaries.  */
+    typedef typename hash_map <map_hash, T *>::iterator map_iterator;
+    for (map_iterator it = m_map.begin (); it != m_map.end (); ++it)
+      release ((*it).second);
+
+    m_released = true;
+  }
+
+  /* Traverses all summarys with a function F called with
+     ARG as argument.  */
+  template<typename Arg, bool (*f)(const T &, Arg)>
+  void traverse (Arg a) const
+  {
+    m_map.traverse <f> (a);
+  }
+
+  /* Basic implementation of removal operation.  */
+  virtual void remove (cgraph_edge *, T *) {}
+
+  /* Basic implementation of duplication operation.  */
+  virtual void duplicate (cgraph_edge *, cgraph_edge *, T *, T *) {}
+
+  /* Allocates new data that are stored within map.  */
+  T* allocate_new ()
+  {
+    /* Call gcc_internal_because we do not want to call finalizer for
+       a type T.  We call dtor explicitly.  */
+    return m_ggc ? new (ggc_internal_alloc (sizeof (T))) T () : new T () ;
+  }
+
+  /* Release an item that is stored within map.  */
+  void release (T *item)
+  {
+    if (m_ggc)
+      {
+	item->~T ();
+	ggc_free (item);
+      }
+    else
+      delete item;
+  }
+
+  /* Getter for summary callgraph edge pointer.  */
+  T* get (cgraph_edge *edge)
+  {
+    return get (hashable_uid (edge));
+  }
+
+  /* Return number of elements handled by data structure.  */
+  size_t elements ()
+  {
+    return m_map.elements ();
+  }
+
+  /* Return true if a summary for the given EDGE already exists.  */
+  bool exists (cgraph_edge *edge)
+  {
+    return m_map.get (hashable_uid (edge)) != NULL;
+  }
+
+  /* Symbol removal hook that is registered to symbol table.  */
+  static void symtab_removal (cgraph_edge *edge, void *data)
+  {
+    call_summary *summary = (call_summary <T *> *) (data);
+
+    int h_uid = summary->hashable_uid (edge);
+    T **v = summary->m_map.get (h_uid);
+
+    if (v)
+      {
+	summary->remove (edge, *v);
+	summary->release (*v);
+	summary->m_map.remove (h_uid);
+      }
+  }
+
+  /* Symbol duplication hook that is registered to symbol table.  */
+  static void symtab_duplication (cgraph_edge *edge1, cgraph_edge *edge2,
+				  void *data)
+  {
+    call_summary *summary = (call_summary <T *> *) (data);
+    T **v = summary->m_map.get (summary->hashable_uid (edge1));
+
+    if (v)
+      {
+	/* This load is necessary, because we insert a new value!  */
+	T *data = *v;
+	T *duplicate = summary->allocate_new ();
+	summary->m_map.put (summary->hashable_uid (edge2), duplicate);
+	summary->duplicate (edge1, edge2, data, duplicate);
+      }
+  }
+
+protected:
+  /* Indication if we use ggc summary.  */
+  bool m_ggc;
+
+private:
+  typedef int_hash <int, 0, -1> map_hash;
+
+  /* Getter for summary callgraph ID.  */
+  T* get (int uid)
+  {
+    bool existed;
+    T **v = &m_map.get_or_insert (uid, &existed);
+    if (!existed)
+      *v = allocate_new ();
+
+    return *v;
+  }
+
+  /* Get a hashable uid of EDGE.  */
+  int hashable_uid (cgraph_edge *edge)
+  {
+    /* Edge uids start at zero which our hash_map does not like.  */
+    return edge->uid + 1;
+  }
+
+  /* Main summary store, where summary ID is used as key.  */
+  hash_map <map_hash, T *> m_map;
+  /* Internal summary removal hook pointer.  */
+  cgraph_edge_hook_list *m_symtab_removal_hook;
+  /* Internal summary duplication hook pointer.  */
+  cgraph_2edge_hook_list *m_symtab_duplication_hook;
+  /* Indicates if the summary is released.  */
+  bool m_released;
+  /* Symbol table the summary is registered to.  */
+  symbol_table *m_symtab;
+
+  template <typename U> friend void gt_ggc_mx (call_summary <U *> * const &);
+  template <typename U> friend void gt_pch_nx (call_summary <U *> * const &);
+  template <typename U> friend void gt_pch_nx (call_summary <U *> * const &,
+      gt_pointer_operator, void *);
+};
+
+template <typename T>
+void
+gt_ggc_mx(call_summary<T *>* const &summary)
+{
+  gcc_checking_assert (summary->m_ggc);
+  gt_ggc_mx (&summary->m_map);
+}
+
+template <typename T>
+void
+gt_pch_nx(call_summary<T *>* const &summary)
+{
+  gcc_checking_assert (summary->m_ggc);
+  gt_pch_nx (&summary->m_map);
+}
+
+template <typename T>
+void
+gt_pch_nx(call_summary<T *>* const& summary, gt_pointer_operator op,
+	  void *cookie)
+{
+  gcc_checking_assert (summary->m_ggc);
+  gt_pch_nx (&summary->m_map, op, cookie);
+}
+
 #endif  /* GCC_SYMBOL_SUMMARY_H  */
-- 
2.11.1

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [RFC PATCH 2/3] Use call_summary in ipa-prop and ipa-cp
  2017-02-27 16:39 [RFC PATCH 0/3] Call summary class Martin Jambor
  2017-02-27 16:39 ` [RFC PATCH 1/3] call_summary to keep info about cgraph_edges Martin Jambor
  2017-02-27 16:39 ` [RFC PATCH 3/3] Remove ipa_update_after_lto_read Martin Jambor
@ 2017-02-27 16:56 ` Martin Jambor
  2017-02-28 11:06   ` jh
  2017-05-03 16:54 ` [RFC PATCH 0/3] Call summary class Martin Jambor
  3 siblings, 1 reply; 9+ messages in thread
From: Martin Jambor @ 2017-02-27 16:56 UTC (permalink / raw)
  To: GCC Patches; +Cc: Jan Hubicka, Martin Liska

Hello,

this is patch is afairly straightforward conversion from use of a
vector indexed by edge->uid to use of the new call_summary from the
previous patch.

The patch is generally a cleanup, hashing is a nicer method of keeping
call-site related information than a gigantic vector that we never
shrink during its lifetime.  Moreover, it should allow further
cleanups and of course is a nice way of testing that the previous
patch works.

Any comments and/or suggestions are welcome,

Martin

2017-02-27  Martin Jambor  <mjambor@suse.cz>

	* ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
	tag.  Added a default constructor and a destructor.
	(ipa_edge_args_sum_t): New class;
	(ipa_edge_args_sum): Declare.
	(ipa_edge_args_vector): Remove declaration.
	(IPA_EDGE_REF): Use ipa_edge_args_sum.
	(ipa_free_edge_args_substructures): Remove declaration.
	(ipa_check_create_edge_args): Use ipa_edge_args_sum.
	(ipa_edge_args_info_available_for_edge_p): Likewise.
	* ipa-prop.c (ipa_edge_args_vector): Removed.
	(edge_removal_hook_holder): Likewise.
	(edge_duplication_hook_holder): Likewise.
	(ipa_edge_args_sum): New variable.
	(ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
	ipa_edge_args_vector.
	(ipa_free_edge_args_substructures): Likewise.
	(ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
	ipa_edge_args_vector.
	(ipa_edge_removal_hook): Turned into method
	ipa_edge_args_sum_t::remove.
	(ipa_edge_duplication_hook): Turned into method
	ipa_edge_args_sum_t::duplicate.
	(ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
	registering edge hooks.
	(ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
	* ipa-inline-analysis.c (estimate_function_body_sizes): Test
	ipa_edge_args_sum instead of ipa_edge_args_vector.
	* ipa-profile.c (ipa_profile): Likewise.
---
 gcc/ipa-inline-analysis.c |  2 +-
 gcc/ipa-profile.c         |  2 +-
 gcc/ipa-prop.c            | 71 +++++++++++------------------------------------
 gcc/ipa-prop.h            | 59 +++++++++++++++++++++++++++++----------
 4 files changed, 63 insertions(+), 71 deletions(-)

diff --git a/gcc/ipa-inline-analysis.c b/gcc/ipa-inline-analysis.c
index 611faab570f..639fc05fccd 100644
--- a/gcc/ipa-inline-analysis.c
+++ b/gcc/ipa-inline-analysis.c
@@ -2937,7 +2937,7 @@ estimate_function_body_sizes (struct cgraph_node *node, bool early)
     {
       if (!early)
         loop_optimizer_finalize ();
-      else if (!ipa_edge_args_vector)
+      else if (!ipa_edge_args_sum)
 	ipa_free_all_node_params ();
       free_dominance_info (CDI_DOMINATORS);
     }
diff --git a/gcc/ipa-profile.c b/gcc/ipa-profile.c
index ae1ca2f3762..0bfbfa1e447 100644
--- a/gcc/ipa-profile.c
+++ b/gcc/ipa-profile.c
@@ -620,7 +620,7 @@ ipa_profile (void)
 				 "Not speculating: target is overwritable "
 				 "and can be discarded.\n");
 		    }
-		  else if (ipa_node_params_sum && ipa_edge_args_vector
+		  else if (ipa_node_params_sum && ipa_edge_args_sum
 			   && (!vec_safe_is_empty
 			       (IPA_NODE_REF (n2)->descriptors))
 			   && ipa_get_param_count (IPA_NODE_REF (n2))
diff --git a/gcc/ipa-prop.c b/gcc/ipa-prop.c
index e4e44ce20c6..6d598763d02 100644
--- a/gcc/ipa-prop.c
+++ b/gcc/ipa-prop.c
@@ -57,12 +57,10 @@ along with GCC; see the file COPYING3.  If not see
 ipa_node_params_t *ipa_node_params_sum = NULL;
 /* Vector of IPA-CP transformation data for each clone.  */
 vec<ipcp_transformation_summary, va_gc> *ipcp_transformations;
-/* Vector where the parameter infos are actually stored. */
-vec<ipa_edge_args, va_gc> *ipa_edge_args_vector;
+/* Edge summary for IPA-CP edge information.  */
+ipa_edge_args_sum_t *ipa_edge_args_sum;
 
 /* Holders of ipa cgraph hooks: */
-static struct cgraph_edge_hook_list *edge_removal_hook_holder;
-static struct cgraph_2edge_hook_list *edge_duplication_hook_holder;
 static struct cgraph_node_hook_list *function_insertion_hook_holder;
 
 /* Description of a reference to an IPA constant.  */
@@ -3537,7 +3535,7 @@ ipa_propagate_indirect_call_infos (struct cgraph_edge *cs,
      (i.e. during early inlining).  */
   if (!ipa_node_params_sum)
     return false;
-  gcc_assert (ipa_edge_args_vector);
+  gcc_assert (ipa_edge_args_sum);
 
   propagate_controlled_uses (cs);
   changed = propagate_info_to_inlined_callees (cs, cs->callee, new_edges);
@@ -3545,31 +3543,16 @@ ipa_propagate_indirect_call_infos (struct cgraph_edge *cs,
   return changed;
 }
 
-/* Frees all dynamically allocated structures that the argument info points
-   to.  */
-
-void
-ipa_free_edge_args_substructures (struct ipa_edge_args *args)
-{
-  vec_free (args->jump_functions);
-  memset (args, 0, sizeof (*args));
-}
-
 /* Free all ipa_edge structures.  */
 
 void
 ipa_free_all_edge_args (void)
 {
-  int i;
-  struct ipa_edge_args *args;
-
-  if (!ipa_edge_args_vector)
+  if (!ipa_edge_args_sum)
     return;
 
-  FOR_EACH_VEC_ELT (*ipa_edge_args_vector, i, args)
-    ipa_free_edge_args_substructures (args);
-
-  vec_free (ipa_edge_args_vector);
+  ipa_edge_args_sum->release ();
+  ipa_edge_args_sum = NULL;
 }
 
 /* Free all ipa_node_params structures.  */
@@ -3601,18 +3584,12 @@ ipa_set_node_agg_value_chain (struct cgraph_node *node,
   (*ipcp_transformations)[node->uid].agg_values = aggvals;
 }
 
-/* Hook that is called by cgraph.c when an edge is removed.  */
+/* Hook that is called by cgraph.c when an edge is removed.  Adjust reference
+   count data structures accordingly.  */
 
-static void
-ipa_edge_removal_hook (struct cgraph_edge *cs, void *data ATTRIBUTE_UNUSED)
+void
+ipa_edge_args_sum_t::remove (cgraph_edge *cs, ipa_edge_args *args)
 {
-  struct ipa_edge_args *args;
-
-  /* During IPA-CP updating we can be called on not-yet analyzed clones.  */
-  if (vec_safe_length (ipa_edge_args_vector) <= (unsigned)cs->uid)
-    return;
-
-  args = IPA_EDGE_REF (cs);
   if (args->jump_functions)
     {
       struct ipa_jump_func *jf;
@@ -3627,24 +3604,17 @@ ipa_edge_removal_hook (struct cgraph_edge *cs, void *data ATTRIBUTE_UNUSED)
 	    rdesc->cs = NULL;
 	}
     }
-
-  ipa_free_edge_args_substructures (IPA_EDGE_REF (cs));
 }
 
-/* Hook that is called by cgraph.c when an edge is duplicated.  */
+/* Method invoked when an edge is duplicated.  Copy ipa_edge_args and adjust
+   reference count data strucutres accordingly.  */
 
-static void
-ipa_edge_duplication_hook (struct cgraph_edge *src, struct cgraph_edge *dst,
-			   void *)
+void
+ipa_edge_args_sum_t::duplicate (cgraph_edge *src, cgraph_edge *dst,
+				ipa_edge_args *old_args, ipa_edge_args *new_args)
 {
-  struct ipa_edge_args *old_args, *new_args;
   unsigned int i;
 
-  ipa_check_create_edge_args ();
-
-  old_args = IPA_EDGE_REF (src);
-  new_args = IPA_EDGE_REF (dst);
-
   new_args->jump_functions = vec_safe_copy (old_args->jump_functions);
   if (old_args->polymorphic_call_contexts)
     new_args->polymorphic_call_contexts
@@ -3811,13 +3781,8 @@ void
 ipa_register_cgraph_hooks (void)
 {
   ipa_check_create_node_params ();
+  ipa_check_create_edge_args ();
 
-  if (!edge_removal_hook_holder)
-    edge_removal_hook_holder =
-      symtab->add_edge_removal_hook (&ipa_edge_removal_hook, NULL);
-  if (!edge_duplication_hook_holder)
-    edge_duplication_hook_holder =
-      symtab->add_edge_duplication_hook (&ipa_edge_duplication_hook, NULL);
   function_insertion_hook_holder =
       symtab->add_cgraph_insertion_hook (&ipa_add_new_function, NULL);
 }
@@ -3827,10 +3792,6 @@ ipa_register_cgraph_hooks (void)
 static void
 ipa_unregister_cgraph_hooks (void)
 {
-  symtab->remove_edge_removal_hook (edge_removal_hook_holder);
-  edge_removal_hook_holder = NULL;
-  symtab->remove_edge_duplication_hook (edge_duplication_hook_holder);
-  edge_duplication_hook_holder = NULL;
   symtab->remove_cgraph_insertion_hook (function_insertion_hook_holder);
   function_insertion_hook_holder = NULL;
 }
diff --git a/gcc/ipa-prop.h b/gcc/ipa-prop.h
index 8f7eb088813..4d8ee0d351d 100644
--- a/gcc/ipa-prop.h
+++ b/gcc/ipa-prop.h
@@ -558,9 +558,24 @@ void ipcp_grow_transformations_if_necessary (void);
 
 /* ipa_edge_args stores information related to a callsite and particularly its
    arguments.  It can be accessed by the IPA_EDGE_REF macro.  */
-struct GTY(()) ipa_edge_args
+
+class GTY((for_user)) ipa_edge_args
 {
-  /* Vector of the callsite's jump function of each parameter.  */
+ public:
+
+  /* Default constructor.  */
+  ipa_edge_args () : jump_functions (NULL), polymorphic_call_contexts (NULL)
+    {}
+
+  /* Destructor.  */
+  ~ipa_edge_args ()
+    {
+      vec_free (jump_functions);
+      vec_free (polymorphic_call_contexts);
+    }
+
+  /* Vectors of the callsite's jump function and polymorphic context
+     information of each parameter.  */
   vec<ipa_jump_func, va_gc> *jump_functions;
   vec<ipa_polymorphic_call_context, va_gc> *polymorphic_call_contexts;
 };
@@ -610,19 +625,35 @@ public:
 			  ipa_node_params *data2);
 };
 
+/* Summary to manange ipa_edge_args structures.  */
+
+class GTY((user)) ipa_edge_args_sum_t : public call_summary <ipa_edge_args *>
+{
+ public:
+  ipa_edge_args_sum_t (symbol_table *table, bool ggc)
+    : call_summary<ipa_edge_args *> (table, ggc) { }
+
+  /* Hook that is called by summary when an edge is duplicated.  */
+  virtual void remove (cgraph_edge *cs, ipa_edge_args *args);
+  /* Hook that is called by summary when an edge is duplicated.  */
+  virtual void duplicate (cgraph_edge *src,
+			  cgraph_edge *dst,
+			  ipa_edge_args *old_args,
+			  ipa_edge_args *new_args);
+};
+
 /* Function summary where the parameter infos are actually stored. */
 extern GTY(()) ipa_node_params_t * ipa_node_params_sum;
+/* Call summary to store information about edges such as jump functions.  */
+extern GTY(()) ipa_edge_args_sum_t *ipa_edge_args_sum;
 
 /* Vector of IPA-CP transformation data for each clone.  */
 extern GTY(()) vec<ipcp_transformation_summary, va_gc> *ipcp_transformations;
-/* Vector where the parameter infos are actually stored. */
-extern GTY(()) vec<ipa_edge_args, va_gc> *ipa_edge_args_vector;
-
 
 /* Return the associated parameter/argument info corresponding to the given
    node/edge.  */
 #define IPA_NODE_REF(NODE) (ipa_node_params_sum->get (NODE))
-#define IPA_EDGE_REF(EDGE) (&(*ipa_edge_args_vector)[(EDGE)->uid])
+#define IPA_EDGE_REF(EDGE) (ipa_edge_args_sum->get (EDGE))
 /* This macro checks validity of index returned by
    ipa_get_param_decl_index function.  */
 #define IS_VALID_JUMP_FUNC_INDEX(I) ((I) != -1)
@@ -630,7 +661,6 @@ extern GTY(()) vec<ipa_edge_args, va_gc> *ipa_edge_args_vector;
 /* Creating and freeing ipa_node_params and ipa_edge_args.  */
 void ipa_create_all_node_params (void);
 void ipa_create_all_edge_args (void);
-void ipa_free_edge_args_substructures (struct ipa_edge_args *);
 void ipa_free_all_node_params (void);
 void ipa_free_all_edge_args (void);
 void ipa_free_all_structures_after_ipa_cp (void);
@@ -657,19 +687,20 @@ ipa_check_create_node_params (void)
 static inline void
 ipa_check_create_edge_args (void)
 {
-  if (vec_safe_length (ipa_edge_args_vector)
-      <= (unsigned) symtab->edges_max_uid)
-    vec_safe_grow_cleared (ipa_edge_args_vector, symtab->edges_max_uid + 1);
+  if (!ipa_edge_args_sum)
+    ipa_edge_args_sum
+      = (new (ggc_cleared_alloc <ipa_edge_args_sum_t> ())
+	 ipa_edge_args_sum_t (symtab, true));
 }
 
-/* Returns true if the array of edge infos is large enough to accommodate an
-   info for EDGE.  The main purpose of this function is that debug dumping
-   function can check info availability without causing reallocations.  */
+/* Returns true if edge summary contains a record for EDGE.  The main purpose
+   of this function is that debug dumping function can check info availability
+   without causing allocations.  */
 
 static inline bool
 ipa_edge_args_info_available_for_edge_p (struct cgraph_edge *edge)
 {
-  return ((unsigned) edge->uid < vec_safe_length (ipa_edge_args_vector));
+  return ipa_edge_args_sum->exists (edge);
 }
 
 static inline ipcp_transformation_summary *
-- 
2.11.1

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [RFC PATCH 1/3] call_summary to keep info about cgraph_edges
  2017-02-27 16:39 ` [RFC PATCH 1/3] call_summary to keep info about cgraph_edges Martin Jambor
@ 2017-02-28 10:48   ` jh
  0 siblings, 0 replies; 9+ messages in thread
From: jh @ 2017-02-28 10:48 UTC (permalink / raw)
  To: Martin Jambor; +Cc: GCC Patches, Jan Hubicka, Martin Liska

Dne 2017-02-27 17:35, Martin Jambor napsal:
> Hello,
> 
> this patch is an actual implementation of the call_summary class (I
> hope the name is a good analogy to the function_summary we have, I am
> opened to other suggestions).  I have kept the implementation close to
> the existing one of function-summary, there are I think only two
> notable differences:
> 
>   - there is no call_graph hook for new edge creation, which is in
>     fact quite natural, so there is no virtual function for that
>     event, and
> 
>   - edge UIDs start with zero, and the inliner does use it to index a
>     vector, so not only did I need to remove asserts that UIDs are
>     non-zero, I also had to actually increment them before using them
>     because apparently our hash-map does not like zero hashes.
> 
> Apart from that, I have introduced a predicate method exists to both
> summaries, in order to figure out whether a summary for a given
> node/edge already exists.  This is useful for example for dumping,
> there is no need for dumping to allocate new stuff by querying.
> 
> Moreover, I have also spotted what I think is a bug in
> function_summary::symtab_removal which does not call the destructor of
> the item it holds when a function is removed and the item is
> garbage-collecotr allocated.  Fixed in a simple hunk below.
> 
> I'll be grateful for any comments and/or suggestions,
> 
> Martin
> 
> 2017-02-24  Martin Jambor  <mjambor@suse.cz>
> 
> 	* symbol-summary.h (function_summary): New method exists.
> 	(function_summary::symtab_removal): Deallocate through release.
> 	(call_summary): New class.
> 	(gt_ggc_mx): New overload.
> 	(gt_pch_nx): Likewise.
> 	(gt_pch_nx): Likewise.

I like this - the way summaries was managed in ipa-prop was always bit 
twisty.
path is OK for next stage1. I guess we may consider putting 
cgraph_indirect_call_info into a summary too
and probably break it into multiple summaries (for example polymorhic 
call info can probably go separately)

Honza
> ---
>  gcc/symbol-summary.h | 214 
> ++++++++++++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 210 insertions(+), 4 deletions(-)
> 
> diff --git a/gcc/symbol-summary.h b/gcc/symbol-summary.h
> index 3bcd14522c8..313082d7d5d 100644
> --- a/gcc/symbol-summary.h
> +++ b/gcc/symbol-summary.h
> @@ -126,6 +126,12 @@ public:
>      return m_map.elements ();
>    }
> 
> +  /* Return true if a summary for the given NODE already exists.  */
> +  bool exists (cgraph_node *node)
> +  {
> +    return m_map.get (node->summary_uid) != NULL;
> +  }
> +
>    /* Enable insertion hook invocation.  */
>    void enable_insertion_hook ()
>    {
> @@ -160,10 +166,7 @@ public:
>      if (v)
>        {
>  	summary->remove (node, *v);
> -
> -	if (!summary->m_ggc)
> -	  delete (*v);
> -
> +	summary->release (*v);
>  	summary->m_map.remove (summary_uid);
>        }
>    }
> @@ -251,4 +254,207 @@ gt_pch_nx(function_summary<T *>* const& summary,
> gt_pointer_operator op,
>    gt_pch_nx (&summary->m_map, op, cookie);
>  }
> 
> +/* An impossible class templated by non-pointers so, which makes sure 
> that only
> +   summaries gathering pointers can be created.  */
> +
> +template <class T>
> +class call_summary
> +{
> +private:
> +  call_summary();
> +};
> +
> +/* Class to store auxiliary information about call graph edges.  */
> +
> +template <class T>
> +class GTY((user)) call_summary <T *>
> +{
> +public:
> +  /* Default construction takes SYMTAB as an argument.  */
> +  call_summary (symbol_table *symtab, bool ggc = false): m_ggc (ggc),
> +    m_map (13, ggc), m_released (false), m_symtab (symtab)
> +  {
> +    m_symtab_removal_hook =
> +      symtab->add_edge_removal_hook
> +      (call_summary::symtab_removal, this);
> +    m_symtab_duplication_hook =
> +      symtab->add_edge_duplication_hook
> +      (call_summary::symtab_duplication, this);
> +  }
> +
> +  /* Destructor.  */
> +  virtual ~call_summary ()
> +  {
> +    release ();
> +  }
> +
> +  /* Destruction method that can be called for GGT purpose.  */
> +  void release ()
> +  {
> +    if (m_released)
> +      return;
> +
> +    m_symtab->remove_edge_removal_hook (m_symtab_removal_hook);
> +    m_symtab->remove_edge_duplication_hook 
> (m_symtab_duplication_hook);
> +
> +    /* Release all summaries.  */
> +    typedef typename hash_map <map_hash, T *>::iterator map_iterator;
> +    for (map_iterator it = m_map.begin (); it != m_map.end (); ++it)
> +      release ((*it).second);
> +
> +    m_released = true;
> +  }
> +
> +  /* Traverses all summarys with a function F called with
> +     ARG as argument.  */
> +  template<typename Arg, bool (*f)(const T &, Arg)>
> +  void traverse (Arg a) const
> +  {
> +    m_map.traverse <f> (a);
> +  }
> +
> +  /* Basic implementation of removal operation.  */
> +  virtual void remove (cgraph_edge *, T *) {}
> +
> +  /* Basic implementation of duplication operation.  */
> +  virtual void duplicate (cgraph_edge *, cgraph_edge *, T *, T *) {}
> +
> +  /* Allocates new data that are stored within map.  */
> +  T* allocate_new ()
> +  {
> +    /* Call gcc_internal_because we do not want to call finalizer for
> +       a type T.  We call dtor explicitly.  */
> +    return m_ggc ? new (ggc_internal_alloc (sizeof (T))) T () : new T 
> () ;
> +  }
> +
> +  /* Release an item that is stored within map.  */
> +  void release (T *item)
> +  {
> +    if (m_ggc)
> +      {
> +	item->~T ();
> +	ggc_free (item);
> +      }
> +    else
> +      delete item;
> +  }
> +
> +  /* Getter for summary callgraph edge pointer.  */
> +  T* get (cgraph_edge *edge)
> +  {
> +    return get (hashable_uid (edge));
> +  }
> +
> +  /* Return number of elements handled by data structure.  */
> +  size_t elements ()
> +  {
> +    return m_map.elements ();
> +  }
> +
> +  /* Return true if a summary for the given EDGE already exists.  */
> +  bool exists (cgraph_edge *edge)
> +  {
> +    return m_map.get (hashable_uid (edge)) != NULL;
> +  }
> +
> +  /* Symbol removal hook that is registered to symbol table.  */
> +  static void symtab_removal (cgraph_edge *edge, void *data)
> +  {
> +    call_summary *summary = (call_summary <T *> *) (data);
> +
> +    int h_uid = summary->hashable_uid (edge);
> +    T **v = summary->m_map.get (h_uid);
> +
> +    if (v)
> +      {
> +	summary->remove (edge, *v);
> +	summary->release (*v);
> +	summary->m_map.remove (h_uid);
> +      }
> +  }
> +
> +  /* Symbol duplication hook that is registered to symbol table.  */
> +  static void symtab_duplication (cgraph_edge *edge1, cgraph_edge 
> *edge2,
> +				  void *data)
> +  {
> +    call_summary *summary = (call_summary <T *> *) (data);
> +    T **v = summary->m_map.get (summary->hashable_uid (edge1));
> +
> +    if (v)
> +      {
> +	/* This load is necessary, because we insert a new value!  */
> +	T *data = *v;
> +	T *duplicate = summary->allocate_new ();
> +	summary->m_map.put (summary->hashable_uid (edge2), duplicate);
> +	summary->duplicate (edge1, edge2, data, duplicate);
> +      }
> +  }
> +
> +protected:
> +  /* Indication if we use ggc summary.  */
> +  bool m_ggc;
> +
> +private:
> +  typedef int_hash <int, 0, -1> map_hash;
> +
> +  /* Getter for summary callgraph ID.  */
> +  T* get (int uid)
> +  {
> +    bool existed;
> +    T **v = &m_map.get_or_insert (uid, &existed);
> +    if (!existed)
> +      *v = allocate_new ();
> +
> +    return *v;
> +  }
> +
> +  /* Get a hashable uid of EDGE.  */
> +  int hashable_uid (cgraph_edge *edge)
> +  {
> +    /* Edge uids start at zero which our hash_map does not like.  */
> +    return edge->uid + 1;
> +  }
> +
> +  /* Main summary store, where summary ID is used as key.  */
> +  hash_map <map_hash, T *> m_map;
> +  /* Internal summary removal hook pointer.  */
> +  cgraph_edge_hook_list *m_symtab_removal_hook;
> +  /* Internal summary duplication hook pointer.  */
> +  cgraph_2edge_hook_list *m_symtab_duplication_hook;
> +  /* Indicates if the summary is released.  */
> +  bool m_released;
> +  /* Symbol table the summary is registered to.  */
> +  symbol_table *m_symtab;
> +
> +  template <typename U> friend void gt_ggc_mx (call_summary <U *> * 
> const &);
> +  template <typename U> friend void gt_pch_nx (call_summary <U *> * 
> const &);
> +  template <typename U> friend void gt_pch_nx (call_summary <U *> * 
> const &,
> +      gt_pointer_operator, void *);
> +};
> +
> +template <typename T>
> +void
> +gt_ggc_mx(call_summary<T *>* const &summary)
> +{
> +  gcc_checking_assert (summary->m_ggc);
> +  gt_ggc_mx (&summary->m_map);
> +}
> +
> +template <typename T>
> +void
> +gt_pch_nx(call_summary<T *>* const &summary)
> +{
> +  gcc_checking_assert (summary->m_ggc);
> +  gt_pch_nx (&summary->m_map);
> +}
> +
> +template <typename T>
> +void
> +gt_pch_nx(call_summary<T *>* const& summary, gt_pointer_operator op,
> +	  void *cookie)
> +{
> +  gcc_checking_assert (summary->m_ggc);
> +  gt_pch_nx (&summary->m_map, op, cookie);
> +}
> +
>  #endif  /* GCC_SYMBOL_SUMMARY_H  */

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [RFC PATCH 3/3] Remove ipa_update_after_lto_read
  2017-02-27 16:39 ` [RFC PATCH 3/3] Remove ipa_update_after_lto_read Martin Jambor
@ 2017-02-28 10:50   ` jh
  0 siblings, 0 replies; 9+ messages in thread
From: jh @ 2017-02-28 10:50 UTC (permalink / raw)
  To: Martin Jambor; +Cc: GCC Patches, Jan Hubicka, Martin Liska

Dne 2017-02-27 17:35, Martin Jambor napsal:
> Hello,
> 
> when working on call summaries, I have found a weird function
> ipa_update_after_lto_read which currently only makes sure that IPA-CP
> function and edge summaries exist, at times when they already have to
> exist.
> 
> So I did some digging in history and found out that originally the
> function was meant for setting a called_with_var_arguments flags where
> appropriate after LTO stream-in.  We have however removed that flag in
> 2011 and the function is basically useless since then (although only
> with call summaries it is definitely a no-OP).  Thus, I'll propose to
> remove it next stage1.
> 
> Thanks,
> 
> Martin
> 
> 2017-02-27  Martin Jambor  <mjambor@suse.cz>
> 
> 	* ipa-prop.c (ipa_update_after_lto_read): Removed.
> 	* ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
> 	* ipa-cp.c (ipcp_propagate_stage): Do not call
> 	ipa_update_after_lto_read.
> 	* ipa-inline.c (ipa_inline): Likewise.

OK, thanks!
Whole streaming infrastructure would deserve major cleanups next stage1 
:)

Honza

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [RFC PATCH 2/3] Use call_summary in ipa-prop and ipa-cp
  2017-02-27 16:56 ` [RFC PATCH 2/3] Use call_summary in ipa-prop and ipa-cp Martin Jambor
@ 2017-02-28 11:06   ` jh
  2017-02-28 17:40     ` Martin Jambor
  0 siblings, 1 reply; 9+ messages in thread
From: jh @ 2017-02-28 11:06 UTC (permalink / raw)
  To: Martin Jambor; +Cc: GCC Patches, Jan Hubicka, Martin Liska

Dne 2017-02-27 17:35, Martin Jambor napsal:
> Hello,
> 
> this is patch is afairly straightforward conversion from use of a
> vector indexed by edge->uid to use of the new call_summary from the
> previous patch.
> 
> The patch is generally a cleanup, hashing is a nicer method of keeping
> call-site related information than a gigantic vector that we never
> shrink during its lifetime.  Moreover, it should allow further
> cleanups and of course is a nice way of testing that the previous
> patch works.
> 
> Any comments and/or suggestions are welcome,
> 
> Martin
> 
> 2017-02-27  Martin Jambor  <mjambor@suse.cz>
> 
> 	* ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
> 	tag.  Added a default constructor and a destructor.
> 	(ipa_edge_args_sum_t): New class;
> 	(ipa_edge_args_sum): Declare.
> 	(ipa_edge_args_vector): Remove declaration.
> 	(IPA_EDGE_REF): Use ipa_edge_args_sum.
> 	(ipa_free_edge_args_substructures): Remove declaration.
> 	(ipa_check_create_edge_args): Use ipa_edge_args_sum.
> 	(ipa_edge_args_info_available_for_edge_p): Likewise.
> 	* ipa-prop.c (ipa_edge_args_vector): Removed.
> 	(edge_removal_hook_holder): Likewise.
> 	(edge_duplication_hook_holder): Likewise.
> 	(ipa_edge_args_sum): New variable.
> 	(ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
> 	ipa_edge_args_vector.
> 	(ipa_free_edge_args_substructures): Likewise.
> 	(ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
> 	ipa_edge_args_vector.
> 	(ipa_edge_removal_hook): Turned into method
> 	ipa_edge_args_sum_t::remove.
> 	(ipa_edge_duplication_hook): Turned into method
> 	ipa_edge_args_sum_t::duplicate.
> 	(ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
> 	registering edge hooks.
> 	(ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
> 	* ipa-inline-analysis.c (estimate_function_body_sizes): Test
> 	ipa_edge_args_sum instead of ipa_edge_args_vector.
> 	* ipa-profile.c (ipa_profile): Likewise.
OK
Did you check what memory consumption consquences it have for Firefox?
Honza

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [RFC PATCH 2/3] Use call_summary in ipa-prop and ipa-cp
  2017-02-28 11:06   ` jh
@ 2017-02-28 17:40     ` Martin Jambor
  0 siblings, 0 replies; 9+ messages in thread
From: Martin Jambor @ 2017-02-28 17:40 UTC (permalink / raw)
  To: jh; +Cc: GCC Patches, Jan Hubicka, Martin Liska

Hi,

On Tue, Feb 28, 2017 at 11:50:01AM +0100, jh wrote:
> Dne 2017-02-27 17:35, Martin Jambor napsal:
> > Hello,
> > 
> > this is patch is afairly straightforward conversion from use of a
> > vector indexed by edge->uid to use of the new call_summary from the
> > previous patch.
> > 
> > The patch is generally a cleanup, hashing is a nicer method of keeping
> > call-site related information than a gigantic vector that we never
> > shrink during its lifetime.  Moreover, it should allow further
> > cleanups and of course is a nice way of testing that the previous
> > patch works.
> > 
> > Any comments and/or suggestions are welcome,
> > 
> > Martin
> > 
> > 2017-02-27  Martin Jambor  <mjambor@suse.cz>
> > 
> > 	* ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
> > 	tag.  Added a default constructor and a destructor.
> > 	(ipa_edge_args_sum_t): New class;
> > 	(ipa_edge_args_sum): Declare.
> > 	(ipa_edge_args_vector): Remove declaration.
> > 	(IPA_EDGE_REF): Use ipa_edge_args_sum.
> > 	(ipa_free_edge_args_substructures): Remove declaration.
> > 	(ipa_check_create_edge_args): Use ipa_edge_args_sum.
> > 	(ipa_edge_args_info_available_for_edge_p): Likewise.
> > 	* ipa-prop.c (ipa_edge_args_vector): Removed.
> > 	(edge_removal_hook_holder): Likewise.
> > 	(edge_duplication_hook_holder): Likewise.
> > 	(ipa_edge_args_sum): New variable.
> > 	(ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
> > 	ipa_edge_args_vector.
> > 	(ipa_free_edge_args_substructures): Likewise.
> > 	(ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
> > 	ipa_edge_args_vector.
> > 	(ipa_edge_removal_hook): Turned into method
> > 	ipa_edge_args_sum_t::remove.
> > 	(ipa_edge_duplication_hook): Turned into method
> > 	ipa_edge_args_sum_t::duplicate.
> > 	(ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
> > 	registering edge hooks.
> > 	(ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
> > 	* ipa-inline-analysis.c (estimate_function_body_sizes): Test
> > 	ipa_edge_args_sum instead of ipa_edge_args_vector.
> > 	* ipa-profile.c (ipa_profile): Likewise.
> OK
> Did you check what memory consumption consquences it have for Firefox?

There is no noticeable difference, unfortunately.

Martin

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [RFC PATCH 0/3] Call summary class
  2017-02-27 16:39 [RFC PATCH 0/3] Call summary class Martin Jambor
                   ` (2 preceding siblings ...)
  2017-02-27 16:56 ` [RFC PATCH 2/3] Use call_summary in ipa-prop and ipa-cp Martin Jambor
@ 2017-05-03 16:54 ` Martin Jambor
  3 siblings, 0 replies; 9+ messages in thread
From: Martin Jambor @ 2017-05-03 16:54 UTC (permalink / raw)
  To: GCC Patches

On Mon, Feb 27, 2017 at 05:36:48PM +0100, Martin Jambor wrote:
> Hello,
> 
> the patch sequence in this thread adds a call_summary class, which is
> analogous to function_summary we already have but which gathers
> information about call graph edges, rather than nodes.
> 
> The first patch implements the class itself, the second modifies
> ipa-prop.[ch] and ipa-cp.c to use it instead of a vector indexed by
> edge->uid and the third patch is a semi-related cleanup I spotted
> along the way.
> 
> I have LTO-bootstrapped and tested the patches on x86_64-linux and
> successfully LTO-built Mozilla Firefox with it.  I'll be grateful for
> any comments, otherwise I'll queue the series for the next stage1.
> 

After Honza re-approved the patches in person, I have committed them
as revisions 247557, 247558 and 247559.

Thanks,

Martin

> 
> Martin Jambor (3):
>   call_summary to keep info about cgraph_edges
>   Use call_summary in ipa-prop and ipa-cp
>   Remove ipa_update_after_lto_read
> 
>  gcc/ipa-cp.c              |   4 -
>  gcc/ipa-inline-analysis.c |   2 +-
>  gcc/ipa-inline.c          |   3 -
>  gcc/ipa-profile.c         |   2 +-
>  gcc/ipa-prop.c            |  82 ++++--------------
>  gcc/ipa-prop.h            |  60 +++++++++----
>  gcc/symbol-summary.h      | 214 +++++++++++++++++++++++++++++++++++++++++++++-
>  7 files changed, 273 insertions(+), 94 deletions(-)
> 
> -- 
> 2.11.1

^ permalink raw reply	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2017-05-03 16:52 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-02-27 16:39 [RFC PATCH 0/3] Call summary class Martin Jambor
2017-02-27 16:39 ` [RFC PATCH 1/3] call_summary to keep info about cgraph_edges Martin Jambor
2017-02-28 10:48   ` jh
2017-02-27 16:39 ` [RFC PATCH 3/3] Remove ipa_update_after_lto_read Martin Jambor
2017-02-28 10:50   ` jh
2017-02-27 16:56 ` [RFC PATCH 2/3] Use call_summary in ipa-prop and ipa-cp Martin Jambor
2017-02-28 11:06   ` jh
2017-02-28 17:40     ` Martin Jambor
2017-05-03 16:54 ` [RFC PATCH 0/3] Call summary class Martin Jambor

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).