This is the mail archive of the gdb-patches@sources.redhat.com mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [RFA/dwarf] Don't process types multiple times


On Fri, Mar 12, 2004 at 06:44:59PM -0500, Jim Blandy wrote:
> 
> Daniel Jacobowitz <drow@false.org> writes:
> > This independent patch is a performance and correctness fix for full symbol
> > processing.  My description from the intercu branch posting:
> > 
> >  There are two ways that we can process a general DIE: process_die or
> >  read_type_die.  Children of particular DIE types may be processed directly,
> >  but these are the only major dispatch points.  It's interesting to notice
> >  that almost everything called from read_type_die starts with "if (die->type)
> >  return": everything but enumeration types and aggregate types, in fact.
> >  This means that if the first reference to an enumeration or aggregate type
> >  is a DW_AT_type or DW_AT_containing_type in a DIE numerically before the
> >  type's DIE, we'll end up calling new_symbol for it twice.
> > 
> >  Fixing this saves about 8% memory and 4% time from gdb -readnow libc.so.6,
> >  a lot of duplicate entries on maint print symbols which I vaguely recall
> >  being confused about but never investigated, and some serious problems for
> >  inter-CU support.  Without this, types could be added to the wrong symbol
> >  table.
> > 
> > Tested i686-pc-linux-gnu, no regressions.  OK to commit?
> 
> So the essential idea here is to make sure that we only create symbols
> when we reach them via the main traversal, not when we happen to reach
> them by following references with die_type.  The current code relies
> on the existing read_structure_scope and read_enumeration functions to
> both read the type and create the symbol, so it can't distinguish
> between the two cases.  Your patch creates a separate function for
> each role, and ensures that the symbol creation functions are only
> called from the main traversal.
> 
> Is that right?  If so, please commit, to trunk and branch.

Precisely.  I've committed it with your and David's suggested cosmetic
changes - I like the revised function names much better.  Here's the
final patch.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

2004-03-14  Daniel Jacobowitz  <drow@mvista.com>

	* dwarf2read.c (read_structure_type): Rename from
	read_structure_scope.  Don't create a symbol or call process_die. 
	Return immediately if die->type is set.  Call read_type_die before
	dwarf2_add_member_fn.
	(process_structure_scope): New function.
	(read_enumeration_type, process_enumeration_scope): New functions,
	broken out from read_enumeration.  Don't create the enumeration
	type if it has already been created.
	(read_enumeration): Removed.
	(process_die): Call read_structure_type, process_structure_scope,
	read_enumeration_type, and process_enumeration_scope.  Just call
	new_symbol for base and subrange types.  Add a comment about other
	type dies.
	(read_type_die): Call read_enumeration_type.
	(add_partial_structure, new_symbol): Update comments.

Index: dwarf2read.c
===================================================================
RCS file: /cvs/src/src/gdb/dwarf2read.c,v
retrieving revision 1.135
diff -u -p -r1.135 dwarf2read.c
--- dwarf2read.c	21 Feb 2004 02:13:35 -0000	1.135
+++ dwarf2read.c	14 Mar 2004 20:48:54 -0000
@@ -804,7 +804,9 @@ static void dwarf2_add_member_fn (struct
 static void dwarf2_attach_fn_fields_to_type (struct field_info *,
 					     struct type *, struct dwarf2_cu *);
 
-static void read_structure_scope (struct die_info *, struct dwarf2_cu *);
+static void read_structure_type (struct die_info *, struct dwarf2_cu *);
+
+static void process_structure_scope (struct die_info *, struct dwarf2_cu *);
 
 static void read_common_block (struct die_info *, struct dwarf2_cu *);
 
@@ -813,7 +815,9 @@ static void read_namespace (struct die_i
 static const char *namespace_name (struct die_info *die,
 				   int *is_anonymous, struct dwarf2_cu *);
 
-static void read_enumeration (struct die_info *, struct dwarf2_cu *);
+static void read_enumeration_type (struct die_info *, struct dwarf2_cu *);
+
+static void process_enumeration_scope (struct die_info *, struct dwarf2_cu *);
 
 static struct type *dwarf_base_type (int, int, struct dwarf2_cu *);
 
@@ -1673,7 +1677,7 @@ add_partial_structure (struct partial_di
 	 what template types look like, because the demangler
 	 frequently doesn't give the same name as the debug info.  We
 	 could fix this by only using the demangled name to get the
-	 prefix (but see comment in read_structure_scope).  */
+	 prefix (but see comment in read_structure_type).  */
 
       /* FIXME: carlton/2004-01-23: If NAMESPACE equals "", we have
 	 the appropriate debug information, so it would be nice to be
@@ -1949,11 +1953,17 @@ process_die (struct die_info *die, struc
     case DW_TAG_class_type:
     case DW_TAG_structure_type:
     case DW_TAG_union_type:
-      read_structure_scope (die, cu);
+      read_structure_type (die, cu);
+      process_structure_scope (die, cu);
       break;
     case DW_TAG_enumeration_type:
-      read_enumeration (die, cu);
+      read_enumeration_type (die, cu);
+      process_enumeration_scope (die, cu);
       break;
+
+    /* FIXME drow/2004-03-14: These initialize die->type, but do not create
+       a symbol or process any children.  Therefore it doesn't do anything
+       that won't be done on-demand by read_type_die.  */
     case DW_TAG_subroutine_type:
       read_subroutine_type (die, cu);
       break;
@@ -1972,21 +1982,19 @@ process_die (struct die_info *die, struc
     case DW_TAG_string_type:
       read_tag_string_type (die, cu);
       break;
+    /* END FIXME */
+
     case DW_TAG_base_type:
       read_base_type (die, cu);
-      if (dwarf2_attr (die, DW_AT_name, cu))
-	{
-	  /* Add a typedef symbol for the base type definition.  */
-	  new_symbol (die, die->type, cu);
-	}
+      /* Add a typedef symbol for the type definition, if it has a
+	 DW_AT_name.  */
+      new_symbol (die, die->type, cu);
       break;
     case DW_TAG_subrange_type:
       read_subrange_type (die, cu);
-      if (dwarf2_attr (die, DW_AT_name, cu))
-       {
-         /* Add a typedef symbol for the base type definition.  */
-         new_symbol (die, die->type, cu);
-       }
+      /* Add a typedef symbol for the type definition, if it has a
+         DW_AT_name.  */
+      new_symbol (die, die->type, cu);
       break;
     case DW_TAG_common_block:
       read_common_block (die, cu);
@@ -2995,7 +3003,7 @@ dwarf2_attach_fn_fields_to_type (struct 
    suppresses creating a symbol table entry itself).  */
 
 static void
-read_structure_scope (struct die_info *die, struct dwarf2_cu *cu)
+read_structure_type (struct die_info *die, struct dwarf2_cu *cu)
 {
   struct objfile *objfile = cu->objfile;
   struct type *type;
@@ -3008,6 +3016,9 @@ read_structure_scope (struct die_info *d
      any.  */
   int need_to_update_name = 0;
 
+  if (die->type)
+    return;
+
   type = alloc_type (objfile);
 
   INIT_CPLUS_SPECIFIC (type);
@@ -3106,7 +3117,7 @@ read_structure_scope (struct die_info *d
 	  else if (child_die->tag == DW_TAG_subprogram)
 	    {
 	      /* C++ member function. */
-	      process_die (child_die, cu);
+	      read_type_die (child_die, cu);
 	      dwarf2_add_member_fn (&fi, child_die, type, cu);
 	      if (need_to_update_name)
 		{
@@ -3149,10 +3160,6 @@ read_structure_scope (struct die_info *d
 	      /* C++ base class field.  */
 	      dwarf2_add_field (&fi, child_die, cu);
 	    }
-	  else
-	    {
-	      process_die (child_die, cu);
-	    }
 	  child_die = sibling_die (child_die);
 	}
 
@@ -3209,8 +3216,6 @@ read_structure_scope (struct die_info *d
 	    }
 	}
 
-      new_symbol (die, type, cu);
-
       do_cleanups (back_to);
     }
   else
@@ -3224,26 +3229,53 @@ read_structure_scope (struct die_info *d
     do_cleanups (back_to);
 }
 
-/* Given a pointer to a die which begins an enumeration, process all
-   the dies that define the members of the enumeration.
+static void
+process_structure_scope (struct die_info *die, struct dwarf2_cu *cu)
+{
+  struct objfile *objfile = cu->objfile;
+  const char *previous_prefix = processing_current_prefix;
 
-   This will be much nicer in draft 6 of the DWARF spec when our
-   members will be dies instead squished into the DW_AT_element_list
-   attribute.
+  if (TYPE_TAG_NAME (die->type) != NULL)
+    processing_current_prefix = TYPE_TAG_NAME (die->type);
 
-   NOTE: We reverse the order of the element list.  */
+  if (die->child != NULL && ! die_is_declaration (die, cu))
+    {
+      struct die_info *child_die;
+
+      child_die = die->child;
+
+      while (child_die && child_die->tag)
+	{
+	  if (child_die->tag == DW_TAG_member
+	      || child_die->tag == DW_TAG_variable
+	      || child_die->tag == DW_TAG_inheritance)
+	    {
+	      /* Do nothing.  */
+	    }
+	  else
+	    process_die (child_die, cu);
+
+	  child_die = sibling_die (child_die);
+	}
+
+      new_symbol (die, die->type, cu);
+    }
+
+  processing_current_prefix = previous_prefix;
+}
+
+/* Given a DW_AT_enumeration_type die, set its type.  We do not
+   complete the type's fields yet, or create any symbols.  */
 
 static void
-read_enumeration (struct die_info *die, struct dwarf2_cu *cu)
+read_enumeration_type (struct die_info *die, struct dwarf2_cu *cu)
 {
   struct objfile *objfile = cu->objfile;
-  struct die_info *child_die;
   struct type *type;
-  struct field *fields;
   struct attribute *attr;
-  struct symbol *sym;
-  int num_fields;
-  int unsigned_enum = 1;
+
+  if (die->type)
+    return;
 
   type = alloc_type (objfile);
 
@@ -3278,6 +3310,26 @@ read_enumeration (struct die_info *die, 
       TYPE_LENGTH (type) = 0;
     }
 
+  die->type = type;
+}
+
+/* Given a pointer to a die which begins an enumeration, process all
+   the dies that define the members of the enumeration, and create the
+   symbol for the enumeration type.
+
+   NOTE: We reverse the order of the element list.  */
+
+static void
+process_enumeration_scope (struct die_info *die, struct dwarf2_cu *cu)
+{
+  struct objfile *objfile = cu->objfile;
+  struct die_info *child_die;
+  struct field *fields;
+  struct attribute *attr;
+  struct symbol *sym;
+  int num_fields;
+  int unsigned_enum = 1;
+
   num_fields = 0;
   fields = NULL;
   if (die->child != NULL)
@@ -3294,7 +3346,7 @@ read_enumeration (struct die_info *die, 
 	      attr = dwarf2_attr (child_die, DW_AT_name, cu);
 	      if (attr)
 		{
-		  sym = new_symbol (child_die, type, cu);
+		  sym = new_symbol (child_die, die->type, cu);
 		  if (SYMBOL_VALUE (sym) < 0)
 		    unsigned_enum = 0;
 
@@ -3321,18 +3373,18 @@ read_enumeration (struct die_info *die, 
 
       if (num_fields)
 	{
-	  TYPE_NFIELDS (type) = num_fields;
-	  TYPE_FIELDS (type) = (struct field *)
-	    TYPE_ALLOC (type, sizeof (struct field) * num_fields);
-	  memcpy (TYPE_FIELDS (type), fields,
+	  TYPE_NFIELDS (die->type) = num_fields;
+	  TYPE_FIELDS (die->type) = (struct field *)
+	    TYPE_ALLOC (die->type, sizeof (struct field) * num_fields);
+	  memcpy (TYPE_FIELDS (die->type), fields,
 		  sizeof (struct field) * num_fields);
 	  xfree (fields);
 	}
       if (unsigned_enum)
-	TYPE_FLAGS (type) |= TYPE_FLAG_UNSIGNED;
+	TYPE_FLAGS (die->type) |= TYPE_FLAG_UNSIGNED;
     }
-  die->type = type;
-  new_symbol (die, type, cu);
+
+  new_symbol (die, die->type, cu);
 }
 
 /* Extract all information from a DW_TAG_array_type DIE and put it in
@@ -5666,7 +5718,7 @@ new_symbol (struct die_info *die, struct
 
 	  /* Make sure that the symbol includes appropriate enclosing
 	     classes/namespaces in its name.  These are calculated in
-	     read_structure_scope, and the correct name is saved in
+	     read_structure_type, and the correct name is saved in
 	     the type.  */
 
 	  if (cu->language == language_cplus)
@@ -6009,10 +6061,10 @@ read_type_die (struct die_info *die, str
     case DW_TAG_class_type:
     case DW_TAG_structure_type:
     case DW_TAG_union_type:
-      read_structure_scope (die, cu);
+      read_structure_type (die, cu);
       break;
     case DW_TAG_enumeration_type:
-      read_enumeration (die, cu);
+      read_enumeration_type (die, cu);
       break;
     case DW_TAG_subprogram:
     case DW_TAG_subroutine_type:


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]