[pushed v2 0/2] Avoid copying in name lookup

[pushed v2 0/2] Avoid copying in name lookup

[Tom Tromey]

Here's v2 of the series to avoid copying in name lookup.

This version addresses all the review comments.  I re-tested it on the
buildbot, and so I'm checking it in.

Tom

[pushed v2 1/2] Avoid some copying in psymtab.c

[Tom Tromey]

I noticed that psymtab.c was always copying the search string in
psymtab_search_name, even when it wasn't necessary.  This patch
removes this function in favor of using the make_ignore_params feature
of lookup_name_info.

Once I had done that, I noticed that lookup_partial_symbol was
creating a lookup_name_info.  However, this function called in loops,
causing even more excess allocation.  This patch further fixes this by
hosting the creation of the lookup_name_info into the callers.

gdb/ChangeLog
2020-04-01  Tom Tromey  <tromey@adacore.com>

	* psymtab.c (psymtab_search_name): Remove function.
	(psym_lookup_symbol): Create search name and lookup name here.
	(lookup_partial_symbol): Remove "name" parameter; add
	lookup_name.
	(psym_expand_symtabs_for_function): Update.
---
 gdb/ChangeLog |  8 +++++++
 gdb/psymtab.c | 59 +++++++++++++++------------------------------------
 2 files changed, 25 insertions(+), 42 deletions(-)

diff --git a/gdb/psymtab.c b/gdb/psymtab.c
index 56c1b68b541..d5e61f88dc6 100644
--- a/gdb/psymtab.c
+++ b/gdb/psymtab.c
@@ -39,7 +39,8 @@
 
 static struct partial_symbol *lookup_partial_symbol (struct objfile *,
 						     struct partial_symtab *,
-						     const char *, int,
+						     const lookup_name_info &,
+						     int,
 						     domain_enum);
 
 static const char *psymtab_to_fullname (struct partial_symtab *ps);
@@ -482,9 +483,12 @@ psym_lookup_symbol (struct objfile *objfile,
 
   lookup_name_info lookup_name (name, symbol_name_match_type::FULL);
 
+  lookup_name_info psym_lookup_name = lookup_name.make_ignore_params ();
+
   for (partial_symtab *ps : require_partial_symbols (objfile, true))
     {
-      if (!ps->readin_p () && lookup_partial_symbol (objfile, ps, name,
+      if (!ps->readin_p () && lookup_partial_symbol (objfile, ps,
+						     psym_lookup_name,
 						     psymtab_index, domain))
 	{
 	  struct symbol *sym, *with_opaque = NULL;
@@ -612,42 +616,14 @@ match_partial_symbol (struct objfile *objfile,
   return NULL;
 }
 
-/* Returns the name used to search psymtabs.  Unlike symtabs, psymtabs do
-   not contain any method/function instance information (since this would
-   force reading type information while reading psymtabs).  Therefore,
-   if NAME contains overload information, it must be stripped before searching
-   psymtabs.  */
-
-static gdb::unique_xmalloc_ptr<char>
-psymtab_search_name (const char *name)
-{
-  switch (current_language->la_language)
-    {
-    case language_cplus:
-      {
-	if (strchr (name, '('))
-	  {
-	    gdb::unique_xmalloc_ptr<char> ret = cp_remove_params (name);
-
-	    if (ret)
-	      return ret;
-	  }
-      }
-      break;
-
-    default:
-      break;
-    }
-
-  return make_unique_xstrdup (name);
-}
-
-/* Look, in partial_symtab PST, for symbol whose natural name is NAME.
-   Check the global symbols if GLOBAL, the static symbols if not.  */
+/* Look, in partial_symtab PST, for symbol whose natural name is
+   LOOKUP_NAME.  Check the global symbols if GLOBAL, the static
+   symbols if not.  */
 
 static struct partial_symbol *
 lookup_partial_symbol (struct objfile *objfile,
-		       struct partial_symtab *pst, const char *name,
+		       struct partial_symtab *pst,
+		       const lookup_name_info &lookup_name,
 		       int global, domain_enum domain)
 {
   struct partial_symbol **start, **psym;
@@ -658,10 +634,6 @@ lookup_partial_symbol (struct objfile *objfile,
   if (length == 0)
     return NULL;
 
-  gdb::unique_xmalloc_ptr<char> search_name = psymtab_search_name (name);
-
-  lookup_name_info lookup_name (search_name.get (), symbol_name_match_type::FULL);
-
   start = (global ?
 	   &objfile->partial_symtabs->global_psymbols[pst->globals_offset] :
 	   &objfile->partial_symtabs->static_psymbols[pst->statics_offset]);
@@ -686,7 +658,7 @@ lookup_partial_symbol (struct objfile *objfile,
 	    internal_error (__FILE__, __LINE__,
 			    _("failed internal consistency check"));
 	  if (strcmp_iw_ordered ((*center)->ginfo.search_name (),
-				 search_name.get ()) >= 0)
+				 lookup_name.name ().c_str ()) >= 0)
 	    {
 	      top = center;
 	    }
@@ -1044,14 +1016,17 @@ static void
 psym_expand_symtabs_for_function (struct objfile *objfile,
 				  const char *func_name)
 {
+  lookup_name_info base_lookup (func_name, symbol_name_match_type::FULL);
+  lookup_name_info lookup_name = base_lookup.make_ignore_params ();
+
   for (partial_symtab *ps : require_partial_symbols (objfile, true))
     {
       if (ps->readin_p ())
 	continue;
 
-      if ((lookup_partial_symbol (objfile, ps, func_name, 1, VAR_DOMAIN)
+      if ((lookup_partial_symbol (objfile, ps, lookup_name, 1, VAR_DOMAIN)
 	   != NULL)
-	  || (lookup_partial_symbol (objfile, ps, func_name, 0, VAR_DOMAIN)
+	  || (lookup_partial_symbol (objfile, ps, lookup_name, 0, VAR_DOMAIN)
 	      != NULL))
 	psymtab_to_symtab (objfile, ps);
     }
-- 
2.21.1

[pushed v2 2/2] Avoid copying in lookup_name_info

[Tom Tromey]

lookup_name_info always copies the name that is passed in.  However,
normally a copy is not needed.  This patch changes this class to avoid
copying.  This required changing the "name" method to return something
else; I chose a gdb::string_view, to avoid excessive calls to strlen
in the code using the lookup_name_info.  However, as this class does
not allow an arbitrary string_view, I've also added a c_str method
that guarantees a \0-terminated result -- a pedantic difference but
one that respects the string_view contract, IMO.

gdb/ChangeLog
2020-04-01  Tom Tromey  <tromey@adacore.com>

	* symtab.h (class lookup_name_info) <lookup_name_info>: Change
	"name" parameter to rvalue reference.  Initialize m_name_holder.
	<lookup_name_info>: New overloads.
	<name>: Return gdb::string_view.
	<c_str>: New method.
	<make_ignore_params>: Update.
	<search_name_hash>: Update.
	<language_lookup_name>: Return const char *.
	<m_name>: Change type.
	* symtab.c (demangle_for_lookup_info::demangle_for_lookup_info)
	(demangle_for_lookup_info::demangle_for_lookup_info): Update.
	(lookup_name_info::match_any): Update.
	* psymtab.c (match_partial_symbol, lookup_partial_symbol):
	Update.
	* minsyms.c (linkage_name_str): Update.
	* language.c (default_symbol_name_matcher): Update.
	* dwarf2/read.c (mapped_index_base::find_name_components_bounds):
	Update.
	* ada-lang.c (ada_fold_name): Change parameter to string_view.
	(ada_lookup_name_info::ada_lookup_name_info): Update.
	(literal_symbol_name_matcher): Update.
---
 gdb/ChangeLog     | 24 +++++++++++++++++++++
 gdb/ada-lang.c    | 36 ++++++++++++++++---------------
 gdb/dwarf2/read.c |  2 +-
 gdb/language.c    |  4 ++--
 gdb/minsyms.c     |  2 +-
 gdb/psymtab.c     |  5 ++---
 gdb/symtab.c      |  8 +++----
 gdb/symtab.h      | 55 ++++++++++++++++++++++++++++++++++++-----------
 8 files changed, 95 insertions(+), 41 deletions(-)

diff --git a/gdb/ada-lang.c b/gdb/ada-lang.c
index 565299a5ca1..029a7912a03 100644
--- a/gdb/ada-lang.c
+++ b/gdb/ada-lang.c
@@ -1016,17 +1016,17 @@ ada_encode (const char *decoded)
    to next call.  */
 
 static char *
-ada_fold_name (const char *name)
+ada_fold_name (gdb::string_view name)
 {
   static char *fold_buffer = NULL;
   static size_t fold_buffer_size = 0;
 
-  int len = strlen (name);
+  int len = name.size ();
   GROW_VECT (fold_buffer, fold_buffer_size, len + 1);
 
   if (name[0] == '\'')
     {
-      strncpy (fold_buffer, name + 1, len - 2);
+      strncpy (fold_buffer, name.data () + 1, len - 2);
       fold_buffer[len - 2] = '\000';
     }
   else
@@ -5657,8 +5657,8 @@ add_nonlocal_symbols (struct obstack *obstackp,
   if (num_defns_collected (obstackp) == 0 && global && !is_wild_match)
     {
       const char *name = ada_lookup_name (lookup_name);
-      lookup_name_info name1 (std::string ("<_ada_") + name + '>',
-			      symbol_name_match_type::FULL);
+      std::string bracket_name = std::string ("<_ada_") + name + '>';
+      lookup_name_info name1 (bracket_name, symbol_name_match_type::FULL);
 
       for (objfile *objfile : current_program_space->objfiles ())
         {
@@ -13946,14 +13946,16 @@ do_exact_match (const char *symbol_search_name,
 
 ada_lookup_name_info::ada_lookup_name_info (const lookup_name_info &lookup_name)
 {
-  const std::string &user_name = lookup_name.name ();
+  gdb::string_view user_name = lookup_name.name ();
 
   if (user_name[0] == '<')
     {
       if (user_name.back () == '>')
-	m_encoded_name = user_name.substr (1, user_name.size () - 2);
+	m_encoded_name
+	  = user_name.substr (1, user_name.size () - 2).to_string ();
       else
-	m_encoded_name = user_name.substr (1, user_name.size () - 1);
+	m_encoded_name
+	  = user_name.substr (1, user_name.size () - 1).to_string ();
       m_encoded_p = true;
       m_verbatim_p = true;
       m_wild_match_p = false;
@@ -13963,19 +13965,19 @@ ada_lookup_name_info::ada_lookup_name_info (const lookup_name_info &lookup_name)
     {
       m_verbatim_p = false;
 
-      m_encoded_p = user_name.find ("__") != std::string::npos;
+      m_encoded_p = user_name.find ("__") != gdb::string_view::npos;
 
       if (!m_encoded_p)
 	{
-	  const char *folded = ada_fold_name (user_name.c_str ());
+	  const char *folded = ada_fold_name (user_name);
 	  const char *encoded = ada_encode_1 (folded, false);
 	  if (encoded != NULL)
 	    m_encoded_name = encoded;
 	  else
-	    m_encoded_name = user_name;
+	    m_encoded_name = user_name.to_string ();
 	}
       else
-	m_encoded_name = user_name;
+	m_encoded_name = user_name.to_string ();
 
       /* Handle the 'package Standard' special case.  See description
 	 of m_standard_p.  */
@@ -14022,12 +14024,12 @@ literal_symbol_name_matcher (const char *symbol_search_name,
 			     const lookup_name_info &lookup_name,
 			     completion_match_result *comp_match_res)
 {
-  const std::string &name = lookup_name.name ();
+  gdb::string_view name_view = lookup_name.name ();
 
-  int cmp = (lookup_name.completion_mode ()
-	     ? strncmp (symbol_search_name, name.c_str (), name.size ())
-	     : strcmp (symbol_search_name, name.c_str ()));
-  if (cmp == 0)
+  if (lookup_name.completion_mode ()
+      ? (strncmp (symbol_search_name, name_view.data (),
+		  name_view.size ()) == 0)
+      : symbol_search_name == name_view)
     {
       if (comp_match_res != NULL)
 	comp_match_res->set_match (symbol_search_name);
diff --git a/gdb/dwarf2/read.c b/gdb/dwarf2/read.c
index 1ec5c1e582b..6ee33fcd6ed 100644
--- a/gdb/dwarf2/read.c
+++ b/gdb/dwarf2/read.c
@@ -3681,7 +3681,7 @@ mapped_index_base::find_name_components_bounds
     = this->name_components_casing == case_sensitive_on ? strcmp : strcasecmp;
 
   const char *lang_name
-    = lookup_name_without_params.language_lookup_name (lang).c_str ();
+    = lookup_name_without_params.language_lookup_name (lang);
 
   /* Comparison function object for lower_bound that matches against a
      given symbol name.  */
diff --git a/gdb/language.c b/gdb/language.c
index 454c6dc45a7..c13fd1a406a 100644
--- a/gdb/language.c
+++ b/gdb/language.c
@@ -699,14 +699,14 @@ default_symbol_name_matcher (const char *symbol_search_name,
 			     const lookup_name_info &lookup_name,
 			     completion_match_result *comp_match_res)
 {
-  const std::string &name = lookup_name.name ();
+  gdb::string_view name = lookup_name.name ();
   completion_match_for_lcd *match_for_lcd
     = (comp_match_res != NULL ? &comp_match_res->match_for_lcd : NULL);
   strncmp_iw_mode mode = (lookup_name.completion_mode ()
 			  ? strncmp_iw_mode::NORMAL
 			  : strncmp_iw_mode::MATCH_PARAMS);
 
-  if (strncmp_iw_with_mode (symbol_search_name, name.c_str (), name.size (),
+  if (strncmp_iw_with_mode (symbol_search_name, name.data (), name.size (),
 			    mode, language_minimal, match_for_lcd) == 0)
     {
       if (comp_match_res != NULL)
diff --git a/gdb/minsyms.c b/gdb/minsyms.c
index e238355dc11..d2ac8172eea 100644
--- a/gdb/minsyms.c
+++ b/gdb/minsyms.c
@@ -467,7 +467,7 @@ linkage_name_str (const lookup_name_info &lookup_name)
   if (current_language->la_language == language_ada)
     return lookup_name.ada ().lookup_name ().c_str ();
 
-  return lookup_name.name ().c_str ();
+  return lookup_name.c_str ();
 }
 
 /* See minsyms.h.  */
diff --git a/gdb/psymtab.c b/gdb/psymtab.c
index d5e61f88dc6..d3569ff0130 100644
--- a/gdb/psymtab.c
+++ b/gdb/psymtab.c
@@ -578,8 +578,7 @@ match_partial_symbol (struct objfile *objfile,
 	  gdb_assert (center < top);
 
 	  enum language lang = (*center)->ginfo.language ();
-	  const char *lang_ln
-	    = name.language_lookup_name (lang).c_str ();
+	  const char *lang_ln = name.language_lookup_name (lang);
 
 	  if (ordered_compare ((*center)->ginfo.search_name (),
 			       lang_ln) >= 0)
@@ -658,7 +657,7 @@ lookup_partial_symbol (struct objfile *objfile,
 	    internal_error (__FILE__, __LINE__,
 			    _("failed internal consistency check"));
 	  if (strcmp_iw_ordered ((*center)->ginfo.search_name (),
-				 lookup_name.name ().c_str ()) >= 0)
+				 lookup_name.c_str ()) >= 0)
 	    {
 	      top = center;
 	    }
diff --git a/gdb/symtab.c b/gdb/symtab.c
index f300d759e03..680280105f5 100644
--- a/gdb/symtab.c
+++ b/gdb/symtab.c
@@ -1790,7 +1790,7 @@ demangle_for_lookup_info::demangle_for_lookup_info
   if (lookup_name.ignore_parameters () && lang == language_cplus)
     {
       gdb::unique_xmalloc_ptr<char> without_params
-	= cp_remove_params_if_any (lookup_name.name ().c_str (),
+	= cp_remove_params_if_any (lookup_name.c_str (),
 				   lookup_name.completion_mode ());
 
       if (without_params != NULL)
@@ -1803,9 +1803,9 @@ demangle_for_lookup_info::demangle_for_lookup_info
     }
 
   if (lookup_name.match_type () == symbol_name_match_type::SEARCH_NAME)
-    m_demangled_name = lookup_name.name ();
+    m_demangled_name = lookup_name.c_str ();
   else
-    m_demangled_name = demangle_for_lookup (lookup_name.name ().c_str (),
+    m_demangled_name = demangle_for_lookup (lookup_name.c_str (),
 					    lang, storage);
 }
 
@@ -1816,7 +1816,7 @@ lookup_name_info::match_any ()
 {
   /* Lookup any symbol that "" would complete.  I.e., this matches all
      symbol names.  */
-  static const lookup_name_info lookup_name ({}, symbol_name_match_type::FULL,
+  static const lookup_name_info lookup_name ("", symbol_name_match_type::FULL,
 					     true);
 
   return lookup_name;
diff --git a/gdb/symtab.h b/gdb/symtab.h
index 18be5d51b85..77f60e6c24a 100644
--- a/gdb/symtab.h
+++ b/gdb/symtab.h
@@ -185,29 +185,58 @@ struct demangle_for_lookup_info final
 class lookup_name_info final
 {
  public:
-  /* Create a new object.  */
-  lookup_name_info (std::string name,
+  /* We delete this overload so that the callers are required to
+     explicitly handle the lifetime of the name.  */
+  lookup_name_info (std::string &&name,
+		    symbol_name_match_type match_type,
+		    bool completion_mode = false,
+		    bool ignore_parameters = false) = delete;
+
+  /* This overload requires that NAME have a lifetime at least as long
+     as the lifetime of this object.  */
+  lookup_name_info (const std::string &name,
+		    symbol_name_match_type match_type,
+		    bool completion_mode = false,
+		    bool ignore_parameters = false)
+    : m_match_type (match_type),
+      m_completion_mode (completion_mode),
+      m_ignore_parameters (ignore_parameters),
+      m_name (name)
+  {}
+
+  /* This overload requires that NAME have a lifetime at least as long
+     as the lifetime of this object.  */
+  lookup_name_info (const char *name,
 		    symbol_name_match_type match_type,
 		    bool completion_mode = false,
 		    bool ignore_parameters = false)
     : m_match_type (match_type),
       m_completion_mode (completion_mode),
       m_ignore_parameters (ignore_parameters),
-      m_name (std::move (name))
+      m_name (name)
   {}
 
   /* Getters.  See description of each corresponding field.  */
   symbol_name_match_type match_type () const { return m_match_type; }
   bool completion_mode () const { return m_completion_mode; }
-  const std::string &name () const { return m_name; }
+  gdb::string_view name () const { return m_name; }
   const bool ignore_parameters () const { return m_ignore_parameters; }
 
+  /* Like the "name" method but guarantees that the returned string is
+     \0-terminated.  */
+  const char *c_str () const
+  {
+    /* Actually this is always guaranteed due to how the class is
+       constructed.  */
+    return m_name.data ();
+  }
+
   /* Return a version of this lookup name that is usable with
      comparisons against symbols have no parameter info, such as
      psymbols and GDB index symbols.  */
   lookup_name_info make_ignore_params () const
   {
-    return lookup_name_info (m_name, m_match_type, m_completion_mode,
+    return lookup_name_info (c_str (), m_match_type, m_completion_mode,
 			     true /* ignore params */);
   }
 
@@ -218,27 +247,27 @@ class lookup_name_info final
     if (!m_demangled_hashes_p[lang])
       {
 	m_demangled_hashes[lang]
-	  = ::search_name_hash (lang, language_lookup_name (lang).c_str ());
+	  = ::search_name_hash (lang, language_lookup_name (lang));
 	m_demangled_hashes_p[lang] = true;
       }
     return m_demangled_hashes[lang];
   }
 
   /* Get the search name for searches in language LANG.  */
-  const std::string &language_lookup_name (language lang) const
+  const char *language_lookup_name (language lang) const
   {
     switch (lang)
       {
       case language_ada:
-	return ada ().lookup_name ();
+	return ada ().lookup_name ().c_str ();
       case language_cplus:
-	return cplus ().lookup_name ();
+	return cplus ().lookup_name ().c_str ();
       case language_d:
-	return d ().lookup_name ();
+	return d ().lookup_name ().c_str ();
       case language_go:
-	return go ().lookup_name ();
+	return go ().lookup_name ().c_str ();
       default:
-	return m_name;
+	return m_name.data ();
       }
   }
 
@@ -287,7 +316,7 @@ class lookup_name_info final
   symbol_name_match_type m_match_type;
   bool m_completion_mode;
   bool m_ignore_parameters;
-  std::string m_name;
+  gdb::string_view m_name;
 
   /* Language-specific info.  These fields are filled lazily the first
      time a lookup is done in the corresponding language.  They're
-- 
2.21.1