From 8d7a7bd9456c6067e874dabf6ee08b0a4e0de7d0 Mon Sep 17 00:00:00 2001
From: Josh Stone <jistone@redhat.com>
Date: Fri, 28 Aug 2009 18:01:51 -0700
Subject: Cache inline instance lookups

We used to call dwarf_func_inline_instances to get the locations where
inlines are used.  This function has to iterate through nearly all DIEs
to find instances, which is a lot of redundant work when we're probing
multiple inline functions.

Now we have our own dwarf iterator to cache all inline instances back to
their origin.  This only needs to be called once for each CU, and all
further inlines are just a map lookup.  Some quick benchmarks:

  stap -l                           Before      After
  kernel.function("*")             25010ms     2110ms
  module("*").function("*")        86550ms    16920ms
  process("stap").function("*")    41330ms      580ms

* dwflpp.cxx (dwflpp::cu_inl_function_caching_callback): Removed.
  (dwflpp::cache_inline_instances): New caching iterator.
  (dwflpp::iterate_over_inline_instances): Cache each CU once.
---
 dwflpp.cxx | 61 ++++++++++++++++++++++++++++++++++++++++++++++++++-----------
 dwflpp.h   |  3 ++-
 2 files changed, 52 insertions(+), 12 deletions(-)

diff --git a/dwflpp.cxx b/dwflpp.cxx
index 7e9894e1..ba8dcec7 100644
--- a/dwflpp.cxx
+++ b/dwflpp.cxx
@@ -502,12 +502,52 @@ dwflpp::func_is_inline()
 }
 
 
-int
-dwflpp::cu_inl_function_caching_callback (Dwarf_Die* func, void *arg)
+void
+dwflpp::cache_inline_instances (Dwarf_Die* die)
 {
-  vector<Dwarf_Die>* v = static_cast<vector<Dwarf_Die>*>(arg);
-  v->push_back (* func);
-  return DWARF_CB_OK;
+  // If this is an inline instance, link it back to its origin
+  Dwarf_Die origin;
+  if (dwarf_tag(die) == DW_TAG_inlined_subroutine &&
+      dwarf_attr_die(die, DW_AT_abstract_origin, &origin))
+    {
+      vector<Dwarf_Die>*& v = cu_inl_function_cache[origin.addr];
+      if (!v)
+        v = new vector<Dwarf_Die>;
+      v->push_back(*die);
+    }
+
+  // Recurse through other scopes that may contain inlines
+  Dwarf_Die child, import;
+  if (dwarf_child(die, &child) == 0)
+    do
+      {
+        switch (dwarf_tag (&child))
+          {
+          // tags that could contain inlines
+          case DW_TAG_compile_unit:
+          case DW_TAG_module:
+          case DW_TAG_lexical_block:
+          case DW_TAG_with_stmt:
+          case DW_TAG_catch_block:
+          case DW_TAG_try_block:
+          case DW_TAG_entry_point:
+          case DW_TAG_inlined_subroutine:
+          case DW_TAG_subprogram:
+            cache_inline_instances(&child);
+            break;
+
+          // imported dies should be followed
+          case DW_TAG_imported_unit:
+            if (dwarf_attr_die(&child, DW_AT_import, &import))
+              cache_inline_instances(&import);
+            break;
+
+          // nothing to do for other tags
+          default:
+            break;
+          }
+      }
+    while (dwarf_siblingof(&child, &child) == 0);
 }
 
 
@@ -518,13 +558,12 @@ dwflpp::iterate_over_inline_instances (int (* callback)(Dwarf_Die * die, void *
   assert (function);
   assert (func_is_inline ());
 
+  if (cu_inl_function_cache_done.insert(cu->addr).second)
+    cache_inline_instances(cu);
+
   vector<Dwarf_Die>* v = cu_inl_function_cache[function->addr];
-  if (v == 0)
-    {
-      v = new vector<Dwarf_Die>;
-      cu_inl_function_cache[function->addr] = v;
-      dwarf_func_inline_instances (function, cu_inl_function_caching_callback, v);
-    }
+  if (!v)
+    return;
 
   for (vector<Dwarf_Die>::iterator i = v->begin(); i != v->end(); ++i)
     {
diff --git a/dwflpp.h b/dwflpp.h
index 073b9bfc..047ad602 100644
--- a/dwflpp.h
+++ b/dwflpp.h
@@ -296,8 +296,9 @@ private:
 
   module_cu_cache_t module_cu_cache;
 
+  std::set<void*> cu_inl_function_cache_done; // CUs that are already cached
   cu_inl_function_cache_t cu_inl_function_cache;
-  static int cu_inl_function_caching_callback (Dwarf_Die* func, void *arg);
+  void cache_inline_instances (Dwarf_Die* die);
 
   /* The global alias cache is used to resolve any DIE found in a
    * module that is stubbed out with DW_AT_declaration with a defining
-- 
cgit