diff --git a/src/profiler/allocation-tracker.cc b/src/profiler/allocation-tracker.cc
index 942068e..791cdf0 100644
--- a/src/profiler/allocation-tracker.cc
+++ b/src/profiler/allocation-tracker.cc
@@ -149,7 +149,8 @@
 
 
 void AddressToTraceMap::Print() {
-  PrintF("[AddressToTraceMap (%" V8PRIuPTR "): \n", ranges_.size());
+  PrintF("[AddressToTraceMap (%" V8_SIZET_PREFIX V8PRIuPTR "): \n",
+         ranges_.size());
   for (RangeMap::iterator it = ranges_.begin(); it != ranges_.end(); ++it) {
     PrintF("[%p - %p] => %u\n", it->second.start, it->first,
         it->second.trace_node_id);
@@ -226,7 +227,7 @@
 
   // Mark the new block as FreeSpace to make sure the heap is iterable
   // while we are capturing stack trace.
-  heap->CreateFillerObjectAt(addr, size);
+  heap->CreateFillerObjectAt(addr, size, ClearRecordedSlots::kNo);
 
   Isolate* isolate = heap->isolate();
   int length = 0;
diff --git a/src/profiler/cpu-profiler.cc b/src/profiler/cpu-profiler.cc
index b6c7945..47585b7 100644
--- a/src/profiler/cpu-profiler.cc
+++ b/src/profiler/cpu-profiler.cc
@@ -211,10 +211,8 @@
   processor_->Enqueue(evt_rec);
 }
 
-
 void CpuProfiler::CodeCreateEvent(Logger::LogEventsAndTags tag,
-                                  Code* code,
-                                  const char* name) {
+                                  AbstractCode* code, const char* name) {
   CodeEventsContainer evt_rec(CodeEventRecord::CODE_CREATION);
   CodeCreateEventRecord* rec = &evt_rec.CodeCreateEventRecord_;
   rec->start = code->address();
@@ -222,14 +220,13 @@
       tag, profiles_->GetFunctionName(name), CodeEntry::kEmptyNamePrefix,
       CodeEntry::kEmptyResourceName, CpuProfileNode::kNoLineNumberInfo,
       CpuProfileNode::kNoColumnNumberInfo, NULL, code->instruction_start());
+  RecordInliningInfo(rec->entry, code);
   rec->size = code->ExecutableSize();
   processor_->Enqueue(evt_rec);
 }
 
-
 void CpuProfiler::CodeCreateEvent(Logger::LogEventsAndTags tag,
-                                  Code* code,
-                                  Name* name) {
+                                  AbstractCode* code, Name* name) {
   CodeEventsContainer evt_rec(CodeEventRecord::CODE_CREATION);
   CodeCreateEventRecord* rec = &evt_rec.CodeCreateEventRecord_;
   rec->start = code->address();
@@ -237,12 +234,13 @@
       tag, profiles_->GetFunctionName(name), CodeEntry::kEmptyNamePrefix,
       CodeEntry::kEmptyResourceName, CpuProfileNode::kNoLineNumberInfo,
       CpuProfileNode::kNoColumnNumberInfo, NULL, code->instruction_start());
+  RecordInliningInfo(rec->entry, code);
   rec->size = code->ExecutableSize();
   processor_->Enqueue(evt_rec);
 }
 
-
-void CpuProfiler::CodeCreateEvent(Logger::LogEventsAndTags tag, Code* code,
+void CpuProfiler::CodeCreateEvent(Logger::LogEventsAndTags tag,
+                                  AbstractCode* code,
                                   SharedFunctionInfo* shared,
                                   CompilationInfo* info, Name* script_name) {
   CodeEventsContainer evt_rec(CodeEventRecord::CODE_CREATION);
@@ -250,9 +248,11 @@
   rec->start = code->address();
   rec->entry = profiles_->NewCodeEntry(
       tag, profiles_->GetFunctionName(shared->DebugName()),
-      CodeEntry::kEmptyNamePrefix, profiles_->GetName(script_name),
+      CodeEntry::kEmptyNamePrefix,
+      profiles_->GetName(InferScriptName(script_name, shared)),
       CpuProfileNode::kNoLineNumberInfo, CpuProfileNode::kNoColumnNumberInfo,
       NULL, code->instruction_start());
+  RecordInliningInfo(rec->entry, code);
   if (info) {
     rec->entry->set_inlined_function_infos(info->inlined_function_infos());
   }
@@ -261,46 +261,65 @@
   processor_->Enqueue(evt_rec);
 }
 
-
-void CpuProfiler::CodeCreateEvent(Logger::LogEventsAndTags tag, Code* code,
+void CpuProfiler::CodeCreateEvent(Logger::LogEventsAndTags tag,
+                                  AbstractCode* abstract_code,
                                   SharedFunctionInfo* shared,
                                   CompilationInfo* info, Name* script_name,
                                   int line, int column) {
   CodeEventsContainer evt_rec(CodeEventRecord::CODE_CREATION);
   CodeCreateEventRecord* rec = &evt_rec.CodeCreateEventRecord_;
-  rec->start = code->address();
+  rec->start = abstract_code->address();
   Script* script = Script::cast(shared->script());
   JITLineInfoTable* line_table = NULL;
   if (script) {
-    line_table = new JITLineInfoTable();
-    for (RelocIterator it(code); !it.done(); it.next()) {
-      RelocInfo::Mode mode = it.rinfo()->rmode();
-      if (RelocInfo::IsPosition(mode)) {
-        int position = static_cast<int>(it.rinfo()->data());
-        if (position >= 0) {
-          int pc_offset = static_cast<int>(it.rinfo()->pc() - code->address());
-          int line_number = script->GetLineNumber(position) + 1;
-          line_table->SetPosition(pc_offset, line_number);
-        }
+    if (abstract_code->IsCode()) {
+      Code* code = abstract_code->GetCode();
+      int start_position = shared->start_position();
+      int end_position = shared->end_position();
+      line_table = new JITLineInfoTable();
+      for (RelocIterator it(code); !it.done(); it.next()) {
+        RelocInfo* reloc_info = it.rinfo();
+        if (!RelocInfo::IsPosition(reloc_info->rmode())) continue;
+        int position = static_cast<int>(reloc_info->data());
+        // TODO(alph): in case of inlining the position may correspond
+        // to an inlined function source code. Do not collect positions
+        // that fall beyond the function source code. There's however a
+        // chance the inlined function has similar positions but in another
+        // script. So the proper fix is to store script_id in some form
+        // along with the inlined function positions.
+        if (position < start_position || position >= end_position) continue;
+        int pc_offset = static_cast<int>(reloc_info->pc() - code->address());
+        int line_number = script->GetLineNumber(position) + 1;
+        line_table->SetPosition(pc_offset, line_number);
+      }
+    } else {
+      BytecodeArray* bytecode = abstract_code->GetBytecodeArray();
+      line_table = new JITLineInfoTable();
+      interpreter::SourcePositionTableIterator it(
+          bytecode->source_position_table());
+      for (; !it.done(); it.Advance()) {
+        int line_number = script->GetLineNumber(it.source_position()) + 1;
+        int pc_offset = it.bytecode_offset() + BytecodeArray::kHeaderSize;
+        line_table->SetPosition(pc_offset, line_number);
       }
     }
   }
   rec->entry = profiles_->NewCodeEntry(
       tag, profiles_->GetFunctionName(shared->DebugName()),
-      CodeEntry::kEmptyNamePrefix, profiles_->GetName(script_name), line,
-      column, line_table, code->instruction_start());
+      CodeEntry::kEmptyNamePrefix,
+      profiles_->GetName(InferScriptName(script_name, shared)), line, column,
+      line_table, abstract_code->instruction_start());
+  RecordInliningInfo(rec->entry, abstract_code);
   if (info) {
     rec->entry->set_inlined_function_infos(info->inlined_function_infos());
   }
   rec->entry->FillFunctionInfo(shared);
-  rec->size = code->ExecutableSize();
+  rec->size = abstract_code->ExecutableSize();
   processor_->Enqueue(evt_rec);
 }
 
-
 void CpuProfiler::CodeCreateEvent(Logger::LogEventsAndTags tag,
-                                  Code* code,
-                                  int args_count) {
+                                  AbstractCode* code, int args_count) {
   CodeEventsContainer evt_rec(CodeEventRecord::CODE_CREATION);
   CodeCreateEventRecord* rec = &evt_rec.CodeCreateEventRecord_;
   rec->start = code->address();
@@ -308,21 +327,21 @@
       tag, profiles_->GetName(args_count), "args_count: ",
       CodeEntry::kEmptyResourceName, CpuProfileNode::kNoLineNumberInfo,
       CpuProfileNode::kNoColumnNumberInfo, NULL, code->instruction_start());
+  RecordInliningInfo(rec->entry, code);
   rec->size = code->ExecutableSize();
   processor_->Enqueue(evt_rec);
 }
 
-
-void CpuProfiler::CodeMoveEvent(Address from, Address to) {
+void CpuProfiler::CodeMoveEvent(AbstractCode* from, Address to) {
   CodeEventsContainer evt_rec(CodeEventRecord::CODE_MOVE);
   CodeMoveEventRecord* rec = &evt_rec.CodeMoveEventRecord_;
-  rec->from = from;
+  rec->from = from->address();
   rec->to = to;
   processor_->Enqueue(evt_rec);
 }
 
-
-void CpuProfiler::CodeDisableOptEvent(Code* code, SharedFunctionInfo* shared) {
+void CpuProfiler::CodeDisableOptEvent(AbstractCode* code,
+                                      SharedFunctionInfo* shared) {
   CodeEventsContainer evt_rec(CodeEventRecord::CODE_DISABLE_OPT);
   CodeDisableOptEventRecord* rec = &evt_rec.CodeDisableOptEventRecord_;
   rec->start = code->address();
@@ -330,7 +349,6 @@
   processor_->Enqueue(evt_rec);
 }
 
-
 void CpuProfiler::CodeDeoptEvent(Code* code, Address pc, int fp_to_sp_delta) {
   CodeEventsContainer evt_rec(CodeEventRecord::CODE_DEOPT);
   CodeDeoptEventRecord* rec = &evt_rec.CodeDeoptEventRecord_;
@@ -343,11 +361,6 @@
   processor_->AddDeoptStack(isolate_, pc, fp_to_sp_delta);
 }
 
-
-void CpuProfiler::CodeDeleteEvent(Address from) {
-}
-
-
 void CpuProfiler::GetterCallbackEvent(Name* name, Address entry_point) {
   CodeEventsContainer evt_rec(CodeEventRecord::CODE_CREATION);
   CodeCreateEventRecord* rec = &evt_rec.CodeCreateEventRecord_;
@@ -360,8 +373,7 @@
   processor_->Enqueue(evt_rec);
 }
 
-
-void CpuProfiler::RegExpCodeCreateEvent(Code* code, String* source) {
+void CpuProfiler::RegExpCodeCreateEvent(AbstractCode* code, String* source) {
   CodeEventsContainer evt_rec(CodeEventRecord::CODE_CREATION);
   CodeCreateEventRecord* rec = &evt_rec.CodeCreateEventRecord_;
   rec->start = code->address();
@@ -386,6 +398,60 @@
   processor_->Enqueue(evt_rec);
 }
 
+Name* CpuProfiler::InferScriptName(Name* name, SharedFunctionInfo* info) {
+  if (name->IsString() && String::cast(name)->length()) return name;
+  if (!info->script()->IsScript()) return name;
+  Object* source_url = Script::cast(info->script())->source_url();
+  return source_url->IsName() ? Name::cast(source_url) : name;
+}
+
+void CpuProfiler::RecordInliningInfo(CodeEntry* entry,
+                                     AbstractCode* abstract_code) {
+  if (!abstract_code->IsCode()) return;
+  Code* code = abstract_code->GetCode();
+  if (code->kind() != Code::OPTIMIZED_FUNCTION) return;
+  DeoptimizationInputData* deopt_input_data =
+      DeoptimizationInputData::cast(code->deoptimization_data());
+  int deopt_count = deopt_input_data->DeoptCount();
+  for (int i = 0; i < deopt_count; i++) {
+    int pc_offset = deopt_input_data->Pc(i)->value();
+    if (pc_offset == -1) continue;
+    int translation_index = deopt_input_data->TranslationIndex(i)->value();
+    TranslationIterator it(deopt_input_data->TranslationByteArray(),
+                           translation_index);
+    Translation::Opcode opcode = static_cast<Translation::Opcode>(it.Next());
+    DCHECK_EQ(Translation::BEGIN, opcode);
+    it.Skip(Translation::NumberOfOperandsFor(opcode));
+    int depth = 0;
+    std::vector<CodeEntry*> inline_stack;
+    while (it.HasNext() &&
+           Translation::BEGIN !=
+               (opcode = static_cast<Translation::Opcode>(it.Next()))) {
+      if (opcode != Translation::JS_FRAME &&
+          opcode != Translation::INTERPRETED_FRAME) {
+        it.Skip(Translation::NumberOfOperandsFor(opcode));
+        continue;
+      }
+      it.Next();  // Skip ast_id
+      int shared_info_id = it.Next();
+      it.Next();  // Skip height
+      SharedFunctionInfo* shared_info = SharedFunctionInfo::cast(
+          deopt_input_data->LiteralArray()->get(shared_info_id));
+      if (!depth++) continue;  // Skip the current function itself.
+      CodeEntry* inline_entry = new CodeEntry(
+          entry->tag(), profiles_->GetFunctionName(shared_info->DebugName()),
+          CodeEntry::kEmptyNamePrefix, entry->resource_name(),
+          CpuProfileNode::kNoLineNumberInfo,
+          CpuProfileNode::kNoColumnNumberInfo, NULL, code->instruction_start());
+      inline_entry->FillFunctionInfo(shared_info);
+      inline_stack.push_back(inline_entry);
+    }
+    if (!inline_stack.empty()) {
+      entry->AddInlineStack(pc_offset, inline_stack);
+      DCHECK(inline_stack.empty());
+    }
+  }
+}
 
 CpuProfiler::CpuProfiler(Isolate* isolate)
     : isolate_(isolate),
diff --git a/src/profiler/cpu-profiler.h b/src/profiler/cpu-profiler.h
index 1a1249c..a04ee3c 100644
--- a/src/profiler/cpu-profiler.h
+++ b/src/profiler/cpu-profiler.h
@@ -201,7 +201,7 @@
               ProfileGenerator* test_generator,
               ProfilerEventsProcessor* test_processor);
 
-  virtual ~CpuProfiler();
+  ~CpuProfiler() override;
 
   void set_sampling_interval(base::TimeDelta value);
   void CollectSample();
@@ -220,29 +220,28 @@
 
   // Must be called via PROFILE macro, otherwise will crash when
   // profiling is not enabled.
-  virtual void CallbackEvent(Name* name, Address entry_point);
-  virtual void CodeCreateEvent(Logger::LogEventsAndTags tag,
-                               Code* code, const char* comment);
-  virtual void CodeCreateEvent(Logger::LogEventsAndTags tag,
-                               Code* code, Name* name);
-  virtual void CodeCreateEvent(Logger::LogEventsAndTags tag, Code* code,
-                               SharedFunctionInfo* shared,
-                               CompilationInfo* info, Name* script_name);
-  virtual void CodeCreateEvent(Logger::LogEventsAndTags tag, Code* code,
-                               SharedFunctionInfo* shared,
-                               CompilationInfo* info, Name* script_name,
-                               int line, int column);
-  virtual void CodeCreateEvent(Logger::LogEventsAndTags tag,
-                               Code* code, int args_count);
-  virtual void CodeMovingGCEvent() {}
-  virtual void CodeMoveEvent(Address from, Address to);
-  virtual void CodeDisableOptEvent(Code* code, SharedFunctionInfo* shared);
-  virtual void CodeDeoptEvent(Code* code, Address pc, int fp_to_sp_delta);
-  virtual void CodeDeleteEvent(Address from);
-  virtual void GetterCallbackEvent(Name* name, Address entry_point);
-  virtual void RegExpCodeCreateEvent(Code* code, String* source);
-  virtual void SetterCallbackEvent(Name* name, Address entry_point);
-  virtual void SharedFunctionInfoMoveEvent(Address from, Address to) {}
+  void CallbackEvent(Name* name, Address entry_point) override;
+  void CodeCreateEvent(Logger::LogEventsAndTags tag, AbstractCode* code,
+                       const char* comment) override;
+  void CodeCreateEvent(Logger::LogEventsAndTags tag, AbstractCode* code,
+                       Name* name) override;
+  void CodeCreateEvent(Logger::LogEventsAndTags tag, AbstractCode* code,
+                       SharedFunctionInfo* shared, CompilationInfo* info,
+                       Name* script_name) override;
+  void CodeCreateEvent(Logger::LogEventsAndTags tag, AbstractCode* code,
+                       SharedFunctionInfo* shared, CompilationInfo* info,
+                       Name* script_name, int line, int column) override;
+  void CodeCreateEvent(Logger::LogEventsAndTags tag, AbstractCode* code,
+                       int args_count) override;
+  void CodeMovingGCEvent() override {}
+  void CodeMoveEvent(AbstractCode* from, Address to) override;
+  void CodeDisableOptEvent(AbstractCode* code,
+                           SharedFunctionInfo* shared) override;
+  void CodeDeoptEvent(Code* code, Address pc, int fp_to_sp_delta);
+  void GetterCallbackEvent(Name* name, Address entry_point) override;
+  void RegExpCodeCreateEvent(AbstractCode* code, String* source) override;
+  void SetterCallbackEvent(Name* name, Address entry_point) override;
+  void SharedFunctionInfoMoveEvent(Address from, Address to) override {}
 
   INLINE(bool is_profiling() const) { return is_profiling_; }
   bool* is_profiling_address() {
@@ -259,6 +258,8 @@
   void StopProcessor();
   void ResetProfiles();
   void LogBuiltins();
+  void RecordInliningInfo(CodeEntry* entry, AbstractCode* abstract_code);
+  Name* InferScriptName(Name* name, SharedFunctionInfo* info);
 
   Isolate* isolate_;
   base::TimeDelta sampling_interval_;
diff --git a/src/profiler/heap-snapshot-generator.cc b/src/profiler/heap-snapshot-generator.cc
index fc43f9f..748f307 100644
--- a/src/profiler/heap-snapshot-generator.cc
+++ b/src/profiler/heap-snapshot-generator.cc
@@ -1383,9 +1383,9 @@
   SetInternalReference(obj, entry,
                        "debug_info", shared->debug_info(),
                        SharedFunctionInfo::kDebugInfoOffset);
-  SetInternalReference(obj, entry,
-                       "inferred_name", shared->inferred_name(),
-                       SharedFunctionInfo::kInferredNameOffset);
+  SetInternalReference(obj, entry, "function_identifier",
+                       shared->function_identifier(),
+                       SharedFunctionInfo::kFunctionIdentifierOffset);
   SetInternalReference(obj, entry,
                        "optimized_code_map", shared->optimized_code_map(),
                        SharedFunctionInfo::kOptimizedCodeMapOffset);
diff --git a/src/profiler/profile-generator.cc b/src/profiler/profile-generator.cc
index 58d06c9..abcd9e5 100644
--- a/src/profiler/profile-generator.cc
+++ b/src/profiler/profile-generator.cc
@@ -50,6 +50,11 @@
 
 CodeEntry::~CodeEntry() {
   delete line_info_;
+  for (auto location : inline_locations_) {
+    for (auto entry : location.second) {
+      delete entry;
+    }
+  }
 }
 
 
@@ -100,6 +105,18 @@
   return v8::CpuProfileNode::kNoLineNumberInfo;
 }
 
+void CodeEntry::AddInlineStack(int pc_offset,
+                               std::vector<CodeEntry*>& inline_stack) {
+  // It's better to use std::move to place the vector into the map,
+  // but it's not supported by the current stdlibc++ on MacOS.
+  inline_locations_.insert(std::make_pair(pc_offset, std::vector<CodeEntry*>()))
+      .first->second.swap(inline_stack);
+}
+
+const std::vector<CodeEntry*>* CodeEntry::GetInlineStack(int pc_offset) const {
+  auto it = inline_locations_.find(pc_offset);
+  return it != inline_locations_.end() ? &it->second : NULL;
+}
 
 void CodeEntry::FillFunctionInfo(SharedFunctionInfo* shared) {
   if (!shared->script()->IsScript()) return;
@@ -109,7 +126,6 @@
   set_bailout_reason(GetBailoutReason(shared->disable_optimization_reason()));
 }
 
-
 CpuProfileDeoptInfo CodeEntry::GetDeoptInfo() {
   DCHECK(has_deopt_info());
 
@@ -274,17 +290,14 @@
   return static_cast<unsigned>(reinterpret_cast<uintptr_t>(entry->value));
 }
 
-ProfileNode* ProfileTree::AddPathFromEnd(const Vector<CodeEntry*>& path,
+ProfileNode* ProfileTree::AddPathFromEnd(const std::vector<CodeEntry*>& path,
                                          int src_line, bool update_stats) {
   ProfileNode* node = root_;
   CodeEntry* last_entry = NULL;
-  for (CodeEntry** entry = path.start() + path.length() - 1;
-       entry != path.start() - 1;
-       --entry) {
-    if (*entry != NULL) {
-      node = node->FindOrAddChild(*entry);
-      last_entry = *entry;
-    }
+  for (auto it = path.rbegin(); it != path.rend(); ++it) {
+    if (*it == NULL) continue;
+    last_entry = *it;
+    node = node->FindOrAddChild(*it);
   }
   if (last_entry && last_entry->has_deopt_info()) {
     node->CollectDeoptInfo(last_entry);
@@ -356,7 +369,7 @@
       top_down_(isolate) {}
 
 void CpuProfile::AddPath(base::TimeTicks timestamp,
-                         const Vector<CodeEntry*>& path, int src_line,
+                         const std::vector<CodeEntry*>& path, int src_line,
                          bool update_stats) {
   ProfileNode* top_frame_node =
       top_down_.AddPathFromEnd(path, src_line, update_stats);
@@ -525,8 +538,8 @@
 }
 
 void CpuProfilesCollection::AddPathToCurrentProfiles(
-    base::TimeTicks timestamp, const Vector<CodeEntry*>& path, int src_line,
-    bool update_stats) {
+    base::TimeTicks timestamp, const std::vector<CodeEntry*>& path,
+    int src_line, bool update_stats) {
   // As starting / stopping profiles is rare relatively to this
   // method, we don't bother minimizing the duration of lock holding,
   // e.g. copying contents of the list to a local vector.
@@ -576,12 +589,10 @@
 
 
 void ProfileGenerator::RecordTickSample(const TickSample& sample) {
-  // Allocate space for stack frames + pc + function + vm-state.
-  ScopedVector<CodeEntry*> entries(sample.frames_count + 3);
-  // As actual number of decoded code entries may vary, initialize
-  // entries vector with NULL values.
-  CodeEntry** entry = entries.start();
-  memset(entry, 0, entries.length() * sizeof(*entry));
+  std::vector<CodeEntry*> entries;
+  // Conservatively reserve space for stack frames + pc + function + vm-state.
+  // There could in fact be more of them because of inlined entries.
+  entries.reserve(sample.frames_count + 3);
 
   // The ProfileNode knows nothing about all versions of generated code for
   // the same JS function. The line number information associated with
@@ -597,13 +608,14 @@
       // Don't use PC when in external callback code, as it can point
       // inside callback's code, and we will erroneously report
       // that a callback calls itself.
-      *entry++ = code_map_.FindEntry(sample.external_callback_entry);
+      entries.push_back(code_map_.FindEntry(sample.external_callback_entry));
     } else {
       CodeEntry* pc_entry = code_map_.FindEntry(sample.pc);
       // If there is no pc_entry we're likely in native code.
       // Find out, if top of stack was pointing inside a JS function
       // meaning that we have encountered a frameless invocation.
       if (!pc_entry && (sample.top_frame_type == StackFrame::JAVA_SCRIPT ||
+                        sample.top_frame_type == StackFrame::INTERPRETED ||
                         sample.top_frame_type == StackFrame::OPTIMIZED)) {
         pc_entry = code_map_.FindEntry(sample.tos);
       }
@@ -619,7 +631,7 @@
           src_line = pc_entry->line_number();
         }
         src_line_not_found = false;
-        *entry++ = pc_entry;
+        entries.push_back(pc_entry);
 
         if (pc_entry->builtin_id() == Builtins::kFunctionPrototypeApply ||
             pc_entry->builtin_id() == Builtins::kFunctionPrototypeCall) {
@@ -630,7 +642,7 @@
           // former case we don't so we simply replace the frame with
           // 'unresolved' entry.
           if (sample.top_frame_type == StackFrame::JAVA_SCRIPT) {
-            *entry++ = unresolved_entry_;
+            entries.push_back(unresolved_entry_);
           }
         }
       }
@@ -639,35 +651,43 @@
     for (const Address *stack_pos = sample.stack,
                        *stack_end = stack_pos + sample.frames_count;
          stack_pos != stack_end; ++stack_pos) {
-      *entry = code_map_.FindEntry(*stack_pos);
+      CodeEntry* entry = code_map_.FindEntry(*stack_pos);
 
-      // Skip unresolved frames (e.g. internal frame) and get source line of
-      // the first JS caller.
-      if (src_line_not_found && *entry) {
+      if (entry) {
+        // Find out if the entry has an inlining stack associated.
         int pc_offset =
-            static_cast<int>(*stack_pos - (*entry)->instruction_start());
-        src_line = (*entry)->GetSourceLine(pc_offset);
-        if (src_line == v8::CpuProfileNode::kNoLineNumberInfo) {
-          src_line = (*entry)->line_number();
+            static_cast<int>(*stack_pos - entry->instruction_start());
+        const std::vector<CodeEntry*>* inline_stack =
+            entry->GetInlineStack(pc_offset);
+        if (inline_stack) {
+          entries.insert(entries.end(), inline_stack->rbegin(),
+                         inline_stack->rend());
         }
-        src_line_not_found = false;
+        // Skip unresolved frames (e.g. internal frame) and get source line of
+        // the first JS caller.
+        if (src_line_not_found) {
+          src_line = entry->GetSourceLine(pc_offset);
+          if (src_line == v8::CpuProfileNode::kNoLineNumberInfo) {
+            src_line = entry->line_number();
+          }
+          src_line_not_found = false;
+        }
       }
-
-      entry++;
+      entries.push_back(entry);
     }
   }
 
   if (FLAG_prof_browser_mode) {
     bool no_symbolized_entries = true;
-    for (CodeEntry** e = entries.start(); e != entry; ++e) {
-      if (*e != NULL) {
+    for (auto e : entries) {
+      if (e != NULL) {
         no_symbolized_entries = false;
         break;
       }
     }
     // If no frames were symbolized, put the VM state entry in.
     if (no_symbolized_entries) {
-      *entry++ = EntryForVMState(sample.state);
+      entries.push_back(EntryForVMState(sample.state));
     }
   }
 
diff --git a/src/profiler/profile-generator.h b/src/profiler/profile-generator.h
index 3c976d6..194b490 100644
--- a/src/profiler/profile-generator.h
+++ b/src/profiler/profile-generator.h
@@ -99,7 +99,11 @@
 
   int GetSourceLine(int pc_offset) const;
 
+  void AddInlineStack(int pc_offset, std::vector<CodeEntry*>& inline_stack);
+  const std::vector<CodeEntry*>* GetInlineStack(int pc_offset) const;
+
   Address instruction_start() const { return instruction_start_; }
+  Logger::LogEventsAndTags tag() const { return TagField::decode(bit_field_); }
 
   static const char* const kEmptyNamePrefix;
   static const char* const kEmptyResourceName;
@@ -109,7 +113,6 @@
  private:
   class TagField : public BitField<Logger::LogEventsAndTags, 0, 8> {};
   class BuiltinIdField : public BitField<Builtins::Name, 8, 8> {};
-  Logger::LogEventsAndTags tag() const { return TagField::decode(bit_field_); }
 
   uint32_t bit_field_;
   const char* name_prefix_;
@@ -125,6 +128,8 @@
   size_t pc_offset_;
   JITLineInfoTable* line_info_;
   Address instruction_start_;
+  // Should be an unordered_map, but it doesn't currently work on Win & MacOS.
+  std::map<int, std::vector<CodeEntry*>> inline_locations_;
 
   std::vector<InlinedFunctionInfo> inlined_function_infos_;
 
@@ -191,7 +196,7 @@
   ~ProfileTree();
 
   ProfileNode* AddPathFromEnd(
-      const Vector<CodeEntry*>& path,
+      const std::vector<CodeEntry*>& path,
       int src_line = v8::CpuProfileNode::kNoLineNumberInfo,
       bool update_stats = true);
   ProfileNode* root() const { return root_; }
@@ -225,7 +230,7 @@
   CpuProfile(Isolate* isolate, const char* title, bool record_samples);
 
   // Add pc -> ... -> main() call path to the profile.
-  void AddPath(base::TimeTicks timestamp, const Vector<CodeEntry*>& path,
+  void AddPath(base::TimeTicks timestamp, const std::vector<CodeEntry*>& path,
                int src_line, bool update_stats);
   void CalculateTotalTicksAndSamplingRate();
 
@@ -334,8 +339,8 @@
 
   // Called from profile generator thread.
   void AddPathToCurrentProfiles(base::TimeTicks timestamp,
-                                const Vector<CodeEntry*>& path, int src_line,
-                                bool update_stats);
+                                const std::vector<CodeEntry*>& path,
+                                int src_line, bool update_stats);
 
   // Limits the number of profiles that can be simultaneously collected.
   static const int kMaxSimultaneousProfiles = 100;
diff --git a/src/profiler/sampler.cc b/src/profiler/sampler.cc
index e331db9..a340424 100644
--- a/src/profiler/sampler.cc
+++ b/src/profiler/sampler.cc
@@ -336,6 +336,14 @@
         reinterpret_cast<Address>(simulator_->get_register(Simulator::sp));
     state->fp =
         reinterpret_cast<Address>(simulator_->get_register(Simulator::fp));
+#elif V8_TARGET_ARCH_S390
+    if (!simulator_->has_bad_pc()) {
+      state->pc = reinterpret_cast<Address>(simulator_->get_pc());
+    }
+    state->sp =
+        reinterpret_cast<Address>(simulator_->get_register(Simulator::sp));
+    state->fp =
+        reinterpret_cast<Address>(simulator_->get_register(Simulator::fp));
 #endif
   }
 
@@ -441,7 +449,7 @@
 #else
   // Extracting the sample from the context is extremely machine dependent.
   ucontext_t* ucontext = reinterpret_cast<ucontext_t*>(context);
-#if !(V8_OS_OPENBSD || (V8_OS_LINUX && V8_HOST_ARCH_PPC))
+#if !(V8_OS_OPENBSD || (V8_OS_LINUX && (V8_HOST_ARCH_PPC || V8_HOST_ARCH_S390)))
   mcontext_t& mcontext = ucontext->uc_mcontext;
 #endif
 #if V8_OS_LINUX
@@ -482,6 +490,17 @@
   state.pc = reinterpret_cast<Address>(ucontext->uc_mcontext.regs->nip);
   state.sp = reinterpret_cast<Address>(ucontext->uc_mcontext.regs->gpr[PT_R1]);
   state.fp = reinterpret_cast<Address>(ucontext->uc_mcontext.regs->gpr[PT_R31]);
+#elif V8_HOST_ARCH_S390
+#if V8_TARGET_ARCH_32_BIT
+  // 31-bit target will have bit 0 (MSB) of the PSW set to denote addressing
+  // mode.  This bit needs to be masked out to resolve actual address.
+  state.pc =
+      reinterpret_cast<Address>(ucontext->uc_mcontext.psw.addr & 0x7FFFFFFF);
+#else
+  state.pc = reinterpret_cast<Address>(ucontext->uc_mcontext.psw.addr);
+#endif  // V8_TARGET_ARCH_32_BIT
+  state.sp = reinterpret_cast<Address>(ucontext->uc_mcontext.gregs[15]);
+  state.fp = reinterpret_cast<Address>(ucontext->uc_mcontext.gregs[11]);
 #endif  // V8_HOST_ARCH_*
 #elif V8_OS_MACOSX
 #if V8_HOST_ARCH_X64
@@ -731,7 +750,18 @@
     frames[i++] = isolate->c_function();
   }
   while (!it.done() && i < frames_limit) {
-    frames[i++] = it.frame()->pc();
+    if (it.frame()->is_interpreted()) {
+      // For interpreted frames use the bytecode array pointer as the pc.
+      InterpretedFrame* frame = static_cast<InterpretedFrame*>(it.frame());
+      // Since the sampler can interrupt execution at any point the
+      // bytecode_array might be garbage, so don't dereference it.
+      Address bytecode_array =
+          reinterpret_cast<Address>(frame->GetBytecodeArray()) - kHeapObjectTag;
+      frames[i++] = bytecode_array + BytecodeArray::kHeaderSize +
+                    frame->GetBytecodeOffset();
+    } else {
+      frames[i++] = it.frame()->pc();
+    }
     it.Advance();
   }
   sample_info->frames_count = i;
diff --git a/src/profiler/sampler.h b/src/profiler/sampler.h
index 8e8ef1c..dcd1255 100644
--- a/src/profiler/sampler.h
+++ b/src/profiler/sampler.h
@@ -58,7 +58,7 @@
   unsigned frames_count : kMaxFramesCountLog2;  // Number of captured frames.
   bool has_external_callback : 1;
   bool update_stats : 1;  // Whether the sample should update aggregated stats.
-  StackFrame::Type top_frame_type : 4;
+  StackFrame::Type top_frame_type : 5;
 };
 
 class Sampler {
diff --git a/src/profiler/sampling-heap-profiler.cc b/src/profiler/sampling-heap-profiler.cc
index c13538c..a32cae3 100644
--- a/src/profiler/sampling-heap-profiler.cc
+++ b/src/profiler/sampling-heap-profiler.cc
@@ -99,7 +99,8 @@
 
   // Mark the new block as FreeSpace to make sure the heap is iterable while we
   // are taking the sample.
-  heap()->CreateFillerObjectAt(soon_object, static_cast<int>(size));
+  heap()->CreateFillerObjectAt(soon_object, static_cast<int>(size),
+                               ClearRecordedSlots::kNo);
 
   Local<v8::Value> loc = v8::Utils::ToLocal(obj);
 
@@ -199,19 +200,22 @@
   int column = v8::AllocationProfile::kNoColumnNumberInfo;
   std::vector<v8::AllocationProfile::Allocation> allocations;
   allocations.reserve(node->allocations_.size());
-  if (node->script_id_ != v8::UnboundScript::kNoScriptId) {
+  if (node->script_id_ != v8::UnboundScript::kNoScriptId &&
+      scripts.find(node->script_id_) != scripts.end()) {
     // Cannot use std::map<T>::at because it is not available on android.
     auto non_const_scripts = const_cast<std::map<int, Script*>&>(scripts);
     Script* script = non_const_scripts[node->script_id_];
-    if (script->name()->IsName()) {
-      Name* name = Name::cast(script->name());
-      script_name = ToApiHandle<v8::String>(
-          isolate_->factory()->InternalizeUtf8String(names_->GetName(name)));
+    if (script) {
+      if (script->name()->IsName()) {
+        Name* name = Name::cast(script->name());
+        script_name = ToApiHandle<v8::String>(
+            isolate_->factory()->InternalizeUtf8String(names_->GetName(name)));
+      }
+      Handle<Script> script_handle(script);
+      line = 1 + Script::GetLineNumber(script_handle, node->script_position_);
+      column =
+          1 + Script::GetColumnNumber(script_handle, node->script_position_);
     }
-    Handle<Script> script_handle(script);
-
-    line = 1 + Script::GetLineNumber(script_handle, node->script_position_);
-    column = 1 + Script::GetColumnNumber(script_handle, node->script_position_);
     for (auto alloc : node->allocations_) {
       allocations.push_back(ScaleSample(alloc.first, alloc.second));
     }