More efficient stack walk in exception throwing.

In the exception handling code, we currently walk down the stack
twice, once to get the stack height which we use to compute frame IDs
(the bottom frame is zero), and once more to find the catch block to
jump to.

For a deep stack, this could result in very slow exception
handling. That is, if have a lot of finally or catch blocks that we
end up jumping to in a deep stack, we need to do a lot of
catch/rethrow chains. Since we'd need to walk down to the bottom each
time to compute frames IDs in each catch/rethrow, we'd need to walk
down O(N^2) frames at the worst case.

Instead of frames IDs ((the bottom frame is zero), we will use the
frame depth (the top frame is zero) and no longer need to walk down
the stack just to get the stack height. We walk down O(N) frames.

This was what was happening with
code.google.gson.functional.CircularReferenceTest. With this change,
the test run time went from ~120s down to ~3s on N5 and it no longer
crashes due to the thread suspension timeout.

(cherry pick commit 649278cec7119cdd1bea3d0b710dbb2aa7c650b6)

Bug: 16800209
Change-Id: Ie815df1e3e8fb9d82e40685d4cc2b8838fd8aa07

diff --git a/runtime/quick_exception_handler.cc b/runtime/quick_exception_handler.cc
index 41d6989..98eeda7 100644 (file)
--- a/runtime/quick_exception_handler.cc
+++ b/runtime/quick_exception_handler.cc
@@ -29,14 +29,14 @@
 namespace art {
 
 static constexpr bool kDebugExceptionDelivery = false;
-static constexpr size_t kInvalidFrameId = 0xffffffff;
+static constexpr size_t kInvalidFrameDepth = 0xffffffff;
 
 QuickExceptionHandler::QuickExceptionHandler(Thread* self, bool is_deoptimization)
   : self_(self), context_(self->GetLongJumpContext()), is_deoptimization_(is_deoptimization),
     method_tracing_active_(is_deoptimization ||
                            Runtime::Current()->GetInstrumentation()->AreExitStubsInstalled()),
     handler_quick_frame_(nullptr), handler_quick_frame_pc_(0), handler_method_(nullptr),
-    handler_dex_pc_(0), clear_exception_(false), handler_frame_id_(kInvalidFrameId) {
+    handler_dex_pc_(0), clear_exception_(false), handler_frame_depth_(kInvalidFrameDepth) {
 }
 
 // Finds catch handler or prepares for deoptimization.
@@ -51,7 +51,7 @@ class CatchBlockStackVisitor FINAL : public StackVisitor {
 
   bool VisitFrame() OVERRIDE SHARED_LOCKS_REQUIRED(Locks::mutator_lock_) {
     mirror::ArtMethod* method = GetMethod();
-    exception_handler_->SetHandlerFrameId(GetFrameId());
+    exception_handler_->SetHandlerFrameDepth(GetFrameDepth());
     if (method == nullptr) {
       // This is the upcall, we remember the frame and last pc so that we may long jump to them.
       exception_handler_->SetHandlerQuickFramePc(GetCurrentQuickFramePc());
@@ -177,7 +177,7 @@ class DeoptimizeStackVisitor FINAL : public StackVisitor {
   }
 
   bool VisitFrame() OVERRIDE SHARED_LOCKS_REQUIRED(Locks::mutator_lock_) {
-    exception_handler_->SetHandlerFrameId(GetFrameId());
+    exception_handler_->SetHandlerFrameDepth(GetFrameDepth());
     mirror::ArtMethod* method = GetMethod();
     if (method == nullptr) {
       // This is the upcall, we remember the frame and last pc so that we may long jump to them.
@@ -295,17 +295,17 @@ void QuickExceptionHandler::DeoptimizeStack() {
 // Unwinds all instrumentation stack frame prior to catch handler or upcall.
 class InstrumentationStackVisitor : public StackVisitor {
  public:
-  InstrumentationStackVisitor(Thread* self, bool is_deoptimization, size_t frame_id)
+  InstrumentationStackVisitor(Thread* self, bool is_deoptimization, size_t frame_depth)
       SHARED_LOCKS_REQUIRED(Locks::mutator_lock_)
       : StackVisitor(self, nullptr),
-        self_(self), frame_id_(frame_id),
+        self_(self), frame_depth_(frame_depth),
         instrumentation_frames_to_pop_(0) {
-    CHECK_NE(frame_id_, kInvalidFrameId);
+    CHECK_NE(frame_depth_, kInvalidFrameDepth);
   }
 
   bool VisitFrame() SHARED_LOCKS_REQUIRED(Locks::mutator_lock_) {
-    size_t current_frame_id = GetFrameId();
-    if (current_frame_id > frame_id_) {
+    size_t current_frame_depth = GetFrameDepth();
+    if (current_frame_depth < frame_depth_) {
       CHECK(GetMethod() != nullptr);
       if (UNLIKELY(GetQuickInstrumentationExitPc() == GetReturnPc())) {
         ++instrumentation_frames_to_pop_;
@@ -323,7 +323,7 @@ class InstrumentationStackVisitor : public StackVisitor {
 
  private:
   Thread* const self_;
-  const size_t frame_id_;
+  const size_t frame_depth_;
   size_t instrumentation_frames_to_pop_;
 
   DISALLOW_COPY_AND_ASSIGN(InstrumentationStackVisitor);
@@ -331,7 +331,7 @@ class InstrumentationStackVisitor : public StackVisitor {
 
 void QuickExceptionHandler::UpdateInstrumentationStack() {
   if (method_tracing_active_) {
-    InstrumentationStackVisitor visitor(self_, is_deoptimization_, handler_frame_id_);
+    InstrumentationStackVisitor visitor(self_, is_deoptimization_, handler_frame_depth_);
     visitor.WalkStack(true);
 
     size_t instrumentation_frames_to_pop = visitor.GetInstrumentationFramesToPop();

diff --git a/runtime/quick_exception_handler.h b/runtime/quick_exception_handler.h
index 1d600ed..b93769c 100644 (file)
--- a/runtime/quick_exception_handler.h
+++ b/runtime/quick_exception_handler.h
@@ -77,8 +77,8 @@ class QuickExceptionHandler {
     clear_exception_ = clear_exception;
   }
 
-  void SetHandlerFrameId(size_t frame_id) {
-    handler_frame_id_ = frame_id;
+  void SetHandlerFrameDepth(size_t frame_depth) {
+    handler_frame_depth_ = frame_depth;
   }
 
  private:
@@ -97,8 +97,8 @@ class QuickExceptionHandler {
   uint32_t handler_dex_pc_;
   // Should the exception be cleared as the catch block has no move-exception?
   bool clear_exception_;
-  // Frame id of the catch handler or the upcall.
-  size_t handler_frame_id_;
+  // Frame depth of the catch handler or the upcall.
+  size_t handler_frame_depth_;
 
   DISALLOW_COPY_AND_ASSIGN(QuickExceptionHandler);
 };

diff --git a/runtime/stack.h b/runtime/stack.h
index 578f569..e58caee 100644 (file)
--- a/runtime/stack.h
+++ b/runtime/stack.h
@@ -553,6 +553,10 @@ class StackVisitor {
     return num_frames_;
   }
 
+  size_t GetFrameDepth() SHARED_LOCKS_REQUIRED(Locks::mutator_lock_) {
+    return cur_depth_;
+  }
+
   // Get the method and dex pc immediately after the one that's currently being visited.
   bool GetNextMethodAndDexPc(mirror::ArtMethod** next_method, uint32_t* next_dex_pc)
       SHARED_LOCKS_REQUIRED(Locks::mutator_lock_);