Introduce helper functions to test parallel recompilation.

src/optimizing-compiler-thread.h

Index: src/optimizing-compiler-thread.h
diff --git a/src/optimizing-compiler-thread.h b/src/optimizing-compiler-thread.h
index d5627266d09898d5bf514512b36cb42261b5152d..8a39b39a47d9002058ef865531b04c660d014a9d 100644
--- a/src/optimizing-compiler-thread.h
+++ b/src/optimizing-compiler-thread.h
@@ -29,8 +29,8 @@
 #define V8_OPTIMIZING_COMPILER_THREAD_H_
 
 #include "atomicops.h"
-#include "platform.h"
 #include "flags.h"
+#include "platform.h"
 #include "unbound-queue.h"
 
 namespace v8 {
@@ -38,6 +38,7 @@ namespace internal {
 
 class HGraphBuilder;
 class OptimizingCompiler;
+class SharedFunctionInfo;
 
 class OptimizingCompilerThread : public Thread {
  public:
@@ -46,6 +47,7 @@ class OptimizingCompilerThread : public Thread {
       isolate_(isolate),
       stop_semaphore_(OS::CreateSemaphore(0)),
       input_queue_semaphore_(OS::CreateSemaphore(0)),
+      output_queue_semaphore_(OS::CreateSemaphore(0)),
       time_spent_compiling_(0),
       time_spent_total_(0) {
     NoBarrier_Store(&stop_thread_, static_cast<AtomicWord>(false));
@@ -57,6 +59,9 @@ class OptimizingCompilerThread : public Thread {
   void QueueForOptimization(OptimizingCompiler* optimizing_compiler);
   void InstallOptimizedFunctions();
 
+  // Wait for the next optimized function and install it.
+  Handle<SharedFunctionInfo> InstallNextOptimizedFunction();
+
   inline bool IsQueueAvailable() {
     // We don't need a barrier since we have a data dependency right
     // after.
@@ -77,6 +82,7 @@ class OptimizingCompilerThread : public Thread {
 
   ~OptimizingCompilerThread() {
     delete input_queue_semaphore_;
+    delete output_queue_semaphore_;  // Only used for manual mode.
     delete stop_semaphore_;
   }
 
@@ -84,6 +90,7 @@ class OptimizingCompilerThread : public Thread {
   Isolate* isolate_;
   Semaphore* stop_semaphore_;
   Semaphore* input_queue_semaphore_;
+  Semaphore* output_queue_semaphore_;
   UnboundQueue<OptimizingCompiler*> input_queue_;
   UnboundQueue<OptimizingCompiler*> output_queue_;
   volatile AtomicWord stop_thread_;