Dramatically reduce time required to complete sync test when running

with lots of threads (by parallelizing shutdown of course)

1 files changed, 9 insertions, 9 deletions

diff --git a/test/cpp/qps/client.h b/test/cpp/qps/client.h
index 4045e13460..8a750196b2 100644
--- a/test/cpp/qps/client.h
+++ b/test/cpp/qps/client.h
@@ -169,6 +169,7 @@ class Client {
   // Must call AwaitThreadsCompletion before destructor to avoid a race
   // between destructor and invocation of virtual ThreadFunc
   void AwaitThreadsCompletion() {
+    gpr_atm_rel_store(&thread_pool_done_, static_cast<gpr_atm>(1));
     DestroyMultithreading();
     std::unique_lock<std::mutex> g(thread_completion_mu_);
     while (threads_remaining_ != 0) {
@@ -180,6 +181,7 @@ class Client {
   bool closed_loop_;
 
   void StartThreads(size_t num_threads) {
+    gpr_atm_rel_store(&thread_pool_done_, static_cast<gpr_atm>(0));
     threads_remaining_ = num_threads;
     for (size_t i = 0; i < num_threads; i++) {
       threads_.emplace_back(new Thread(this, i));
@@ -241,16 +243,11 @@ class Client {
   class Thread {
    public:
     Thread(Client* client, size_t idx)
-        : done_(false),
-          client_(client),
+        : client_(client),
           idx_(idx),
           impl_(&Thread::ThreadFunc, this) {}
 
     ~Thread() {
-      {
-        std::lock_guard<std::mutex> g(mu_);
-        done_ = true;
-      }
       impl_.join();
     }
 
@@ -280,11 +277,14 @@ class Client {
         if (entry.used()) {
           histogram_.Add(entry.value());
         }
+	bool done = false;
         if (!thread_still_ok) {
           gpr_log(GPR_ERROR, "Finishing client thread due to RPC error");
-          done_ = true;
+          done = true;
         }
-        if (done_) {
+	done = done || (gpr_atm_acq_load(&client_->thread_pool_done_) !=
+			static_cast<gpr_atm>(0));
+        if (done) {
           client_->CompleteThread();
           return;
         }
@@ -292,7 +292,6 @@ class Client {
     }
 
     std::mutex mu_;
-    bool done_;
     Histogram histogram_;
     Client* client_;
     const size_t idx_;
@@ -305,6 +304,7 @@ class Client {
   InterarrivalTimer interarrival_timer_;
   std::vector<gpr_timespec> next_time_;
 
+  gpr_atm thread_pool_done_;
   std::mutex thread_completion_mu_;
   size_t threads_remaining_;
   std::condition_variable threads_complete_;