*** empty log message ***

14 files changed, 472 insertions, 366 deletions

diff --git a/TAO/performance-tests/Cubit/TAO/MT_Cubit/README b/TAO/performance-tests/Cubit/TAO/MT_Cubit/README
index 8a545ac722c..e0b79a019f3 100644
--- a/TAO/performance-tests/Cubit/TAO/MT_Cubit/README
+++ b/TAO/performance-tests/Cubit/TAO/MT_Cubit/README
@@ -69,6 +69,10 @@ Options are case sensitive, e.g., "-U" has a different meaning than "-u".
 	Default Value: Disabled. To enable, specify "-U" on the
                        command line.
 
+[-r] turn on the thread-per-rate tests.  (Priority inversion tests)
+
+	Default Value: Disabled.
+
 The server will print the IOR of the servants, but it is not required
 by the client.
 
@@ -84,10 +88,10 @@ by the client.
          [-r]			// use thread-per-rate test.
          [-o]			// makes client use oneway calls instead
          [-x]			// makes a call to servant to shutdown
-         [-u]			// run the client utilization test
-         [-U]			// run the servant utilization test
+         [-u <seconds> ]	// run the client utilization test for <seconds> seconds
+         [-U <seconds> ]	// run the servant utilization test for <seconds> seconds
          [-1]			// use one to n test.
-         [-g]			// choose the granularity of timing calls
+         [-g <granularity>]	// choose the granularity of timing calls
          [-c]			// obtain number of context switches.
          [-s]			// makes client *NOT* use the name service
          [-m]			// use multiple priorities for the low priority client.
@@ -116,7 +120,7 @@ by the client.
 
 	Default Value: nil, which means not to write the iors to disk.
 
-[-r] turn on the thread-per-rate tests.
+[-r] turn on the thread-per-rate tests.  (Priority inversion tests)
 
 	Default Value: Disabled.
 
diff --git a/TAO/performance-tests/Cubit/TAO/MT_Cubit/Task_Client.cpp b/TAO/performance-tests/Cubit/TAO/MT_Cubit/Task_Client.cpp
index d4c95c2c70d..29cb263d76d 100644
--- a/TAO/performance-tests/Cubit/TAO/MT_Cubit/Task_Client.cpp
+++ b/TAO/performance-tests/Cubit/TAO/MT_Cubit/Task_Client.cpp
@@ -30,7 +30,7 @@ Task_State::Task_State (int argc, char **argv)
     run_server_utilization_test_ (0),
     util_time_ (0)
 {
-  ACE_Get_Opt opts (argc, argv, "Umu:sn:t:d:rxof:g:1c");
+  ACE_Get_Opt opts (argc, argv, "U:mu:sn:t:d:rxof:g:1c");
   int c;
   int datatype;
 
@@ -43,6 +43,7 @@ Task_State::Task_State (int argc, char **argv)
       break;
     case 'U':
       run_server_utilization_test_ = 1;
+      util_time_ = ACE_OS::atoi (opts.optarg);
       break;
     case 'm':
       use_multiple_priority_ = 1;
@@ -648,6 +649,7 @@ Client::run_tests (Cubit_ptr cb,
   double sleep_time = (1 / frequency) * ACE_ONE_SECOND_IN_USECS * ts_->granularity_; // usec
   double delta = 0;
 
+  // time to wait for utilization tests to know when to stop.
   ACE_Time_Value max_wait_time (ts_->util_time_, 0);
   ACE_Countdown_Time countdown (&max_wait_time);
 
@@ -671,22 +673,27 @@ Client::run_tests (Cubit_ptr cb,
 	 // lowest frequency thread.
 	 (ts_->thread_per_rate_ == 1 && id_ < (ts_->thread_count_ - 1)) ||
 	 // continous loop if we are running the utilization test
-	 (ts_->use_utilization_test_ == 1); 
+	 (ts_->use_utilization_test_ == 1) ||
+	 // continous loop if we are running the SERVER utilization test
+	 (ts_->run_server_utilization_test_ == 1); 
        i++)
     {
       // Elapsed time will be in microseconds.
       ACE_Time_Value delta_t;
 
-      if ( (i % ts_->granularity_) == 0)
+      // start timing a call
+      if ( (i % ts_->granularity_) == 0 && 
+	   (ts_->use_utilization_test_ == 0) &&
+	   (ts_->run_server_utilization_test_ == 0)
+	   )
         {
-          if (ts_->use_utilization_test_ == 0 && ts_->run_server_utilization_test_ == 0)
-            {
-              ACE_Time_Value tv (0,
-                                 (u_long) ((sleep_time - delta) < 0
-                                           ? 0
-                                           : (sleep_time - delta)));
-	      ACE_OS::sleep (tv);
-            }
+	  // delay a sufficient amount of time to be able to enforce
+	  // the calling frequency (i.e., 20Hz, 10Hz, 5Hz, 1Hz).
+	  ACE_Time_Value tv (0,
+			     (u_long) ((sleep_time - delta) < 0
+				       ? 0
+				       : (sleep_time - delta)));
+	  ACE_OS::sleep (tv);
 
 #if defined (CHORUS)
           pstartTime = pccTime1Get();
@@ -885,7 +892,11 @@ Client::run_tests (Cubit_ptr cb,
             }
         }
 
-      if ((i % ts_->granularity_) == (ts_->granularity_ - 1))
+      // stop the timer
+      if ( (i % ts_->granularity_) == (ts_->granularity_ - 1) && 
+	   (ts_->use_utilization_test_ == 0) &&
+	   (ts_->run_server_utilization_test_ == 0)
+	   )
         {
 #if defined (CHORUS)
           pstopTime = pccTime1Get();
@@ -909,27 +920,47 @@ Client::run_tests (Cubit_ptr cb,
           my_jitter_array [i/ts_->granularity_] = real_time; // in units of microseconds.
           // update the latency array, correcting the index using the granularity
 #else /* ACE_LACKS_FLOATING_POINT */
-          // Store the time in secs.
-
-// These comments are to temporarily fix what seems a bug in
-// the ACE_Long_Long class that is used to calc the elapsed
-// time.
-// I'll leave these here to debug it later.
- double tmp = (double)delta_t.sec ();
- double tmp2 = (double)delta_t.usec ();
- if (tmp > 100000) {tmp=0.0; tmp2 = 2000.0; fprintf(stderr, "tmp > 100000!, delta_t.usec ()=%f\n", (double)delta_t.usec ());}
- if (tmp2 < 100 && tmp == 0) {tmp=0.0; tmp2 = 2000.0; /*fprintf(stderr, "tmp2 < 100!, delta_t.usec ()=%d , delta_t.usec ()=%d\n", delta_t.sec (), delta_t.usec ());*/ }
 
- real_time = tmp + tmp2 / (double)ACE_ONE_SECOND_IN_USECS;
- //          real_time = (double)delta_t.sec () + (double)delta_t.usec () / (double)ACE_ONE_SECOND_IN_USECS;
+          // Store the time in secs.
 
+#if defined (VXWORKS)
+	  // @@ David, these comments are to temporarily fix what
+	  // seems a bug in the ACE_Long_Long class that is used to
+	  // calc the elapsed time.  It seems that subtraction of two
+	  // ACE_Long_Long are not done correctly when the least
+	  // significant value has wrapped around.  For example to
+	  // subtract these values: 00ff1001:00000001 minus
+	  // 00ff1000:ffffffff would give a huge number, instead of
+	  // giving 2.
+
+	  // This is only occuring in VxWorks.
+	  // I'll leave these here to debug it later.
+	  double tmp = (double)delta_t.sec ();
+	  double tmp2 = (double)delta_t.usec ();
+	  if (tmp > 100000) 
+	    { 
+	      tmp = 0.0; 
+	      tmp2 = 2000.0; 
+	      fprintf (stderr, "tmp > 100000!, delta_t.usec ()=%u\n", delta_t.usec ());
+	    }
+	  
+	  real_time = tmp + tmp2 / (double)ACE_ONE_SECOND_IN_USECS;
+#else	  	  
+	  real_time = ((double) delta_t.sec () + 
+		       (double) delta_t.usec () / (double) ACE_ONE_SECOND_IN_USECS);
+#endif /* VXWORKS */  
+	  
           real_time /= ts_->granularity_;
-
+	  
           delta = ((0.4 * fabs (real_time * ACE_ONE_SECOND_IN_USECS)) + (0.6 * delta)); // pow(10,6)
           latency += (real_time * ts_->granularity_);
-	  my_jitter_array [i/ts_->granularity_] = real_time * ACE_ONE_SECOND_IN_MSECS;
+	  my_jitter_array [i/ts_->granularity_] = real_time * ACE_ONE_SECOND_IN_MSECS; 
 #endif /* !ACE_LACKS_FLOATING_POINT */
-        }
+        } // END OF IF   :
+          //       if ( (i % ts_->granularity_) == (ts_->granularity_ - 1) && 
+          // 	   (ts_->use_utilization_test_ == 0) &&
+          // 	   (ts_->run_server_utilization_test_ == 0)
+          // 	   )
 
       if ( ts_->thread_per_rate_ == 1 && id_ < (ts_->thread_count_ - 1) )
 	{
@@ -951,11 +982,15 @@ Client::run_tests (Cubit_ptr cb,
 	      }
 	  }
 	  
-      if (ts_->use_utilization_test_ == 1)
+      if (ts_->use_utilization_test_ == 1 || 
+	  ts_->run_server_utilization_test_ == 1)
 	{
 	  countdown.update ();
 	  if (max_wait_time == ACE_Time_Value::zero)
-	    break;
+	    {
+	      ts_->loop_count_ = call_count;
+	      break;
+	    }
 	}
 
     } /* end of for () */
@@ -963,7 +998,12 @@ Client::run_tests (Cubit_ptr cb,
   if (id_ == 0)
     ts_->high_priority_loop_count_ = call_count;
 
-  if (call_count > 0)
+  // perform latency stats onlt if we are not running the utilization
+  // tests.
+  if (call_count > 0 && 
+      (ts_->use_utilization_test_ == 0) &&
+      (ts_->run_server_utilization_test_ == 0)
+      )
     {
       if (error_count == 0)
         {
diff --git a/TAO/performance-tests/Cubit/TAO/MT_Cubit/Util_Thread.cpp b/TAO/performance-tests/Cubit/TAO/MT_Cubit/Util_Thread.cpp
index 4d3c2a84e94..8a33b046b0b 100644
--- a/TAO/performance-tests/Cubit/TAO/MT_Cubit/Util_Thread.cpp
+++ b/TAO/performance-tests/Cubit/TAO/MT_Cubit/Util_Thread.cpp
@@ -51,7 +51,7 @@ Util_Thread::svc (void)
   return 0;
 }
 
-double
+u_long
 Util_Thread::get_number_of_computations (void)
 {
   return this->number_of_computations_;
diff --git a/TAO/performance-tests/Cubit/TAO/MT_Cubit/Util_Thread.h b/TAO/performance-tests/Cubit/TAO/MT_Cubit/Util_Thread.h
index 8f199da4b5a..5509de3d32b 100644
--- a/TAO/performance-tests/Cubit/TAO/MT_Cubit/Util_Thread.h
+++ b/TAO/performance-tests/Cubit/TAO/MT_Cubit/Util_Thread.h
@@ -31,7 +31,7 @@ public:
   virtual int svc (void);
   // The thread entry point function.
 
-  double get_number_of_computations (void);
+  u_long get_number_of_computations (void);
   // Gets the number of computations the thread was able to perform.
 
   void computation (void);
@@ -44,7 +44,7 @@ private:
   int run_computations (void);
   // Run the computations.
 
-  double number_of_computations_;
+  u_long number_of_computations_;
   // This stores the number of computations performed
   // by the Util_Thread so far
 
diff --git a/TAO/performance-tests/Cubit/TAO/MT_Cubit/client.cpp b/TAO/performance-tests/Cubit/TAO/MT_Cubit/client.cpp
index 69dea293097..0aa507cac31 100644
--- a/TAO/performance-tests/Cubit/TAO/MT_Cubit/client.cpp
+++ b/TAO/performance-tests/Cubit/TAO/MT_Cubit/client.cpp
@@ -150,13 +150,14 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
   int i;
   u_int j;
 
+  char * task_id = 0;
+
   ACE_High_Res_Timer timer_;
   ACE_Time_Value delta_t;
 
   timer_.start ();
 
 #if defined (VXWORKS) 
-  char * task_id = 0;
   ctx = 0;
   ACE_NEW_RETURN (task_id,
                   char [32],
@@ -236,7 +237,7 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
   priority = ACE_THR_PRI_FIFO_DEF;
 #elif defined (ACE_WIN32)
   priority = ACE_Sched_Params::priority_max (ACE_SCHED_FIFO,
-                                    ACE_SCOPE_THREAD);
+					     ACE_SCOPE_THREAD);
 #else  /* ! VXWORKS */
   priority = ACE_THR_PRI_FIFO_DEF + 25;
 #endif /* ! ACE_WIN32 */
@@ -248,17 +249,13 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
   if (high_priority_client.activate (THR_BOUND | ACE_SCHED_FIFO,
                                      1,
                                      0,
-#if !defined (VXWORKS) 
-                                     priority) == -1)
-#else
                                      priority,
  				     -1,
  				     0,
  				     0,
  				     0,
  				     0,
- 				     &task_id) == -1)
-#endif /* VXWORKS */
+ 				     (ACE_thread_t*)task_id) == -1)
     ACE_ERROR ((LM_ERROR,
                 "%p; priority is %d\n",
                 "activate failed",
@@ -271,6 +268,9 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
 
   number_of_low_priority_client = ts->thread_count_ - 1;
 
+  // mechanism to distribute the available priorities among the
+  // threads when there are not enough different priorities for all
+  // threads.
   if (ts->use_multiple_priority_ == 1)
     {    
       ACE_Sched_Priority_Iterator priority_iterator (ACE_SCHED_FIFO,
@@ -286,23 +286,24 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
       // 1 priority is exclusive for the high priority client.
       number_of_priorities --;
 
-      // if utilization thread is present, reduce in 1 the available priorities.
+      // if utilization thread is present, reduce in 1 the available
+      // priorities.
       if (ts->use_utilization_test_ == 1)
 	{
 	  number_of_priorities --;
 	}
 
-      // Drop the priority, so that the priority of clients will increase
-      // with increasing client number.
-      for (j = 0; j < number_of_low_priority_client + 1; j++)
+      // Drop the priority, so that the priority of clients will
+      // increase with increasing client number.
+      for (j = 0; j < number_of_low_priority_client; j++)
 	priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
 							priority,
 							ACE_SCOPE_THREAD);
 
-      // if the lowest priority of the "low priority clients" is the minimum, 
-      // and we are running the utilization thread, increment the priority, 
-      // since we don't want the utlization thread and a "low priority thread" 
-      // to have the same priority.
+      // if the lowest priority of the "low priority clients" is the
+      // minimum, and we are running the utilization thread, increment
+      // the priority, since we don't want the utlization thread and a
+      // "low priority thread" to have the same priority.
       if ( priority == ACE_Sched_Params::priority_min (ACE_SCHED_FIFO,
 						       ACE_SCOPE_THREAD) &&
 	   ts->use_utilization_test_ == 1)
@@ -310,10 +311,10 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
                                                     priority,
                                                     ACE_SCOPE_THREAD);
 	
-      // granularity of the assignment of the priorities.  Some OSs have 
-      // fewer levels of priorities than we have threads in our test, so 
-      // with this mechanism we assign priorities to groups of threads when 
-      // there are more threads than priorities.
+      // granularity of the assignment of the priorities.  Some OSs
+      // have fewer levels of priorities than we have threads in our
+      // test, so with this mechanism we assign priorities to groups
+      // of threads when there are more threads than priorities.
       grain = number_of_low_priority_client / number_of_priorities;
       counter = 0;
 
@@ -359,18 +360,13 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
       if (low_priority_client[i - 1]->activate (THR_BOUND | ACE_SCHED_FIFO,
 						1,
 						0,
-#if !defined (VXWORKS) 
-						priority) == -1)
-#else
 						priority,    // These are constructor defaults.     
 						-1,	     //  int grp_id = -1,                   
 						0,	     //  ACE_Task_Base *task = 0,           
 						0,	     //  ACE_hthread_t thread_handles[] = 0,
 						0,	     //  void *stack[] = 0,                 
 						0,	     //  size_t stack_size[] = 0,           
-						&task_id) == -1)
-#endif /* VXWORKS */
-
+						(ACE_thread_t*)task_id) == -1)
         ACE_ERROR ((LM_ERROR,
 	            "%p; priority is %d\n",
 	            "activate failed",
@@ -380,14 +376,15 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
         {    
           counter = (counter + 1) % grain;
           if ( (counter == 0) && 
-               //Just so when we distribute the priorities among the threads, we make sure we don't go overboard.
+               //Just so when we distribute the priorities among the
+               //threads, we make sure we don't go overboard.
                ((number_of_priorities * grain) > (number_of_low_priority_client - (i - 1))) )
-              {
-          	      // Get the next higher priority.
-          	      priority = ACE_Sched_Params::next_priority (ACE_SCHED_FIFO,
-          						     priority,
-          						     ACE_SCOPE_THREAD);
-              }
+	    {
+	      // Get the next higher priority.
+	      priority = ACE_Sched_Params::next_priority (ACE_SCHED_FIFO,
+							  priority,
+							  ACE_SCOPE_THREAD);
+	    }
           
         }          
     } /* end of for () */
@@ -443,12 +440,13 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
 #if defined (VXWORKS) 
   if (ts->context_switch_test_ == 1)
     {
-     fprintf(stderr, "Adding the context switch hook!\n");
-     taskSwitchHookAdd ((FUNCPTR)&switchHook);
+      fprintf(stderr, "Adding the context switch hook!\n");
+      taskSwitchHookAdd ((FUNCPTR)&switchHook);
     }
 #endif
   
-  // Wait for all the client threads to exit (except the utilization thread).
+  // Wait for all the client threads to exit (except the utilization
+  // thread).
   thread_manager->wait ();
 
 #if defined (NO_ACE_QUANTIFY)
@@ -461,9 +459,9 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
   timer_.elapsed_time (delta_t);
 
   if (ts->use_utilization_test_ == 1)
-    // signal the utilization thread to finish with its work..
-    // only if utilization test was specified.  See
-    // description of this variable in header file.
+    // signal the utilization thread to finish with its work..  only
+    // if utilization test was specified.  See description of this
+    // variable in header file.
     {
       util_thread.done_ = 1;
 
@@ -479,7 +477,8 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
 #if defined (ACE_HAS_PRUSAGE_T)
       timer_for_context_switch.stop ();
       timer_for_context_switch.get_rusage (usage);
-      // Add up the voluntary context switches & involuntary context switches
+      // Add up the voluntary context switches & involuntary context
+      // switches
       context_switch = usage.pr_vctx + usage.pr_ictx - context_switch;
       ACE_DEBUG ((LM_DEBUG, 
 		  "Voluntary context switches=%d, Involuntary context switches=%d\n",
@@ -488,7 +487,8 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
 #elif defined (ACE_HAS_GETRUSAGE) && !defined (ACE_WIN32)
       timer_for_context_switch.stop ();
       timer_for_context_switch.get_rusage (usage);
-      // Add up the voluntary context switches & involuntary context switches
+      // Add up the voluntary context switches & involuntary context
+      // switches
       context_switch = usage.ru_nvcsw + usage.ru_nivcsw - context_switch;
       ACE_DEBUG ((LM_DEBUG, 
 		  "Voluntary context switches=%d, Involuntary context switches=%d\n",
@@ -496,88 +496,95 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
 		  usage.ru_nivcsw));
 #elif defined (VXWORKS) /* ACE_HAS_GETRUSAGE */
       taskSwitchHookDelete ((FUNCPTR)&switchHook);
-       ACE_DEBUG ((LM_DEBUG, 
+      ACE_DEBUG ((LM_DEBUG, 
  		  "Context switches=%d\n",
  		  ctx));
 #endif
     }
 
+  // if running the utilization test, don't report latency nor jitter.
+  if (ts->use_utilization_test_ == 0 && 
+      ts->run_server_utilization_test_ == 0)
+    {
 #if defined (VXWORKS)
-  ACE_DEBUG ((LM_DEBUG, "Test done.\n"
-              "High priority client latency : %f msec, jitter: %f msec\n"
-              "Low priority client latency : %f msec, jitter: %f msec\n",
-              high_priority_client.get_high_priority_latency (),
-              high_priority_client.get_high_priority_jitter (),
-              low_priority_client[0]->get_low_priority_latency (),
-              low_priority_client[0]->get_low_priority_jitter ()));
-  // output the latency values to a file, tab separated, to import it
-  // to Excel to calculate jitter, in the mean time we come up with
-  // the sqrt() function.
-  output_latency (ts);
-
+      ACE_DEBUG ((LM_DEBUG, "Test done.\n"
+		  "High priority client latency : %f msec, jitter: %f msec\n"
+		  "Low priority client latency : %f msec, jitter: %f msec\n",
+		  high_priority_client.get_high_priority_latency (),
+		  high_priority_client.get_high_priority_jitter (),
+		  low_priority_client[0]->get_low_priority_latency (),
+		  low_priority_client[0]->get_low_priority_jitter ()));
+      // output the latency values to a file, tab separated, to import
+      // it to Excel to calculate jitter, in the mean time we come up
+      // with the sqrt() function.
+      output_latency (ts);
 #elif defined (CHORUS)
-  ACE_DEBUG ((LM_DEBUG, 
-	      "Test done.\n"
-	      "High priority client latency : %u usec\n"
-	      "Low priority client latency : %u usec\n",
-	      high_priority_client.get_high_priority_latency (),
-	      low_priority_client[0]->get_low_priority_latency () ));
-
-  // output the latency values to a file, tab separated, to import it
-  // to Excel to calculate jitter, in the mean time we come up with
-  // the sqrt() function.
-  output_latency (ts);
+      ACE_DEBUG ((LM_DEBUG, 
+		  "Test done.\n"
+		  "High priority client latency : %u usec\n"
+		  "Low priority client latency : %u usec\n",
+		  high_priority_client.get_high_priority_latency (),
+		  low_priority_client[0]->get_low_priority_latency () ));
+
+      // output the latency values to a file, tab separated, to import
+      // it to Excel to calculate jitter, in the mean time we come up
+      // with the sqrt() function.
+      output_latency (ts);
 #else /* !CHORUS */
-  ACE_DEBUG ((LM_DEBUG, "Test done.\n"
-              "High priority client latency : %f msec, jitter: %f msec\n"
-              "Low priority client latency : %f msec, jitter: %f msec\n"
-              "# of context switches: %d, context_switch_time: %f msec\n"
-              "total context switch time: %f msec\n",
-              high_priority_client.get_high_priority_latency (),
-              high_priority_client.get_high_priority_jitter (),
-              low_priority_client[0]->get_low_priority_latency (),
-              low_priority_client[0]->get_low_priority_jitter (),
-              context_switch,
-              csw/1000,
-              csw * context_switch/1000 ));
-//  output_latency (ts);
+      ACE_DEBUG ((LM_DEBUG, "Test done.\n"
+		  "High priority client latency : %f msec, jitter: %f msec\n"
+		  "Low priority client latency : %f msec, jitter: %f msec\n"
+		  "# of context switches: %d, context_switch_time: %f msec\n"
+		  "total context switch time: %f msec\n",
+		  high_priority_client.get_high_priority_latency (),
+		  high_priority_client.get_high_priority_jitter (),
+		  low_priority_client[0]->get_low_priority_latency (),
+		  low_priority_client[0]->get_low_priority_jitter (),
+		  context_switch,
+		  csw/1000,
+		  csw * context_switch/1000 ));
+      // output_latency (ts);
 #endif /* !VXWORKS && !CHORUS */
-
+    }
 
   if (ts->use_utilization_test_ == 1)
-  {
-    total_util_task_duration = util_task_duration * util_thread.get_number_of_computations ();
+    {
+      total_util_task_duration = util_task_duration * util_thread.get_number_of_computations ();
     
-    total_latency = (delta_t.sec () * 
-                     ACE_ONE_SECOND_IN_MSECS + 
-                     (double)delta_t.usec () / ACE_ONE_SECOND_IN_MSECS);    
-
-    total_latency_high = total_latency - total_util_task_duration;
-
-    // Calc and print the CPU percentage. I add 0.5 to round to the
-    // nearest integer before casting it to int.
-    ACE_DEBUG ((LM_DEBUG, 
-  		"\t%% ORB Client CPU utilization: %u %%\n"
-  		"\t%% Idle time: %u %%\n",
-  		(u_int) (total_latency_high * 100 / total_latency + 0.5),
-  		(u_int) (total_util_task_duration * 100 / total_latency + 0.5) ));
+      total_latency = (delta_t.sec () * 
+		       ACE_ONE_SECOND_IN_MSECS + 
+		       (double)delta_t.usec () / ACE_ONE_SECOND_IN_MSECS);    
+
+      total_latency_high = total_latency - total_util_task_duration;
+
+      // Calc and print the CPU percentage. I add 0.5 to round to the
+      // nearest integer before casting it to int.
+      ACE_DEBUG ((LM_DEBUG, 
+		  "\t%% ORB Client CPU utilization: %u %%\n"
+		  "\t%% Idle time: %u %%\n\n",
+		  (u_int) (total_latency_high * 100 / total_latency + 0.5),
+		  (u_int) (total_util_task_duration * 100 / total_latency + 0.5) ));
     
-  #if defined (ACE_LACKS_FLOATING_POINT)
-    ACE_DEBUG ((LM_DEBUG,
-  		"(%t) utilization task performed %u computations\n",
-  		util_thread.get_number_of_computations ()));
-    ACE_DEBUG ((LM_DEBUG,
-  		"(%t) utilization computation time is %u usecs\n", 
-  		util_task_duration));
-  #else
-    ACE_DEBUG ((LM_DEBUG,
-  		"(%t) utilization task performed %g computations\n",
-  		util_thread.get_number_of_computations ()));
-    ACE_DEBUG ((LM_DEBUG,
-  		"(%t) utilization computation time is %f msecs\n", 
-  		util_task_duration));
-  #endif /* ! ACE_LACKS_FLOATING_POINT */
-  }
+      ACE_DEBUG ((LM_DEBUG,
+		  "(%t) UTILIZATION task performed \t%u computations\n"
+		  "(%t) CLIENT task performed \t\t%u CORBA calls\n"
+		  "\t Ratio of computations to CORBA calls is %u.%u:1\n\n",
+		  util_thread.get_number_of_computations (),
+		  ts->loop_count_,
+		  util_thread.get_number_of_computations () / ts->loop_count_,
+		  (util_thread.get_number_of_computations () % ts->loop_count_) * 100 / ts->loop_count_
+		  ));
+
+#if defined (ACE_LACKS_FLOATING_POINT)
+      ACE_DEBUG ((LM_DEBUG,
+		  "(%t) utilization computation time is %u usecs\n", 
+		  util_task_duration));
+#else
+      ACE_DEBUG ((LM_DEBUG,
+		  "(%t) utilization computation time is %f msecs\n", 
+		  util_task_duration));
+#endif /* ! ACE_LACKS_FLOATING_POINT */
+    }
 
 #if defined (VXWORKS)
   delete task_id;
@@ -608,25 +615,20 @@ do_thread_per_rate_test (ACE_Thread_Manager *thread_manager,
   ACE_DEBUG ((LM_DEBUG,
 	      "Creating 20 Hz client with priority %d\n",
 	      priority));
+
   if (CB_20Hz_client.activate (THR_BOUND | ACE_SCHED_FIFO, 1, 1, priority) == -1)
-    ACE_ERROR ((LM_ERROR,
-		"%p\n"
-		"activate failed"));
+    ACE_ERROR ((LM_ERROR, "(%P|%t) errno = %p: activate failed\n"));
 
   priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
 						  priority,
 						  ACE_SCOPE_THREAD);
 
-  priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
-						  priority,
-						  ACE_SCOPE_THREAD);
   ACE_DEBUG ((LM_DEBUG,
 	      "Creating 10 Hz client with priority %d\n",
 	      priority));
+
   if (CB_10Hz_client.activate (THR_BOUND | ACE_SCHED_FIFO, 1, 1, priority) == -1)
-    ACE_ERROR ((LM_ERROR,
-		"%p\n"
-		"activate failed"));
+    ACE_ERROR ((LM_ERROR, "(%P|%t) errno = %p: activate failed\n"));
 
   priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
 						  priority,
@@ -634,10 +636,9 @@ do_thread_per_rate_test (ACE_Thread_Manager *thread_manager,
   ACE_DEBUG ((LM_DEBUG,
 	      "Creating 5 Hz client with priority %d\n",
 	      priority));
+
   if (CB_5Hz_client.activate (THR_BOUND | ACE_SCHED_FIFO, 1, 1, priority) == -1)
-    ACE_ERROR ((LM_ERROR,
-		"%p\n"
-		"activate failed"));
+    ACE_ERROR ((LM_ERROR, "(%P|%t) errno = %p: activate failed\n"));
 
   priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
 						  priority,
@@ -645,10 +646,9 @@ do_thread_per_rate_test (ACE_Thread_Manager *thread_manager,
   ACE_DEBUG ((LM_DEBUG,
 	      "Creating 1 Hz client with priority %d\n",
 	      priority));
+
   if (CB_1Hz_client.activate (THR_BOUND | ACE_SCHED_FIFO, 1, 1, priority) == -1)
-    ACE_ERROR ((LM_ERROR,
-		"%p\n"
-		"activate failed"));
+    ACE_ERROR ((LM_ERROR, "(%P|%t) errno = %p: activate failed\n"));
 
   // Wait for all the threads to exit.
   thread_manager->wait ();
diff --git a/TAO/performance-tests/Cubit/TAO/MT_Cubit/cubit_i.cpp b/TAO/performance-tests/Cubit/TAO/MT_Cubit/cubit_i.cpp
index 4ed26dce816..1c4d2799ac4 100644
--- a/TAO/performance-tests/Cubit/TAO/MT_Cubit/cubit_i.cpp
+++ b/TAO/performance-tests/Cubit/TAO/MT_Cubit/cubit_i.cpp
@@ -25,16 +25,7 @@ Cubit_i::cube_octet (CORBA::Octet o,
       ts_->barrier_->wait ();
     }
   
-//   if (ts_->run_server_utilization_test_ == 0)
-//     {
-//       ACE_DEBUG ((LM_DEBUG,
-// 		  "(%t) )))))))) "
-// 		  "test STARTED at %D\n"));
-      
-//       ts_->run_server_utilization_test_ = 1;
-//       this->util_started_ = 1;
-//       this->ts_->timer_.start ();
-//     }
+  ts_->loop_count_++;
 
   return (CORBA::Octet) (o * o * o);
 }
@@ -75,6 +66,9 @@ void Cubit_i::shutdown (CORBA::Environment &)
 {
   ACE_DEBUG ((LM_DEBUG, 
 	      "(%t) Calling TAO_ORB_Core_instance ()->orb ()->shutdown ()\n"));
+
   TAO_ORB_Core_instance ()->orb ()->shutdown ();
+
+  // this is to signal the utilization thread to exit its loop.
   ts_->utilization_task_started_ = 0;
 }
diff --git a/TAO/performance-tests/Cubit/TAO/MT_Cubit/server.cpp b/TAO/performance-tests/Cubit/TAO/MT_Cubit/server.cpp
index 2a9e62169db..0e0ae8f8824 100644
--- a/TAO/performance-tests/Cubit/TAO/MT_Cubit/server.cpp
+++ b/TAO/performance-tests/Cubit/TAO/MT_Cubit/server.cpp
@@ -543,8 +543,8 @@ start_servants (ACE_Thread_Manager *serv_thr_mgr, ACE_Barrier &start_barrier, Ta
                                     0,
                                     priority) == -1)
     {
-      ACE_ERROR ((LM_ERROR, "(%P|%t) %p\n",
-                  "high_priority_task->activate"));
+      ACE_ERROR ((LM_ERROR, "(%P|%t) %p\n"
+                  "\thigh_priority_task->activate failed"));
     }
 
   // Create an array to hold pointers to the low priority tasks.
@@ -580,7 +580,7 @@ start_servants (ACE_Thread_Manager *serv_thr_mgr, ACE_Barrier &start_barrier, Ta
 
       // Drop the priority, so that the priority of clients will increase
       // with increasing client number.
-      for (i = 0; i < number_of_low_priority_servants + 1; i++)
+      for (i = 0; i < number_of_low_priority_servants; i++)
         priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
                                                         priority,
                                                         ACE_SCOPE_THREAD);
@@ -643,8 +643,8 @@ start_servants (ACE_Thread_Manager *serv_thr_mgr, ACE_Barrier &start_barrier, Ta
                                            0,
                                            priority) == -1)
         {
-          ACE_ERROR ((LM_ERROR, "(%P|%t; %p\n",
-                      "low_priority_task[i]->activate"));
+          ACE_ERROR ((LM_ERROR, "(%P|%t) %p\n"
+                      "\tlow_priority_task[i]->activate"));
         }
 
       ACE_DEBUG ((LM_DEBUG,
@@ -657,7 +657,8 @@ start_servants (ACE_Thread_Manager *serv_thr_mgr, ACE_Barrier &start_barrier, Ta
         {
           counter = (counter + 1) % grain;
           if ( (counter == 0) &&
-               //Just so when we distribute the priorities among the threads, we make sure we don't go overboard.
+               //Just so when we distribute the priorities among the
+               //threads, we make sure we don't go overboard.
                ((number_of_priorities * grain) > (number_of_low_priority_servants - (i - 1))) )
               {
                       // Get the next higher priority.
@@ -724,8 +725,8 @@ start_utilization (ACE_Thread_Manager *util_thr_mgr, Task_State *ts)
 			   0,
 			   priority) == -1)
     {
-      ACE_ERROR ((LM_ERROR, "(%P|%t) %p\n",
-                  "util_task->activate"));
+      ACE_ERROR ((LM_ERROR, "(%P|%t) %p\n"
+                  "\tutil_task->activate failed"));
     }
 
   return util_task;
@@ -784,7 +785,10 @@ main (int argc, char *argv[])
   Task_State ts ( _argc, _argv);
 
   if (run_utilization_test == 1)
-    ts.run_server_utilization_test_ = 1;
+    {
+      ts.run_server_utilization_test_ = 1;
+      ts.loop_count_ = 0;
+     }
 
   Util_Thread * util_task = 0;
 
@@ -873,11 +877,22 @@ main (int argc, char *argv[])
 			      
       total_latency_servants = total_latency - total_util_task_duration;
     
+      ACE_DEBUG ((LM_DEBUG, 
+		  "-------------------------- Stats -------------------------------\n"));
+
       ACE_DEBUG ((LM_DEBUG,
-		  "(%t) utilization task performed %g computations\n"
-		  "(%t) each computation had a duration of %f msecs\n"
-		  "(%t) elapsed time is %f msecs\n",
+		  "(%t) UTILIZATION task performed \t%u computations\n"
+		  "(%t) SERVANT task serviced \t\t%u CORBA calls\n"
+		  "\t Ratio of computations to CORBA calls is %u.%u:1\n\n",
 		  util_task->get_number_of_computations (),
+		  ts.loop_count_,
+		  util_task->get_number_of_computations () / ts.loop_count_,
+		  (util_task->get_number_of_computations () % ts.loop_count_) * 100 / ts.loop_count_
+		  ));
+
+      ACE_DEBUG ((LM_DEBUG,
+		  "(%t) Each computation had a duration of %f msecs\n"
+		  "(%t) Total elapsed time of test is %f msecs\n",
 		  util_task_duration / 1000,
 		  total_latency / 1000));    
 
diff --git a/TAO/tests/Cubit/TAO/MT_Cubit/README b/TAO/tests/Cubit/TAO/MT_Cubit/README
index 8a545ac722c..e0b79a019f3 100644
--- a/TAO/tests/Cubit/TAO/MT_Cubit/README
+++ b/TAO/tests/Cubit/TAO/MT_Cubit/README
@@ -69,6 +69,10 @@ Options are case sensitive, e.g., "-U" has a different meaning than "-u".
 	Default Value: Disabled. To enable, specify "-U" on the
                        command line.
 
+[-r] turn on the thread-per-rate tests.  (Priority inversion tests)
+
+	Default Value: Disabled.
+
 The server will print the IOR of the servants, but it is not required
 by the client.
 
@@ -84,10 +88,10 @@ by the client.
          [-r]			// use thread-per-rate test.
          [-o]			// makes client use oneway calls instead
          [-x]			// makes a call to servant to shutdown
-         [-u]			// run the client utilization test
-         [-U]			// run the servant utilization test
+         [-u <seconds> ]	// run the client utilization test for <seconds> seconds
+         [-U <seconds> ]	// run the servant utilization test for <seconds> seconds
          [-1]			// use one to n test.
-         [-g]			// choose the granularity of timing calls
+         [-g <granularity>]	// choose the granularity of timing calls
          [-c]			// obtain number of context switches.
          [-s]			// makes client *NOT* use the name service
          [-m]			// use multiple priorities for the low priority client.
@@ -116,7 +120,7 @@ by the client.
 
 	Default Value: nil, which means not to write the iors to disk.
 
-[-r] turn on the thread-per-rate tests.
+[-r] turn on the thread-per-rate tests.  (Priority inversion tests)
 
 	Default Value: Disabled.
 
diff --git a/TAO/tests/Cubit/TAO/MT_Cubit/Task_Client.cpp b/TAO/tests/Cubit/TAO/MT_Cubit/Task_Client.cpp
index d4c95c2c70d..29cb263d76d 100644
--- a/TAO/tests/Cubit/TAO/MT_Cubit/Task_Client.cpp
+++ b/TAO/tests/Cubit/TAO/MT_Cubit/Task_Client.cpp
@@ -30,7 +30,7 @@ Task_State::Task_State (int argc, char **argv)
     run_server_utilization_test_ (0),
     util_time_ (0)
 {
-  ACE_Get_Opt opts (argc, argv, "Umu:sn:t:d:rxof:g:1c");
+  ACE_Get_Opt opts (argc, argv, "U:mu:sn:t:d:rxof:g:1c");
   int c;
   int datatype;
 
@@ -43,6 +43,7 @@ Task_State::Task_State (int argc, char **argv)
       break;
     case 'U':
       run_server_utilization_test_ = 1;
+      util_time_ = ACE_OS::atoi (opts.optarg);
       break;
     case 'm':
       use_multiple_priority_ = 1;
@@ -648,6 +649,7 @@ Client::run_tests (Cubit_ptr cb,
   double sleep_time = (1 / frequency) * ACE_ONE_SECOND_IN_USECS * ts_->granularity_; // usec
   double delta = 0;
 
+  // time to wait for utilization tests to know when to stop.
   ACE_Time_Value max_wait_time (ts_->util_time_, 0);
   ACE_Countdown_Time countdown (&max_wait_time);
 
@@ -671,22 +673,27 @@ Client::run_tests (Cubit_ptr cb,
 	 // lowest frequency thread.
 	 (ts_->thread_per_rate_ == 1 && id_ < (ts_->thread_count_ - 1)) ||
 	 // continous loop if we are running the utilization test
-	 (ts_->use_utilization_test_ == 1); 
+	 (ts_->use_utilization_test_ == 1) ||
+	 // continous loop if we are running the SERVER utilization test
+	 (ts_->run_server_utilization_test_ == 1); 
        i++)
     {
       // Elapsed time will be in microseconds.
       ACE_Time_Value delta_t;
 
-      if ( (i % ts_->granularity_) == 0)
+      // start timing a call
+      if ( (i % ts_->granularity_) == 0 && 
+	   (ts_->use_utilization_test_ == 0) &&
+	   (ts_->run_server_utilization_test_ == 0)
+	   )
         {
-          if (ts_->use_utilization_test_ == 0 && ts_->run_server_utilization_test_ == 0)
-            {
-              ACE_Time_Value tv (0,
-                                 (u_long) ((sleep_time - delta) < 0
-                                           ? 0
-                                           : (sleep_time - delta)));
-	      ACE_OS::sleep (tv);
-            }
+	  // delay a sufficient amount of time to be able to enforce
+	  // the calling frequency (i.e., 20Hz, 10Hz, 5Hz, 1Hz).
+	  ACE_Time_Value tv (0,
+			     (u_long) ((sleep_time - delta) < 0
+				       ? 0
+				       : (sleep_time - delta)));
+	  ACE_OS::sleep (tv);
 
 #if defined (CHORUS)
           pstartTime = pccTime1Get();
@@ -885,7 +892,11 @@ Client::run_tests (Cubit_ptr cb,
             }
         }
 
-      if ((i % ts_->granularity_) == (ts_->granularity_ - 1))
+      // stop the timer
+      if ( (i % ts_->granularity_) == (ts_->granularity_ - 1) && 
+	   (ts_->use_utilization_test_ == 0) &&
+	   (ts_->run_server_utilization_test_ == 0)
+	   )
         {
 #if defined (CHORUS)
           pstopTime = pccTime1Get();
@@ -909,27 +920,47 @@ Client::run_tests (Cubit_ptr cb,
           my_jitter_array [i/ts_->granularity_] = real_time; // in units of microseconds.
           // update the latency array, correcting the index using the granularity
 #else /* ACE_LACKS_FLOATING_POINT */
-          // Store the time in secs.
-
-// These comments are to temporarily fix what seems a bug in
-// the ACE_Long_Long class that is used to calc the elapsed
-// time.
-// I'll leave these here to debug it later.
- double tmp = (double)delta_t.sec ();
- double tmp2 = (double)delta_t.usec ();
- if (tmp > 100000) {tmp=0.0; tmp2 = 2000.0; fprintf(stderr, "tmp > 100000!, delta_t.usec ()=%f\n", (double)delta_t.usec ());}
- if (tmp2 < 100 && tmp == 0) {tmp=0.0; tmp2 = 2000.0; /*fprintf(stderr, "tmp2 < 100!, delta_t.usec ()=%d , delta_t.usec ()=%d\n", delta_t.sec (), delta_t.usec ());*/ }
 
- real_time = tmp + tmp2 / (double)ACE_ONE_SECOND_IN_USECS;
- //          real_time = (double)delta_t.sec () + (double)delta_t.usec () / (double)ACE_ONE_SECOND_IN_USECS;
+          // Store the time in secs.
 
+#if defined (VXWORKS)
+	  // @@ David, these comments are to temporarily fix what
+	  // seems a bug in the ACE_Long_Long class that is used to
+	  // calc the elapsed time.  It seems that subtraction of two
+	  // ACE_Long_Long are not done correctly when the least
+	  // significant value has wrapped around.  For example to
+	  // subtract these values: 00ff1001:00000001 minus
+	  // 00ff1000:ffffffff would give a huge number, instead of
+	  // giving 2.
+
+	  // This is only occuring in VxWorks.
+	  // I'll leave these here to debug it later.
+	  double tmp = (double)delta_t.sec ();
+	  double tmp2 = (double)delta_t.usec ();
+	  if (tmp > 100000) 
+	    { 
+	      tmp = 0.0; 
+	      tmp2 = 2000.0; 
+	      fprintf (stderr, "tmp > 100000!, delta_t.usec ()=%u\n", delta_t.usec ());
+	    }
+	  
+	  real_time = tmp + tmp2 / (double)ACE_ONE_SECOND_IN_USECS;
+#else	  	  
+	  real_time = ((double) delta_t.sec () + 
+		       (double) delta_t.usec () / (double) ACE_ONE_SECOND_IN_USECS);
+#endif /* VXWORKS */  
+	  
           real_time /= ts_->granularity_;
-
+	  
           delta = ((0.4 * fabs (real_time * ACE_ONE_SECOND_IN_USECS)) + (0.6 * delta)); // pow(10,6)
           latency += (real_time * ts_->granularity_);
-	  my_jitter_array [i/ts_->granularity_] = real_time * ACE_ONE_SECOND_IN_MSECS;
+	  my_jitter_array [i/ts_->granularity_] = real_time * ACE_ONE_SECOND_IN_MSECS; 
 #endif /* !ACE_LACKS_FLOATING_POINT */
-        }
+        } // END OF IF   :
+          //       if ( (i % ts_->granularity_) == (ts_->granularity_ - 1) && 
+          // 	   (ts_->use_utilization_test_ == 0) &&
+          // 	   (ts_->run_server_utilization_test_ == 0)
+          // 	   )
 
       if ( ts_->thread_per_rate_ == 1 && id_ < (ts_->thread_count_ - 1) )
 	{
@@ -951,11 +982,15 @@ Client::run_tests (Cubit_ptr cb,
 	      }
 	  }
 	  
-      if (ts_->use_utilization_test_ == 1)
+      if (ts_->use_utilization_test_ == 1 || 
+	  ts_->run_server_utilization_test_ == 1)
 	{
 	  countdown.update ();
 	  if (max_wait_time == ACE_Time_Value::zero)
-	    break;
+	    {
+	      ts_->loop_count_ = call_count;
+	      break;
+	    }
 	}
 
     } /* end of for () */
@@ -963,7 +998,12 @@ Client::run_tests (Cubit_ptr cb,
   if (id_ == 0)
     ts_->high_priority_loop_count_ = call_count;
 
-  if (call_count > 0)
+  // perform latency stats onlt if we are not running the utilization
+  // tests.
+  if (call_count > 0 && 
+      (ts_->use_utilization_test_ == 0) &&
+      (ts_->run_server_utilization_test_ == 0)
+      )
     {
       if (error_count == 0)
         {
diff --git a/TAO/tests/Cubit/TAO/MT_Cubit/Util_Thread.cpp b/TAO/tests/Cubit/TAO/MT_Cubit/Util_Thread.cpp
index 4d3c2a84e94..8a33b046b0b 100644
--- a/TAO/tests/Cubit/TAO/MT_Cubit/Util_Thread.cpp
+++ b/TAO/tests/Cubit/TAO/MT_Cubit/Util_Thread.cpp
@@ -51,7 +51,7 @@ Util_Thread::svc (void)
   return 0;
 }
 
-double
+u_long
 Util_Thread::get_number_of_computations (void)
 {
   return this->number_of_computations_;
diff --git a/TAO/tests/Cubit/TAO/MT_Cubit/Util_Thread.h b/TAO/tests/Cubit/TAO/MT_Cubit/Util_Thread.h
index 8f199da4b5a..5509de3d32b 100644
--- a/TAO/tests/Cubit/TAO/MT_Cubit/Util_Thread.h
+++ b/TAO/tests/Cubit/TAO/MT_Cubit/Util_Thread.h
@@ -31,7 +31,7 @@ public:
   virtual int svc (void);
   // The thread entry point function.
 
-  double get_number_of_computations (void);
+  u_long get_number_of_computations (void);
   // Gets the number of computations the thread was able to perform.
 
   void computation (void);
@@ -44,7 +44,7 @@ private:
   int run_computations (void);
   // Run the computations.
 
-  double number_of_computations_;
+  u_long number_of_computations_;
   // This stores the number of computations performed
   // by the Util_Thread so far
 
diff --git a/TAO/tests/Cubit/TAO/MT_Cubit/client.cpp b/TAO/tests/Cubit/TAO/MT_Cubit/client.cpp
index 69dea293097..0aa507cac31 100644
--- a/TAO/tests/Cubit/TAO/MT_Cubit/client.cpp
+++ b/TAO/tests/Cubit/TAO/MT_Cubit/client.cpp
@@ -150,13 +150,14 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
   int i;
   u_int j;
 
+  char * task_id = 0;
+
   ACE_High_Res_Timer timer_;
   ACE_Time_Value delta_t;
 
   timer_.start ();
 
 #if defined (VXWORKS) 
-  char * task_id = 0;
   ctx = 0;
   ACE_NEW_RETURN (task_id,
                   char [32],
@@ -236,7 +237,7 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
   priority = ACE_THR_PRI_FIFO_DEF;
 #elif defined (ACE_WIN32)
   priority = ACE_Sched_Params::priority_max (ACE_SCHED_FIFO,
-                                    ACE_SCOPE_THREAD);
+					     ACE_SCOPE_THREAD);
 #else  /* ! VXWORKS */
   priority = ACE_THR_PRI_FIFO_DEF + 25;
 #endif /* ! ACE_WIN32 */
@@ -248,17 +249,13 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
   if (high_priority_client.activate (THR_BOUND | ACE_SCHED_FIFO,
                                      1,
                                      0,
-#if !defined (VXWORKS) 
-                                     priority) == -1)
-#else
                                      priority,
  				     -1,
  				     0,
  				     0,
  				     0,
  				     0,
- 				     &task_id) == -1)
-#endif /* VXWORKS */
+ 				     (ACE_thread_t*)task_id) == -1)
     ACE_ERROR ((LM_ERROR,
                 "%p; priority is %d\n",
                 "activate failed",
@@ -271,6 +268,9 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
 
   number_of_low_priority_client = ts->thread_count_ - 1;
 
+  // mechanism to distribute the available priorities among the
+  // threads when there are not enough different priorities for all
+  // threads.
   if (ts->use_multiple_priority_ == 1)
     {    
       ACE_Sched_Priority_Iterator priority_iterator (ACE_SCHED_FIFO,
@@ -286,23 +286,24 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
       // 1 priority is exclusive for the high priority client.
       number_of_priorities --;
 
-      // if utilization thread is present, reduce in 1 the available priorities.
+      // if utilization thread is present, reduce in 1 the available
+      // priorities.
       if (ts->use_utilization_test_ == 1)
 	{
 	  number_of_priorities --;
 	}
 
-      // Drop the priority, so that the priority of clients will increase
-      // with increasing client number.
-      for (j = 0; j < number_of_low_priority_client + 1; j++)
+      // Drop the priority, so that the priority of clients will
+      // increase with increasing client number.
+      for (j = 0; j < number_of_low_priority_client; j++)
 	priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
 							priority,
 							ACE_SCOPE_THREAD);
 
-      // if the lowest priority of the "low priority clients" is the minimum, 
-      // and we are running the utilization thread, increment the priority, 
-      // since we don't want the utlization thread and a "low priority thread" 
-      // to have the same priority.
+      // if the lowest priority of the "low priority clients" is the
+      // minimum, and we are running the utilization thread, increment
+      // the priority, since we don't want the utlization thread and a
+      // "low priority thread" to have the same priority.
       if ( priority == ACE_Sched_Params::priority_min (ACE_SCHED_FIFO,
 						       ACE_SCOPE_THREAD) &&
 	   ts->use_utilization_test_ == 1)
@@ -310,10 +311,10 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
                                                     priority,
                                                     ACE_SCOPE_THREAD);
 	
-      // granularity of the assignment of the priorities.  Some OSs have 
-      // fewer levels of priorities than we have threads in our test, so 
-      // with this mechanism we assign priorities to groups of threads when 
-      // there are more threads than priorities.
+      // granularity of the assignment of the priorities.  Some OSs
+      // have fewer levels of priorities than we have threads in our
+      // test, so with this mechanism we assign priorities to groups
+      // of threads when there are more threads than priorities.
       grain = number_of_low_priority_client / number_of_priorities;
       counter = 0;
 
@@ -359,18 +360,13 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
       if (low_priority_client[i - 1]->activate (THR_BOUND | ACE_SCHED_FIFO,
 						1,
 						0,
-#if !defined (VXWORKS) 
-						priority) == -1)
-#else
 						priority,    // These are constructor defaults.     
 						-1,	     //  int grp_id = -1,                   
 						0,	     //  ACE_Task_Base *task = 0,           
 						0,	     //  ACE_hthread_t thread_handles[] = 0,
 						0,	     //  void *stack[] = 0,                 
 						0,	     //  size_t stack_size[] = 0,           
-						&task_id) == -1)
-#endif /* VXWORKS */
-
+						(ACE_thread_t*)task_id) == -1)
         ACE_ERROR ((LM_ERROR,
 	            "%p; priority is %d\n",
 	            "activate failed",
@@ -380,14 +376,15 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
         {    
           counter = (counter + 1) % grain;
           if ( (counter == 0) && 
-               //Just so when we distribute the priorities among the threads, we make sure we don't go overboard.
+               //Just so when we distribute the priorities among the
+               //threads, we make sure we don't go overboard.
                ((number_of_priorities * grain) > (number_of_low_priority_client - (i - 1))) )
-              {
-          	      // Get the next higher priority.
-          	      priority = ACE_Sched_Params::next_priority (ACE_SCHED_FIFO,
-          						     priority,
-          						     ACE_SCOPE_THREAD);
-              }
+	    {
+	      // Get the next higher priority.
+	      priority = ACE_Sched_Params::next_priority (ACE_SCHED_FIFO,
+							  priority,
+							  ACE_SCOPE_THREAD);
+	    }
           
         }          
     } /* end of for () */
@@ -443,12 +440,13 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
 #if defined (VXWORKS) 
   if (ts->context_switch_test_ == 1)
     {
-     fprintf(stderr, "Adding the context switch hook!\n");
-     taskSwitchHookAdd ((FUNCPTR)&switchHook);
+      fprintf(stderr, "Adding the context switch hook!\n");
+      taskSwitchHookAdd ((FUNCPTR)&switchHook);
     }
 #endif
   
-  // Wait for all the client threads to exit (except the utilization thread).
+  // Wait for all the client threads to exit (except the utilization
+  // thread).
   thread_manager->wait ();
 
 #if defined (NO_ACE_QUANTIFY)
@@ -461,9 +459,9 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
   timer_.elapsed_time (delta_t);
 
   if (ts->use_utilization_test_ == 1)
-    // signal the utilization thread to finish with its work..
-    // only if utilization test was specified.  See
-    // description of this variable in header file.
+    // signal the utilization thread to finish with its work..  only
+    // if utilization test was specified.  See description of this
+    // variable in header file.
     {
       util_thread.done_ = 1;
 
@@ -479,7 +477,8 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
 #if defined (ACE_HAS_PRUSAGE_T)
       timer_for_context_switch.stop ();
       timer_for_context_switch.get_rusage (usage);
-      // Add up the voluntary context switches & involuntary context switches
+      // Add up the voluntary context switches & involuntary context
+      // switches
       context_switch = usage.pr_vctx + usage.pr_ictx - context_switch;
       ACE_DEBUG ((LM_DEBUG, 
 		  "Voluntary context switches=%d, Involuntary context switches=%d\n",
@@ -488,7 +487,8 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
 #elif defined (ACE_HAS_GETRUSAGE) && !defined (ACE_WIN32)
       timer_for_context_switch.stop ();
       timer_for_context_switch.get_rusage (usage);
-      // Add up the voluntary context switches & involuntary context switches
+      // Add up the voluntary context switches & involuntary context
+      // switches
       context_switch = usage.ru_nvcsw + usage.ru_nivcsw - context_switch;
       ACE_DEBUG ((LM_DEBUG, 
 		  "Voluntary context switches=%d, Involuntary context switches=%d\n",
@@ -496,88 +496,95 @@ do_priority_inversion_test (ACE_Thread_Manager *thread_manager,
 		  usage.ru_nivcsw));
 #elif defined (VXWORKS) /* ACE_HAS_GETRUSAGE */
       taskSwitchHookDelete ((FUNCPTR)&switchHook);
-       ACE_DEBUG ((LM_DEBUG, 
+      ACE_DEBUG ((LM_DEBUG, 
  		  "Context switches=%d\n",
  		  ctx));
 #endif
     }
 
+  // if running the utilization test, don't report latency nor jitter.
+  if (ts->use_utilization_test_ == 0 && 
+      ts->run_server_utilization_test_ == 0)
+    {
 #if defined (VXWORKS)
-  ACE_DEBUG ((LM_DEBUG, "Test done.\n"
-              "High priority client latency : %f msec, jitter: %f msec\n"
-              "Low priority client latency : %f msec, jitter: %f msec\n",
-              high_priority_client.get_high_priority_latency (),
-              high_priority_client.get_high_priority_jitter (),
-              low_priority_client[0]->get_low_priority_latency (),
-              low_priority_client[0]->get_low_priority_jitter ()));
-  // output the latency values to a file, tab separated, to import it
-  // to Excel to calculate jitter, in the mean time we come up with
-  // the sqrt() function.
-  output_latency (ts);
-
+      ACE_DEBUG ((LM_DEBUG, "Test done.\n"
+		  "High priority client latency : %f msec, jitter: %f msec\n"
+		  "Low priority client latency : %f msec, jitter: %f msec\n",
+		  high_priority_client.get_high_priority_latency (),
+		  high_priority_client.get_high_priority_jitter (),
+		  low_priority_client[0]->get_low_priority_latency (),
+		  low_priority_client[0]->get_low_priority_jitter ()));
+      // output the latency values to a file, tab separated, to import
+      // it to Excel to calculate jitter, in the mean time we come up
+      // with the sqrt() function.
+      output_latency (ts);
 #elif defined (CHORUS)
-  ACE_DEBUG ((LM_DEBUG, 
-	      "Test done.\n"
-	      "High priority client latency : %u usec\n"
-	      "Low priority client latency : %u usec\n",
-	      high_priority_client.get_high_priority_latency (),
-	      low_priority_client[0]->get_low_priority_latency () ));
-
-  // output the latency values to a file, tab separated, to import it
-  // to Excel to calculate jitter, in the mean time we come up with
-  // the sqrt() function.
-  output_latency (ts);
+      ACE_DEBUG ((LM_DEBUG, 
+		  "Test done.\n"
+		  "High priority client latency : %u usec\n"
+		  "Low priority client latency : %u usec\n",
+		  high_priority_client.get_high_priority_latency (),
+		  low_priority_client[0]->get_low_priority_latency () ));
+
+      // output the latency values to a file, tab separated, to import
+      // it to Excel to calculate jitter, in the mean time we come up
+      // with the sqrt() function.
+      output_latency (ts);
 #else /* !CHORUS */
-  ACE_DEBUG ((LM_DEBUG, "Test done.\n"
-              "High priority client latency : %f msec, jitter: %f msec\n"
-              "Low priority client latency : %f msec, jitter: %f msec\n"
-              "# of context switches: %d, context_switch_time: %f msec\n"
-              "total context switch time: %f msec\n",
-              high_priority_client.get_high_priority_latency (),
-              high_priority_client.get_high_priority_jitter (),
-              low_priority_client[0]->get_low_priority_latency (),
-              low_priority_client[0]->get_low_priority_jitter (),
-              context_switch,
-              csw/1000,
-              csw * context_switch/1000 ));
-//  output_latency (ts);
+      ACE_DEBUG ((LM_DEBUG, "Test done.\n"
+		  "High priority client latency : %f msec, jitter: %f msec\n"
+		  "Low priority client latency : %f msec, jitter: %f msec\n"
+		  "# of context switches: %d, context_switch_time: %f msec\n"
+		  "total context switch time: %f msec\n",
+		  high_priority_client.get_high_priority_latency (),
+		  high_priority_client.get_high_priority_jitter (),
+		  low_priority_client[0]->get_low_priority_latency (),
+		  low_priority_client[0]->get_low_priority_jitter (),
+		  context_switch,
+		  csw/1000,
+		  csw * context_switch/1000 ));
+      // output_latency (ts);
 #endif /* !VXWORKS && !CHORUS */
-
+    }
 
   if (ts->use_utilization_test_ == 1)
-  {
-    total_util_task_duration = util_task_duration * util_thread.get_number_of_computations ();
+    {
+      total_util_task_duration = util_task_duration * util_thread.get_number_of_computations ();
     
-    total_latency = (delta_t.sec () * 
-                     ACE_ONE_SECOND_IN_MSECS + 
-                     (double)delta_t.usec () / ACE_ONE_SECOND_IN_MSECS);    
-
-    total_latency_high = total_latency - total_util_task_duration;
-
-    // Calc and print the CPU percentage. I add 0.5 to round to the
-    // nearest integer before casting it to int.
-    ACE_DEBUG ((LM_DEBUG, 
-  		"\t%% ORB Client CPU utilization: %u %%\n"
-  		"\t%% Idle time: %u %%\n",
-  		(u_int) (total_latency_high * 100 / total_latency + 0.5),
-  		(u_int) (total_util_task_duration * 100 / total_latency + 0.5) ));
+      total_latency = (delta_t.sec () * 
+		       ACE_ONE_SECOND_IN_MSECS + 
+		       (double)delta_t.usec () / ACE_ONE_SECOND_IN_MSECS);    
+
+      total_latency_high = total_latency - total_util_task_duration;
+
+      // Calc and print the CPU percentage. I add 0.5 to round to the
+      // nearest integer before casting it to int.
+      ACE_DEBUG ((LM_DEBUG, 
+		  "\t%% ORB Client CPU utilization: %u %%\n"
+		  "\t%% Idle time: %u %%\n\n",
+		  (u_int) (total_latency_high * 100 / total_latency + 0.5),
+		  (u_int) (total_util_task_duration * 100 / total_latency + 0.5) ));
     
-  #if defined (ACE_LACKS_FLOATING_POINT)
-    ACE_DEBUG ((LM_DEBUG,
-  		"(%t) utilization task performed %u computations\n",
-  		util_thread.get_number_of_computations ()));
-    ACE_DEBUG ((LM_DEBUG,
-  		"(%t) utilization computation time is %u usecs\n", 
-  		util_task_duration));
-  #else
-    ACE_DEBUG ((LM_DEBUG,
-  		"(%t) utilization task performed %g computations\n",
-  		util_thread.get_number_of_computations ()));
-    ACE_DEBUG ((LM_DEBUG,
-  		"(%t) utilization computation time is %f msecs\n", 
-  		util_task_duration));
-  #endif /* ! ACE_LACKS_FLOATING_POINT */
-  }
+      ACE_DEBUG ((LM_DEBUG,
+		  "(%t) UTILIZATION task performed \t%u computations\n"
+		  "(%t) CLIENT task performed \t\t%u CORBA calls\n"
+		  "\t Ratio of computations to CORBA calls is %u.%u:1\n\n",
+		  util_thread.get_number_of_computations (),
+		  ts->loop_count_,
+		  util_thread.get_number_of_computations () / ts->loop_count_,
+		  (util_thread.get_number_of_computations () % ts->loop_count_) * 100 / ts->loop_count_
+		  ));
+
+#if defined (ACE_LACKS_FLOATING_POINT)
+      ACE_DEBUG ((LM_DEBUG,
+		  "(%t) utilization computation time is %u usecs\n", 
+		  util_task_duration));
+#else
+      ACE_DEBUG ((LM_DEBUG,
+		  "(%t) utilization computation time is %f msecs\n", 
+		  util_task_duration));
+#endif /* ! ACE_LACKS_FLOATING_POINT */
+    }
 
 #if defined (VXWORKS)
   delete task_id;
@@ -608,25 +615,20 @@ do_thread_per_rate_test (ACE_Thread_Manager *thread_manager,
   ACE_DEBUG ((LM_DEBUG,
 	      "Creating 20 Hz client with priority %d\n",
 	      priority));
+
   if (CB_20Hz_client.activate (THR_BOUND | ACE_SCHED_FIFO, 1, 1, priority) == -1)
-    ACE_ERROR ((LM_ERROR,
-		"%p\n"
-		"activate failed"));
+    ACE_ERROR ((LM_ERROR, "(%P|%t) errno = %p: activate failed\n"));
 
   priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
 						  priority,
 						  ACE_SCOPE_THREAD);
 
-  priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
-						  priority,
-						  ACE_SCOPE_THREAD);
   ACE_DEBUG ((LM_DEBUG,
 	      "Creating 10 Hz client with priority %d\n",
 	      priority));
+
   if (CB_10Hz_client.activate (THR_BOUND | ACE_SCHED_FIFO, 1, 1, priority) == -1)
-    ACE_ERROR ((LM_ERROR,
-		"%p\n"
-		"activate failed"));
+    ACE_ERROR ((LM_ERROR, "(%P|%t) errno = %p: activate failed\n"));
 
   priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
 						  priority,
@@ -634,10 +636,9 @@ do_thread_per_rate_test (ACE_Thread_Manager *thread_manager,
   ACE_DEBUG ((LM_DEBUG,
 	      "Creating 5 Hz client with priority %d\n",
 	      priority));
+
   if (CB_5Hz_client.activate (THR_BOUND | ACE_SCHED_FIFO, 1, 1, priority) == -1)
-    ACE_ERROR ((LM_ERROR,
-		"%p\n"
-		"activate failed"));
+    ACE_ERROR ((LM_ERROR, "(%P|%t) errno = %p: activate failed\n"));
 
   priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
 						  priority,
@@ -645,10 +646,9 @@ do_thread_per_rate_test (ACE_Thread_Manager *thread_manager,
   ACE_DEBUG ((LM_DEBUG,
 	      "Creating 1 Hz client with priority %d\n",
 	      priority));
+
   if (CB_1Hz_client.activate (THR_BOUND | ACE_SCHED_FIFO, 1, 1, priority) == -1)
-    ACE_ERROR ((LM_ERROR,
-		"%p\n"
-		"activate failed"));
+    ACE_ERROR ((LM_ERROR, "(%P|%t) errno = %p: activate failed\n"));
 
   // Wait for all the threads to exit.
   thread_manager->wait ();
diff --git a/TAO/tests/Cubit/TAO/MT_Cubit/cubit_i.cpp b/TAO/tests/Cubit/TAO/MT_Cubit/cubit_i.cpp
index 4ed26dce816..1c4d2799ac4 100644
--- a/TAO/tests/Cubit/TAO/MT_Cubit/cubit_i.cpp
+++ b/TAO/tests/Cubit/TAO/MT_Cubit/cubit_i.cpp
@@ -25,16 +25,7 @@ Cubit_i::cube_octet (CORBA::Octet o,
       ts_->barrier_->wait ();
     }
   
-//   if (ts_->run_server_utilization_test_ == 0)
-//     {
-//       ACE_DEBUG ((LM_DEBUG,
-// 		  "(%t) )))))))) "
-// 		  "test STARTED at %D\n"));
-      
-//       ts_->run_server_utilization_test_ = 1;
-//       this->util_started_ = 1;
-//       this->ts_->timer_.start ();
-//     }
+  ts_->loop_count_++;
 
   return (CORBA::Octet) (o * o * o);
 }
@@ -75,6 +66,9 @@ void Cubit_i::shutdown (CORBA::Environment &)
 {
   ACE_DEBUG ((LM_DEBUG, 
 	      "(%t) Calling TAO_ORB_Core_instance ()->orb ()->shutdown ()\n"));
+
   TAO_ORB_Core_instance ()->orb ()->shutdown ();
+
+  // this is to signal the utilization thread to exit its loop.
   ts_->utilization_task_started_ = 0;
 }
diff --git a/TAO/tests/Cubit/TAO/MT_Cubit/server.cpp b/TAO/tests/Cubit/TAO/MT_Cubit/server.cpp
index 2a9e62169db..0e0ae8f8824 100644
--- a/TAO/tests/Cubit/TAO/MT_Cubit/server.cpp
+++ b/TAO/tests/Cubit/TAO/MT_Cubit/server.cpp
@@ -543,8 +543,8 @@ start_servants (ACE_Thread_Manager *serv_thr_mgr, ACE_Barrier &start_barrier, Ta
                                     0,
                                     priority) == -1)
     {
-      ACE_ERROR ((LM_ERROR, "(%P|%t) %p\n",
-                  "high_priority_task->activate"));
+      ACE_ERROR ((LM_ERROR, "(%P|%t) %p\n"
+                  "\thigh_priority_task->activate failed"));
     }
 
   // Create an array to hold pointers to the low priority tasks.
@@ -580,7 +580,7 @@ start_servants (ACE_Thread_Manager *serv_thr_mgr, ACE_Barrier &start_barrier, Ta
 
       // Drop the priority, so that the priority of clients will increase
       // with increasing client number.
-      for (i = 0; i < number_of_low_priority_servants + 1; i++)
+      for (i = 0; i < number_of_low_priority_servants; i++)
         priority = ACE_Sched_Params::previous_priority (ACE_SCHED_FIFO,
                                                         priority,
                                                         ACE_SCOPE_THREAD);
@@ -643,8 +643,8 @@ start_servants (ACE_Thread_Manager *serv_thr_mgr, ACE_Barrier &start_barrier, Ta
                                            0,
                                            priority) == -1)
         {
-          ACE_ERROR ((LM_ERROR, "(%P|%t; %p\n",
-                      "low_priority_task[i]->activate"));
+          ACE_ERROR ((LM_ERROR, "(%P|%t) %p\n"
+                      "\tlow_priority_task[i]->activate"));
         }
 
       ACE_DEBUG ((LM_DEBUG,
@@ -657,7 +657,8 @@ start_servants (ACE_Thread_Manager *serv_thr_mgr, ACE_Barrier &start_barrier, Ta
         {
           counter = (counter + 1) % grain;
           if ( (counter == 0) &&
-               //Just so when we distribute the priorities among the threads, we make sure we don't go overboard.
+               //Just so when we distribute the priorities among the
+               //threads, we make sure we don't go overboard.
                ((number_of_priorities * grain) > (number_of_low_priority_servants - (i - 1))) )
               {
                       // Get the next higher priority.
@@ -724,8 +725,8 @@ start_utilization (ACE_Thread_Manager *util_thr_mgr, Task_State *ts)
 			   0,
 			   priority) == -1)
     {
-      ACE_ERROR ((LM_ERROR, "(%P|%t) %p\n",
-                  "util_task->activate"));
+      ACE_ERROR ((LM_ERROR, "(%P|%t) %p\n"
+                  "\tutil_task->activate failed"));
     }
 
   return util_task;
@@ -784,7 +785,10 @@ main (int argc, char *argv[])
   Task_State ts ( _argc, _argv);
 
   if (run_utilization_test == 1)
-    ts.run_server_utilization_test_ = 1;
+    {
+      ts.run_server_utilization_test_ = 1;
+      ts.loop_count_ = 0;
+     }
 
   Util_Thread * util_task = 0;
 
@@ -873,11 +877,22 @@ main (int argc, char *argv[])
 			      
       total_latency_servants = total_latency - total_util_task_duration;
     
+      ACE_DEBUG ((LM_DEBUG, 
+		  "-------------------------- Stats -------------------------------\n"));
+
       ACE_DEBUG ((LM_DEBUG,
-		  "(%t) utilization task performed %g computations\n"
-		  "(%t) each computation had a duration of %f msecs\n"
-		  "(%t) elapsed time is %f msecs\n",
+		  "(%t) UTILIZATION task performed \t%u computations\n"
+		  "(%t) SERVANT task serviced \t\t%u CORBA calls\n"
+		  "\t Ratio of computations to CORBA calls is %u.%u:1\n\n",
 		  util_task->get_number_of_computations (),
+		  ts.loop_count_,
+		  util_task->get_number_of_computations () / ts.loop_count_,
+		  (util_task->get_number_of_computations () % ts.loop_count_) * 100 / ts.loop_count_
+		  ));
+
+      ACE_DEBUG ((LM_DEBUG,
+		  "(%t) Each computation had a duration of %f msecs\n"
+		  "(%t) Total elapsed time of test is %f msecs\n",
 		  util_task_duration / 1000,
 		  total_latency / 1000));