Don't create a marker for start event in hipModuleLaunchKernel

And also don't optimize the case where start==stop event to compute
elapsed time since the command can be a NDRange one.
HIP directed test will need to be fixed for that.

Change-Id: I64fadd6ab8ab1a490e7a2b7165a591df5a5cf3a2

hipamd/vdi/hip_event.cpp

@@ -86,12 +86,8 @@ hipError_t Event::elapsedTime(Event& eStop, float& ms) {
     return hipErrorNotReady;
   }
 
-  if (event_ != eStop.event_) {
-    ms = static_cast<float>(static_cast<int64_t>(eStop.event_->profilingInfo().end_ -
-                            event_->profilingInfo().start_))/1000000.f;
-  } else {
-    ms = 0.f;
-  }
+  ms = static_cast<float>(static_cast<int64_t>(eStop.event_->profilingInfo().end_ -
+                          event_->profilingInfo().start_))/1000000.f;
 
   return hipSuccess;
 }

hipamd/vdi/hip_module.cpp

@@ -324,12 +324,6 @@ hipError_t ihipModuleLaunchKernel(hipFunction_t f,
     }
   }
 
-  if(startEvent != nullptr) {
-    amd::Command* startCommand = new hip::TimerMarker(*queue);
-    startCommand->enqueue();
-    eStart->addMarker(queue, startCommand);
-  }
-
   amd::NDRangeKernelCommand* command = new amd::NDRangeKernelCommand(
     *queue, waitList, *kernel, ndrange, sharedMemBytes,
     params, gridId, numGrids, prevGridSum, allGridSum, firstDevice);
@@ -345,11 +339,14 @@ hipError_t ihipModuleLaunchKernel(hipFunction_t f,
 
   command->enqueue();
 
+  if(startEvent != nullptr) {
+    eStart->addMarker(queue, command);
+    command->retain();
+  }
   if(stopEvent != nullptr) {
     eStop->addMarker(queue, command);
     command->retain();
   }
-
   command->release();
 
   return hipSuccess;