Merge branch 'privatestaging' into p2p

Conflicts:
	include/hcc_detail/hip_hcc.h
	src/hip_hcc.cpp

include/hcc_detail/hip_hcc.h

@@ -451,7 +451,7 @@ private:
     void                        waitCopy(LockedAccessor_StreamCrit_t &crit, ihipSignal_t *signal);
 
     // The unsigned return is hipMemcpyKind
-    unsigned resolveMemcpyDirection(bool srcInDeviceMem, bool dstInDeviceMem);
+    unsigned resolveMemcpyDirection(bool srcTracked, bool dstTracked, bool srcInDeviceMem, bool dstInDeviceMem);
     void setAsyncCopyAgents(unsigned kind, ihipCommand_t *commandType, hsa_agent_t *srcAgent, hsa_agent_t *dstAgent);
 
     unsigned                    _device_index;       // index into the g_device array 

include/hcc_detail/hip_runtime.h

@@ -401,6 +401,13 @@ __device__ int __all(  int input);
 __device__ int __any( int input);
 __device__  unsigned long long int __ballot( int input);
 
+// __ldg function
+template <typename T>
+__device__ __forceinline__ T __ldg( const T * addr)
+{
+	return *addr;
+}
+
 // warp shuffle functions
 #ifdef __cplusplus