diff --git a/Eigen/src/Core/MathFunctions.h b/Eigen/src/Core/MathFunctions.h
index 36a6eb650..a7e75b4a9 100644
--- a/Eigen/src/Core/MathFunctions.h
+++ b/Eigen/src/Core/MathFunctions.h
@@ -395,7 +395,44 @@ inline NewType cast(const OldType& x)
 }
 
 /****************************************************************************
-* Implementation of logp1                                                *
+* Implementation of round                                                   *
+****************************************************************************/
+// In C++11 we can specialize round_impl for real Scalars
+// Let's be conservative and enable the default C++11 implementation only if we are sure it exists
+#if (__cplusplus >= 201103L) && (EIGEN_COMP_GNUC_STRICT || EIGEN_COMP_CLANG || EIGEN_COMP_MSVC || EIGEN_COMP_ICC)  \
+    && (EIGEN_ARCH_i386_OR_x86_64) && (EIGEN_OS_GNULINUX || EIGEN_OS_WIN_STRICT || EIGEN_OS_MAC)
+  template<typename Scalar>
+  struct round_impl {
+    static inline Scalar run(const Scalar& x)
+    {
+      EIGEN_STATIC_ASSERT((!NumTraits<Scalar>::IsComplex), NUMERIC_TYPE_MUST_BE_REAL)
+      using std::round;
+      return round(x);
+    }
+  };
+// No C++11, use our own implementation
+#else
+  template<typename Scalar>
+  struct round_impl
+  {
+    static inline Scalar run(const Scalar& x)
+    {
+      EIGEN_STATIC_ASSERT((!NumTraits<Scalar>::IsComplex), NUMERIC_TYPE_MUST_BE_REAL)
+      using std::floor;
+      using std::ceil;
+      return (x > 0.0) ? floor(x + 0.5) : ceil(x - 0.5);
+    }
+  };
+#endif
+
+template<typename Scalar>
+struct round_retval
+{
+  typedef Scalar type;
+};
+
+/****************************************************************************
+* Implementation of log1p                                                   *
 ****************************************************************************/
 template<typename Scalar, bool isComplex = NumTraits<Scalar>::IsComplex >
 struct log1p_impl
@@ -775,13 +812,11 @@ bool (isInf)(const std::complex<T>& x)
   return isinf(real(x)) || isinf(imag(x));
 }
 
-template<typename T>
+template<typename Scalar>
 EIGEN_DEVICE_FUNC
-T (round)(const T& x)
+inline EIGEN_MATHFUNC_RETVAL(round, Scalar) round(const Scalar& x)
 {
-  using std::floor;
-  using std::ceil;
-  return (x > 0.0) ? floor(x + 0.5) : ceil(x - 0.5);
+  return EIGEN_MATHFUNC_IMPL(round, Scalar)::run(x);
 }
 
 template<typename T>