diff --git a/Eigen/src/Core/GenericPacketMath.h b/Eigen/src/Core/GenericPacketMath.h
index 9618b489c..dba1d5a42 100644
--- a/Eigen/src/Core/GenericPacketMath.h
+++ b/Eigen/src/Core/GenericPacketMath.h
@@ -559,7 +559,8 @@ pblend(const Selector<unpacket_traits<Packet>::size>& ifPacket, const Packet& th
 }
 
 /** \internal \returns \a a with last coefficients replaced by the scalar b */
-template<typename Packet> EIGEN_DEVICE_FUNC Packet pinsertlast(const Packet& a, typename unpacket_traits<Packet>::type b)
+template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
+pinsertlast(const Packet& a, typename unpacket_traits<Packet>::type b)
 {
   // Default implementation based on pblend.
   // It must be specialized for higher performance.
diff --git a/Eigen/src/Core/arch/AVX/PacketMath.h b/Eigen/src/Core/arch/AVX/PacketMath.h
index 8fb844982..0efdde722 100644
--- a/Eigen/src/Core/arch/AVX/PacketMath.h
+++ b/Eigen/src/Core/arch/AVX/PacketMath.h
@@ -604,12 +604,12 @@ template<> EIGEN_STRONG_INLINE Packet4d pblend(const Selector<4>& ifPacket, cons
   return _mm256_blendv_pd(thenPacket, elsePacket, false_mask);
 }
 
-template<> EIGEN_DEVICE_FUNC Packet8f pinsertlast(const Packet8f& a, float b)
+template<> EIGEN_STRONG_INLINE Packet8f pinsertlast(const Packet8f& a, float b)
 {
   return _mm256_blend_ps(a,pset1<Packet8f>(b),(1<<7));
 }
 
-template<> EIGEN_DEVICE_FUNC Packet4d pinsertlast(const Packet4d& a, double b)
+template<> EIGEN_STRONG_INLINE Packet4d pinsertlast(const Packet4d& a, double b)
 {
   return _mm256_blend_pd(a,pset1<Packet4d>(b),(1<<3));
 }
diff --git a/Eigen/src/Core/arch/SSE/PacketMath.h b/Eigen/src/Core/arch/SSE/PacketMath.h
index 13cb51b8d..b519e3748 100755
--- a/Eigen/src/Core/arch/SSE/PacketMath.h
+++ b/Eigen/src/Core/arch/SSE/PacketMath.h
@@ -818,7 +818,7 @@ template<> EIGEN_STRONG_INLINE Packet2d pblend(const Selector<2>& ifPacket, cons
 #endif
 }
 
-template<> EIGEN_DEVICE_FUNC Packet4f pinsertlast(const Packet4f& a, float b)
+template<> EIGEN_STRONG_INLINE Packet4f pinsertlast(const Packet4f& a, float b)
 {
 #ifdef EIGEN_VECTORIZE_SSE4_1
   return _mm_blend_ps(a,pset1<Packet4f>(b),(1<<3));
@@ -828,7 +828,7 @@ template<> EIGEN_DEVICE_FUNC Packet4f pinsertlast(const Packet4f& a, float b)
 #endif
 }
 
-template<> EIGEN_DEVICE_FUNC Packet2d pinsertlast(const Packet2d& a, double b)
+template<> EIGEN_STRONG_INLINE Packet2d pinsertlast(const Packet2d& a, double b)
 {
 #ifdef EIGEN_VECTORIZE_SSE4_1
   return _mm_blend_pd(a,pset1<Packet2d>(b),(1<<1));