diff --git a/Grid/tensors/Tensor_class.h b/Grid/tensors/Tensor_class.h
index 1b182e32..b48cefcc 100644
--- a/Grid/tensors/Tensor_class.h
+++ b/Grid/tensors/Tensor_class.h
@@ -64,6 +64,8 @@ public:
   
   enum { TensorLevel = GridTypeMapper<vtype>::TensorLevel + 1 };
 
+  static constexpr accelerator_inline int Nsimd(void) { return sizeof(vector_type)/sizeof(scalar_type); } 
+
   // Scalar no action
   accelerator iScalar() = default;
 
@@ -194,6 +196,8 @@ public:
 
   // get double precision version
   typedef iVector<typename GridTypeMapper<vtype>::DoublePrecision, N> DoublePrecision;
+
+  static constexpr accelerator_inline int Nsimd(void) { return sizeof(vector_type)/sizeof(scalar_type); } 
   
   template <class T, typename std::enable_if<!isGridTensor<T>::value, T>::type * = nullptr>
   accelerator_inline auto operator=(T arg) -> iVector<vtype, N> {
@@ -304,6 +308,8 @@ public:
 
   enum { TensorLevel = GridTypeMapper<vtype>::TensorLevel + 1 };
 
+  static constexpr accelerator_inline int Nsimd(void) { return sizeof(vector_type)/sizeof(scalar_type); } 
+
   accelerator_inline iMatrix(const Zero &z) { zeroit(*this); };
   accelerator iMatrix() = default;
 
diff --git a/Grid/tensors/Tensors.h b/Grid/tensors/Tensors.h
index c40700c1..43072a24 100644
--- a/Grid/tensors/Tensors.h
+++ b/Grid/tensors/Tensors.h
@@ -48,4 +48,7 @@ Author: neo <cossu@post.kek.jp>
 #include <Grid/tensors/Tensor_extract_merge.h>
 #include <Grid/tensors/Tensor_logical.h>
 
+// Maybe expect different thread on each SIMD lane
+#include <Grid/tensors/Tensor_SIMT.h>
+
 #endif