diff --git a/src/CUDAStream.cu b/src/CUDAStream.cu
new file mode 100644
index 0000000..d95f9a1
--- /dev/null
+++ b/src/CUDAStream.cu
@@ -0,0 +1,29 @@
+
+#include "CUDAStream.h"
+
+template <class T>
+void CUDAStream<T>::copy()
+{
+  return;
+}
+
+template <class T>
+void CUDAStream<T>::mul()
+{
+  return;
+}
+
+template <class T>
+void CUDAStream<T>::add()
+{
+  return;
+}
+
+template <class T>
+void CUDAStream<T>::triad()
+{
+  return;
+}
+
+template class CUDAStream<float>;
+template class CUDAStream<double>;
diff --git a/src/CUDAStream.h b/src/CUDAStream.h
new file mode 100644
index 0000000..bfea7ed
--- /dev/null
+++ b/src/CUDAStream.h
@@ -0,0 +1,13 @@
+
+
+#include "Stream.h"
+
+template <class T>
+class CUDAStream : public Stream<T>
+{
+  public:
+    void copy();
+    void add();
+    void mul();
+    void triad();
+};
diff --git a/src/Stream.h b/src/Stream.h
new file mode 100644
index 0000000..e283437
--- /dev/null
+++ b/src/Stream.h
@@ -0,0 +1,23 @@
+
+#pragma once
+
+#include <vector>
+
+template <class T>
+class Stream
+{
+  public:
+      // Kernels
+      // These must be blocking calls
+      virtual void copy() = 0;
+      virtual void mul() = 0;
+      virtual void add() = 0;
+      virtual void triad() = 0;
+
+
+      // Implementation specific device functions
+      static std::vector<int> getDeviceList();
+      static std::vector<int> getDeviceName();
+      static std::vector<int> getDeviceDriver();
+
+};
diff --git a/src/common.h b/src/common.h
new file mode 100644
index 0000000..02d43a2
--- /dev/null
+++ b/src/common.h
@@ -0,0 +1,2 @@
+
+#define VERSION_STRING "2.0"
diff --git a/src/main.cpp b/src/main.cpp
new file mode 100644
index 0000000..9e22b89
--- /dev/null
+++ b/src/main.cpp
@@ -0,0 +1,22 @@
+
+#include <iostream>
+
+#include "common.h"
+#include "Stream.h"
+#include "CUDAStream.h"
+
+
+
+int main(int argc, char *argv[])
+{
+  std::cout
+    << "GPU-STREAM" << std::endl
+    << "Version:" << VERSION_STRING << std::endl
+    << "Implementation: " << std::endl;
+
+  Stream<double> *stream;
+  stream = new CUDAStream<double>();
+
+  delete[] stream;
+
+}