diff --git a/llvm/projects/hpvm-tensor-rt/CMakeLists_cpu.txt b/llvm/projects/hpvm-tensor-rt/CMakeLists_cpu.txt
new file mode 100644
index 0000000000000000000000000000000000000000..cff0129c2aa02b9776ed7bba8e92029d2c2560e8
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/CMakeLists_cpu.txt
@@ -0,0 +1,19 @@
+cmake_minimum_required (VERSION 2.6)
+project (approxhpvm-tensorRt-cpu)
+
+
+# Addresses a bug where code is not compiled as C++11 in non-CUDA code and older g++ versions
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 " )
+ 
+
+# Adding new rule for building a cuDNN runtime library
+add_library(tensor_cpu_runtime tensor_runtime/src/tensor_cpu_runtime.cc)
+target_link_libraries(tensor_cpu_runtime)
+
+
+#**** CPU sources
+add_executable(fc2_cpu  dnn_sources/src/fc2_cpu.cc)
+target_link_libraries(fc2_cpu  tensor_cpu_runtime)
+
+
+
diff --git a/llvm/projects/hpvm-tensor-rt/dnn_sources/include/utils_cpu.h b/llvm/projects/hpvm-tensor-rt/dnn_sources/include/utils_cpu.h
index f0358556022d81e914af79f4080f6594aa61c924..c5981d752c7562e1a12fdf41095813e8ac21ec17 100644
--- a/llvm/projects/hpvm-tensor-rt/dnn_sources/include/utils_cpu.h
+++ b/llvm/projects/hpvm-tensor-rt/dnn_sources/include/utils_cpu.h
@@ -7,7 +7,7 @@
 #include <sstream>
 #include <vector>
 #include <bits/stdc++.h>
-#include "../../tensor_runtime/include/tensor.h"
+#include "../../tensor_runtime/include/tensor_cpu.h"
 #include "types.h"
 #include <cmath>
 
@@ -88,10 +88,10 @@ struct Tensor* readTrainedWeightsCPU(const char* file_name, int data_type,
   printf("size in bytes = %lu, bytes read = %lu \n", size_in_bytes, bytes_read);
 
   fclose(file);
-  
+
   
   struct Tensor* weights = (struct Tensor*) create4DTensorCPU(data_type, nchw, dim1_size, dim2_size,
-					                   dim3_size, dim4_size);
+					                      dim3_size, dim4_size);
   
   initTensorData(weights, tensor_data, size_in_bytes);
   //compareValues(weights, tensor_data, num_elems);
diff --git a/llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_cpu.h b/llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_cpu.h
new file mode 100644
index 0000000000000000000000000000000000000000..c812a516ae785de6ae36017a904f1a0dace5ac4f
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_cpu.h
@@ -0,0 +1,25 @@
+
+
+#ifndef TENSOR_HEADER
+#define TENSOR_HEADER
+
+
+struct Dimension{
+  int num_dims;
+  size_t* dim_sizes;
+};
+
+
+struct Tensor{
+  int data_type;
+  int data_format;
+  void* host_data;
+  void* gpu_data; // The pointers should not be device specific per se - TODO: Better design needed
+  size_t num_elems; // Total elements
+  size_t size_in_bytes; // Total size in bytes
+  struct Dimension dims;
+};
+
+
+#endif
+
diff --git a/llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_cpu_runtime.h b/llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_cpu_runtime.h
index dfa9511055f04d703696d65f5fd24094bc202f3e..cc84094111b2deb945cc8e65239a7624049f836f 100644
--- a/llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_cpu_runtime.h
+++ b/llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_cpu_runtime.h
@@ -4,7 +4,6 @@
 #include <cmath>
 #include <memory>
 #include <string>
-//#include "runtime_types.h"
 
 
 #ifndef CUDNN_HEADER
@@ -41,10 +40,12 @@ extern "C"{
 			 int vertical_stride, int horizontal_stride);
 
   void* tensorGemmCPU(void* lhs, void* rhs);
-  // NOTE: In place operation
+
   void* tensorAddCPU(void* x, void* bias);
-  // NOTE: In-place operation
+
   void* tensorReluCPU(void* input);
+
+  void* tensorRelu2CPU(void* input);
   
   void* tensorTanhCPU(void* input);
   
diff --git a/llvm/projects/hpvm-tensor-rt/tensor_runtime/src/tensor_cpu_runtime.cc b/llvm/projects/hpvm-tensor-rt/tensor_runtime/src/tensor_cpu_runtime.cc
index 336c98b767d416e98de6e848d395e80f148d118a..8d2591372a65486e92512b620ef931bed74f0cf3 100644
--- a/llvm/projects/hpvm-tensor-rt/tensor_runtime/src/tensor_cpu_runtime.cc
+++ b/llvm/projects/hpvm-tensor-rt/tensor_runtime/src/tensor_cpu_runtime.cc
@@ -23,7 +23,8 @@
 
 // Tensor runtime header files
 #include "../include/tensor_cpu_runtime.h"
-#include "../include/tensor.h"
+#include "../include/tensor_cpu.h"
+
 
 
 void llvm_hpvm_initTensorRt(int gpuid){