diff --git a/CMakeLists.txt b/CMakeLists.txt
index 5551c54..c2bc5d7 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -58,24 +58,6 @@ else()
   message("Skipping HIP...")
 endif()
 
-#-------------------------------------------------------------------------------
-# HCC
-#-------------------------------------------------------------------------------
-find_program(HCC_BINARY hcc HINTS ${HCC_PATH}/bin PATHS ${HCC_PATH}/bin)
-if(EXISTS ${HCC_BINARY})
-  #can the following be tied to the target only?
-  set(CMAKE_CXX_COMPILER ${HCC_BINARY})
-  # list(APPEND CMAKE_CXX_FLAGS -hc -I /usr/include/c++/v1 -std=c++11 -stdlib=libc++ -I/opt/rocm/hcc-lc/include)
-  # list(APPEND CMAKE_EXE_LINKER_FLAGS -hc -std=c++11 -L/opt/rocm/hcc-lc/lib -Wl,--rpath=/opt/rocm/hcc-lc/lib -lc++ -lc++abi -ldl -lpthread -Wl,--whole-archive -lmcwamp -Wl,--no-whole-archive  -mcmodel=small )
-  add_executable(gpu-stream-hc main.cpp HCStream.cpp)
-  set_target_properties(gpu-stream-hc PROPERTIES
-    COMPILE_FLAGS "-hc -I /usr/include/c++/v1 -std=c++11 -stdlib=libc++ -I/opt/rocm/hcc-lc/include"
-    LINK_FLAGS "-hc -std=c++11 -L/opt/rocm/hcc-lc/lib -Wl,--rpath=/opt/rocm/hcc-lc/lib -lc++ -lc++abi -ldl -lpthread -Wl,--whole-archive -lmcwamp -Wl,--no-whole-archive  -mcmodel=small"
-    )
-  target_compile_definitions(gpu-stream-hc PUBLIC HC)
-else()
-  message("Skipping HC...")
-endif()
 
 
 #-------------------------------------------------------------------------------
@@ -110,6 +92,9 @@ else ()
   message("Skipping OpenCL...")
 endif ()
 
+
+
+
 #-------------------------------------------------------------------------------
 # OpenACC
 #-------------------------------------------------------------------------------
@@ -187,6 +172,37 @@ else()
   message("Skipping Kokkos... (use -DKOKKOS_PATH=/path/to/kokkos to opt in)")
 endif()
 
+#-------------------------------------------------------------------------------
+# HCC
+#-------------------------------------------------------------------------------
+find_program(HCC_BINARY hcc HINTS ${HCC_PATH}/bin PATHS ${HCC_PATH}/bin)
+if(EXISTS ${HCC_BINARY})
+  #can the following be tied to the target only?
+  set(OLD_CMAKE_CXX_COMPILER ${CMAKE_CXX_COMPILER})
+  set(CMAKE_CXX_COMPILER ${HCC_BINARY})
+  
+  # list(APPEND CMAKE_CXX_FLAGS -hc -I /usr/include/c++/v1 -std=c++11 -stdlib=libc++ -I/opt/rocm/hcc-lc/include)
+  # list(APPEND CMAKE_EXE_LINKER_FLAGS -hc -std=c++11 -L/opt/rocm/hcc-lc/lib -Wl,--rpath=/opt/rocm/hcc-lc/lib -lc++ -lc++abi -ldl -lpthread -Wl,--whole-archive -lmcwamp -Wl,--no-whole-archive  -mcmodel=small )
+  add_executable(gpu-stream-hc main.cpp HCStream.cpp)
+  set_target_properties(gpu-stream-hc PROPERTIES
+    COMPILE_FLAGS "-hc -I/usr/include/c++/v1 -std=c++11 -stdlib=libc++ -I/opt/rocm/hcc-lc/include"
+    LINK_FLAGS "-hc -std=c++11 -L/opt/rocm/hcc-lc/lib -Wl,--rpath=/opt/rocm/hcc-lc/lib -lc++ -lc++abi -ldl -lpthread -Wl,--whole-archive -lmcwamp -Wl,--no-whole-archive  -mcmodel=small"
+    )
+  message("OpenCL magic: ${OpenCL_LIBRARY} ${OpenCL_INCLUDE_DIR}")
+  set_property(TARGET gpu-stream-ocl APPEND PROPERTY COMPILE_FLAGS "-I/opt/rocm/opencl/include/ -stdlib=libc++ -I/opt/rocm/hcc-lc/include")
+  set_property(TARGET gpu-stream-ocl APPEND PROPERTY LINK_FLAGS "-L/opt/rocm/opencl/lib/x86_64 -std=c++11 -L/opt/rocm/hcc-lc/lib -Wl,--rpath=/opt/rocm/hcc-lc/lib -lc++ -lc++abi ")
+  # set_target_properties(gpu-stream-ocl PROPERTIES
+  #   # COMPILE_FLAGS "-I/opt/rocm/opencl/include/opencl1.2"
+  #   LINK_FLAGS "-I /opt/rocm/opencl/include/opencl1.2 -L /opt/rocm/opencl/lib/x86_64 -lOpenCL -std=c++11"
+  #   )
+  target_compile_definitions(gpu-stream-hc PUBLIC HC)
+  if(TARGET gpu-stream-ocl)
+    unset(gpu-stream-ocl)
+  endif()
+else()
+  message("Skipping HC...")
+endif()
+
 #-------------------------------------------------------------------------------
 # SYCL
 #-------------------------------------------------------------------------------
diff --git a/HCStream.cpp b/HCStream.cpp
index 0c55af3..8cea1dc 100644
--- a/HCStream.cpp
+++ b/HCStream.cpp
@@ -72,7 +72,9 @@ HCStream<T>::HCStream(const unsigned int ARRAY_SIZE, const int device_index):
   // // Set device
   std::vector<hc::accelerator> accs = hc::accelerator::get_all();
   auto current = accs[device_index];
-  
+
+  hc::accelerator::set_default(current.get_device_path());
+
   std::cout << "Using HC device " << getDeviceName(current) << std::endl;
   
   // // The array size must be divisible by TBSIZE for kernel launches
@@ -142,11 +144,15 @@ void HCStream<T>::read_arrays(std::vector<T>& a, std::vector<T>& b, std::vector<
 template <class T>
 void HCStream<T>::copy()
 {
+
+  hc::array<T,1>& device_a = this->d_a;
+  hc::array<T,1>& device_c = this->d_c;
+
   try{
-  // launch a GPU kernel to compute the saxpy in parallel 
+  // launch a GPU kernel to compute the saxpy in parallel
     hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size)
-								, [&](hc::index<1> i) [[hc]] {
-								  d_c[i] = d_a[i];
+								, [&](hc::index<1> index) [[hc]] {
+                                  device_c[index] = device_a[index];
 								});
     future_kernel.wait();
   }
@@ -160,11 +166,14 @@ template <class T>
 void HCStream<T>::mul()
 {
   const T scalar = 0.3;
+  hc::array<T,1>& device_b = this->d_b;
+  hc::array<T,1>& device_c = this->d_c;
+
   try{
   // launch a GPU kernel to compute the saxpy in parallel 
     hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size)
 								, [&](hc::index<1> i) [[hc]] {
-								  d_b[i] = scalar*d_c[i];
+                                  device_b[i] = scalar*device_c[i];
 								});
     future_kernel.wait();
   }
@@ -177,11 +186,16 @@ void HCStream<T>::mul()
 template <class T>
 void HCStream<T>::add()
 {
+
+  hc::array<T,1>& device_a = this->d_a;
+  hc::array<T,1>& device_b = this->d_b;
+  hc::array<T,1>& device_c = this->d_c;
+
   try{
     // launch a GPU kernel to compute the saxpy in parallel 
     hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size)
 								, [&](hc::index<1> i) [[hc]] {
-								  d_c[i] = d_a[i]+d_b[i];
+                                  device_c[i] = device_a[i]+device_b[i];
 								});
     future_kernel.wait();
   }
@@ -195,11 +209,15 @@ template <class T>
 void HCStream<T>::triad()
 {
   const T scalar = 0.3;
+  hc::array<T,1>& device_a = this->d_a;
+  hc::array<T,1>& device_b = this->d_b;
+  hc::array<T,1>& device_c = this->d_c;
+
   try{
     // launch a GPU kernel to compute the saxpy in parallel 
     hc::completion_future future_kernel = hc::parallel_for_each(hc::extent<1>(array_size)
 								, [&](hc::index<1> i) [[hc]] {
-								  d_a[i] = d_b[i] + scalar*d_c[i];
+                                  device_a[i] = device_b[i] + scalar*device_c[i];
 								});
     future_kernel.wait();
   }