diff --git a/.gitmodules b/.gitmodules
index 963fe98..f6a315b 100644
--- a/.gitmodules
+++ b/.gitmodules
@@ -1,3 +1,6 @@
 [submodule "external/moodycamel"]
 	path = external/moodycamel
 	url = https://github.com/cameron314/concurrentqueue
+[submodule "test/Hetero-Mark"]
+	path = test/Hetero-Mark
+	url = https://github.com/NUCAR-DEV/Hetero-Mark
diff --git a/CMakeLists.txt b/CMakeLists.txt
index c214e35..a5bc849 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -52,3 +52,5 @@ set(GCC_COVERAGE_LINK_FLAGS
 
 add_subdirectory(compilation)
 add_subdirectory(runtime)
+enable_testing()
+add_subdirectory(tests)
diff --git a/test/Hetero-Mark b/test/Hetero-Mark
new file mode 160000
index 0000000..b8ea324
--- /dev/null
+++ b/test/Hetero-Mark
@@ -0,0 +1 @@
+Subproject commit b8ea324d2394401776487b3520338d03f8c27676
diff --git a/tests/CMakeLists.txt b/tests/CMakeLists.txt
new file mode 100644
index 0000000..9737d7e
--- /dev/null
+++ b/tests/CMakeLists.txt
@@ -0,0 +1,14 @@
+set(TESTCASES aes)
+foreach(case ${TESTCASES})
+  message(STATUS "Add test: ${case}")
+  message(STATUS "${CMAKE_CURRENT_SOURCE_DIR}")
+
+  message(STATUS "${CMAKE_BINARY_DIR}")
+  find_program(BASH_PROGRAM bash)
+  execute_process(COMMAND find . -name "*test.cc" -type f -delete)
+  add_test(
+    NAME "${case}"
+    COMMAND
+      ${BASH_PROGRAM} ${CMAKE_CURRENT_SOURCE_DIR}/runHeteroMark.sh ${case}
+      ${CUDA_PATH} ${CMAKE_CURRENT_SOURCE_DIR}/Hetero-Mark ${CMAKE_BINARY_DIR})
+endforeach()
diff --git a/tests/runHeteroMark.sh b/tests/runHeteroMark.sh
new file mode 100644
index 0000000..8c5af9f
--- /dev/null
+++ b/tests/runHeteroMark.sh
@@ -0,0 +1,32 @@
+#!/bin/bash
+set -e
+TestCase=$1
+CUDA_PATH=$2
+HeteroMark_PATH=$3
+CuPBoP_BUILD_PATH=$4
+DATASET_PATH=/tmp/data
+if [ ! -d $DATASET_PATH ]; then
+    echo "Download Hetero Mark dataset"
+    cd /tmp
+    wget https://www.dropbox.com/s/ie2hcxw9lfoghg8/data.tar.gz?dl=1
+    tar -xzf 'data.tar.gz?dl=1'
+fi
+
+mkdir -p $TestCase
+cd $TestCase
+clang++ -std=c++11 $HeteroMark_PATH/src/$1/cuda/$1_cuda_benchmark.cu -I$HeteroMark_PATH \
+        --cuda-path=$CUDA_PATH  \
+        --cuda-gpu-arch=sm_50 -L$CUDA_PATH/lib64 \
+        -lcudart_static -ldl -lrt -pthread -save-temps -v  || true
+export LD_LIBRARY_PATH=$CuPBoP_BUILD_PATH/runtime:$CuPBoP_BUILD_PATH/runtime/threadPool:$LD_LIBRARY_PATH
+export PATH=$CuPBoP_BUILD_PATH/compilation:$PATH
+kernelTranslator $1_cuda_benchmark-cuda-nvptx64-nvidia-cuda-sm_50.bc kernel.bc
+hostTranslator $1_cuda_benchmark-host-x86_64-unknown-linux-gnu.bc host.bc
+llc --relocation-model=pic --filetype=obj  kernel.bc
+llc --relocation-model=pic --filetype=obj  host.bc
+g++ -o $1 -fPIC -no-pie \
+    $HeteroMark_PATH/src/$1/cuda/main.cc host.o kernel.o $HeteroMark_PATH/src/$1/*.cc  $HeteroMark_PATH/src/common/benchmark/*.cc \
+    $HeteroMark_PATH/src/common/command_line_option/*.cc  $HeteroMark_PATH/src/common/time_measurement/*.cc \
+    -L$CuPBoP_BUILD_PATH/runtime   -L$CuPBoP_BUILD_PATH/runtime/threadPool \
+    -I$HeteroMark_PATH -lpthread -lc -lx86Runtime -lthreadPool
+./$1 -i $DATASET_PATH/aes/1KB.data -k $DATASET_PATH/aes/key.data -q -v