diff --git a/source/lib/src/cuda/coord.cu b/source/lib/src/cuda/coord.cu
index 4f7714ca50..317a8d42c9 100644
--- a/source/lib/src/cuda/coord.cu
+++ b/source/lib/src/cuda/coord.cu
@@ -286,8 +286,6 @@ void compute_int_data(
     const int * nat_end=cell_info+3;
     const int * ext_stt=cell_info+6;
     const int * ext_end=cell_info+9;
-    const int * ngcell=cell_info+12;
-    const FPTYPE * boxt = region.boxt;
     const FPTYPE * rec_boxt = region.rec_boxt;
     
     const int nblock_loc=(nloc+TPB-1)/TPB;
diff --git a/source/lib/tests/test_coord.cc b/source/lib/tests/test_coord.cc
index e3b0552b1d..8dee80745e 100644
--- a/source/lib/tests/test_coord.cc
+++ b/source/lib/tests/test_coord.cc
@@ -473,3 +473,276 @@ TEST_F(TestCopyCoord, gpu_lessmem)
   // 	    << nall << std::endl;
 }
 #endif //GOOGLE_CUDA
+
+class TestCopyCoordMoreCell : public ::testing::Test
+{
+protected:
+  std::vector<double > posi = {
+    0.041, 0.072, 0.100,
+    4.053, 0.041, 0.068,
+    1.130, 0.014, 0.109,
+    0.018, 1.134, 0.139,
+    5.120, 0.087, 0.002,
+    4.067, 1.141, 0.055
+  };
+  std::vector<int > atype = {0, 0, 1, 1, 1, 1};
+  std::vector<double > _expected_posi_cpy = {0.041, 0.072, 0.1, 4.053, 0.041, 0.068, 1.13, 0.014, 0.109, 0.018, 1.134, 0.139, 5.12, 0.087, 0.002, 4.067, 1.141, 0.055, 0.018, 1.134, 20.139, 0.018, 21.134, 0.139, 0.018, 21.134, 20.139, 0.041, 0.072, 20.1, 0.041, 20.072, 0.1, 0.041, 20.072, 20.1, 1.13, 0.014, 20.109, 1.13, 20.014, 0.109, 1.13, 20.014, 20.109, 4.053, 0.041, 20.068,
+  4.053, 20.041, 0.068, 4.053, 20.041, 20.068, 4.067, 1.141, 20.055, 4.067, 21.141, 0.055, 4.067, 21.141, 20.055, 5.12, 0.087, 20.002, 5.12, 20.087, 0.002, 5.12, 20.087, 20.002, 20.018, 1.134, 0.139, 20.018, 1.134, 20.139, 20.018, 21.134, 0.139, 20.018, 21.134, 20.139, 20.041, 0.072, 0.1, 20.041, 0.072, 20.1, 20.041, 20.072, 0.1, 20.041, 20.072, 20.1,
+  21.13, 0.014, 0.109, 21.13, 0.014, 20.109, 21.13, 20.014, 0.109, 21.13, 20.014, 20.109, 24.053, 0.041, 0.068, 24.053, 0.041, 20.068, 24.053, 20.041, 0.068, 24.053, 20.041, 20.068, 24.067,1.141, 0.055, 24.067, 1.141, 20.055, 24.067, 21.141, 0.055, 24.067, 21.141, 20.055, 25.12, 0.087, 0.002, 25.12, 0.087, 20.002, 25.12, 20.087, 0.002, 25.12, 20.087, 20.002,
+  };
+  std::vector<double > expected_posi_cpy;
+  std::vector<int > _expected_atype_cpy = {
+    0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 
+  };
+  std::vector<int > expected_atype_cpy;
+  std::vector<int > _expected_mapping = {
+    0, 1, 2, 3, 4, 5, 3, 3, 3, 0, 0, 0, 2, 2, 2, 1, 1, 1, 5, 5, 5, 4, 4, 4, 3, 3, 3, 3, 0, 0, 0, 0, 2, 2, 2, 2, 1, 1, 1, 1, 5, 5, 5, 5, 4, 4, 4, 4,
+  };
+  std::vector<int > expected_mapping;
+  int ntypes = 2;  
+  int nloc, expected_nall;
+  double rc = 4;
+  std::vector<double> boxt = {20., 0., 0., 0., 20., 0., 0., 0., 20.};
+  std::vector<int> ncell, ngcell;
+
+  void SetUp() override {
+    nloc = posi.size() / 3;
+    expected_nall = _expected_posi_cpy.size() / 3;
+    EXPECT_EQ(expected_nall, _expected_atype_cpy.size());
+    EXPECT_EQ(expected_nall, _expected_mapping.size());
+    // sort the atoms between nloc and nall, to remove the uncertainty of the ordering
+    sort_atoms(
+	expected_posi_cpy,
+	expected_atype_cpy,
+	expected_mapping,
+	_expected_posi_cpy,
+	_expected_atype_cpy,
+	_expected_mapping,
+	nloc,
+	expected_nall);
+  }  
+};
+
+TEST_F(TestCopyCoordMoreCell, cpu)
+{
+  int mem_size = 1000;
+  std::vector<double > out_c(mem_size * 3);
+  std::vector<int > out_t(mem_size);
+  std::vector<int > mapping(mem_size);
+  int nall;
+  deepmd::Region<double> region;
+  init_region_cpu(region, &boxt[0]);
+  
+  int ret = copy_coord_cpu(
+      &out_c[0],
+      &out_t[0],
+      &mapping[0],
+      &nall,
+      &posi[0],
+      &atype[0],
+      nloc,
+      mem_size,
+      rc,
+      region);
+  EXPECT_EQ(ret, 0);
+  EXPECT_EQ(nall, expected_nall);
+  // std::cout << "---------------------" 
+  // 	    << nloc << " " 
+  // 	    << nall << std::endl;
+  
+  out_c.resize(nall*3);
+  out_t.resize(nall);
+  mapping.resize(nall);
+  
+  std::vector<double > out_c_1(mem_size * 3);
+  std::vector<int > out_t_1(mem_size);
+  std::vector<int > mapping_1(mem_size);
+  sort_atoms(out_c_1, out_t_1, mapping_1, out_c, out_t, mapping, nloc, nall);
+  for(int ii = 0; ii < expected_nall; ++ii){
+    for(int dd = 0; dd < 3; ++dd){
+      EXPECT_LT(fabs(out_c_1[ii*3+dd] - expected_posi_cpy[ii*3+dd]), 1e-12);
+    }
+    EXPECT_EQ(out_t_1[ii], expected_atype_cpy[ii]);
+    EXPECT_EQ(mapping_1[ii], expected_mapping[ii]);
+  }  
+}
+
+TEST_F(TestCopyCoordMoreCell, cpu_lessmem)
+{
+  int mem_size = 40;
+  std::vector<double > out_c(mem_size * 3);
+  std::vector<int > out_t(mem_size);
+  std::vector<int > mapping(mem_size);
+  int nall;
+  deepmd::Region<double> region;
+  init_region_cpu(region, &boxt[0]);
+  
+  int ret = copy_coord_cpu(
+      &out_c[0],
+      &out_t[0],
+      &mapping[0],
+      &nall,
+      &posi[0],
+      &atype[0],
+      nloc,
+      mem_size,
+      rc,
+      region);
+  EXPECT_EQ(ret, 1);
+  // EXPECT_EQ(nall, expected_nall);
+  // std::cout << "---------------------" 
+  // 	    << nloc << " " 
+  // 	    << nall << std::endl;
+}
+
+#if GOOGLE_CUDA
+TEST_F(TestCopyCoordMoreCell, gpu)
+{
+  int mem_size = 1000;
+  std::vector<double > out_c(mem_size * 3);
+  std::vector<int > out_t(mem_size);
+  std::vector<int > mapping(mem_size);
+  int nall;
+  std::vector<int> cell_info;
+  cell_info.resize(23);
+  deepmd::Region<double> region;
+  deepmd::Region<double> region_dev;
+  double * new_boxt = region_dev.boxt;
+  double * new_rec_boxt = region_dev.rec_boxt;
+  init_region_cpu(region, &boxt[0]);
+  deepmd::compute_cell_info(&cell_info[0], rc, region);
+  std::vector<double> box_info;
+  box_info.resize(18);
+  memcpy(&box_info[0], &boxt[0], sizeof(double)*9);
+  memcpy(&box_info[9], region.rec_boxt, sizeof(double)*9);
+  const int loc_cellnum=cell_info[21];
+  const int total_cellnum=cell_info[22];
+  int * cell_info_dev=NULL;
+  double * box_info_dev=NULL;
+  double * out_c_dev=NULL, * in_c_dev=NULL;
+  int * out_t_dev=NULL, * in_t_dev=NULL, * mapping_dev=NULL, * int_data_dev=NULL;
+  deepmd::malloc_device_memory_sync(cell_info_dev, cell_info);
+  deepmd::malloc_device_memory_sync(box_info_dev, box_info);
+  deepmd::malloc_device_memory_sync(in_c_dev, posi);
+  deepmd::malloc_device_memory_sync(in_t_dev, atype);
+  deepmd::malloc_device_memory(out_c_dev, mem_size * 3);
+  deepmd::malloc_device_memory(out_t_dev, mem_size);
+  deepmd::malloc_device_memory(mapping_dev, mem_size);
+  deepmd::malloc_device_memory(int_data_dev, nloc*3+loc_cellnum+total_cellnum*3+total_cellnum*3+loc_cellnum+1+total_cellnum+1+nloc);
+  region_dev.boxt = box_info_dev;
+  region_dev.rec_boxt = box_info_dev + 9;
+  int ret = deepmd::copy_coord_gpu(
+      out_c_dev, 
+      out_t_dev, 
+      mapping_dev, 
+      &nall,
+      int_data_dev,
+      in_c_dev,
+      in_t_dev,
+      nloc,
+      mem_size,
+      loc_cellnum,
+      total_cellnum,
+      cell_info_dev,
+      region_dev);
+  region_dev.boxt = new_boxt;
+  region_dev.rec_boxt = new_rec_boxt;
+  deepmd::memcpy_device_to_host(out_c_dev, out_c);
+  deepmd::memcpy_device_to_host(out_t_dev, out_t);
+  deepmd::memcpy_device_to_host(mapping_dev, mapping);
+  deepmd::delete_device_memory(cell_info_dev);
+  deepmd::delete_device_memory(box_info_dev);
+  deepmd::delete_device_memory(in_c_dev);
+  deepmd::delete_device_memory(in_t_dev);
+  deepmd::delete_device_memory(out_c_dev);
+  deepmd::delete_device_memory(out_t_dev);
+  deepmd::delete_device_memory(mapping_dev);
+  deepmd::delete_device_memory(int_data_dev);
+  EXPECT_EQ(ret, 0);
+  EXPECT_EQ(nall, expected_nall);
+  out_c.resize(nall*3);
+  out_t.resize(nall);
+  mapping.resize(nall);
+
+  std::vector<double > out_c_1(mem_size * 3);
+  std::vector<int > out_t_1(mem_size);
+  std::vector<int > mapping_1(mem_size);
+  sort_atoms(out_c_1, out_t_1, mapping_1, out_c, out_t, mapping, nloc, nall);
+  for(int ii = 0; ii < expected_nall; ++ii){
+    for(int dd = 0; dd < 3; ++dd){
+      EXPECT_LT(fabs(out_c_1[ii*3+dd] - expected_posi_cpy[ii*3+dd]), 1e-12);
+    }
+    EXPECT_EQ(out_t_1[ii], expected_atype_cpy[ii]);
+    EXPECT_EQ(mapping_1[ii], expected_mapping[ii]);
+  }  
+}
+
+TEST_F(TestCopyCoordMoreCell, gpu_lessmem)
+{
+  int mem_size = 40;
+  std::vector<double > out_c(mem_size * 3);
+  std::vector<int > out_t(mem_size);
+  std::vector<int > mapping(mem_size);
+  int nall;
+  std::vector<int> cell_info;
+  cell_info.resize(23);
+  deepmd::Region<double> region;
+  deepmd::Region<double> region_dev;
+  double * new_boxt = region_dev.boxt;
+  double * new_rec_boxt = region_dev.rec_boxt;
+  init_region_cpu(region, &boxt[0]);
+  deepmd::compute_cell_info(&cell_info[0], rc, region);
+  std::vector<double> box_info;
+  box_info.resize(18);
+  memcpy(&box_info[0], &boxt[0], sizeof(double)*9);
+  memcpy(&box_info[9], region.rec_boxt, sizeof(double)*9);
+  const int loc_cellnum=cell_info[21];
+  const int total_cellnum=cell_info[22];
+  int * cell_info_dev=NULL;
+  double * box_info_dev=NULL;
+  double * out_c_dev=NULL, * in_c_dev=NULL;
+  int * out_t_dev=NULL, * in_t_dev=NULL, * mapping_dev=NULL, * int_data_dev=NULL;
+  deepmd::malloc_device_memory_sync(cell_info_dev, cell_info);
+  deepmd::malloc_device_memory_sync(box_info_dev, box_info);
+  deepmd::malloc_device_memory_sync(in_c_dev, posi);
+  deepmd::malloc_device_memory_sync(in_t_dev, atype);
+  deepmd::malloc_device_memory(out_c_dev, mem_size * 3);
+  deepmd::malloc_device_memory(out_t_dev, mem_size);
+  deepmd::malloc_device_memory(mapping_dev, mem_size);
+  deepmd::malloc_device_memory(int_data_dev, nloc*3+loc_cellnum+total_cellnum*3+total_cellnum*3+loc_cellnum+1+total_cellnum+1+nloc);
+  region_dev.boxt = box_info_dev;
+  region_dev.rec_boxt = box_info_dev + 9;
+  int ret = deepmd::copy_coord_gpu(
+      out_c_dev, 
+      out_t_dev, 
+      mapping_dev, 
+      &nall,
+      int_data_dev,
+      in_c_dev,
+      in_t_dev,
+      nloc,
+      mem_size,
+      loc_cellnum,
+      total_cellnum,
+      cell_info_dev,
+      region_dev);
+  region_dev.boxt = new_boxt;
+  region_dev.rec_boxt = new_rec_boxt;
+  deepmd::memcpy_device_to_host(out_c_dev, out_c);
+  deepmd::memcpy_device_to_host(out_t_dev, out_t);
+  deepmd::memcpy_device_to_host(mapping_dev, mapping);
+  deepmd::delete_device_memory(cell_info_dev);
+  deepmd::delete_device_memory(box_info_dev);
+  deepmd::delete_device_memory(in_c_dev);
+  deepmd::delete_device_memory(in_t_dev);
+  deepmd::delete_device_memory(out_c_dev);
+  deepmd::delete_device_memory(out_t_dev);
+  deepmd::delete_device_memory(mapping_dev);
+  deepmd::delete_device_memory(int_data_dev);
+  EXPECT_EQ(ret, 1);
+  // EXPECT_EQ(nall, expected_nall);
+  // std::cout << "---------------------" 
+  // 	    << nloc << " " 
+  // 	    << nall << std::endl;
+}
+#endif //GOOGLE_CUDA
\ No newline at end of file