fix broadcast_op_test and reduce_op_test

CMakeLists.txt

@@ -40,7 +40,7 @@ option(WITH_AMD_GPU     "Compile PaddlePaddle with AMD GPU"             OFF)
 option(WITH_AVX         "Compile PaddlePaddle with AVX intrinsics"      ${AVX_FOUND})
 option(WITH_MKL         "Compile PaddlePaddle with MKL support."        ${AVX_FOUND})
 option(WITH_DSO         "Compile PaddlePaddle with dynamic linked CUDA" ON)
-option(WITH_TESTING     "Compile PaddlePaddle with unit testing"        OFF)
+option(WITH_TESTING     "Compile PaddlePaddle with unit testing"        ON)
 option(WITH_SWIG_PY     "Compile PaddlePaddle with inference api"       ON)
 option(WITH_STYLE_CHECK "Compile PaddlePaddle with style check"         ON)
 option(WITH_PYTHON      "Compile PaddlePaddle with python interpreter"  ON)

paddle/fluid/framework/details/broadcast_op_handle_test.cc

@@ -139,7 +139,7 @@ struct TestBroadcastOpHandle {
       PADDLE_ENFORCE_EQ(out_tensor.lod(), lod, "lod is not equal.");
 
       f::Tensor result_tensor;
-      f::TensorCopy(out_tensor, cpu_place, *(ctxs_[j]), &result_tensor);
+      f::TensorCopy(out_tensor, cpu_place, *(ctxs_[j]), &result_tensor, true);
       float* ct = result_tensor.mutable_data<float>(cpu_place);
 
       for (int64_t i = 0; i < f::product(kDims); ++i) {
@@ -185,7 +185,7 @@ struct TestBroadcastOpHandle {
       }
 
       f::Tensor result_tensor;
-      f::TensorCopy(rt, cpu_place, *(ctxs_[j]), &result_tensor);
+      f::TensorCopy(rt, cpu_place, *(ctxs_[j]), &result_tensor, true);
       float* ct = result_tensor.data<float>();
 
       for (int64_t i = 0; i < f::product(kDims); ++i) {

paddle/fluid/framework/details/reduce_op_handle_test.cc

@@ -194,7 +194,8 @@ struct TestReduceOpHandle {
     }
 
     f::Tensor result_tensor;
-    f::TensorCopy(rt, cpu_place, *(ctxs_[output_scope_idx]), &result_tensor);
+    f::TensorCopy(rt, cpu_place, *(ctxs_[output_scope_idx]), &result_tensor,
+                  true);
     float *ct = result_tensor.data<float>();
 
     for (int64_t j = 0; j < f::product(result_tensor.dims()); ++j) {
@@ -239,7 +240,8 @@ struct TestReduceOpHandle {
     auto &rt = out_var->Get<f::LoDTensor>();
 
     f::Tensor result_tensor;
-    f::TensorCopy(rt, cpu_place, *(ctxs_[output_scope_idx]), &result_tensor);
+    f::TensorCopy(rt, cpu_place, *(ctxs_[output_scope_idx]), &result_tensor,
+                  true);
     float *ct = result_tensor.data<float>();
 
     for (int64_t j = 0; j < f::product(result_tensor.dims()); ++j) {