diff --git a/lite/api/cxx_api_impl.cc b/lite/api/cxx_api_impl.cc
index 3e6e10103e9f3af51923459a5921f9781431f352..81ea60eac66849f8ce42fb8cb210226d18bbfa9b 100644
--- a/lite/api/cxx_api_impl.cc
+++ b/lite/api/cxx_api_impl.cc
@@ -42,11 +42,11 @@ void CxxPaddleApiImpl::Init(const lite_api::CxxConfig &config) {
 
 #if (defined LITE_WITH_X86) && (defined PADDLE_WITH_MKLML) && \
     !(defined LITE_ON_MODEL_OPTIMIZE_TOOL)
-  int num_threads = config.cpu_math_library_num_threads();
+  int num_threads = config.x86_math_library_num_threads();
   int real_num_threads = num_threads > 1 ? num_threads : 1;
   paddle::lite::x86::MKL_Set_Num_Threads(real_num_threads);
   omp_set_num_threads(real_num_threads);
-  VLOG(3) << "set_cpu_math_library_math_threads() is set successfully and the "
+  VLOG(3) << "set_x86_math_library_math_threads() is set successfully and the "
              "number of threads is:"
           << num_threads;
 #endif
diff --git a/lite/api/paddle_api.h b/lite/api/paddle_api.h
index a014719c5783bcec3988bba65c53cc8cf52f0b4c..6308699ac91900d161a55ee121e4d9777947fede 100644
--- a/lite/api/paddle_api.h
+++ b/lite/api/paddle_api.h
@@ -133,7 +133,9 @@ class LITE_API CxxConfig : public ConfigBase {
   std::string model_file_;
   std::string param_file_;
   bool model_from_memory_{false};
-  int cpu_math_library_math_threads_ = 1;
+#ifdef LITE_WITH_X86
+  int x86_math_library_math_threads_ = 1;
+#endif
 
  public:
   void set_valid_places(const std::vector<Place>& x) { valid_places_ = x; }
@@ -153,12 +155,14 @@ class LITE_API CxxConfig : public ConfigBase {
   std::string param_file() const { return param_file_; }
   bool model_from_memory() const { return model_from_memory_; }
 
-  void set_cpu_math_library_num_threads(int threads) {
-    cpu_math_library_math_threads_ = threads;
+#ifdef LITE_WITH_X86
+  void set_x86_math_library_num_threads(int threads) {
+    x86_math_library_math_threads_ = threads;
   }
-  int cpu_math_library_num_threads() const {
-    return cpu_math_library_math_threads_;
+  int x86_math_library_num_threads() const {
+    return x86_math_library_math_threads_;
   }
+#endif
 };
 
 /// MobileConfig is the config for the light weight predictor, it will skip
diff --git a/lite/api/test_step_rnn_lite_x86.cc b/lite/api/test_step_rnn_lite_x86.cc
index 075d314df6f46ab9dc8531b26c23d05d24e63bb4..013fd82b19bc22ace22184389249a7b2d9bf237e 100644
--- a/lite/api/test_step_rnn_lite_x86.cc
+++ b/lite/api/test_step_rnn_lite_x86.cc
@@ -30,7 +30,9 @@ TEST(Step_rnn, test_step_rnn_lite_x86) {
   std::string model_dir = FLAGS_model_dir;
   lite_api::CxxConfig config;
   config.set_model_dir(model_dir);
-  config.set_cpu_math_library_num_threads(1);
+#ifdef LITE_WITH_X86
+  config.set_x86_math_library_num_threads(1);
+#endif
   config.set_valid_places({lite_api::Place{TARGET(kX86), PRECISION(kInt64)},
                            lite_api::Place{TARGET(kX86), PRECISION(kFloat)},
                            lite_api::Place{TARGET(kHost), PRECISION(kFloat)}});