use multi-thread to speedup CI tests

test=develop

paddle/fluid/inference/tests/api/CMakeLists.txt

@@ -30,6 +30,7 @@ function(inference_analysis_api_int8_test target model_dir data_dir filename)
              --infer_data=${data_dir}/data.bin
              --warmup_batch_size=100
              --batch_size=50
+             --paddle_num_threads=${CPU_NUM_THREADS_ON_CI}
 	     --iterations=2)
 endfunction()
 
@@ -120,7 +121,8 @@ set(TRANSFORMER_INSTALL_DIR "${INFERENCE_DEMO_INSTALL_DIR}/transformer")
 download_model_and_data(${TRANSFORMER_INSTALL_DIR} "temp%2Ftransformer_model.tar.gz" "temp%2Ftransformer_data.txt.tar.gz")
 inference_analysis_test(test_analyzer_transformer SRCS analyzer_transformer_tester.cc 
   EXTRA_DEPS ${INFERENCE_EXTRA_DEPS}
-  ARGS --infer_model=${TRANSFORMER_INSTALL_DIR}/model --infer_data=${TRANSFORMER_INSTALL_DIR}/data.txt --batch_size=8 SERIAL)
+  ARGS --infer_model=${TRANSFORMER_INSTALL_DIR}/model --infer_data=${TRANSFORMER_INSTALL_DIR}/data.txt --batch_size=8 
+       --paddle_num_threads=${CPU_NUM_THREADS_ON_CI} SERIAL)
 
 # ocr
 set(OCR_INSTALL_DIR "${INFERENCE_DEMO_INSTALL_DIR}/ocr")

paddle/fluid/inference/tests/api/config_printer.h

@@ -51,8 +51,6 @@ std::ostream &operator<<(std::ostream &os, const NativeConfig &config) {
      << "fraction_of_gpu_memory: " << config.fraction_of_gpu_memory << "\n";
   os << GenSpaces(num_spaces)
      << "specify_input_name: " << config.specify_input_name << "\n";
-  os << GenSpaces(num_spaces)
-     << "cpu_num_threads: " << config.cpu_math_library_num_threads() << "\n";
   num_spaces--;
   os << GenSpaces(num_spaces) << "}\n";
   return os;
@@ -72,8 +70,8 @@ std::ostream &operator<<(std::ostream &os, const AnalysisConfig &config) {
   }
   os << GenSpaces(num_spaces) << "enable_ir_optim: " << config.ir_optim()
      << "\n";
-  os << GenSpaces(num_spaces) << "enable_ir_optim: " << config.ir_optim()
-     << "\n";
+  os << GenSpaces(num_spaces)
+     << "cpu_num_threads: " << config.cpu_math_library_num_threads() << "\n";
   os << GenSpaces(num_spaces)
      << "use_feed_fetch_ops: " << config.use_feed_fetch_ops_enabled() << "\n";
   os << GenSpaces(num_spaces)

paddle/fluid/inference/tests/test.cmake

@@ -2,6 +2,7 @@ include(ExternalProject)
 set(INFERENCE_URL "http://paddle-inference-dist.bj.bcebos.com" CACHE STRING "inference download url")
 set(INFERENCE_DEMO_INSTALL_DIR "${THIRD_PARTY_PATH}/inference_demo" CACHE STRING
     "A path setting inference demo download directories.")
+set(CPU_NUM_THREADS_ON_CI 4 CACHE STRING "Run multi-threads on CI to reduce CI time.")
 
 function(inference_download INSTALL_DIR URL FILENAME)
   message(STATUS "Download inference test stuff from ${URL}/${FILENAME}")

python/paddle/fluid/contrib/tests/CMakeLists.txt

@@ -7,7 +7,7 @@ endif()
 
 foreach(src ${TEST_OPS})
     if(src MATCHES "test_calibration")
-        py_test(${src} SRCS ${src}.py ENVS FLAGS_use_mkldnn=true)
+        py_test(${src} SRCS ${src}.py ENVS FLAGS_use_mkldnn=true FLAGS_OMP_NUM_THREADS=${CPU_NUM_THREADS_ON_CI})
     else()
         py_test(${src} SRCS ${src}.py)
     endif()