Merge pull request #10436 from alalek:test_threads

modules/dnn/perf/opencl/perf_convolution.cpp

@@ -77,8 +77,6 @@ OCL_PERF_TEST_P( ConvolutionPerfTest, perf, Combine(
     std::vector<Mat*> inpBlobs(1, &inpBlob);
     std::vector<Mat> outBlobs, internalBlobs;
 
-    cv::setNumThreads(cv::getNumberOfCPUs());
-
     Ptr<Layer> layer = cv::dnn::LayerFactory::createLayerInstance("Convolution", lp);
     std::vector<MatShape> inputShapes(1, shape(inpBlob)), outShapes, internals;
     layer->getMemoryShapes(inputShapes, 0, outShapes, internals);
@@ -99,7 +97,7 @@ OCL_PERF_TEST_P( ConvolutionPerfTest, perf, Combine(
     Mat inpBlob2D = inpBlob.reshape(1, outCn);
     Mat wgtBlob2D = wgtBlob.reshape(1, outCn*(inpCn/groups));
     Mat outBlob2D = outBlobs[0].reshape(1, outBlobs[0].size[0]);
-    declare.in(inpBlob2D, wgtBlob2D, WARMUP_RNG).out(outBlob2D).tbb_threads(cv::getNumThreads());
+    declare.in(inpBlob2D, wgtBlob2D, WARMUP_RNG).out(outBlob2D);
 
     // warmup
     layer->forward(inpBlobs, outBlobs, internalBlobs);

modules/dnn/perf/perf_convolution.cpp

@@ -60,8 +60,6 @@ PERF_TEST_P( ConvolutionPerfTest, perf, Combine(
     std::vector<Mat*> inpBlobs(1, &inpBlob);
     std::vector<Mat> outBlobs, internalBlobs;
 
-    cv::setNumThreads(cv::getNumberOfCPUs());
-
     Ptr<Layer> layer = cv::dnn::LayerFactory::createLayerInstance("Convolution", lp);
     std::vector<MatShape> inputShapes(1, shape(inpBlob)), outShapes, internals;
     layer->getMemoryShapes(inputShapes, 0, outShapes, internals);
@@ -81,7 +79,7 @@ PERF_TEST_P( ConvolutionPerfTest, perf, Combine(
     Mat inpBlob2D = inpBlob.reshape(1, outCn);
     Mat wgtBlob2D = wgtBlob.reshape(1, outCn*(inpCn/groups));
     Mat outBlob2D = outBlobs[0].reshape(1, outBlobs[0].size[0]);
-    declare.in(inpBlob2D, wgtBlob2D, WARMUP_RNG).out(outBlob2D).tbb_threads(cv::getNumThreads());
+    declare.in(inpBlob2D, wgtBlob2D, WARMUP_RNG).out(outBlob2D);
 
     layer->forward(inpBlobs, outBlobs, internalBlobs); /// warmup
 

modules/dnn/test/test_layers.cpp

@@ -107,8 +107,6 @@ void testLayerUsingCaffeModels(String basename, int targetId = DNN_TARGET_CPU,
     String inpfile = (useCommonInputBlob) ? _tf("blob.npy") : _tf(basename + ".input.npy");
     String outfile = _tf(basename + ".npy");
 
-    cv::setNumThreads(cv::getNumberOfCPUs());
-
     Net net = readNetFromCaffe(prototxt, (useCaffeModel) ? caffemodel : String());
     ASSERT_FALSE(net.empty());
 
@@ -537,8 +535,6 @@ void testLayerUsingDarknetModels(String basename, bool useDarknetModel = false,
     String inpfile = (useCommonInputBlob) ? _tf("blob.npy") : _tf(basename + ".input.npy");
     String outfile = _tf(basename + ".npy");
 
-    cv::setNumThreads(cv::getNumberOfCPUs());
-
     Net net = readNetFromDarknet(cfg, (useDarknetModel) ? weights : String());
     ASSERT_FALSE(net.empty());
 

modules/ts/include/opencv2/ts/ts_ext.hpp

@@ -11,11 +11,13 @@
 namespace cvtest {
 void checkIppStatus();
 extern bool skipUnstableTests;
+extern int testThreads;
 }
 
 #define CV__TEST_INIT \
     cv::ipp::setIppStatus(0); \
-    cv::theRNG().state = cvtest::param_seed;
+    cv::theRNG().state = cvtest::param_seed; \
+    cv::setNumThreads(cvtest::testThreads);
 #define CV__TEST_CLEANUP ::cvtest::checkIppStatus();
 #define CV__TEST_BODY_IMPL(name) \
     { \

modules/ts/src/ts.cpp

@@ -695,12 +695,14 @@ void checkIppStatus()
 }
 
 bool skipUnstableTests = false;
+int testThreads = 0;
 
 void parseCustomOptions(int argc, char **argv)
 {
     const char * const command_line_keys =
         "{ ipp test_ipp_check |false    |check whether IPP works without failures }"
         "{ test_seed          |809564   |seed for random numbers generator }"
+        "{ test_threads       |-1       |the number of worker threads, if parallel execution is enabled}"
         "{ skip_unstable      |false    |skip unstable tests }"
         "{ h   help           |false    |print help info                          }";
 
@@ -721,6 +723,8 @@ void parseCustomOptions(int argc, char **argv)
 
     param_seed = parser.get<unsigned int>("test_seed");
 
+    testThreads = parser.get<int>("test_threads");
+
     skipUnstableTests = parser.get<bool>("skip_unstable");
 }
 

modules/ts/src/ts_perf.cpp

@@ -39,7 +39,6 @@ static double       param_max_deviation;
 static unsigned int param_min_samples;
 static unsigned int param_force_samples;
 static double       param_time_limit;
-static int          param_threads;
 static bool         param_write_sanity;
 static bool         param_verify_sanity;
 #ifdef CV_COLLECT_IMPL_DATA
@@ -1042,7 +1041,7 @@ void TestBase::Init(const std::vector<std::string> & availableImpls,
 #ifdef HAVE_IPP
     test_ipp_check      = !args.get<bool>("perf_ipp_check") ? getenv("OPENCV_IPP_CHECK") != NULL : true;
 #endif
-    param_threads       = args.get<int>("perf_threads");
+    testThreads         = args.get<int>("perf_threads");
 #ifdef CV_COLLECT_IMPL_DATA
     param_collect_impl  = args.get<bool>("perf_collect_impl");
 #endif
@@ -1160,7 +1159,7 @@ void TestBase::Init(const std::vector<std::string> & availableImpls,
 void TestBase::RecordRunParameters()
 {
     ::testing::Test::RecordProperty("cv_implementation", param_impl);
-    ::testing::Test::RecordProperty("cv_num_threads", param_threads);
+    ::testing::Test::RecordProperty("cv_num_threads", testThreads);
 
 #ifdef HAVE_CUDA
     if (param_impl == "cuda")
@@ -1851,8 +1850,8 @@ void TestBase::SetUp()
 {
     cv::theRNG().state = param_seed; // this rng should generate same numbers for each run
 
-    if (param_threads >= 0)
+    if (testThreads >= 0)
-        cv::setNumThreads(param_threads);
+        cv::setNumThreads(testThreads);
     else
         cv::setNumThreads(-1);