added initializers into GPU perf. tests

samples/gpu/performance/performance.cpp

@@ -4,41 +4,35 @@
 using namespace std;
 using namespace cv;
 
-void Test::gen(Mat& mat, int rows, int cols, int type, Scalar low, Scalar high)
-{
-    mat.create(rows, cols, type);
-
-    RNG rng(0);
-    rng.fill(mat, RNG::UNIFORM, low, high);
-}
-
-
 void TestSystem::run()
 {
+    // Run initializers
+    vector<Runnable*>::iterator it = inits_.begin();
+    for (; it != inits_.end(); ++it)
+        (*it)->run();
+
     cout << setiosflags(ios_base::left);
     cout << "    " << setw(10) << "CPU, ms" << setw(10) << "GPU, ms" 
         << setw(10) << "SPEEDUP" << "DESCRIPTION\n";
     cout << resetiosflags(ios_base::left);
 
-    vector<Test*>::iterator it = tests_.begin();
+    // Run tests
+    it = tests_.begin();
     for (; it != tests_.end(); ++it)
     {
-        Test* test = *it;
+        cout << endl << (*it)->name() << ":\n";
-
+        (*it)->run();
-        cout << endl << test->name() << ":\n";
+        flush_subtest_data();
-        test->run();
-
-        flush();
     }
 
     cout << setiosflags(ios_base::fixed | ios_base::left);
-    cout << "\nAverage GPU Speedup: x" << setprecision(3) 
+    cout << "\naverage GPU speedup: x" << setprecision(3) 
         << speedup_total_ / num_subtests_called_ << endl;
     cout << resetiosflags(ios_base::fixed | ios_base::left);
 }
 
 
-void TestSystem::flush()
+void TestSystem::flush_subtest_data()
 {
     if (!can_flush_)
         return;
@@ -75,6 +69,14 @@ void TestSystem::flush()
 }
 
 
+void gen(Mat& mat, int rows, int cols, int type, Scalar low, Scalar high)
+{
+    mat.create(rows, cols, type);
+    RNG rng(0);
+    rng.fill(mat, RNG::UNIFORM, low, high);
+}
+
+
 int main()
 {
     TestSystem::instance()->run();

samples/gpu/performance/performance.h

@@ -7,16 +7,13 @@
 #include <string>
 #include <opencv2/core/core.hpp>
 
-class Test
+class Runnable
 {
 public:
-    explicit Test(const std::string& name): name_(name) {}
+    explicit Runnable(const std::string& name): name_(name) {}
 
     const std::string& name() const { return name_; }
 
-    void gen(cv::Mat& mat, int rows, int cols, int type, 
-             cv::Scalar low, cv::Scalar high);
-
     virtual void run() = 0;
 
 private:
@@ -33,7 +30,8 @@ public:
         return &me;
     }
 
-    void add(Test* test) { tests_.push_back(test); }
+    void addInit(Runnable* init) { inits_.push_back(init); }
+    void addTest(Runnable* test) { tests_.push_back(test); }
 
     void run();
 
@@ -58,7 +56,7 @@ public:
     // Ends current subtest and starts new one
     std::stringstream& subtest()
     {
-        flush();
+        flush_subtest_data();
         return description_;
     }
 
@@ -66,9 +64,10 @@ private:
     TestSystem(): can_flush_(false), cpu_elapsed_(0), gpu_elapsed_(0), 
                   speedup_total_(0.0), num_subtests_called_(0) {};
 
-    void flush();
+    void flush_subtest_data();
 
-    std::vector<Test*> tests_;
+    std::vector<Runnable*> inits_;
+    std::vector<Runnable*> tests_;
 
     // Current test (subtest) description
     std::stringstream description_;
@@ -84,14 +83,27 @@ private:
 
 
 #define TEST(name) \
-    struct name##_test: Test \
+    struct name##_test: Runnable \
     { \
-        name##_test(): Test(#name) { TestSystem::instance()->add(this); } \
+        name##_test(): Runnable(#name) { \
+            TestSystem::instance()->addTest(this); \
+        } \
         void run(); \
     } name##_test_instance; \
     void name##_test::run()
 
 
+#define INIT(name) \
+    struct name##_init: Runnable \
+    { \
+        name##_init(): Runnable(#name) { \
+            TestSystem::instance()->addInit(this); \
+        } \
+        void run(); \
+    } name##_init_instance; \
+    void name##_init::run()
+
+
 #define CPU_ON TestSystem::instance()->cpuOn()
 #define GPU_ON TestSystem::instance()->gpuOn()
 #define CPU_OFF TestSystem::instance()->cpuOff()
@@ -99,4 +111,7 @@ private:
 #define SUBTEST TestSystem::instance()->subtest()
 #define DESCRIPTION TestSystem::instance()->subtest()
 
+void gen(cv::Mat& mat, int rows, int cols, int type, cv::Scalar low, 
+         cv::Scalar high);
+
 #endif // OPENCV_GPU_SAMPLE_PERFORMANCE_H_

samples/gpu/performance/tests.cpp

@@ -6,21 +6,36 @@
 using namespace std;
 using namespace cv;
 
+// This code calls CUFFT DFT and initializes that lib
+INIT(CUFFT_library)
+{
+    Mat src, templ;
+    gen(src, 500, 500, CV_32F, 0, 1);
+    gen(templ, 500, 500, CV_32F, 0, 1);
+
+    gpu::GpuMat d_src(src);
+    gpu::GpuMat d_templ(templ);
+    gpu::GpuMat d_result;
+
+    gpu::matchTemplate(d_src, d_templ, d_result, CV_TM_CCORR);
+}
+
+
 TEST(matchTemplate)
 {
-    Mat image, templ, result;
+    Mat src, templ, result;
-    gen(image, 3000, 3000, CV_32F, 0, 1);
+    gen(src, 3000, 3000, CV_32F, 0, 1);
 
-    gpu::GpuMat d_image(image), d_templ, d_result;
+    gpu::GpuMat d_image(src), d_templ, d_result;
 
     for (int templ_size = 5; templ_size <= 1000; templ_size *= 2)
     {
-        SUBTEST << "img " << image.rows << ", templ " << templ_size << ", 32F, CCORR";
+        SUBTEST << "src " << src.rows << ", templ " << templ_size << ", 32F, CCORR";
 
         gen(templ, templ_size, templ_size, CV_32F, 0, 1);
 
         CPU_ON;
-        matchTemplate(image, templ, result, CV_TM_CCORR);
+        matchTemplate(src, templ, result, CV_TM_CCORR);
         CPU_OFF;
 
         d_templ = templ;
@@ -42,7 +57,7 @@ TEST(minMaxLoc)
 
     for (int size = 2000; size <= 8000; size *= 2)
     {
-        SUBTEST << "img " << size << ", 32F, no mask";
+        SUBTEST << "src " << size << ", 32F, no mask";
 
         gen(src, size, size, CV_32F, 0, 1);
 
@@ -66,7 +81,7 @@ TEST(remap)
 
     for (int size = 1000; size <= 8000; size *= 2)
     {
-        SUBTEST << "img " << size << " and 8UC1, 32FC1 maps";
+        SUBTEST << "src " << size << " and 8UC1, 32FC1 maps";
 
         gen(src, size, size, CV_8UC1, 0, 256);
         gen(xmap, size, size, CV_32F, 0, size);