Merge pull request #671 from cuda-geek/static_assert

Static assert

modules/calib3d/test/test_fundam.cpp

@@ -1079,7 +1079,7 @@ protected:
     void run_func();
     void prepare_to_validation( int );
 
-    double sampson_error(const double* f, double x1, double y1, double x2, double y2); 
+    double sampson_error(const double* f, double x1, double y1, double x2, double y2);
 
     int method;
     int img_size;
@@ -1145,9 +1145,8 @@ void CV_EssentialMatTest::get_test_array_types_and_sizes( int /*test_case_idx*/,
     int pt_count = MAX(5, cvRound(exp(pt_count_exp)));
 
     dims = cvtest::randInt(rng) % 2 + 2;
-    dims = 2; 
+    dims = 2;
     method = CV_LMEDS << (cvtest::randInt(rng) % 2);
-    
 
     types[INPUT][0] = CV_MAKETYPE(pt_depth, 1);
 
@@ -1192,11 +1191,11 @@ void CV_EssentialMatTest::get_test_array_types_and_sizes( int /*test_case_idx*/,
     sizes[OUTPUT][0] = sizes[REF_OUTPUT][0] = cvSize(3,1);
     types[OUTPUT][0] = types[REF_OUTPUT][0] = CV_64FC1;
     sizes[OUTPUT][1] = sizes[REF_OUTPUT][1] = cvSize(pt_count,1);
-    types[OUTPUT][1] = types[REF_OUTPUT][1] = CV_8UC1;       
+    types[OUTPUT][1] = types[REF_OUTPUT][1] = CV_8UC1;
     sizes[OUTPUT][2] = sizes[REF_OUTPUT][2] = cvSize(1,1);
     types[OUTPUT][2] = types[REF_OUTPUT][2] = CV_64FC1;
     sizes[OUTPUT][3] = sizes[REF_OUTPUT][3] = cvSize(1,1);
-    types[OUTPUT][3] = types[REF_OUTPUT][3] = CV_8UC1;       
+    types[OUTPUT][3] = types[REF_OUTPUT][3] = CV_8UC1;
 
 }
 
@@ -1289,46 +1288,46 @@ int CV_EssentialMatTest::prepare_test_case( int test_case_idx )
 void CV_EssentialMatTest::run_func()
 {
     Mat _input0(test_mat[INPUT][0]), _input1(test_mat[INPUT][1]);
-    Mat K(test_mat[INPUT][4]); 
-    double focal(K.at<double>(0, 0)); 
-    cv::Point2d pp(K.at<double>(0, 2), K.at<double>(1, 2)); 
+    Mat K(test_mat[INPUT][4]);
+    double focal(K.at<double>(0, 0));
+    cv::Point2d pp(K.at<double>(0, 2), K.at<double>(1, 2));
 
     RNG& rng = ts->get_rng();
     Mat E, mask1(test_mat[TEMP][1]);
-    E = cv::findEssentialMat( _input0, _input1, focal, pp, method, 0.99, MAX(sigma*3, 0.0001), mask1 ); 
-    if (E.rows > 3) 
+    E = cv::findEssentialMat( _input0, _input1, focal, pp, method, 0.99, MAX(sigma*3, 0.0001), mask1 );
+    if (E.rows > 3)
     {
-        int count = E.rows / 3; 
-        int row = (cvtest::randInt(rng) % count) * 3; 
-        E = E.rowRange(row, row + 3) * 1.0; 
+        int count = E.rows / 3;
+        int row = (cvtest::randInt(rng) % count) * 3;
+        E = E.rowRange(row, row + 3) * 1.0;
     }
 
-    E.copyTo(test_mat[TEMP][0]); 
+    E.copyTo(test_mat[TEMP][0]);
 
-    Mat R, t, mask2; 
-    recoverPose( E, _input0, _input1, R, t, focal, pp, mask2 ); 
-    R.copyTo(test_mat[TEMP][2]); 
-    t.copyTo(test_mat[TEMP][3]); 
-    mask2.copyTo(test_mat[TEMP][4]); 
+    Mat R, t, mask2;
+    recoverPose( E, _input0, _input1, R, t, focal, pp, mask2 );
+    R.copyTo(test_mat[TEMP][2]);
+    t.copyTo(test_mat[TEMP][3]);
+    mask2.copyTo(test_mat[TEMP][4]);
 }
 
 double CV_EssentialMatTest::sampson_error(const double * f, double x1, double y1, double x2, double y2)
 {
     double Fx1[3] = {
-        f[0] * x1 + f[1] * y1 + f[2], 
-        f[3] * x1 + f[4] * y1 + f[5], 
+        f[0] * x1 + f[1] * y1 + f[2],
+        f[3] * x1 + f[4] * y1 + f[5],
         f[6] * x1 + f[7] * y1 + f[8]
-    }; 
+    };
     double Ftx2[3] = {
-        f[0] * x2 + f[3] * y2 + f[6], 
-        f[1] * x2 + f[4] * y2 + f[7], 
+        f[0] * x2 + f[3] * y2 + f[6],
+        f[1] * x2 + f[4] * y2 + f[7],
         f[2] * x2 + f[5] * y2 + f[8]
-    }; 
-    double x2tFx1 = Fx1[0] * x2 + Fx1[1] * y2 + Fx1[2]; 
+    };
+    double x2tFx1 = Fx1[0] * x2 + Fx1[1] * y2 + Fx1[2];
 
-    double error = x2tFx1 * x2tFx1 / (Fx1[0] * Fx1[0] + Fx1[1] * Fx1[1] + Ftx2[0] * Ftx2[0] + Ftx2[1] * Ftx2[1]); 
-    error = sqrt(error); 
-    return error; 
+    double error = x2tFx1 * x2tFx1 / (Fx1[0] * Fx1[0] + Fx1[1] * Fx1[1] + Ftx2[0] * Ftx2[0] + Ftx2[1] * Ftx2[1]);
+    error = sqrt(error);
+    return error;
 
 }
 
@@ -1338,7 +1337,7 @@ void CV_EssentialMatTest::prepare_to_validation( int test_case_idx )
     const Mat& A = test_mat[INPUT][4];
     double f0[9], f[9], e[9];
     Mat F0(3, 3, CV_64FC1, f0), F(3, 3, CV_64F, f);
-    Mat E(3, 3, CV_64F, e); 
+    Mat E(3, 3, CV_64F, e);
 
     Mat invA, R=Rt0.colRange(0, 3), T1, T2;
 
@@ -1362,7 +1361,7 @@ void CV_EssentialMatTest::prepare_to_validation( int test_case_idx )
     uchar* mtfm2 = test_mat[OUTPUT][1].data;
     double* e_prop1 = (double*)test_mat[REF_OUTPUT][0].data;
     double* e_prop2 = (double*)test_mat[OUTPUT][0].data;
-    Mat E_prop2 = Mat(3, 1, CV_64F, e_prop2); 
+    Mat E_prop2 = Mat(3, 1, CV_64F, e_prop2);
 
     int i, pt_count = test_mat[INPUT][2].cols;
     Mat p1( 1, pt_count, CV_64FC2 );
@@ -1381,8 +1380,8 @@ void CV_EssentialMatTest::prepare_to_validation( int test_case_idx )
         double y1 = p1.at<Point2d>(i).y;
         double x2 = p2.at<Point2d>(i).x;
         double y2 = p2.at<Point2d>(i).y;
-//        double t0 = sampson_error(f0, x1, y1, x2, y2); 
-//        double t = sampson_error(f, x1, y1, x2, y2); 
+//        double t0 = sampson_error(f0, x1, y1, x2, y2);
+//        double t = sampson_error(f, x1, y1, x2, y2);
         double n1 = 1./sqrt(x1*x1 + y1*y1 + 1);
         double n2 = 1./sqrt(x2*x2 + y2*y2 + 1);
         double t0 = fabs(f0[0]*x2*x1 + f0[1]*x2*y1 + f0[2]*x2 +
@@ -1394,7 +1393,7 @@ void CV_EssentialMatTest::prepare_to_validation( int test_case_idx )
         mtfm1[i] = 1;
         mtfm2[i] = !status[i] || t0 > err_level || t < err_level;
     }
-    
+
     e_prop1[0] = sqrt(0.5);
     e_prop1[1] = sqrt(0.5);
     e_prop1[2] = 0;
@@ -1402,26 +1401,26 @@ void CV_EssentialMatTest::prepare_to_validation( int test_case_idx )
     e_prop2[0] = 0;
     e_prop2[1] = 0;
     e_prop2[2] = 0;
-    SVD::compute(E, E_prop2); 
+    SVD::compute(E, E_prop2);
 
 
 
-    double* pose_prop1 = (double*)test_mat[REF_OUTPUT][2].data; 
-    double* pose_prop2 = (double*)test_mat[OUTPUT][2].data; 
-    double terr1 = norm(Rt0.col(3) / norm(Rt0.col(3)) + test_mat[TEMP][3]); 
-    double terr2 = norm(Rt0.col(3) / norm(Rt0.col(3)) - test_mat[TEMP][3]); 
-    Mat rvec; 
-    Rodrigues(Rt0.colRange(0, 3), rvec); 
-    pose_prop1[0] = 0; 
-    // No check for CV_LMeDS on translation. Since it 
-    // involves with some degraded problem, when data is exact inliers. 
-    pose_prop2[0] = method == CV_LMEDS || pt_count == 5 ? 0 : MIN(terr1, terr2); 
+    double* pose_prop1 = (double*)test_mat[REF_OUTPUT][2].data;
+    double* pose_prop2 = (double*)test_mat[OUTPUT][2].data;
+    double terr1 = norm(Rt0.col(3) / norm(Rt0.col(3)) + test_mat[TEMP][3]);
+    double terr2 = norm(Rt0.col(3) / norm(Rt0.col(3)) - test_mat[TEMP][3]);
+    Mat rvec;
+    Rodrigues(Rt0.colRange(0, 3), rvec);
+    pose_prop1[0] = 0;
+    // No check for CV_LMeDS on translation. Since it
+    // involves with some degraded problem, when data is exact inliers.
+    pose_prop2[0] = method == CV_LMEDS || pt_count == 5 ? 0 : MIN(terr1, terr2);
 
 
-//    int inliers_count = countNonZero(test_mat[TEMP][1]); 
-//    int good_count = countNonZero(test_mat[TEMP][4]); 
-    test_mat[OUTPUT][3] = true; //good_count >= inliers_count / 2; 
-    test_mat[REF_OUTPUT][3] = true; 
+//    int inliers_count = countNonZero(test_mat[TEMP][1]);
+//    int good_count = countNonZero(test_mat[TEMP][4]);
+    test_mat[OUTPUT][3] = true; //good_count >= inliers_count / 2;
+    test_mat[REF_OUTPUT][3] = true;
 
 
 }

modules/core/include/opencv2/core/affine.hpp

@@ -68,7 +68,7 @@ namespace cv
 
         //Rodrigues vector
         Affine3(const Vec3& rvec, const Vec3& t = Vec3::all(0));
-        
+
         //Combines all contructors above. Supports 4x4, 3x3, 1x3, 3x1 sizes of data matrix
         explicit Affine3(const cv::Mat& data, const Vec3& t = Vec3::all(0));
 
@@ -79,13 +79,13 @@ namespace cv
 
         //Rotation matrix
         void rotation(const Mat3& R);
-        
+
         //Rodrigues vector
         void rotation(const Vec3& rvec);
-        
+
         //Combines rotation methods above. Suports 3x3, 1x3, 3x1 sizes of data matrix;
         void rotation(const Mat& data);
-        
+
         //Euler angles
         void rotation(float_type alpha, float_type beta, float_type gamma);
 
@@ -218,7 +218,7 @@ template<typename T> inline void cv::Affine3<T>::rotation(const Vec3& rvec)
 template<typename T> inline void cv::Affine3<T>::rotation(const cv::Mat& data)
 {
     CV_Assert(data.type() == cv::DataType<T>::type);
-    
+
     if (data.cols == 3 && data.rows == 3)
     {
         Mat3 R;

modules/core/include/opencv2/core/mat.hpp

@@ -764,7 +764,7 @@ inline void SVD::solveZ( InputArray m, OutputArray _dst )
 template<typename _Tp, int m, int n, int nm> inline void
     SVD::compute( const Matx<_Tp, m, n>& a, Matx<_Tp, nm, 1>& w, Matx<_Tp, m, nm>& u, Matx<_Tp, n, nm>& vt )
 {
-    assert( nm == MIN(m, n));
+    CV_StaticAssert( nm == MIN(m, n), "Invalid size of output vector.");
     Mat _a(a, false), _u(u, false), _w(w, false), _vt(vt, false);
     SVD::compute(_a, _w, _u, _vt);
     CV_Assert(_w.data == (uchar*)&w.val[0] && _u.data == (uchar*)&u.val[0] && _vt.data == (uchar*)&vt.val[0]);
@@ -773,7 +773,7 @@ template<typename _Tp, int m, int n, int nm> inline void
 template<typename _Tp, int m, int n, int nm> inline void
 SVD::compute( const Matx<_Tp, m, n>& a, Matx<_Tp, nm, 1>& w )
 {
-    assert( nm == MIN(m, n));
+    CV_StaticAssert( nm == MIN(m, n), "Invalid size of output vector.");
     Mat _a(a, false), _w(w, false);
     SVD::compute(_a, _w);
     CV_Assert(_w.data == (uchar*)&w.val[0]);
@@ -784,7 +784,7 @@ SVD::backSubst( const Matx<_Tp, nm, 1>& w, const Matx<_Tp, m, nm>& u,
                 const Matx<_Tp, n, nm>& vt, const Matx<_Tp, m, nb>& rhs,
                 Matx<_Tp, n, nb>& dst )
 {
-    assert( nm == MIN(m, n));
+    CV_StaticAssert( nm == MIN(m, n), "Invalid size of output vector.");
     Mat _u(u, false), _w(w, false), _vt(vt, false), _rhs(rhs, false), _dst(dst, false);
     SVD::backSubst(_w, _u, _vt, _rhs, _dst);
     CV_Assert(_dst.data == (uchar*)&dst.val[0]);

modules/core/include/opencv2/core/operations.hpp

@@ -58,6 +58,43 @@
 # pragma warning(disable:4127) //conditional expression is constant
 #endif
 
+//////////////// static assert /////////////////
+
+#define CVAUX_CONCAT_EXP(a, b) a##b
+#define CVAUX_CONCAT(a, b) CVAUX_CONCAT_EXP(a,b)
+
+#ifdef __cplusplus
+#  if defined(__clang__)
+#    ifndef __has_extension
+#      define __has_extension __has_feature /* compatibility, for older versions of clang */
+#    endif
+#    if __has_extension(cxx_static_assert)
+#      define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#    endif
+#  elif defined(__GNUC__)
+#    if (defined(__GXX_EXPERIMENTAL_CXX0X__) || __cplusplus >= 201103L)
+#      define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#    endif
+#  elif defined(_MSC_VER)
+#    if _MSC_VER >= 1600 /* MSVC 10 */
+#      define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#    endif
+#  endif
+#  ifndef CV_StaticAssert
+#    if defined(__GNUC__) && (__GNUC__ > 3) && (__GNUC_MINOR__ > 2)
+#      define CV_StaticAssert(condition, reason) ({ extern int __attribute__((error("CV_StaticAssert: " reason " " #condition))) CV_StaticAssert(); ((condition) ? 0 : CV_StaticAssert()); })
+#    else
+       namespace cv {
+         template <bool x> struct CV_StaticAssert_failed;
+         template <> struct CV_StaticAssert_failed<true> { enum { val = 1 }; };
+         template<int x> struct CV_StaticAssert_test{};
+       }
+#      define CV_StaticAssert(condition, reason)\
+         typedef cv::CV_StaticAssert_test< sizeof(cv::CV_StaticAssert_failed< static_cast<bool>(condition) >) > CVAUX_CONCAT(CV_StaticAssert_failed_at_, __LINE__)
+#    endif
+#  endif
+#endif
+
 namespace cv
 {
 
@@ -164,28 +201,28 @@ template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0)
 
 template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1)
 {
-    assert(channels >= 2);
+    CV_StaticAssert(channels >= 2, "Matx should have at least 2 elaments.");
     val[0] = v0; val[1] = v1;
     for(int i = 2; i < channels; i++) val[i] = _Tp(0);
 }
 
 template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2)
 {
-    assert(channels >= 3);
+    CV_StaticAssert(channels >= 3, "Matx should have at least 3 elaments.");
     val[0] = v0; val[1] = v1; val[2] = v2;
     for(int i = 3; i < channels; i++) val[i] = _Tp(0);
 }
 
 template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3)
 {
-    assert(channels >= 4);
+    CV_StaticAssert(channels >= 4, "Matx should have at least 4 elaments.");
     val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
     for(int i = 4; i < channels; i++) val[i] = _Tp(0);
 }
 
 template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3, _Tp v4)
 {
-    assert(channels >= 5);
+    CV_StaticAssert(channels >= 5, "Matx should have at least 5 elaments.");
     val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3; val[4] = v4;
     for(int i = 5; i < channels; i++) val[i] = _Tp(0);
 }
@@ -193,7 +230,7 @@ template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1
 template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
                                                         _Tp v4, _Tp v5)
 {
-    assert(channels >= 6);
+    CV_StaticAssert(channels >= 6, "Matx should have at least 6 elaments.");
     val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
     val[4] = v4; val[5] = v5;
     for(int i = 6; i < channels; i++) val[i] = _Tp(0);
@@ -202,7 +239,7 @@ template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1
 template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
                                                         _Tp v4, _Tp v5, _Tp v6)
 {
-    assert(channels >= 7);
+    CV_StaticAssert(channels >= 7, "Matx should have at least 7 elaments.");
     val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
     val[4] = v4; val[5] = v5; val[6] = v6;
     for(int i = 7; i < channels; i++) val[i] = _Tp(0);
@@ -211,7 +248,7 @@ template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1
 template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
                                                         _Tp v4, _Tp v5, _Tp v6, _Tp v7)
 {
-    assert(channels >= 8);
+    CV_StaticAssert(channels >= 8, "Matx should have at least 8 elaments.");
     val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
     val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
     for(int i = 8; i < channels; i++) val[i] = _Tp(0);
@@ -221,7 +258,7 @@ template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1
                                                         _Tp v4, _Tp v5, _Tp v6, _Tp v7,
                                                         _Tp v8)
 {
-    assert(channels >= 9);
+    CV_StaticAssert(channels >= 9, "Matx should have at least 9 elaments.");
     val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
     val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
     val[8] = v8;
@@ -232,7 +269,7 @@ template<typename _Tp, int m, int n> inline Matx<_Tp, m, n>::Matx(_Tp v0, _Tp v1
                                                         _Tp v4, _Tp v5, _Tp v6, _Tp v7,
                                                         _Tp v8, _Tp v9)
 {
-    assert(channels >= 10);
+    CV_StaticAssert(channels >= 10, "Matx should have at least 10 elaments.");
     val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
     val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
     val[8] = v8; val[9] = v9;
@@ -245,7 +282,7 @@ inline Matx<_Tp,m,n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
                             _Tp v4, _Tp v5, _Tp v6, _Tp v7,
                             _Tp v8, _Tp v9, _Tp v10, _Tp v11)
 {
-    assert(channels == 12);
+    CV_StaticAssert(channels == 12, "Matx should have at least 12 elaments.");
     val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
     val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
     val[8] = v8; val[9] = v9; val[10] = v10; val[11] = v11;
@@ -257,7 +294,7 @@ inline Matx<_Tp,m,n>::Matx(_Tp v0, _Tp v1, _Tp v2, _Tp v3,
                            _Tp v8, _Tp v9, _Tp v10, _Tp v11,
                            _Tp v12, _Tp v13, _Tp v14, _Tp v15)
 {
-    assert(channels == 16);
+    CV_StaticAssert(channels == 16, "Matx should have at least 16 elaments.");
     val[0] = v0; val[1] = v1; val[2] = v2; val[3] = v3;
     val[4] = v4; val[5] = v5; val[6] = v6; val[7] = v7;
     val[8] = v8; val[9] = v9; val[10] = v10; val[11] = v11;

modules/features2d/src/fast.cpp

@@ -100,61 +100,61 @@ void FAST_t(InputArray _img, std::vector<KeyPoint>& keypoints, int threshold, bo
     #if CV_SSE2
             if( patternSize == 16 )
             {
-            for(; j < img.cols - 16 - 3; j += 16, ptr += 16)
-            {
-                __m128i m0, m1;
-                __m128i v0 = _mm_loadu_si128((const __m128i*)ptr);
-                __m128i v1 = _mm_xor_si128(_mm_subs_epu8(v0, t), delta);
-                v0 = _mm_xor_si128(_mm_adds_epu8(v0, t), delta);
-
-                __m128i x0 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[0])), delta);
-                __m128i x1 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[quarterPatternSize])), delta);
-                __m128i x2 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[2*quarterPatternSize])), delta);
-                __m128i x3 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[3*quarterPatternSize])), delta);
-                m0 = _mm_and_si128(_mm_cmpgt_epi8(x0, v0), _mm_cmpgt_epi8(x1, v0));
-                m1 = _mm_and_si128(_mm_cmpgt_epi8(v1, x0), _mm_cmpgt_epi8(v1, x1));
-                m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x1, v0), _mm_cmpgt_epi8(x2, v0)));
-                m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x1), _mm_cmpgt_epi8(v1, x2)));
-                m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x2, v0), _mm_cmpgt_epi8(x3, v0)));
-                m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x2), _mm_cmpgt_epi8(v1, x3)));
-                m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x3, v0), _mm_cmpgt_epi8(x0, v0)));
-                m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x3), _mm_cmpgt_epi8(v1, x0)));
-                m0 = _mm_or_si128(m0, m1);
-                int mask = _mm_movemask_epi8(m0);
-                if( mask == 0 )
-                    continue;
-                if( (mask & 255) == 0 )
+                for(; j < img.cols - 16 - 3; j += 16, ptr += 16)
                 {
-                    j -= 8;
-                    ptr -= 8;
-                    continue;
-                }
+                    __m128i m0, m1;
+                    __m128i v0 = _mm_loadu_si128((const __m128i*)ptr);
+                    __m128i v1 = _mm_xor_si128(_mm_subs_epu8(v0, t), delta);
+                    v0 = _mm_xor_si128(_mm_adds_epu8(v0, t), delta);
+
+                    __m128i x0 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[0])), delta);
+                    __m128i x1 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[quarterPatternSize])), delta);
+                    __m128i x2 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[2*quarterPatternSize])), delta);
+                    __m128i x3 = _mm_sub_epi8(_mm_loadu_si128((const __m128i*)(ptr + pixel[3*quarterPatternSize])), delta);
+                    m0 = _mm_and_si128(_mm_cmpgt_epi8(x0, v0), _mm_cmpgt_epi8(x1, v0));
+                    m1 = _mm_and_si128(_mm_cmpgt_epi8(v1, x0), _mm_cmpgt_epi8(v1, x1));
+                    m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x1, v0), _mm_cmpgt_epi8(x2, v0)));
+                    m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x1), _mm_cmpgt_epi8(v1, x2)));
+                    m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x2, v0), _mm_cmpgt_epi8(x3, v0)));
+                    m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x2), _mm_cmpgt_epi8(v1, x3)));
+                    m0 = _mm_or_si128(m0, _mm_and_si128(_mm_cmpgt_epi8(x3, v0), _mm_cmpgt_epi8(x0, v0)));
+                    m1 = _mm_or_si128(m1, _mm_and_si128(_mm_cmpgt_epi8(v1, x3), _mm_cmpgt_epi8(v1, x0)));
+                    m0 = _mm_or_si128(m0, m1);
+                    int mask = _mm_movemask_epi8(m0);
+                    if( mask == 0 )
+                        continue;
+                    if( (mask & 255) == 0 )
+                    {
+                        j -= 8;
+                        ptr -= 8;
+                        continue;
+                    }
 
-                __m128i c0 = _mm_setzero_si128(), c1 = c0, max0 = c0, max1 = c0;
-                for( k = 0; k < N; k++ )
-                {
-                    __m128i x = _mm_xor_si128(_mm_loadu_si128((const __m128i*)(ptr + pixel[k])), delta);
-                    m0 = _mm_cmpgt_epi8(x, v0);
-                    m1 = _mm_cmpgt_epi8(v1, x);
+                    __m128i c0 = _mm_setzero_si128(), c1 = c0, max0 = c0, max1 = c0;
+                    for( k = 0; k < N; k++ )
+                    {
+                        __m128i x = _mm_xor_si128(_mm_loadu_si128((const __m128i*)(ptr + pixel[k])), delta);
+                        m0 = _mm_cmpgt_epi8(x, v0);
+                        m1 = _mm_cmpgt_epi8(v1, x);
 
-                    c0 = _mm_and_si128(_mm_sub_epi8(c0, m0), m0);
-                    c1 = _mm_and_si128(_mm_sub_epi8(c1, m1), m1);
+                        c0 = _mm_and_si128(_mm_sub_epi8(c0, m0), m0);
+                        c1 = _mm_and_si128(_mm_sub_epi8(c1, m1), m1);
 
-                    max0 = _mm_max_epu8(max0, c0);
-                    max1 = _mm_max_epu8(max1, c1);
-                }
+                        max0 = _mm_max_epu8(max0, c0);
+                        max1 = _mm_max_epu8(max1, c1);
+                    }
 
-                max0 = _mm_max_epu8(max0, max1);
-                int m = _mm_movemask_epi8(_mm_cmpgt_epi8(max0, K16));
+                    max0 = _mm_max_epu8(max0, max1);
+                    int m = _mm_movemask_epi8(_mm_cmpgt_epi8(max0, K16));
 
-                for( k = 0; m > 0 && k < 16; k++, m >>= 1 )
-                    if(m & 1)
-                    {
-                        cornerpos[ncorners++] = j+k;
-                        if(nonmax_suppression)
-                            curr[j+k] = (uchar)cornerScore<patternSize>(ptr+k, pixel, threshold);
-                    }
-            }
+                    for( k = 0; m > 0 && k < 16; k++, m >>= 1 )
+                        if(m & 1)
+                        {
+                            cornerpos[ncorners++] = j+k;
+                            if(nonmax_suppression)
+                                curr[j+k] = (uchar)cornerScore<patternSize>(ptr+k, pixel, threshold);
+                        }
+                }
             }
     #endif
             for( ; j < img.cols - 3; j++, ptr++ )

modules/ts/include/opencv2/ts/ts_perf.hpp

@@ -208,7 +208,6 @@ private:
 #define SANITY_CHECK_KEYPOINTS(array, ...) ::perf::Regression::addKeypoints(this, #array, array , ## __VA_ARGS__)
 #define SANITY_CHECK_MATCHES(array, ...) ::perf::Regression::addMatches(this, #array, array , ## __VA_ARGS__)
 
-#ifdef HAVE_CUDA
 class CV_EXPORTS GpuPerf
 {
 public:
@@ -216,9 +215,6 @@ public:
 };
 
 # define PERF_RUN_GPU()  ::perf::GpuPerf::targetDevice()
-#else
-# define PERF_RUN_GPU()  false
-#endif
 
 
 /*****************************************************************************************\

modules/ts/src/ts_perf.cpp

@@ -1324,12 +1324,14 @@ void perf::sort(std::vector<cv::KeyPoint>& pts, cv::InputOutputArray descriptors
 /*****************************************************************************************\
 *                                  ::perf::GpuPerf
 \*****************************************************************************************/
-#ifdef HAVE_CUDA
 bool perf::GpuPerf::targetDevice()
 {
+#ifdef HAVE_CUDA
     return !param_run_cpu;
-}
+#else
+    return false;
 #endif
+}
 
 /*****************************************************************************************\
 *                                  ::perf::PrintTo