core: rework code locality

- to reduce binaries size of FFmpeg Windows wrapper
- MinGW linker doesn't support -ffunction-sections (used for FFmpeg Windows wrapper)
- move code to improve locality with its used dependencies
- move UMat::dot() to matmul.dispatch.cpp (Mat::dot() is already there)
- move UMat::inv() to lapack.cpp
- move UMat::mul() to arithm.cpp
- move UMat:eye() to matrix_operations.cpp (near setIdentity() implementation)
- move normalize(): convert_scale.cpp => norm.cpp
- move convertAndUnrollScalar(): arithm.cpp => copy.cpp
- move scalarToRawData(): array.cpp => copy.cpp
- move transpose(): matrix_operations.cpp => matrix_transform.cpp
- move flip(), rotate(): copy.cpp => matrix_transform.cpp (rotate90 uses flip and transpose)
- add 'OPENCV_CORE_EXCLUDE_C_API' CMake variable to exclude compilation of C-API functions from the core module
- matrix_wrap.cpp: add compile-time checks for CUDA/OpenGL calls
- the steps above allow to reduce FFmpeg wrapper size for ~1.5Mb (initial size of OpenCV part is about 3Mb)

modules/core/CMakeLists.txt

@@ -145,6 +145,10 @@ ocv_target_link_libraries(${the_module} PRIVATE
     "${OPENCV_HAL_LINKER_LIBS}"
 )
 
+if(OPENCV_CORE_EXCLUDE_C_API)
+  ocv_target_compile_definitions(${the_module} PRIVATE "OPENCV_EXCLUDE_C_API=1")
+endif()
+
 if(HAVE_HPX)
   ocv_target_link_libraries(${the_module} LINK_PRIVATE "${HPX_LIBRARIES}")
 endif()

modules/core/src/arithm.cpp

@@ -57,26 +57,6 @@ namespace cv
 *                                   logical operations                                   *
 \****************************************************************************************/
 
-void convertAndUnrollScalar( const Mat& sc, int buftype, uchar* scbuf, size_t blocksize )
-{
-    int scn = (int)sc.total(), cn = CV_MAT_CN(buftype);
-    size_t esz = CV_ELEM_SIZE(buftype);
-    BinaryFunc cvtFn = getConvertFunc(sc.depth(), buftype);
-    CV_Assert(cvtFn);
-    cvtFn(sc.ptr(), 1, 0, 1, scbuf, 1, Size(std::min(cn, scn), 1), 0);
-    // unroll the scalar
-    if( scn < cn )
-    {
-        CV_Assert( scn == 1 );
-        size_t esz1 = CV_ELEM_SIZE1(buftype);
-        for( size_t i = esz1; i < esz; i++ )
-            scbuf[i] = scbuf[i - esz1];
-    }
-    for( size_t i = esz; i < blocksize*esz; i++ )
-        scbuf[i] = scbuf[i - esz];
-}
-
-
 enum { OCL_OP_ADD=0, OCL_OP_SUB=1, OCL_OP_RSUB=2, OCL_OP_ABSDIFF=3, OCL_OP_MUL=4,
        OCL_OP_MUL_SCALE=5, OCL_OP_DIV_SCALE=6, OCL_OP_RECIP_SCALE=7, OCL_OP_ADDW=8,
        OCL_OP_AND=9, OCL_OP_OR=10, OCL_OP_XOR=11, OCL_OP_NOT=12, OCL_OP_MIN=13, OCL_OP_MAX=14,
@@ -1002,9 +982,7 @@ static BinaryFuncC* getRecipTab()
     return recipTab;
 }
 
-}
-
-void cv::multiply(InputArray src1, InputArray src2,
+void multiply(InputArray src1, InputArray src2,
                   OutputArray dst, double scale, int dtype)
 {
     CV_INSTRUMENT_REGION();
@@ -1013,7 +991,7 @@ void cv::multiply(InputArray src1, InputArray src2,
               true, &scale, std::abs(scale - 1.0) < DBL_EPSILON ? OCL_OP_MUL : OCL_OP_MUL_SCALE);
 }
 
-void cv::divide(InputArray src1, InputArray src2,
+void divide(InputArray src1, InputArray src2,
                 OutputArray dst, double scale, int dtype)
 {
     CV_INSTRUMENT_REGION();
@@ -1021,7 +999,7 @@ void cv::divide(InputArray src1, InputArray src2,
     arithm_op(src1, src2, dst, noArray(), dtype, getDivTab(), true, &scale, OCL_OP_DIV_SCALE);
 }
 
-void cv::divide(double scale, InputArray src2,
+void divide(double scale, InputArray src2,
                 OutputArray dst, int dtype)
 {
     CV_INSTRUMENT_REGION();
@@ -1029,13 +1007,17 @@ void cv::divide(double scale, InputArray src2,
     arithm_op(src2, src2, dst, noArray(), dtype, getRecipTab(), true, &scale, OCL_OP_RECIP_SCALE);
 }
 
+UMat UMat::mul(InputArray m, double scale) const
+{
+    UMat dst;
+    multiply(*this, m, dst, scale);
+    return dst;
+}
+
 /****************************************************************************************\
 *                                      addWeighted                                       *
 \****************************************************************************************/
 
-namespace cv
-{
-
 static BinaryFuncC* getAddWeightedTab()
 {
     static BinaryFuncC addWeightedTab[] =
@@ -1849,6 +1831,9 @@ void cv::inRange(InputArray _src, InputArray _lowerb,
     }
 }
 
+
+#ifndef OPENCV_EXCLUDE_C_API
+
 /****************************************************************************************\
 *                                Earlier API: cvAdd etc.                                 *
 \****************************************************************************************/
@@ -2111,4 +2096,5 @@ cvMaxS( const void* srcarr1, double value, void* dstarr )
     cv::max( src1, value, dst );
 }
 
+#endif  // OPENCV_EXCLUDE_C_API
 /* End of file. */

modules/core/src/array.cpp

@@ -48,6 +48,8 @@
 
 #include "precomp.hpp"
 
+#ifndef OPENCV_EXCLUDE_C_API
+
 #define  CV_ORIGIN_TL  0
 #define  CV_ORIGIN_BL  1
 
@@ -3211,53 +3213,6 @@ void DefaultDeleter<CvMatND>::operator ()(CvMatND* obj) const { cvReleaseMatND(&
 void DefaultDeleter<CvSparseMat>::operator ()(CvSparseMat* obj) const { cvReleaseSparseMat(&obj); }
 void DefaultDeleter<CvMemStorage>::operator ()(CvMemStorage* obj) const { cvReleaseMemStorage(&obj); }
 
-template <typename T> static inline
-void scalarToRawData_(const Scalar& s, T * const buf, const int cn, const int unroll_to)
-{
-    int i = 0;
-    for(; i < cn; i++)
-        buf[i] = saturate_cast<T>(s.val[i]);
-    for(; i < unroll_to; i++)
-        buf[i] = buf[i-cn];
-}
-
-void scalarToRawData(const Scalar& s, void* _buf, int type, int unroll_to)
-{
-    CV_INSTRUMENT_REGION();
-
-    const int depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type);
-    CV_Assert(cn <= 4);
-    switch(depth)
-    {
-    case CV_8U:
-        scalarToRawData_<uchar>(s, (uchar*)_buf, cn, unroll_to);
-        break;
-    case CV_8S:
-        scalarToRawData_<schar>(s, (schar*)_buf, cn, unroll_to);
-        break;
-    case CV_16U:
-        scalarToRawData_<ushort>(s, (ushort*)_buf, cn, unroll_to);
-        break;
-    case CV_16S:
-        scalarToRawData_<short>(s, (short*)_buf, cn, unroll_to);
-        break;
-    case CV_32S:
-        scalarToRawData_<int>(s, (int*)_buf, cn, unroll_to);
-        break;
-    case CV_32F:
-        scalarToRawData_<float>(s, (float*)_buf, cn, unroll_to);
-        break;
-    case CV_64F:
-        scalarToRawData_<double>(s, (double*)_buf, cn, unroll_to);
-        break;
-    case CV_16F:
-        scalarToRawData_<float16_t>(s, (float16_t*)_buf, cn, unroll_to);
-        break;
-    default:
-        CV_Error(CV_StsUnsupportedFormat,"");
-    }
-}
-
 } // cv::
 
 
@@ -3295,4 +3250,5 @@ void* cvClone( const void* struct_ptr )
 }
 
 
+#endif  // OPENCV_EXCLUDE_C_API
 /* End of file. */

modules/core/src/convert_c.cpp

@@ -5,6 +5,7 @@
 
 #include "precomp.hpp"
 
+#ifndef OPENCV_EXCLUDE_C_API
 
 CV_IMPL void
 cvSplit( const void* srcarr, void* dstarr0, void* dstarr1, void* dstarr2, void* dstarr3 )
@@ -132,3 +133,5 @@ CV_IMPL void cvNormalize( const CvArr* srcarr, CvArr* dstarr,
     CV_Assert( dst.size() == src.size() && src.channels() == dst.channels() );
     cv::normalize( src, dst, a, b, norm_type, dst.type(), mask );
 }
+
+#endif  // OPENCV_EXCLUDE_C_API

modules/core/src/convert_scale.dispatch.cpp

@@ -9,7 +9,6 @@
 #include "convert_scale.simd.hpp"
 #include "convert_scale.simd_declarations.hpp" // defines CV_CPU_DISPATCH_MODES_ALL=AVX2,...,BASELINE based on CMakeLists.txt content
 
-
 namespace cv
 {
 
@@ -117,143 +116,4 @@ void convertScaleAbs(InputArray _src, OutputArray _dst, double alpha, double bet
     }
 }
 
-//==================================================================================================
-
-#ifdef HAVE_OPENCL
-
-static bool ocl_normalize( InputArray _src, InputOutputArray _dst, InputArray _mask, int dtype,
-                           double scale, double delta )
-{
-    UMat src = _src.getUMat();
-
-    if( _mask.empty() )
-        src.convertTo( _dst, dtype, scale, delta );
-    else if (src.channels() <= 4)
-    {
-        const ocl::Device & dev = ocl::Device::getDefault();
-
-        int stype = _src.type(), sdepth = CV_MAT_DEPTH(stype), cn = CV_MAT_CN(stype),
-                ddepth = CV_MAT_DEPTH(dtype), wdepth = std::max(CV_32F, std::max(sdepth, ddepth)),
-                rowsPerWI = dev.isIntel() ? 4 : 1;
-
-        float fscale = static_cast<float>(scale), fdelta = static_cast<float>(delta);
-        bool haveScale = std::fabs(scale - 1) > DBL_EPSILON,
-                haveZeroScale = !(std::fabs(scale) > DBL_EPSILON),
-                haveDelta = std::fabs(delta) > DBL_EPSILON,
-                doubleSupport = dev.doubleFPConfig() > 0;
-
-        if (!haveScale && !haveDelta && stype == dtype)
-        {
-            _src.copyTo(_dst, _mask);
-            return true;
-        }
-        if (haveZeroScale)
-        {
-            _dst.setTo(Scalar(delta), _mask);
-            return true;
-        }
-
-        if ((sdepth == CV_64F || ddepth == CV_64F) && !doubleSupport)
-            return false;
-
-        char cvt[2][40];
-        String opts = format("-D srcT=%s -D dstT=%s -D convertToWT=%s -D cn=%d -D rowsPerWI=%d"
-                             " -D convertToDT=%s -D workT=%s%s%s%s -D srcT1=%s -D dstT1=%s",
-                             ocl::typeToStr(stype), ocl::typeToStr(dtype),
-                             ocl::convertTypeStr(sdepth, wdepth, cn, cvt[0]), cn,
-                             rowsPerWI, ocl::convertTypeStr(wdepth, ddepth, cn, cvt[1]),
-                             ocl::typeToStr(CV_MAKE_TYPE(wdepth, cn)),
-                             doubleSupport ? " -D DOUBLE_SUPPORT" : "",
-                             haveScale ? " -D HAVE_SCALE" : "",
-                             haveDelta ? " -D HAVE_DELTA" : "",
-                             ocl::typeToStr(sdepth), ocl::typeToStr(ddepth));
-
-        ocl::Kernel k("normalizek", ocl::core::normalize_oclsrc, opts);
-        if (k.empty())
-            return false;
-
-        UMat mask = _mask.getUMat(), dst = _dst.getUMat();
-
-        ocl::KernelArg srcarg = ocl::KernelArg::ReadOnlyNoSize(src),
-                maskarg = ocl::KernelArg::ReadOnlyNoSize(mask),
-                dstarg = ocl::KernelArg::ReadWrite(dst);
-
-        if (haveScale)
-        {
-            if (haveDelta)
-                k.args(srcarg, maskarg, dstarg, fscale, fdelta);
-            else
-                k.args(srcarg, maskarg, dstarg, fscale);
-        }
-        else
-        {
-            if (haveDelta)
-                k.args(srcarg, maskarg, dstarg, fdelta);
-            else
-                k.args(srcarg, maskarg, dstarg);
-        }
-
-        size_t globalsize[2] = { (size_t)src.cols, ((size_t)src.rows + rowsPerWI - 1) / rowsPerWI };
-        return k.run(2, globalsize, NULL, false);
-    }
-    else
-    {
-        UMat temp;
-        src.convertTo( temp, dtype, scale, delta );
-        temp.copyTo( _dst, _mask );
-    }
-
-    return true;
-}
-
-#endif
-
-void normalize(InputArray _src, InputOutputArray _dst, double a, double b,
-               int norm_type, int rtype, InputArray _mask)
-{
-    CV_INSTRUMENT_REGION();
-
-    double scale = 1, shift = 0;
-    int type = _src.type(), depth = CV_MAT_DEPTH(type);
-
-    if( rtype < 0 )
-        rtype = _dst.fixedType() ? _dst.depth() : depth;
-
-    if( norm_type == CV_MINMAX )
-    {
-        double smin = 0, smax = 0;
-        double dmin = MIN( a, b ), dmax = MAX( a, b );
-        minMaxIdx( _src, &smin, &smax, 0, 0, _mask );
-        scale = (dmax - dmin)*(smax - smin > DBL_EPSILON ? 1./(smax - smin) : 0);
-        if( rtype == CV_32F )
-        {
-            scale = (float)scale;
-            shift = (float)dmin - (float)(smin*scale);
-        }
-        else
-            shift = dmin - smin*scale;
-    }
-    else if( norm_type == CV_L2 || norm_type == CV_L1 || norm_type == CV_C )
-    {
-        scale = norm( _src, norm_type, _mask );
-        scale = scale > DBL_EPSILON ? a/scale : 0.;
-        shift = 0;
-    }
-    else
-        CV_Error( CV_StsBadArg, "Unknown/unsupported norm type" );
-
-    CV_OCL_RUN(_dst.isUMat(),
-               ocl_normalize(_src, _dst, _mask, rtype, scale, shift))
-
-    Mat src = _src.getMat();
-    if( _mask.empty() )
-        src.convertTo( _dst, rtype, scale, shift );
-    else
-    {
-        Mat temp;
-        src.convertTo( temp, rtype, scale, shift );
-        temp.copyTo( _dst, _mask );
-    }
-}
-
 } // namespace

modules/core/src/copy.cpp

@@ -53,6 +53,72 @@
 namespace cv
 {
 
+template <typename T> static inline
+void scalarToRawData_(const Scalar& s, T * const buf, const int cn, const int unroll_to)
+{
+    int i = 0;
+    for(; i < cn; i++)
+        buf[i] = saturate_cast<T>(s.val[i]);
+    for(; i < unroll_to; i++)
+        buf[i] = buf[i-cn];
+}
+
+void scalarToRawData(const Scalar& s, void* _buf, int type, int unroll_to)
+{
+    CV_INSTRUMENT_REGION();
+
+    const int depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type);
+    CV_Assert(cn <= 4);
+    switch(depth)
+    {
+    case CV_8U:
+        scalarToRawData_<uchar>(s, (uchar*)_buf, cn, unroll_to);
+        break;
+    case CV_8S:
+        scalarToRawData_<schar>(s, (schar*)_buf, cn, unroll_to);
+        break;
+    case CV_16U:
+        scalarToRawData_<ushort>(s, (ushort*)_buf, cn, unroll_to);
+        break;
+    case CV_16S:
+        scalarToRawData_<short>(s, (short*)_buf, cn, unroll_to);
+        break;
+    case CV_32S:
+        scalarToRawData_<int>(s, (int*)_buf, cn, unroll_to);
+        break;
+    case CV_32F:
+        scalarToRawData_<float>(s, (float*)_buf, cn, unroll_to);
+        break;
+    case CV_64F:
+        scalarToRawData_<double>(s, (double*)_buf, cn, unroll_to);
+        break;
+    case CV_16F:
+        scalarToRawData_<float16_t>(s, (float16_t*)_buf, cn, unroll_to);
+        break;
+    default:
+        CV_Error(CV_StsUnsupportedFormat,"");
+    }
+}
+
+void convertAndUnrollScalar( const Mat& sc, int buftype, uchar* scbuf, size_t blocksize )
+{
+    int scn = (int)sc.total(), cn = CV_MAT_CN(buftype);
+    size_t esz = CV_ELEM_SIZE(buftype);
+    BinaryFunc cvtFn = getConvertFunc(sc.depth(), buftype);
+    CV_Assert(cvtFn);
+    cvtFn(sc.ptr(), 1, 0, 1, scbuf, 1, Size(std::min(cn, scn), 1), 0);
+    // unroll the scalar
+    if( scn < cn )
+    {
+        CV_Assert( scn == 1 );
+        size_t esz1 = CV_ELEM_SIZE1(buftype);
+        for( size_t i = esz1; i < esz; i++ )
+            scbuf[i] = scbuf[i - esz1];
+    }
+    for( size_t i = esz; i < blocksize*esz; i++ )
+        scbuf[i] = scbuf[i - esz];
+}
+
 template<typename T> static void
 copyMask_(const uchar* _src, size_t sstep, const uchar* mask, size_t mstep, uchar* _dst, size_t dstep, Size size)
 {
@@ -594,490 +660,6 @@ Mat& Mat::setTo(InputArray _value, InputArray _mask)
     return *this;
 }
 
-#if CV_SIMD128
-template<typename V> CV_ALWAYS_INLINE void flipHoriz_single( const uchar* src, size_t sstep, uchar* dst, size_t dstep, Size size, size_t esz )
-{
-    typedef typename V::lane_type T;
-    int end = (int)(size.width*esz);
-    int width = (end + 1)/2;
-    int width_1 = width & -v_uint8x16::nlanes;
-    int i, j;
-
-#if CV_STRONG_ALIGNMENT
-    CV_Assert(isAligned<sizeof(T)>(src, dst));
-#endif
-
-    for( ; size.height--; src += sstep, dst += dstep )
-    {
-        for( i = 0, j = end; i < width_1; i += v_uint8x16::nlanes, j -= v_uint8x16::nlanes )
-        {
-            V t0, t1;
-
-            t0 = v_load((T*)((uchar*)src + i));
-            t1 = v_load((T*)((uchar*)src + j - v_uint8x16::nlanes));
-            t0 = v_reverse(t0);
-            t1 = v_reverse(t1);
-            v_store((T*)(dst + j - v_uint8x16::nlanes), t0);
-            v_store((T*)(dst + i), t1);
-        }
-        if (isAligned<sizeof(T)>(src, dst))
-        {
-            for ( ; i < width; i += sizeof(T), j -= sizeof(T) )
-            {
-                T t0, t1;
-
-                t0 = *((T*)((uchar*)src + i));
-                t1 = *((T*)((uchar*)src + j - sizeof(T)));
-                *((T*)(dst + j - sizeof(T))) = t0;
-                *((T*)(dst + i)) = t1;
-            }
-        }
-        else
-        {
-            for ( ; i < width; i += sizeof(T), j -= sizeof(T) )
-            {
-                for (int k = 0; k < (int)sizeof(T); k++)
-                {
-                    uchar t0, t1;
-
-                    t0 = *((uchar*)src + i + k);
-                    t1 = *((uchar*)src + j + k - sizeof(T));
-                    *(dst + j + k - sizeof(T)) = t0;
-                    *(dst + i + k) = t1;
-                }
-            }
-        }
-    }
-}
-
-template<typename T1, typename T2> CV_ALWAYS_INLINE void flipHoriz_double( const uchar* src, size_t sstep, uchar* dst, size_t dstep, Size size, size_t esz )
-{
-    int end = (int)(size.width*esz);
-    int width = (end + 1)/2;
-
-#if CV_STRONG_ALIGNMENT
-    CV_Assert(isAligned<sizeof(T1)>(src, dst));
-    CV_Assert(isAligned<sizeof(T2)>(src, dst));
-#endif
-
-    for( ; size.height--; src += sstep, dst += dstep )
-    {
-        for ( int i = 0, j = end; i < width; i += sizeof(T1) + sizeof(T2), j -= sizeof(T1) + sizeof(T2) )
-        {
-            T1 t0, t1;
-            T2 t2, t3;
-
-            t0 = *((T1*)((uchar*)src + i));
-            t2 = *((T2*)((uchar*)src + i + sizeof(T1)));
-            t1 = *((T1*)((uchar*)src + j - sizeof(T1) - sizeof(T2)));
-            t3 = *((T2*)((uchar*)src + j - sizeof(T2)));
-            *((T1*)(dst + j - sizeof(T1) - sizeof(T2))) = t0;
-            *((T2*)(dst + j - sizeof(T2))) = t2;
-            *((T1*)(dst + i)) = t1;
-            *((T2*)(dst + i + sizeof(T1))) = t3;
-        }
-    }
-}
-#endif
-
-static void
-flipHoriz( const uchar* src, size_t sstep, uchar* dst, size_t dstep, Size size, size_t esz )
-{
-#if CV_SIMD
-#if CV_STRONG_ALIGNMENT
-    size_t alignmentMark = ((size_t)src)|((size_t)dst)|sstep|dstep;
-#endif
-    if (esz == 2 * v_uint8x16::nlanes)
-    {
-        int end = (int)(size.width*esz);
-        int width = end/2;
-
-        for( ; size.height--; src += sstep, dst += dstep )
-        {
-            for( int i = 0, j = end - 2 * v_uint8x16::nlanes; i < width; i += 2 * v_uint8x16::nlanes, j -= 2 * v_uint8x16::nlanes )
-            {
-#if CV_SIMD256
-                v_uint8x32 t0, t1;
-
-                t0 = v256_load((uchar*)src + i);
-                t1 = v256_load((uchar*)src + j);
-                v_store(dst + j, t0);
-                v_store(dst + i, t1);
-#else
-                v_uint8x16 t0, t1, t2, t3;
-
-                t0 = v_load((uchar*)src + i);
-                t1 = v_load((uchar*)src + i + v_uint8x16::nlanes);
-                t2 = v_load((uchar*)src + j);
-                t3 = v_load((uchar*)src + j + v_uint8x16::nlanes);
-                v_store(dst + j, t0);
-                v_store(dst + j + v_uint8x16::nlanes, t1);
-                v_store(dst + i, t2);
-                v_store(dst + i + v_uint8x16::nlanes, t3);
-#endif
-            }
-        }
-    }
-    else if (esz == v_uint8x16::nlanes)
-    {
-        int end = (int)(size.width*esz);
-        int width = end/2;
-
-        for( ; size.height--; src += sstep, dst += dstep )
-        {
-            for( int i = 0, j = end - v_uint8x16::nlanes; i < width; i += v_uint8x16::nlanes, j -= v_uint8x16::nlanes )
-            {
-                v_uint8x16 t0, t1;
-
-                t0 = v_load((uchar*)src + i);
-                t1 = v_load((uchar*)src + j);
-                v_store(dst + j, t0);
-                v_store(dst + i, t1);
-            }
-        }
-    }
-    else if (esz == 8
-#if CV_STRONG_ALIGNMENT
-            && isAligned<sizeof(uint64)>(alignmentMark)
-#endif
-    )
-    {
-        flipHoriz_single<v_uint64x2>(src, sstep, dst, dstep, size, esz);
-    }
-    else if (esz == 4
-#if CV_STRONG_ALIGNMENT
-            && isAligned<sizeof(unsigned)>(alignmentMark)
-#endif
-    )
-    {
-        flipHoriz_single<v_uint32x4>(src, sstep, dst, dstep, size, esz);
-    }
-    else if (esz == 2
-#if CV_STRONG_ALIGNMENT
-            && isAligned<sizeof(ushort)>(alignmentMark)
-#endif
-    )
-    {
-        flipHoriz_single<v_uint16x8>(src, sstep, dst, dstep, size, esz);
-    }
-    else if (esz == 1)
-    {
-        flipHoriz_single<v_uint8x16>(src, sstep, dst, dstep, size, esz);
-    }
-    else if (esz == 24
-#if CV_STRONG_ALIGNMENT
-            && isAligned<sizeof(uint64_t)>(alignmentMark)
-#endif
-    )
-    {
-        int end = (int)(size.width*esz);
-        int width = (end + 1)/2;
-
-        for( ; size.height--; src += sstep, dst += dstep )
-        {
-            for ( int i = 0, j = end; i < width; i += v_uint8x16::nlanes + sizeof(uint64_t), j -= v_uint8x16::nlanes + sizeof(uint64_t) )
-            {
-                v_uint8x16 t0, t1;
-                uint64_t t2, t3;
-
-                t0 = v_load((uchar*)src + i);
-                t2 = *((uint64_t*)((uchar*)src + i + v_uint8x16::nlanes));
-                t1 = v_load((uchar*)src + j - v_uint8x16::nlanes - sizeof(uint64_t));
-                t3 = *((uint64_t*)((uchar*)src + j - sizeof(uint64_t)));
-                v_store(dst + j - v_uint8x16::nlanes - sizeof(uint64_t), t0);
-                *((uint64_t*)(dst + j - sizeof(uint64_t))) = t2;
-                v_store(dst + i, t1);
-                *((uint64_t*)(dst + i + v_uint8x16::nlanes)) = t3;
-            }
-        }
-    }
-#if !CV_STRONG_ALIGNMENT
-    else if (esz == 12)
-    {
-        flipHoriz_double<uint64_t,uint>(src, sstep, dst, dstep, size, esz);
-    }
-    else if (esz == 6)
-    {
-        flipHoriz_double<uint,ushort>(src, sstep, dst, dstep, size, esz);
-    }
-    else if (esz == 3)
-    {
-        flipHoriz_double<ushort,uchar>(src, sstep, dst, dstep, size, esz);
-    }
-#endif
-    else
-#endif // CV_SIMD
-    {
-        int i, j, limit = (int)(((size.width + 1)/2)*esz);
-        AutoBuffer<int> _tab(size.width*esz);
-        int* tab = _tab.data();
-
-        for( i = 0; i < size.width; i++ )
-            for( size_t k = 0; k < esz; k++ )
-                tab[i*esz + k] = (int)((size.width - i - 1)*esz + k);
-
-        for( ; size.height--; src += sstep, dst += dstep )
-        {
-            for( i = 0; i < limit; i++ )
-            {
-                j = tab[i];
-                uchar t0 = src[i], t1 = src[j];
-                dst[i] = t1; dst[j] = t0;
-            }
-        }
-    }
-}
-
-static void
-flipVert( const uchar* src0, size_t sstep, uchar* dst0, size_t dstep, Size size, size_t esz )
-{
-    const uchar* src1 = src0 + (size.height - 1)*sstep;
-    uchar* dst1 = dst0 + (size.height - 1)*dstep;
-    size.width *= (int)esz;
-
-    for( int y = 0; y < (size.height + 1)/2; y++, src0 += sstep, src1 -= sstep,
-                                                  dst0 += dstep, dst1 -= dstep )
-    {
-        int i = 0;
-#if CV_SIMD
-#if CV_STRONG_ALIGNMENT
-        if (isAligned<sizeof(int)>(src0, src1, dst0, dst1))
-#endif
-        {
-            for (; i <= size.width - CV_SIMD_WIDTH; i += CV_SIMD_WIDTH)
-            {
-                v_int32 t0 = vx_load((int*)(src0 + i));
-                v_int32 t1 = vx_load((int*)(src1 + i));
-                vx_store((int*)(dst0 + i), t1);
-                vx_store((int*)(dst1 + i), t0);
-            }
-        }
-#if CV_STRONG_ALIGNMENT
-        else
-        {
-            for (; i <= size.width - CV_SIMD_WIDTH; i += CV_SIMD_WIDTH)
-            {
-                v_uint8 t0 = vx_load(src0 + i);
-                v_uint8 t1 = vx_load(src1 + i);
-                vx_store(dst0 + i, t1);
-                vx_store(dst1 + i, t0);
-            }
-        }
-#endif
-#endif
-
-        if (isAligned<sizeof(int)>(src0, src1, dst0, dst1))
-        {
-            for( ; i <= size.width - 16; i += 16 )
-            {
-                int t0 = ((int*)(src0 + i))[0];
-                int t1 = ((int*)(src1 + i))[0];
-
-                ((int*)(dst0 + i))[0] = t1;
-                ((int*)(dst1 + i))[0] = t0;
-
-                t0 = ((int*)(src0 + i))[1];
-                t1 = ((int*)(src1 + i))[1];
-
-                ((int*)(dst0 + i))[1] = t1;
-                ((int*)(dst1 + i))[1] = t0;
-
-                t0 = ((int*)(src0 + i))[2];
-                t1 = ((int*)(src1 + i))[2];
-
-                ((int*)(dst0 + i))[2] = t1;
-                ((int*)(dst1 + i))[2] = t0;
-
-                t0 = ((int*)(src0 + i))[3];
-                t1 = ((int*)(src1 + i))[3];
-
-                ((int*)(dst0 + i))[3] = t1;
-                ((int*)(dst1 + i))[3] = t0;
-            }
-
-            for( ; i <= size.width - 4; i += 4 )
-            {
-                int t0 = ((int*)(src0 + i))[0];
-                int t1 = ((int*)(src1 + i))[0];
-
-                ((int*)(dst0 + i))[0] = t1;
-                ((int*)(dst1 + i))[0] = t0;
-            }
-        }
-
-        for( ; i < size.width; i++ )
-        {
-            uchar t0 = src0[i];
-            uchar t1 = src1[i];
-
-            dst0[i] = t1;
-            dst1[i] = t0;
-        }
-    }
-}
-
-#ifdef HAVE_OPENCL
-
-enum { FLIP_COLS = 1 << 0, FLIP_ROWS = 1 << 1, FLIP_BOTH = FLIP_ROWS | FLIP_COLS };
-
-static bool ocl_flip(InputArray _src, OutputArray _dst, int flipCode )
-{
-    CV_Assert(flipCode >= -1 && flipCode <= 1);
-
-    const ocl::Device & dev = ocl::Device::getDefault();
-    int type = _src.type(), depth = CV_MAT_DEPTH(type), cn = CV_MAT_CN(type),
-            flipType, kercn = std::min(ocl::predictOptimalVectorWidth(_src, _dst), 4);
-
-    bool doubleSupport = dev.doubleFPConfig() > 0;
-    if (!doubleSupport && depth == CV_64F)
-        kercn = cn;
-
-    if (cn > 4)
-        return false;
-
-    const char * kernelName;
-    if (flipCode == 0)
-        kernelName = "arithm_flip_rows", flipType = FLIP_ROWS;
-    else if (flipCode > 0)
-        kernelName = "arithm_flip_cols", flipType = FLIP_COLS;
-    else
-        kernelName = "arithm_flip_rows_cols", flipType = FLIP_BOTH;
-
-    int pxPerWIy = (dev.isIntel() && (dev.type() & ocl::Device::TYPE_GPU)) ? 4 : 1;
-    kercn = (cn!=3 || flipType == FLIP_ROWS) ? std::max(kercn, cn) : cn;
-
-    ocl::Kernel k(kernelName, ocl::core::flip_oclsrc,
-        format( "-D T=%s -D T1=%s -D DEPTH=%d -D cn=%d -D PIX_PER_WI_Y=%d -D kercn=%d",
-                kercn != cn ? ocl::typeToStr(CV_MAKE_TYPE(depth, kercn)) : ocl::vecopTypeToStr(CV_MAKE_TYPE(depth, kercn)),
-                kercn != cn ? ocl::typeToStr(depth) : ocl::vecopTypeToStr(depth), depth, cn, pxPerWIy, kercn));
-    if (k.empty())
-        return false;
-
-    Size size = _src.size();
-    _dst.create(size, type);
-    UMat src = _src.getUMat(), dst = _dst.getUMat();
-
-    int cols = size.width * cn / kercn, rows = size.height;
-    cols = flipType == FLIP_COLS ? (cols + 1) >> 1 : cols;
-    rows = flipType & FLIP_ROWS ? (rows + 1) >> 1 : rows;
-
-    k.args(ocl::KernelArg::ReadOnlyNoSize(src),
-           ocl::KernelArg::WriteOnly(dst, cn, kercn), rows, cols);
-
-    size_t maxWorkGroupSize = dev.maxWorkGroupSize();
-    CV_Assert(maxWorkGroupSize % 4 == 0);
-
-    size_t globalsize[2] = { (size_t)cols, ((size_t)rows + pxPerWIy - 1) / pxPerWIy },
-            localsize[2] = { maxWorkGroupSize / 4, 4 };
-    return k.run(2, globalsize, (flipType == FLIP_COLS) && !dev.isIntel() ? localsize : NULL, false);
-}
-
-#endif
-
-#if defined HAVE_IPP
-static bool ipp_flip(Mat &src, Mat &dst, int flip_mode)
-{
-#ifdef HAVE_IPP_IW
-    CV_INSTRUMENT_REGION_IPP();
-
-    // Details: https://github.com/opencv/opencv/issues/12943
-    if (flip_mode <= 0 /* swap rows */
-        && cv::ipp::getIppTopFeatures() != ippCPUID_SSE42
-        && (int64_t)(src.total()) * src.elemSize() >= CV_BIG_INT(0x80000000)/*2Gb*/
-    )
-        return false;
-
-    IppiAxis ippMode;
-    if(flip_mode < 0)
-        ippMode = ippAxsBoth;
-    else if(flip_mode == 0)
-        ippMode = ippAxsHorizontal;
-    else
-        ippMode = ippAxsVertical;
-
-    try
-    {
-        ::ipp::IwiImage iwSrc = ippiGetImage(src);
-        ::ipp::IwiImage iwDst = ippiGetImage(dst);
-
-        CV_INSTRUMENT_FUN_IPP(::ipp::iwiMirror, iwSrc, iwDst, ippMode);
-    }
-    catch(const ::ipp::IwException &)
-    {
-        return false;
-    }
-
-    return true;
-#else
-    CV_UNUSED(src); CV_UNUSED(dst); CV_UNUSED(flip_mode);
-    return false;
-#endif
-}
-#endif
-
-
-void flip( InputArray _src, OutputArray _dst, int flip_mode )
-{
-    CV_INSTRUMENT_REGION();
-
-    CV_Assert( _src.dims() <= 2 );
-    Size size = _src.size();
-
-    if (flip_mode < 0)
-    {
-        if (size.width == 1)
-            flip_mode = 0;
-        if (size.height == 1)
-            flip_mode = 1;
-    }
-
-    if ((size.width == 1 && flip_mode > 0) ||
-        (size.height == 1 && flip_mode == 0))
-    {
-        return _src.copyTo(_dst);
-    }
-
-    CV_OCL_RUN( _dst.isUMat(), ocl_flip(_src, _dst, flip_mode))
-
-    Mat src = _src.getMat();
-    int type = src.type();
-    _dst.create( size, type );
-    Mat dst = _dst.getMat();
-
-    CV_IPP_RUN_FAST(ipp_flip(src, dst, flip_mode));
-
-    size_t esz = CV_ELEM_SIZE(type);
-
-    if( flip_mode <= 0 )
-        flipVert( src.ptr(), src.step, dst.ptr(), dst.step, src.size(), esz );
-    else
-        flipHoriz( src.ptr(), src.step, dst.ptr(), dst.step, src.size(), esz );
-
-    if( flip_mode < 0 )
-        flipHoriz( dst.ptr(), dst.step, dst.ptr(), dst.step, dst.size(), esz );
-}
-
-void rotate(InputArray _src, OutputArray _dst, int rotateMode)
-{
-    CV_Assert(_src.dims() <= 2);
-
-    switch (rotateMode)
-    {
-    case ROTATE_90_CLOCKWISE:
-        transpose(_src, _dst);
-        flip(_dst, _dst, 1);
-        break;
-    case ROTATE_180:
-        flip(_src, _dst, -1);
-        break;
-    case ROTATE_90_COUNTERCLOCKWISE:
-        transpose(_src, _dst);
-        flip(_dst, _dst, 0);
-        break;
-    default:
-        break;
-    }
-}
 
 #if defined HAVE_OPENCL && !defined __APPLE__
 
@@ -1499,6 +1081,9 @@ void cv::copyMakeBorder( InputArray _src, OutputArray _dst, int top, int bottom,
     }
 }
 
+
+#ifndef OPENCV_EXCLUDE_C_API
+
 /* dst = src */
 CV_IMPL void
 cvCopy( const void* srcarr, void* dstarr, const void* maskarr )
@@ -1614,4 +1199,5 @@ cvRepeat( const CvArr* srcarr, CvArr* dstarr )
     cv::repeat(src, dst.rows/src.rows, dst.cols/src.cols, dst);
 }
 
+#endif  // OPENCV_EXCLUDE_C_API
 /* End of file. */

modules/core/src/datastructs.cpp

@@ -40,6 +40,8 @@
 //M*/
 #include "precomp.hpp"
 
+#ifndef OPENCV_EXCLUDE_C_API
+
 /* default alignment for dynamic data strucutures, resided in storages. */
 #define  CV_STRUCT_ALIGN    ((int)sizeof(double))
 
@@ -3585,4 +3587,5 @@ void  seqInsertSlice( CvSeq* seq, int before_index, const CvArr* from_arr )
 
 }
 
+#endif  // OPENCV_EXCLUDE_C_API
 /* End of file. */

modules/core/src/dxt.cpp

@@ -4640,6 +4640,9 @@ int cv::getOptimalDFTSize( int size0 )
     return optimalDFTSizeTab[b];
 }
 
+
+#ifndef OPENCV_EXCLUDE_C_API
+
 CV_IMPL void
 cvDFT( const CvArr* srcarr, CvArr* dstarr, int flags, int nonzero_rows )
 {
@@ -4695,4 +4698,5 @@ cvGetOptimalDFTSize( int size0 )
     return cv::getOptimalDFTSize(size0);
 }
 
+#endif  // OPENCV_EXCLUDE_C_API
 /* End of file. */

modules/core/src/lapack.cpp

@@ -753,8 +753,6 @@ SVBkSb( int m, int n, const double* w, size_t wstep,
                 (double*)alignPtr(buffer, sizeof(double)), DBL_EPSILON*2 );
 }
 
-}
-
 /****************************************************************************************\
 *                                 Determinant of the matrix                              *
 \****************************************************************************************/
@@ -764,7 +762,7 @@ SVBkSb( int m, int n, const double* w, size_t wstep,
                    m(0,1)*((double)m(1,0)*m(2,2) - (double)m(1,2)*m(2,0)) +  \
                    m(0,2)*((double)m(1,0)*m(2,1) - (double)m(1,1)*m(2,0)))
 
-double cv::determinant( InputArray _mat )
+double determinant( InputArray _mat )
 {
     CV_INSTRUMENT_REGION();
 
@@ -842,7 +840,7 @@ double cv::determinant( InputArray _mat )
 #define Df( y, x ) ((float*)(dstdata + y*dststep))[x]
 #define Dd( y, x ) ((double*)(dstdata + y*dststep))[x]
 
-double cv::invert( InputArray _src, OutputArray _dst, int method )
+double invert( InputArray _src, OutputArray _dst, int method )
 {
     CV_INSTRUMENT_REGION();
 
@@ -1069,13 +1067,19 @@ double cv::invert( InputArray _src, OutputArray _dst, int method )
     return result;
 }
 
+UMat UMat::inv(int method) const
+{
+    UMat m;
+    invert(*this, m, method);
+    return m;
+}
 
 
 /****************************************************************************************\
 *                              Solving a linear system                                   *
 \****************************************************************************************/
 
-bool cv::solve( InputArray _src, InputArray _src2arg, OutputArray _dst, int method )
+bool solve( InputArray _src, InputArray _src2arg, OutputArray _dst, int method )
 {
     CV_INSTRUMENT_REGION();
 
@@ -1374,7 +1378,7 @@ bool cv::solve( InputArray _src, InputArray _src2arg, OutputArray _dst, int meth
 
 /////////////////// finding eigenvalues and eigenvectors of a symmetric matrix ///////////////
 
-bool cv::eigen( InputArray _src, OutputArray _evals, OutputArray _evects )
+bool eigen( InputArray _src, OutputArray _evals, OutputArray _evects )
 {
     CV_INSTRUMENT_REGION();
 
@@ -1396,7 +1400,7 @@ bool cv::eigen( InputArray _src, OutputArray _evals, OutputArray _evects )
     const bool evecNeeded = _evects.needed();
     const int esOptions = evecNeeded ? Eigen::ComputeEigenvectors : Eigen::EigenvaluesOnly;
     _evals.create(n, 1, type);
-    cv::Mat evals = _evals.getMat();
+    Mat evals = _evals.getMat();
     if ( type == CV_64F )
     {
         Eigen::MatrixXd src_eig, zeros_eig;
@@ -1448,9 +1452,6 @@ bool cv::eigen( InputArray _src, OutputArray _evals, OutputArray _evects )
 #endif
 }
 
-namespace cv
-{
-
 static void _SVDcompute( InputArray _aarr, OutputArray _w,
                          OutputArray _u, OutputArray _vt, int flags )
 {
@@ -1598,6 +1599,9 @@ void cv::SVBackSubst(InputArray w, InputArray u, InputArray vt, InputArray rhs,
 }
 
 
+
+#ifndef OPENCV_EXCLUDE_C_API
+
 CV_IMPL double
 cvDet( const CvArr* arr )
 {
@@ -1789,3 +1793,4 @@ cvSVBkSb( const CvArr* warr, const CvArr* uarr,
     cv::SVD::backSubst(w, u, v, rhs, dst);
     CV_Assert( dst.data == dst0.data );
 }
+#endif  // OPENCV_EXCLUDE_C_API

modules/core/src/mathfuncs.cpp

@@ -1638,6 +1638,9 @@ void patchNaNs( InputOutputArray _a, double _val )
 
 }
 
+
+#ifndef OPENCV_EXCLUDE_C_API
+
 CV_IMPL float cvCbrt(float value) { return cv::cubeRoot(value); }
 CV_IMPL float cvFastArctan(float y, float x) { return cv::fastAtan2(y, x); }
 
@@ -1721,6 +1724,7 @@ CV_IMPL int cvCheckArr( const CvArr* arr, int flags,
     return cv::checkRange(cv::cvarrToMat(arr), (flags & CV_CHECK_QUIET) != 0, 0, minVal, maxVal );
 }
 
+#endif  // OPENCV_EXCLUDE_C_API
 
 /*
   Finds real roots of cubic, quadratic or linear equation.
@@ -2016,6 +2020,8 @@ double cv::solvePoly( InputArray _coeffs0, OutputArray _roots0, int maxIters )
 }
 
 
+#ifndef OPENCV_EXCLUDE_C_API
+
 CV_IMPL int
 cvSolveCubic( const CvMat* coeffs, CvMat* roots )
 {
@@ -2035,6 +2041,7 @@ void cvSolvePoly(const CvMat* a, CvMat *r, int maxiter, int)
     CV_Assert( _r.data == _r0.data ); // check that the array of roots was not reallocated
 }
 
+#endif  // OPENCV_EXCLUDE_C_API
 
 
 // Common constants for dispatched code

modules/core/src/matmul.dispatch.cpp

@@ -999,8 +999,79 @@ double Mat::dot(InputArray _mat) const
     return r;
 }
 
+
+#ifdef HAVE_OPENCL
+
+static bool ocl_dot( InputArray _src1, InputArray _src2, double & res )
+{
+    UMat src1 = _src1.getUMat().reshape(1), src2 = _src2.getUMat().reshape(1);
+
+    int type = src1.type(), depth = CV_MAT_DEPTH(type),
+            kercn = ocl::predictOptimalVectorWidth(src1, src2);
+    bool doubleSupport = ocl::Device::getDefault().doubleFPConfig() > 0;
+
+    if ( !doubleSupport && depth == CV_64F )
+        return false;
+
+    int dbsize = ocl::Device::getDefault().maxComputeUnits();
+    size_t wgs = ocl::Device::getDefault().maxWorkGroupSize();
+    int ddepth = std::max(CV_32F, depth);
+
+    int wgs2_aligned = 1;
+    while (wgs2_aligned < (int)wgs)
+        wgs2_aligned <<= 1;
+    wgs2_aligned >>= 1;
+
+    char cvt[40];
+    ocl::Kernel k("reduce", ocl::core::reduce_oclsrc,
+                  format("-D srcT=%s -D srcT1=%s -D dstT=%s -D dstTK=%s -D ddepth=%d -D convertToDT=%s -D OP_DOT "
+                         "-D WGS=%d -D WGS2_ALIGNED=%d%s%s%s -D kercn=%d",
+                         ocl::typeToStr(CV_MAKE_TYPE(depth, kercn)), ocl::typeToStr(depth),
+                         ocl::typeToStr(ddepth), ocl::typeToStr(CV_MAKE_TYPE(ddepth, kercn)),
+                         ddepth, ocl::convertTypeStr(depth, ddepth, kercn, cvt),
+                         (int)wgs, wgs2_aligned, doubleSupport ? " -D DOUBLE_SUPPORT" : "",
+                         _src1.isContinuous() ? " -D HAVE_SRC_CONT" : "",
+                         _src2.isContinuous() ? " -D HAVE_SRC2_CONT" : "", kercn));
+    if (k.empty())
+        return false;
+
+    UMat db(1, dbsize, ddepth);
+
+    ocl::KernelArg src1arg = ocl::KernelArg::ReadOnlyNoSize(src1),
+            src2arg = ocl::KernelArg::ReadOnlyNoSize(src2),
+            dbarg = ocl::KernelArg::PtrWriteOnly(db);
+
+    k.args(src1arg, src1.cols, (int)src1.total(), dbsize, dbarg, src2arg);
+
+    size_t globalsize = dbsize * wgs;
+    if (k.run(1, &globalsize, &wgs, true))
+    {
+        res = sum(db.getMat(ACCESS_READ))[0];
+        return true;
+    }
+    return false;
+}
+
+#endif
+
+double UMat::dot(InputArray m) const
+{
+    CV_INSTRUMENT_REGION();
+
+    CV_Assert(m.sameSize(*this) && m.type() == type());
+
+#ifdef HAVE_OPENCL
+    double r = 0;
+    CV_OCL_RUN_(dims <= 2, ocl_dot(*this, m, r), r)
+#endif
+
+    return getMat(ACCESS_READ).dot(m);
+}
+
 }  // namespace cv::
 
+
+#ifndef OPENCV_EXCLUDE_C_API
 /****************************************************************************************\
 *                                    Earlier API                                         *
 \****************************************************************************************/
@@ -1225,4 +1296,6 @@ cvBackProjectPCA( const CvArr* proj_arr, const CvArr* avg_arr,
     CV_Assert(dst0.data == dst.data);
 }
 
+#endif  // OPENCV_EXCLUDE_C_API
+
 /* End of file. */

modules/core/src/matrix_c.cpp

@@ -6,6 +6,7 @@
 #include "opencv2/core/mat.hpp"
 #include "opencv2/core/types_c.h"
 
+#ifndef OPENCV_EXCLUDE_C_API
 // glue
 
 CvMatND cvMatND(const cv::Mat& m)
@@ -360,7 +361,6 @@ cvSort( const CvArr* _src, CvArr* _dst, CvArr* _idx, int flags )
     }
 }
 
-
 CV_IMPL int
 cvKMeans2( const CvArr* _samples, int cluster_count, CvArr* _labels,
            CvTermCriteria termcrit, int attempts, CvRNG*,
@@ -389,3 +389,5 @@ cvKMeans2( const CvArr* _samples, int cluster_count, CvArr* _labels,
         *_compactness = compactness;
     return 1;
 }
+
+#endif  // OPENCV_EXCLUDE_C_API

modules/core/src/matrix_operations.cpp

@@ -226,6 +226,23 @@ void cv::setIdentity( InputOutputArray _m, const Scalar& s )
     }
 }
 
+
+namespace cv {
+
+UMat UMat::eye(int rows, int cols, int type)
+{
+    return UMat::eye(Size(cols, rows), type);
+}
+
+UMat UMat::eye(Size size, int type)
+{
+    UMat m(size, type);
+    setIdentity(m);
+    return m;
+}
+
+}  // namespace
+
 //////////////////////////////////////////// trace ///////////////////////////////////////////
 
 cv::Scalar cv::trace( InputArray _m )
@@ -260,285 +277,6 @@ cv::Scalar cv::trace( InputArray _m )
     return cv::sum(m.diag());
 }
 
-////////////////////////////////////// transpose /////////////////////////////////////////
-
-namespace cv
-{
-
-template<typename T> static void
-transpose_( const uchar* src, size_t sstep, uchar* dst, size_t dstep, Size sz )
-{
-    int i=0, j, m = sz.width, n = sz.height;
-
-    #if CV_ENABLE_UNROLLED
-    for(; i <= m - 4; i += 4 )
-    {
-        T* d0 = (T*)(dst + dstep*i);
-        T* d1 = (T*)(dst + dstep*(i+1));
-        T* d2 = (T*)(dst + dstep*(i+2));
-        T* d3 = (T*)(dst + dstep*(i+3));
-
-        for( j = 0; j <= n - 4; j += 4 )
-        {
-            const T* s0 = (const T*)(src + i*sizeof(T) + sstep*j);
-            const T* s1 = (const T*)(src + i*sizeof(T) + sstep*(j+1));
-            const T* s2 = (const T*)(src + i*sizeof(T) + sstep*(j+2));
-            const T* s3 = (const T*)(src + i*sizeof(T) + sstep*(j+3));
-
-            d0[j] = s0[0]; d0[j+1] = s1[0]; d0[j+2] = s2[0]; d0[j+3] = s3[0];
-            d1[j] = s0[1]; d1[j+1] = s1[1]; d1[j+2] = s2[1]; d1[j+3] = s3[1];
-            d2[j] = s0[2]; d2[j+1] = s1[2]; d2[j+2] = s2[2]; d2[j+3] = s3[2];
-            d3[j] = s0[3]; d3[j+1] = s1[3]; d3[j+2] = s2[3]; d3[j+3] = s3[3];
-        }
-
-        for( ; j < n; j++ )
-        {
-            const T* s0 = (const T*)(src + i*sizeof(T) + j*sstep);
-            d0[j] = s0[0]; d1[j] = s0[1]; d2[j] = s0[2]; d3[j] = s0[3];
-        }
-    }
-    #endif
-    for( ; i < m; i++ )
-    {
-        T* d0 = (T*)(dst + dstep*i);
-        j = 0;
-        #if CV_ENABLE_UNROLLED
-        for(; j <= n - 4; j += 4 )
-        {
-            const T* s0 = (const T*)(src + i*sizeof(T) + sstep*j);
-            const T* s1 = (const T*)(src + i*sizeof(T) + sstep*(j+1));
-            const T* s2 = (const T*)(src + i*sizeof(T) + sstep*(j+2));
-            const T* s3 = (const T*)(src + i*sizeof(T) + sstep*(j+3));
-
-            d0[j] = s0[0]; d0[j+1] = s1[0]; d0[j+2] = s2[0]; d0[j+3] = s3[0];
-        }
-        #endif
-        for( ; j < n; j++ )
-        {
-            const T* s0 = (const T*)(src + i*sizeof(T) + j*sstep);
-            d0[j] = s0[0];
-        }
-    }
-}
-
-template<typename T> static void
-transposeI_( uchar* data, size_t step, int n )
-{
-    for( int i = 0; i < n; i++ )
-    {
-        T* row = (T*)(data + step*i);
-        uchar* data1 = data + i*sizeof(T);
-        for( int j = i+1; j < n; j++ )
-            std::swap( row[j], *(T*)(data1 + step*j) );
-    }
-}
-
-typedef void (*TransposeFunc)( const uchar* src, size_t sstep, uchar* dst, size_t dstep, Size sz );
-typedef void (*TransposeInplaceFunc)( uchar* data, size_t step, int n );
-
-#define DEF_TRANSPOSE_FUNC(suffix, type) \
-static void transpose_##suffix( const uchar* src, size_t sstep, uchar* dst, size_t dstep, Size sz ) \
-{ transpose_<type>(src, sstep, dst, dstep, sz); } \
-\
-static void transposeI_##suffix( uchar* data, size_t step, int n ) \
-{ transposeI_<type>(data, step, n); }
-
-DEF_TRANSPOSE_FUNC(8u, uchar)
-DEF_TRANSPOSE_FUNC(16u, ushort)
-DEF_TRANSPOSE_FUNC(8uC3, Vec3b)
-DEF_TRANSPOSE_FUNC(32s, int)
-DEF_TRANSPOSE_FUNC(16uC3, Vec3s)
-DEF_TRANSPOSE_FUNC(32sC2, Vec2i)
-DEF_TRANSPOSE_FUNC(32sC3, Vec3i)
-DEF_TRANSPOSE_FUNC(32sC4, Vec4i)
-DEF_TRANSPOSE_FUNC(32sC6, Vec6i)
-DEF_TRANSPOSE_FUNC(32sC8, Vec8i)
-
-static TransposeFunc transposeTab[] =
-{
-    0, transpose_8u, transpose_16u, transpose_8uC3, transpose_32s, 0, transpose_16uC3, 0,
-    transpose_32sC2, 0, 0, 0, transpose_32sC3, 0, 0, 0, transpose_32sC4,
-    0, 0, 0, 0, 0, 0, 0, transpose_32sC6, 0, 0, 0, 0, 0, 0, 0, transpose_32sC8
-};
-
-static TransposeInplaceFunc transposeInplaceTab[] =
-{
-    0, transposeI_8u, transposeI_16u, transposeI_8uC3, transposeI_32s, 0, transposeI_16uC3, 0,
-    transposeI_32sC2, 0, 0, 0, transposeI_32sC3, 0, 0, 0, transposeI_32sC4,
-    0, 0, 0, 0, 0, 0, 0, transposeI_32sC6, 0, 0, 0, 0, 0, 0, 0, transposeI_32sC8
-};
-
-#ifdef HAVE_OPENCL
-
-static bool ocl_transpose( InputArray _src, OutputArray _dst )
-{
-    const ocl::Device & dev = ocl::Device::getDefault();
-    const int TILE_DIM = 32, BLOCK_ROWS = 8;
-    int type = _src.type(), cn = CV_MAT_CN(type), depth = CV_MAT_DEPTH(type),
-        rowsPerWI = dev.isIntel() ? 4 : 1;
-
-    UMat src = _src.getUMat();
-    _dst.create(src.cols, src.rows, type);
-    UMat dst = _dst.getUMat();
-
-    String kernelName("transpose");
-    bool inplace = dst.u == src.u;
-
-    if (inplace)
-    {
-        CV_Assert(dst.cols == dst.rows);
-        kernelName += "_inplace";
-    }
-    else
-    {
-        // check required local memory size
-        size_t required_local_memory = (size_t) TILE_DIM*(TILE_DIM+1)*CV_ELEM_SIZE(type);
-        if (required_local_memory > ocl::Device::getDefault().localMemSize())
-            return false;
-    }
-
-    ocl::Kernel k(kernelName.c_str(), ocl::core::transpose_oclsrc,
-                  format("-D T=%s -D T1=%s -D cn=%d -D TILE_DIM=%d -D BLOCK_ROWS=%d -D rowsPerWI=%d%s",
-                         ocl::memopTypeToStr(type), ocl::memopTypeToStr(depth),
-                         cn, TILE_DIM, BLOCK_ROWS, rowsPerWI, inplace ? " -D INPLACE" : ""));
-    if (k.empty())
-        return false;
-
-    if (inplace)
-        k.args(ocl::KernelArg::ReadWriteNoSize(dst), dst.rows);
-    else
-        k.args(ocl::KernelArg::ReadOnly(src),
-               ocl::KernelArg::WriteOnlyNoSize(dst));
-
-    size_t localsize[2]  = { TILE_DIM, BLOCK_ROWS };
-    size_t globalsize[2] = { (size_t)src.cols, inplace ? ((size_t)src.rows + rowsPerWI - 1) / rowsPerWI : (divUp((size_t)src.rows, TILE_DIM) * BLOCK_ROWS) };
-
-    if (inplace && dev.isIntel())
-    {
-        localsize[0] = 16;
-        localsize[1] = dev.maxWorkGroupSize() / localsize[0];
-    }
-
-    return k.run(2, globalsize, localsize, false);
-}
-
-#endif
-
-#ifdef HAVE_IPP
-static bool ipp_transpose( Mat &src, Mat &dst )
-{
-    CV_INSTRUMENT_REGION_IPP();
-
-    int type = src.type();
-    typedef IppStatus (CV_STDCALL * IppiTranspose)(const void * pSrc, int srcStep, void * pDst, int dstStep, IppiSize roiSize);
-    typedef IppStatus (CV_STDCALL * IppiTransposeI)(const void * pSrcDst, int srcDstStep, IppiSize roiSize);
-    IppiTranspose ippiTranspose = 0;
-    IppiTransposeI ippiTranspose_I = 0;
-
-    if (dst.data == src.data && dst.cols == dst.rows)
-    {
-        CV_SUPPRESS_DEPRECATED_START
-        ippiTranspose_I =
-            type == CV_8UC1 ? (IppiTransposeI)ippiTranspose_8u_C1IR :
-            type == CV_8UC3 ? (IppiTransposeI)ippiTranspose_8u_C3IR :
-            type == CV_8UC4 ? (IppiTransposeI)ippiTranspose_8u_C4IR :
-            type == CV_16UC1 ? (IppiTransposeI)ippiTranspose_16u_C1IR :
-            type == CV_16UC3 ? (IppiTransposeI)ippiTranspose_16u_C3IR :
-            type == CV_16UC4 ? (IppiTransposeI)ippiTranspose_16u_C4IR :
-            type == CV_16SC1 ? (IppiTransposeI)ippiTranspose_16s_C1IR :
-            type == CV_16SC3 ? (IppiTransposeI)ippiTranspose_16s_C3IR :
-            type == CV_16SC4 ? (IppiTransposeI)ippiTranspose_16s_C4IR :
-            type == CV_32SC1 ? (IppiTransposeI)ippiTranspose_32s_C1IR :
-            type == CV_32SC3 ? (IppiTransposeI)ippiTranspose_32s_C3IR :
-            type == CV_32SC4 ? (IppiTransposeI)ippiTranspose_32s_C4IR :
-            type == CV_32FC1 ? (IppiTransposeI)ippiTranspose_32f_C1IR :
-            type == CV_32FC3 ? (IppiTransposeI)ippiTranspose_32f_C3IR :
-            type == CV_32FC4 ? (IppiTransposeI)ippiTranspose_32f_C4IR : 0;
-        CV_SUPPRESS_DEPRECATED_END
-    }
-    else
-    {
-        ippiTranspose =
-            type == CV_8UC1 ? (IppiTranspose)ippiTranspose_8u_C1R :
-            type == CV_8UC3 ? (IppiTranspose)ippiTranspose_8u_C3R :
-            type == CV_8UC4 ? (IppiTranspose)ippiTranspose_8u_C4R :
-            type == CV_16UC1 ? (IppiTranspose)ippiTranspose_16u_C1R :
-            type == CV_16UC3 ? (IppiTranspose)ippiTranspose_16u_C3R :
-            type == CV_16UC4 ? (IppiTranspose)ippiTranspose_16u_C4R :
-            type == CV_16SC1 ? (IppiTranspose)ippiTranspose_16s_C1R :
-            type == CV_16SC3 ? (IppiTranspose)ippiTranspose_16s_C3R :
-            type == CV_16SC4 ? (IppiTranspose)ippiTranspose_16s_C4R :
-            type == CV_32SC1 ? (IppiTranspose)ippiTranspose_32s_C1R :
-            type == CV_32SC3 ? (IppiTranspose)ippiTranspose_32s_C3R :
-            type == CV_32SC4 ? (IppiTranspose)ippiTranspose_32s_C4R :
-            type == CV_32FC1 ? (IppiTranspose)ippiTranspose_32f_C1R :
-            type == CV_32FC3 ? (IppiTranspose)ippiTranspose_32f_C3R :
-            type == CV_32FC4 ? (IppiTranspose)ippiTranspose_32f_C4R : 0;
-    }
-
-    IppiSize roiSize = { src.cols, src.rows };
-    if (ippiTranspose != 0)
-    {
-        if (CV_INSTRUMENT_FUN_IPP(ippiTranspose, src.ptr(), (int)src.step, dst.ptr(), (int)dst.step, roiSize) >= 0)
-            return true;
-    }
-    else if (ippiTranspose_I != 0)
-    {
-        if (CV_INSTRUMENT_FUN_IPP(ippiTranspose_I, dst.ptr(), (int)dst.step, roiSize) >= 0)
-            return true;
-    }
-    return false;
-}
-#endif
-
-}
-
-
-void cv::transpose( InputArray _src, OutputArray _dst )
-{
-    CV_INSTRUMENT_REGION();
-
-    int type = _src.type(), esz = CV_ELEM_SIZE(type);
-    CV_Assert( _src.dims() <= 2 && esz <= 32 );
-
-    CV_OCL_RUN(_dst.isUMat(),
-               ocl_transpose(_src, _dst))
-
-    Mat src = _src.getMat();
-    if( src.empty() )
-    {
-        _dst.release();
-        return;
-    }
-
-    _dst.create(src.cols, src.rows, src.type());
-    Mat dst = _dst.getMat();
-
-    // handle the case of single-column/single-row matrices, stored in STL vectors.
-    if( src.rows != dst.cols || src.cols != dst.rows )
-    {
-        CV_Assert( src.size() == dst.size() && (src.cols == 1 || src.rows == 1) );
-        src.copyTo(dst);
-        return;
-    }
-
-    CV_IPP_RUN_FAST(ipp_transpose(src, dst))
-
-    if( dst.data == src.data )
-    {
-        TransposeInplaceFunc func = transposeInplaceTab[esz];
-        CV_Assert( func != 0 );
-        CV_Assert( dst.cols == dst.rows );
-        func( dst.ptr(), dst.step, dst.rows );
-    }
-    else
-    {
-        TransposeFunc func = transposeTab[esz];
-        CV_Assert( func != 0 );
-        func( src.ptr(), src.step, dst.ptr(), dst.step, src.size() );
-    }
-}
-
 
 ////////////////////////////////////// completeSymm /////////////////////////////////////////
 

modules/core/src/matrix_wrap.cpp

@@ -316,6 +316,7 @@ void _InputArray::getUMatVector(std::vector<UMat>& umv) const
 
 cuda::GpuMat _InputArray::getGpuMat() const
 {
+#ifdef HAVE_CUDA
     _InputArray::KindFlag k = kind();
 
     if (k == CUDA_GPU_MAT)
@@ -339,14 +340,22 @@ cuda::GpuMat _InputArray::getGpuMat() const
         return cuda::GpuMat();
 
     CV_Error(cv::Error::StsNotImplemented, "getGpuMat is available only for cuda::GpuMat and cuda::HostMem");
+#else
+    CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
 }
 void _InputArray::getGpuMatVector(std::vector<cuda::GpuMat>& gpumv) const
 {
+#ifdef HAVE_CUDA
     _InputArray::KindFlag k = kind();
     if (k == STD_VECTOR_CUDA_GPU_MAT)
     {
         gpumv = *(std::vector<cuda::GpuMat>*)obj;
     }
+#else
+    CV_UNUSED(gpumv);
+    CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
 }
 ogl::Buffer _InputArray::getOGlBuffer() const
 {
@@ -453,11 +462,15 @@ Size _InputArray::size(int i) const
 
     if (k == STD_VECTOR_CUDA_GPU_MAT)
     {
+#ifdef HAVE_CUDA
         const std::vector<cuda::GpuMat>& vv = *(const std::vector<cuda::GpuMat>*)obj;
         if (i < 0)
             return vv.empty() ? Size() : Size((int)vv.size(), 1);
         CV_Assert(i < (int)vv.size());
         return vv[i].size();
+#else
+        CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
     }
 
     if( k == STD_VECTOR_UMAT )
@@ -792,6 +805,7 @@ int _InputArray::type(int i) const
 
     if (k == STD_VECTOR_CUDA_GPU_MAT)
     {
+#ifdef HAVE_CUDA
         const std::vector<cuda::GpuMat>& vv = *(const std::vector<cuda::GpuMat>*)obj;
         if (vv.empty())
         {
@@ -800,6 +814,9 @@ int _InputArray::type(int i) const
         }
         CV_Assert(i < (int)vv.size());
         return vv[i >= 0 ? i : 0].type();
+#else
+        CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
     }
 
     if( k == OPENGL_BUFFER )
@@ -1161,22 +1178,34 @@ void _OutputArray::create(Size _sz, int mtype, int i, bool allowTransposed, _Out
     {
         CV_Assert(!fixedSize() || ((cuda::GpuMat*)obj)->size() == _sz);
         CV_Assert(!fixedType() || ((cuda::GpuMat*)obj)->type() == mtype);
+#ifdef HAVE_CUDA
         ((cuda::GpuMat*)obj)->create(_sz, mtype);
         return;
+#else
+        CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
     }
     if( k == OPENGL_BUFFER && i < 0 && !allowTransposed && fixedDepthMask == 0 )
     {
         CV_Assert(!fixedSize() || ((ogl::Buffer*)obj)->size() == _sz);
         CV_Assert(!fixedType() || ((ogl::Buffer*)obj)->type() == mtype);
+#ifdef HAVE_OPENGL
         ((ogl::Buffer*)obj)->create(_sz, mtype);
         return;
+#else
+        CV_Error(Error::StsNotImplemented, "OpenGL support is not enabled in this OpenCV build (missing HAVE_OPENGL)");
+#endif
     }
     if( k == CUDA_HOST_MEM && i < 0 && !allowTransposed && fixedDepthMask == 0 )
     {
         CV_Assert(!fixedSize() || ((cuda::HostMem*)obj)->size() == _sz);
         CV_Assert(!fixedType() || ((cuda::HostMem*)obj)->type() == mtype);
+#ifdef HAVE_CUDA
         ((cuda::HostMem*)obj)->create(_sz, mtype);
         return;
+#else
+        CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
     }
     int sizes[] = {_sz.height, _sz.width};
     create(2, sizes, mtype, i, allowTransposed, fixedDepthMask);
@@ -1203,22 +1232,34 @@ void _OutputArray::create(int _rows, int _cols, int mtype, int i, bool allowTran
     {
         CV_Assert(!fixedSize() || ((cuda::GpuMat*)obj)->size() == Size(_cols, _rows));
         CV_Assert(!fixedType() || ((cuda::GpuMat*)obj)->type() == mtype);
+#ifdef HAVE_CUDA
         ((cuda::GpuMat*)obj)->create(_rows, _cols, mtype);
         return;
+#else
+        CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
     }
     if( k == OPENGL_BUFFER && i < 0 && !allowTransposed && fixedDepthMask == 0 )
     {
         CV_Assert(!fixedSize() || ((ogl::Buffer*)obj)->size() == Size(_cols, _rows));
         CV_Assert(!fixedType() || ((ogl::Buffer*)obj)->type() == mtype);
+#ifdef HAVE_OPENGL
         ((ogl::Buffer*)obj)->create(_rows, _cols, mtype);
         return;
+#else
+        CV_Error(Error::StsNotImplemented, "OpenGL support is not enabled in this OpenCV build (missing HAVE_OPENGL)");
+#endif
     }
     if( k == CUDA_HOST_MEM && i < 0 && !allowTransposed && fixedDepthMask == 0 )
     {
         CV_Assert(!fixedSize() || ((cuda::HostMem*)obj)->size() == Size(_cols, _rows));
         CV_Assert(!fixedType() || ((cuda::HostMem*)obj)->type() == mtype);
+#ifdef HAVE_CUDA
         ((cuda::HostMem*)obj)->create(_rows, _cols, mtype);
         return;
+#else
+        CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
     }
     int sizes[] = {_rows, _cols};
     create(2, sizes, mtype, i, allowTransposed, fixedDepthMask);
@@ -1641,20 +1682,32 @@ void _OutputArray::release() const
 
     if( k == CUDA_GPU_MAT )
     {
+#ifdef HAVE_CUDA
         ((cuda::GpuMat*)obj)->release();
         return;
+#else
+        CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
     }
 
     if( k == CUDA_HOST_MEM )
     {
+#ifdef HAVE_CUDA
         ((cuda::HostMem*)obj)->release();
         return;
+#else
+        CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
     }
 
     if( k == OPENGL_BUFFER )
     {
+#ifdef HAVE_OPENGL
         ((ogl::Buffer*)obj)->release();
         return;
+#else
+        CV_Error(Error::StsNotImplemented, "OpenGL support is not enabled in this OpenCV build (missing HAVE_OPENGL)");
+#endif
     }
 
     if( k == NONE )
@@ -1685,8 +1738,12 @@ void _OutputArray::release() const
     }
     if (k == STD_VECTOR_CUDA_GPU_MAT)
     {
+#ifdef HAVE_CUDA
         ((std::vector<cuda::GpuMat>*)obj)->clear();
         return;
+#else
+        CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
     }
     CV_Error(Error::StsNotImplemented, "Unknown/unsupported array type");
 }
@@ -1794,9 +1851,13 @@ void _OutputArray::setTo(const _InputArray& arr, const _InputArray & mask) const
         ((UMat*)obj)->setTo(arr, mask);
     else if( k == CUDA_GPU_MAT )
     {
+#ifdef HAVE_CUDA
         Mat value = arr.getMat();
         CV_Assert( checkScalar(value, type(), arr.kind(), _InputArray::CUDA_GPU_MAT) );
         ((cuda::GpuMat*)obj)->setTo(Scalar(Vec<double, 4>(value.ptr<double>())), mask);
+#else
+        CV_Error(Error::StsNotImplemented, "CUDA support is not enabled in this OpenCV build (missing HAVE_CUDA)");
+#endif
     }
     else
         CV_Error(Error::StsNotImplemented, "");

modules/core/src/norm.cpp

@@ -205,13 +205,10 @@ int normL1_(const uchar* a, const uchar* b, int n)
     return d;
 }
 
-}} //cv::hal
+} //cv::hal
 
 //==================================================================================================
 
-namespace cv
-{
-
 template<typename T, typename ST> int
 normInf_(const T* src, const uchar* mask, ST* _result, int len, int cn)
 {
@@ -594,12 +591,10 @@ static bool ipp_norm(Mat &src, int normType, Mat &mask, double &result)
     CV_UNUSED(src); CV_UNUSED(normType); CV_UNUSED(mask); CV_UNUSED(result);
 #endif
     return false;
-}
-#endif
-
-} // cv::
+}  // ipp_norm()
+#endif  // HAVE_IPP
 
-double cv::norm( InputArray _src, int normType, InputArray _mask )
+double norm( InputArray _src, int normType, InputArray _mask )
 {
     CV_INSTRUMENT_REGION();
 
@@ -792,9 +787,6 @@ double cv::norm( InputArray _src, int normType, InputArray _mask )
 //==================================================================================================
 
 #ifdef HAVE_OPENCL
-
-namespace cv {
-
 static bool ocl_norm( InputArray _src1, InputArray _src2, int normType, InputArray _mask, double & result )
 {
 #ifdef __ANDROID__
@@ -849,15 +841,10 @@ static bool ocl_norm( InputArray _src1, InputArray _src2, int normType, InputArr
         result /= (s2 + DBL_EPSILON);
 
     return true;
-}
-
-}
-
-#endif
+}  // ocl_norm()
+#endif  // HAVE_OPENCL
 
 #ifdef HAVE_IPP
-namespace cv
-{
 static bool ipp_norm(InputArray _src1, InputArray _src2, int normType, InputArray _mask, double &result)
 {
     CV_INSTRUMENT_REGION_IPP();
@@ -1083,12 +1070,11 @@ static bool ipp_norm(InputArray _src1, InputArray _src2, int normType, InputArra
     CV_UNUSED(_src1); CV_UNUSED(_src2); CV_UNUSED(normType); CV_UNUSED(_mask); CV_UNUSED(result);
 #endif
     return false;
-}
-}
-#endif
+}  // ipp_norm
+#endif  // HAVE_IPP
 
 
-double cv::norm( InputArray _src1, InputArray _src2, int normType, InputArray _mask )
+double norm( InputArray _src1, InputArray _src2, int normType, InputArray _mask )
 {
     CV_INSTRUMENT_REGION();
 
@@ -1280,12 +1266,12 @@ double cv::norm( InputArray _src1, InputArray _src2, int normType, InputArray _m
     return result.d;
 }
 
-cv::Hamming::ResultType cv::Hamming::operator()( const unsigned char* a, const unsigned char* b, int size ) const
+cv::Hamming::ResultType Hamming::operator()( const unsigned char* a, const unsigned char* b, int size ) const
 {
     return cv::hal::normHamming(a, b, size);
 }
 
-double cv::PSNR(InputArray _src1, InputArray _src2, double R)
+double PSNR(InputArray _src1, InputArray _src2, double R)
 {
     CV_INSTRUMENT_REGION();
 
@@ -1295,3 +1281,141 @@ double cv::PSNR(InputArray _src1, InputArray _src2, double R)
     double diff = std::sqrt(norm(_src1, _src2, NORM_L2SQR)/(_src1.total()*_src1.channels()));
     return 20*log10(R/(diff+DBL_EPSILON));
 }
+
+
+#ifdef HAVE_OPENCL
+static bool ocl_normalize( InputArray _src, InputOutputArray _dst, InputArray _mask, int dtype,
+                           double scale, double delta )
+{
+    UMat src = _src.getUMat();
+
+    if( _mask.empty() )
+        src.convertTo( _dst, dtype, scale, delta );
+    else if (src.channels() <= 4)
+    {
+        const ocl::Device & dev = ocl::Device::getDefault();
+
+        int stype = _src.type(), sdepth = CV_MAT_DEPTH(stype), cn = CV_MAT_CN(stype),
+                ddepth = CV_MAT_DEPTH(dtype), wdepth = std::max(CV_32F, std::max(sdepth, ddepth)),
+                rowsPerWI = dev.isIntel() ? 4 : 1;
+
+        float fscale = static_cast<float>(scale), fdelta = static_cast<float>(delta);
+        bool haveScale = std::fabs(scale - 1) > DBL_EPSILON,
+                haveZeroScale = !(std::fabs(scale) > DBL_EPSILON),
+                haveDelta = std::fabs(delta) > DBL_EPSILON,
+                doubleSupport = dev.doubleFPConfig() > 0;
+
+        if (!haveScale && !haveDelta && stype == dtype)
+        {
+            _src.copyTo(_dst, _mask);
+            return true;
+        }
+        if (haveZeroScale)
+        {
+            _dst.setTo(Scalar(delta), _mask);
+            return true;
+        }
+
+        if ((sdepth == CV_64F || ddepth == CV_64F) && !doubleSupport)
+            return false;
+
+        char cvt[2][40];
+        String opts = format("-D srcT=%s -D dstT=%s -D convertToWT=%s -D cn=%d -D rowsPerWI=%d"
+                             " -D convertToDT=%s -D workT=%s%s%s%s -D srcT1=%s -D dstT1=%s",
+                             ocl::typeToStr(stype), ocl::typeToStr(dtype),
+                             ocl::convertTypeStr(sdepth, wdepth, cn, cvt[0]), cn,
+                             rowsPerWI, ocl::convertTypeStr(wdepth, ddepth, cn, cvt[1]),
+                             ocl::typeToStr(CV_MAKE_TYPE(wdepth, cn)),
+                             doubleSupport ? " -D DOUBLE_SUPPORT" : "",
+                             haveScale ? " -D HAVE_SCALE" : "",
+                             haveDelta ? " -D HAVE_DELTA" : "",
+                             ocl::typeToStr(sdepth), ocl::typeToStr(ddepth));
+
+        ocl::Kernel k("normalizek", ocl::core::normalize_oclsrc, opts);
+        if (k.empty())
+            return false;
+
+        UMat mask = _mask.getUMat(), dst = _dst.getUMat();
+
+        ocl::KernelArg srcarg = ocl::KernelArg::ReadOnlyNoSize(src),
+                maskarg = ocl::KernelArg::ReadOnlyNoSize(mask),
+                dstarg = ocl::KernelArg::ReadWrite(dst);
+
+        if (haveScale)
+        {
+            if (haveDelta)
+                k.args(srcarg, maskarg, dstarg, fscale, fdelta);
+            else
+                k.args(srcarg, maskarg, dstarg, fscale);
+        }
+        else
+        {
+            if (haveDelta)
+                k.args(srcarg, maskarg, dstarg, fdelta);
+            else
+                k.args(srcarg, maskarg, dstarg);
+        }
+
+        size_t globalsize[2] = { (size_t)src.cols, ((size_t)src.rows + rowsPerWI - 1) / rowsPerWI };
+        return k.run(2, globalsize, NULL, false);
+    }
+    else
+    {
+        UMat temp;
+        src.convertTo( temp, dtype, scale, delta );
+        temp.copyTo( _dst, _mask );
+    }
+
+    return true;
+}  // ocl_normalize
+#endif  // HAVE_OPENCL
+
+void normalize(InputArray _src, InputOutputArray _dst, double a, double b,
+               int norm_type, int rtype, InputArray _mask)
+{
+    CV_INSTRUMENT_REGION();
+
+    double scale = 1, shift = 0;
+    int type = _src.type(), depth = CV_MAT_DEPTH(type);
+
+    if( rtype < 0 )
+        rtype = _dst.fixedType() ? _dst.depth() : depth;
+
+    if( norm_type == CV_MINMAX )
+    {
+        double smin = 0, smax = 0;
+        double dmin = MIN( a, b ), dmax = MAX( a, b );
+        minMaxIdx( _src, &smin, &smax, 0, 0, _mask );
+        scale = (dmax - dmin)*(smax - smin > DBL_EPSILON ? 1./(smax - smin) : 0);
+        if( rtype == CV_32F )
+        {
+            scale = (float)scale;
+            shift = (float)dmin - (float)(smin*scale);
+        }
+        else
+            shift = dmin - smin*scale;
+    }
+    else if( norm_type == CV_L2 || norm_type == CV_L1 || norm_type == CV_C )
+    {
+        scale = norm( _src, norm_type, _mask );
+        scale = scale > DBL_EPSILON ? a/scale : 0.;
+        shift = 0;
+    }
+    else
+        CV_Error( CV_StsBadArg, "Unknown/unsupported norm type" );
+
+    CV_OCL_RUN(_dst.isUMat(),
+               ocl_normalize(_src, _dst, _mask, rtype, scale, shift))
+
+    Mat src = _src.getMat();
+    if( _mask.empty() )
+        src.convertTo( _dst, rtype, scale, shift );
+    else
+    {
+        Mat temp;
+        src.convertTo( temp, rtype, scale, shift );
+        temp.copyTo( _dst, _mask );
+    }
+}
+
+}  // namespace

modules/core/src/rand.cpp

@@ -750,6 +750,9 @@ void cv::randShuffle( InputOutputArray _dst, double iterFactor, RNG* _rng )
     func( dst, rng, iterFactor );
 }
 
+
+#ifndef OPENCV_EXCLUDE_C_API
+
 CV_IMPL void
 cvRandArr( CvRNG* _rng, CvArr* arr, int disttype, CvScalar param1, CvScalar param2 )
 {
@@ -767,6 +770,9 @@ CV_IMPL void cvRandShuffle( CvArr* arr, CvRNG* _rng, double iter_factor )
     cv::randShuffle( dst, iter_factor, &rng );
 }
 
+#endif  // OPENCV_EXCLUDE_C_API
+
+
 // Mersenne Twister random number generator.
 // Inspired by http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/MT2002/CODES/mt19937ar.c
 

modules/core/src/stat_c.cpp

@@ -5,6 +5,8 @@
 
 #include "precomp.hpp"
 
+#ifndef OPENCV_EXCLUDE_C_API
+
 CV_IMPL CvScalar cvSum( const CvArr* srcarr )
 {
     cv::Scalar sum = cv::sum(cv::cvarrToMat(srcarr, false, true, 1));
@@ -117,3 +119,5 @@ cvNorm( const void* imgA, const void* imgB, int normType, const void* maskarr )
 
     return !maskarr ? cv::norm(a, b, normType) : cv::norm(a, b, normType, mask);
 }
+
+#endif  // OPENCV_EXCLUDE_C_API

modules/core/src/umatrix.cpp

@@ -1318,88 +1318,6 @@ UMat UMat::t() const
     return m;
 }
 
-UMat UMat::inv(int method) const
-{
-    UMat m;
-    invert(*this, m, method);
-    return m;
-}
-
-UMat UMat::mul(InputArray m, double scale) const
-{
-    UMat dst;
-    multiply(*this, m, dst, scale);
-    return dst;
-}
-
-#ifdef HAVE_OPENCL
-
-static bool ocl_dot( InputArray _src1, InputArray _src2, double & res )
-{
-    UMat src1 = _src1.getUMat().reshape(1), src2 = _src2.getUMat().reshape(1);
-
-    int type = src1.type(), depth = CV_MAT_DEPTH(type),
-            kercn = ocl::predictOptimalVectorWidth(src1, src2);
-    bool doubleSupport = ocl::Device::getDefault().doubleFPConfig() > 0;
-
-    if ( !doubleSupport && depth == CV_64F )
-        return false;
-
-    int dbsize = ocl::Device::getDefault().maxComputeUnits();
-    size_t wgs = ocl::Device::getDefault().maxWorkGroupSize();
-    int ddepth = std::max(CV_32F, depth);
-
-    int wgs2_aligned = 1;
-    while (wgs2_aligned < (int)wgs)
-        wgs2_aligned <<= 1;
-    wgs2_aligned >>= 1;
-
-    char cvt[40];
-    ocl::Kernel k("reduce", ocl::core::reduce_oclsrc,
-                  format("-D srcT=%s -D srcT1=%s -D dstT=%s -D dstTK=%s -D ddepth=%d -D convertToDT=%s -D OP_DOT "
-                         "-D WGS=%d -D WGS2_ALIGNED=%d%s%s%s -D kercn=%d",
-                         ocl::typeToStr(CV_MAKE_TYPE(depth, kercn)), ocl::typeToStr(depth),
-                         ocl::typeToStr(ddepth), ocl::typeToStr(CV_MAKE_TYPE(ddepth, kercn)),
-                         ddepth, ocl::convertTypeStr(depth, ddepth, kercn, cvt),
-                         (int)wgs, wgs2_aligned, doubleSupport ? " -D DOUBLE_SUPPORT" : "",
-                         _src1.isContinuous() ? " -D HAVE_SRC_CONT" : "",
-                         _src2.isContinuous() ? " -D HAVE_SRC2_CONT" : "", kercn));
-    if (k.empty())
-        return false;
-
-    UMat db(1, dbsize, ddepth);
-
-    ocl::KernelArg src1arg = ocl::KernelArg::ReadOnlyNoSize(src1),
-            src2arg = ocl::KernelArg::ReadOnlyNoSize(src2),
-            dbarg = ocl::KernelArg::PtrWriteOnly(db);
-
-    k.args(src1arg, src1.cols, (int)src1.total(), dbsize, dbarg, src2arg);
-
-    size_t globalsize = dbsize * wgs;
-    if (k.run(1, &globalsize, &wgs, true))
-    {
-        res = sum(db.getMat(ACCESS_READ))[0];
-        return true;
-    }
-    return false;
-}
-
-#endif
-
-double UMat::dot(InputArray m) const
-{
-    CV_INSTRUMENT_REGION();
-
-    CV_Assert(m.sameSize(*this) && m.type() == type());
-
-#ifdef HAVE_OPENCL
-    double r = 0;
-    CV_OCL_RUN_(dims <= 2, ocl_dot(*this, m, r), r)
-#endif
-
-    return getMat(ACCESS_READ).dot(m);
-}
-
 UMat UMat::zeros(int rows, int cols, int type)
 {
     return UMat(rows, cols, type, Scalar::all(0));
@@ -1430,18 +1348,6 @@ UMat UMat::ones(int ndims, const int* sz, int type)
     return UMat(ndims, sz, type, Scalar(1));
 }
 
-UMat UMat::eye(int rows, int cols, int type)
-{
-    return UMat::eye(Size(cols, rows), type);
-}
-
-UMat UMat::eye(Size size, int type)
-{
-    UMat m(size, type);
-    setIdentity(m);
-    return m;
-}
-
 }
 
 /* End of file. */