fixed gpu arithm functions (mismatch with cpu version)

modules/gpu/src/cuda/element_operations.cu

@@ -488,11 +488,29 @@ namespace cv { namespace gpu { namespace device
 
     template <typename T, typename D> struct Multiply : binary_function<T, T, D>
     {
-        Multiply(double scale_) : scale(scale_) {}
+        Multiply(float scale_) : scale(scale_) {}
         __device__ __forceinline__ D operator ()(T a, T b) const
         {
             return saturate_cast<D>(scale * a * b);
         }
+        const float scale;
+    };
+    template <typename T> struct Multiply<T, double> : binary_function<T, T, double>
+    {
+        Multiply(double scale_) : scale(scale_) {}
+        __device__ __forceinline__ double operator ()(T a, T b) const
+        {
+            return scale * a * b;
+        }
+        const double scale;
+    };
+    template <> struct Multiply<int, int> : binary_function<int, int, int>
+    {
+        Multiply(double scale_) : scale(scale_) {}
+        __device__ __forceinline__ int operator ()(int a, int b) const
+        {
+            return saturate_cast<int>(scale * a * b);
+        }
         const double scale;
     };
 
@@ -517,11 +535,36 @@ namespace cv { namespace gpu { namespace device
         enum { smart_shift = 4 };
     };
 
+    template <typename T, typename D> struct MultiplyCaller
+    {
+        static void call(const DevMem2Db& src1, const DevMem2Db& src2, const DevMem2Db& dst, double scale, cudaStream_t stream)
+        {
+            Multiply<T, D> op(static_cast<float>(scale));
+            cv::gpu::device::transform((DevMem2D_<T>)src1, (DevMem2D_<T>)src2, (DevMem2D_<D>)dst, op, WithOutMask(), stream);
+        }
+    };
+    template <typename T> struct MultiplyCaller<T, double>
+    {
+        static void call(const DevMem2Db& src1, const DevMem2Db& src2, const DevMem2Db& dst, double scale, cudaStream_t stream)
+        {
+            cudaSafeCall( cudaSetDoubleForDevice(&scale) );
+            Multiply<T, double> op(scale);
+            cv::gpu::device::transform((DevMem2D_<T>)src1, (DevMem2D_<T>)src2, (DevMem2D_<double>)dst, op, WithOutMask(), stream);
+        }
+    };
+    template <> struct MultiplyCaller<int, int>
+    {
+        static void call(const DevMem2Db& src1, const DevMem2Db& src2, const DevMem2Db& dst, double scale, cudaStream_t stream)
+        {
+            cudaSafeCall( cudaSetDoubleForDevice(&scale) );
+            Multiply<int, int> op(scale);
+            cv::gpu::device::transform((DevMem2D_<int>)src1, (DevMem2D_<int>)src2, (DevMem2D_<int>)dst, op, WithOutMask(), stream);
+        }
+    };
+
     template <typename T, typename D> void multiply_gpu(const DevMem2Db& src1, const DevMem2Db& src2, const DevMem2Db& dst, double scale, cudaStream_t stream)
     {
-        cudaSafeCall( cudaSetDoubleForDevice(&scale) );
-        Multiply<T, D> op(scale);
-        cv::gpu::device::transform((DevMem2D_<T>)src1, (DevMem2D_<T>)src2, (DevMem2D_<D>)dst, op, WithOutMask(), stream);
+        MultiplyCaller<T, D>::call(src1, src2, dst, scale, stream);
     }
 
     template void multiply_gpu<uchar, uchar >(const DevMem2Db& src1, const DevMem2Db& src2, const DevMem2Db& dst, double scale, cudaStream_t stream);
@@ -729,7 +772,7 @@ namespace cv { namespace gpu { namespace device
         Divide(double scale_) : scale(scale_) {}
         __device__ __forceinline__ D operator ()(T a, T b) const
         {
-            return b != 0 ? saturate_cast<D>(scale * a / b) : 0;
+            return b != 0 ? saturate_cast<D>(a * scale / b) : 0;
         }
         const double scale;
     };

modules/gpu/src/stereobm.cpp

@@ -55,7 +55,7 @@ void cv::gpu::StereoBM_GPU::operator() ( const GpuMat&, const GpuMat&, GpuMat&,
 
 #else /* !defined (HAVE_CUDA) */
 
-namespace cv { namespace gpu { namespace device 
+namespace cv { namespace gpu { namespace device
 {
     namespace stereobm
     {
@@ -65,10 +65,13 @@ namespace cv { namespace gpu { namespace device
     }
 }}}
 
-const float defaultAvgTexThreshold = 3;
+namespace
+{
+    const float defaultAvgTexThreshold = 3;
+}
 
 cv::gpu::StereoBM_GPU::StereoBM_GPU()
-    : preset(BASIC_PRESET), ndisp(DEFAULT_NDISP), winSize(DEFAULT_WINSZ), avergeTexThreshold(defaultAvgTexThreshold)  
+    : preset(BASIC_PRESET), ndisp(DEFAULT_NDISP), winSize(DEFAULT_WINSZ), avergeTexThreshold(defaultAvgTexThreshold)
 {
 }
 
@@ -100,9 +103,9 @@ namespace
     {
         using namespace ::cv::gpu::device::stereobm;
 
-        CV_DbgAssert(left.rows == right.rows && left.cols == right.cols);
-        CV_DbgAssert(left.type() == CV_8UC1);
-        CV_DbgAssert(right.type() == CV_8UC1);
+        CV_Assert(left.rows == right.rows && left.cols == right.cols);
+        CV_Assert(left.type() == CV_8UC1);
+        CV_Assert(right.type() == CV_8UC1);
 
         disparity.create(left.size(), CV_8U);
         minSSD.create(left.size(), CV_32S);
@@ -115,7 +118,7 @@ namespace
             leBuf.create( left.size(),  left.type());
             riBuf.create(right.size(), right.type());
 
-		    prefilter_xsobel( left, leBuf, 31, stream);
+            prefilter_xsobel( left, leBuf, 31, stream);
             prefilter_xsobel(right, riBuf, 31, stream);
 
             le_for_bm = leBuf;

modules/highgui/src/window_gtk.cpp

@@ -254,7 +254,7 @@ static void cvImageWidget_set_size(GtkWidget * widget, int max_width, int max_he
 }
 
 static void
-cvImageWidget_size_allocate (GtkWidget     *widget,  
+cvImageWidget_size_allocate (GtkWidget     *widget,
                         GtkAllocation *allocation)
 {
   CvImageWidget *image_widget;
@@ -719,7 +719,7 @@ namespace
         void generateBitmapFont(const std::string& family, int height, int weight, bool italic, bool underline, int start, int count, int base) const;
 
         bool isGlContextInitialized() const;
-        
+
         PFNGLGENBUFFERSPROC    glGenBuffersExt;
         PFNGLDELETEBUFFERSPROC glDeleteBuffersExt;
 
@@ -866,22 +866,22 @@ namespace
 
         CV_FUNCNAME( "GlFuncTab_GTK::generateBitmapFont" );
 
-        __BEGIN__;        
-        
+        __BEGIN__;
+
         fontDecr = pango_font_description_new();
-        
+
         pango_font_description_set_size(fontDecr, height);
-        
+
         pango_font_description_set_family_static(fontDecr, family.c_str());
-        
+
         pango_font_description_set_weight(fontDecr, static_cast<PangoWeight>(weight));
-        
+
         pango_font_description_set_style(fontDecr, italic ? PANGO_STYLE_ITALIC : PANGO_STYLE_NORMAL);
-                
+
         pangoFont = gdk_gl_font_use_pango_font(fontDecr, start, count, base);
-        
+
         pango_font_description_free(fontDecr);
-        
+
         if (!pangoFont)
             CV_ERROR(CV_OpenGlApiCallError, "Can't create font");
 
@@ -960,13 +960,13 @@ namespace
 
     void releaseGlContext(CvWindow* window)
     {
-        CV_FUNCNAME( "releaseGlContext" );
+        //CV_FUNCNAME( "releaseGlContext" );
 
-        __BEGIN__;
+        //__BEGIN__;
 
         window->useGl = false;
 
-        __END__;
+        //__END__;
     }
 
     void drawGl(CvWindow* window)