fixed remap under cc < 2.0

diff --git a/modules/gpu/src/cuda/remap.cu b/modules/gpu/src/cuda/remap.cu
index 1ab8447..2f5f3cf 100644 (file)
--- a/modules/gpu/src/cuda/remap.cu
+++ b/modules/gpu/src/cuda/remap.cu
@@ -69,7 +69,8 @@ namespace cv { namespace gpu { namespace imgproc
 \r
     template <template <typename> class Filter, template <typename> class B, typename T> struct RemapDispatcherStream\r
     {\r
-        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, const float* borderValue, cudaStream_t stream)\r
+        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, \r
+            const float* borderValue, cudaStream_t stream, int)\r
         {\r
             typedef typename TypeVec<float, VecTraits<T>::cn>::vec_type work_type; \r
             \r
@@ -87,7 +88,7 @@ namespace cv { namespace gpu { namespace imgproc
     \r
     template <template <typename> class Filter, template <typename> class B, typename T> struct RemapDispatcherNonStream\r
     {\r
-        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, const float* borderValue)\r
+        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, const float* borderValue, int)\r
         {\r
             typedef typename TypeVec<float, VecTraits<T>::cn>::vec_type work_type; \r
             \r
@@ -118,10 +119,10 @@ namespace cv { namespace gpu { namespace imgproc
     }; \\r
     template <template <typename> class Filter, template <typename> class B> struct RemapDispatcherNonStream<Filter, B, type> \\r
     { \\r
-        static void call(const DevMem2D_< type >& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_< type >& dst, const float* borderValue) \\r
+        static void call(const DevMem2D_< type >& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_< type >& dst, const float* borderValue, int cc) \\r
         { \\r
             typedef typename TypeVec<float, VecTraits< type >::cn>::vec_type work_type; \\r
-            dim3 block(32, 8); \\r
+            dim3 block(32, cc >= 20 ? 8 : 4); \\r
             dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y)); \\r
             TextureBinder texHandler(&tex_remap_ ## type , src); \\r
             tex_remap_ ## type ##_reader texSrc; \\r
@@ -135,7 +136,7 @@ namespace cv { namespace gpu { namespace imgproc
     }; \\r
     template <template <typename> class Filter> struct RemapDispatcherNonStream<Filter, BrdReplicate, type> \\r
     { \\r
-        static void call(const DevMem2D_< type >& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_< type >& dst, const float*) \\r
+        static void call(const DevMem2D_< type >& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_< type >& dst, const float*, int) \\r
         { \\r
             dim3 block(32, 8); \\r
             dim3 grid(divUp(dst.cols, block.x), divUp(dst.rows, block.y)); \\r
@@ -176,18 +177,21 @@ namespace cv { namespace gpu { namespace imgproc
 \r
     template <template <typename> class Filter, template <typename> class B, typename T> struct RemapDispatcher\r
     { \r
-        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, const float* borderValue, cudaStream_t stream)\r
+        static void call(const DevMem2D_<T>& src, const DevMem2Df& mapx, const DevMem2Df& mapy, const DevMem2D_<T>& dst, \r
+            const float* borderValue, cudaStream_t stream, int cc)\r
         {\r
             if (stream == 0)\r
-                RemapDispatcherNonStream<Filter, B, T>::call(src, mapx, mapy, dst, borderValue);\r
+                RemapDispatcherNonStream<Filter, B, T>::call(src, mapx, mapy, dst, borderValue, cc);\r
             else\r
-                RemapDispatcherStream<Filter, B, T>::call(src, mapx, mapy, dst, borderValue, stream);\r
+                RemapDispatcherStream<Filter, B, T>::call(src, mapx, mapy, dst, borderValue, stream, cc);\r
         }\r
     };\r
 \r
-    template <typename T> void remap_gpu(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream)\r
+    template <typename T> void remap_gpu(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, \r
+        int borderMode, const float* borderValue, cudaStream_t stream, int cc)\r
     {\r
-        typedef void (*caller_t)(const DevMem2D_<T>& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2D_<T>& dst, const float* borderValue, cudaStream_t stream);\r
+        typedef void (*caller_t)(const DevMem2D_<T>& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2D_<T>& dst, \r
+            const float* borderValue, cudaStream_t stream, int cc);\r
 \r
         static const caller_t callers[3][5] = \r
         {\r
@@ -214,36 +218,36 @@ namespace cv { namespace gpu { namespace imgproc
             }\r
         };\r
 \r
-        callers[interpolation][borderMode](static_cast< DevMem2D_<T> >(src), xmap, ymap, static_cast< DevMem2D_<T> >(dst), borderValue, stream);\r
+        callers[interpolation][borderMode](static_cast< DevMem2D_<T> >(src), xmap, ymap, static_cast< DevMem2D_<T> >(dst), borderValue, stream, cc);\r
     }\r
 \r
-    template void remap_gpu<uchar >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    //template void remap_gpu<uchar2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    template void remap_gpu<uchar3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    template void remap_gpu<uchar4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
+    template void remap_gpu<uchar >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    //template void remap_gpu<uchar2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    template void remap_gpu<uchar3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    template void remap_gpu<uchar4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
     \r
-    //template void remap_gpu<schar>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    //template void remap_gpu<char2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    //template void remap_gpu<char3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    //template void remap_gpu<char4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
+    //template void remap_gpu<schar>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    //template void remap_gpu<char2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    //template void remap_gpu<char3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    //template void remap_gpu<char4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
     \r
-    template void remap_gpu<ushort >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    //template void remap_gpu<ushort2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    template void remap_gpu<ushort3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    template void remap_gpu<ushort4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
+    template void remap_gpu<ushort >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    //template void remap_gpu<ushort2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    template void remap_gpu<ushort3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    template void remap_gpu<ushort4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
     \r
-    template void remap_gpu<short >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    //template void remap_gpu<short2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    template void remap_gpu<short3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    template void remap_gpu<short4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
+    template void remap_gpu<short >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    //template void remap_gpu<short2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    template void remap_gpu<short3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    template void remap_gpu<short4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
     \r
-    //template void remap_gpu<int >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    //template void remap_gpu<int2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    //template void remap_gpu<int3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    //template void remap_gpu<int4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
+    //template void remap_gpu<int >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    //template void remap_gpu<int2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    //template void remap_gpu<int3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    //template void remap_gpu<int4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
     \r
-    template void remap_gpu<float >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    //template void remap_gpu<float2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    template void remap_gpu<float3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
-    template void remap_gpu<float4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
+    template void remap_gpu<float >(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    //template void remap_gpu<float2>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    template void remap_gpu<float3>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
+    template void remap_gpu<float4>(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
 }}}\r

diff --git a/modules/gpu/src/imgproc.cpp b/modules/gpu/src/imgproc.cpp
index 528ac65..3b47447 100644 (file)
--- a/modules/gpu/src/imgproc.cpp
+++ b/modules/gpu/src/imgproc.cpp
@@ -110,14 +110,15 @@ void cv::gpu::CannyBuf::release() { throw_nogpu(); }
 namespace cv { namespace gpu {  namespace imgproc\r
 {\r
     template <typename T> void remap_gpu(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, \r
-                                         int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
+                                         int interpolation, int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
 }}}\r
 \r
 void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const GpuMat& ymap, int interpolation, int borderMode, const Scalar& borderValue, Stream& stream)\r
 {\r
     using namespace cv::gpu::imgproc;\r
 \r
-    typedef void (*caller_t)(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, int borderMode, const float* borderValue, cudaStream_t stream);\r
+    typedef void (*caller_t)(const DevMem2Db& src, const DevMem2Df& xmap, const DevMem2Df& ymap, const DevMem2Db& dst, int interpolation, \r
+        int borderMode, const float* borderValue, cudaStream_t stream, int cc);\r
     static const caller_t callers[6][4] = \r
     {\r
         {remap_gpu<uchar>, 0/*remap_gpu<uchar2>*/, remap_gpu<uchar3>, remap_gpu<uchar4>},\r
@@ -145,7 +146,10 @@ void cv::gpu::remap(const GpuMat& src, GpuMat& dst, const GpuMat& xmap, const Gp
     Scalar_<float> borderValueFloat;\r
     borderValueFloat = borderValue;\r
 \r
-    func(src, xmap, ymap, dst, interpolation, gpuBorderType, borderValueFloat.val, StreamAccessor::getStream(stream));\r
+    DeviceInfo info;\r
+    int cc = info.majorVersion() * 10 + info.minorVersion();\r
+\r
+    func(src, xmap, ymap, dst, interpolation, gpuBorderType, borderValueFloat.val, StreamAccessor::getStream(stream), cc);\r
 }\r
 \r
 ////////////////////////////////////////////////////////////////////////\r