Merge pull request #3800 from vrabaud:cuda

Get CUDA code to compile with clang CUDA and without CUDA #3800

Changelist:
- there are some syntactic changes: `<< <` -> `<<<`. For some reason, I do not need to change all those in the code.
-  `::min` -> `std::min` in `__host__` code
- `modules/cudaimgproc/src/moments.cpp` needs to have the CUDA code in the `#ifdef`
- The signature of `cv::cuda::swapChannels` is not exactly the same as the C++ one in `modules/cudaimgproc/src/color.cpp` 
- `cv::cuda::FarnebackOpticalFlow::create` needs to be explicit about which FarnebackOpticalFlow it returns

### Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch

Author: vrabaud

modules/cudaarithm/src/arithm.cpp

@@ -54,6 +54,8 @@ void cv::cuda::mulAndScaleSpectrums(InputArray, InputArray, OutputArray, int, fl
 
 void cv::cuda::dft(InputArray, OutputArray, Size, int, Stream&) { throw_no_cuda(); }
 
+Ptr<DFT> cv::cuda::createDFT(Size, int) { throw_no_cuda(); return Ptr<DFT>(); }
+
 Ptr<Convolution> cv::cuda::createConvolution(Size) { throw_no_cuda(); return Ptr<Convolution>(); }
 
 #else /* !defined (HAVE_CUDA) */

modules/cudaarithm/src/cuda/polar_cart.cu

@@ -289,9 +289,9 @@ namespace
         const T scale = angleInDegrees ? static_cast<T>(CV_PI / 180.0) : static_cast<T>(1.0);
 
         if (mag.empty())
-            polarToCartImpl_<T, false> << <grid, block, 0, stream >> >(mag, angle, x, y, scale);
+            polarToCartImpl_<T, false> <<<grid, block, 0, stream >>>(mag, angle, x, y, scale);
         else
-            polarToCartImpl_<T, true> << <grid, block, 0, stream >> >(mag, angle, x, y, scale);
+            polarToCartImpl_<T, true> <<<grid, block, 0, stream >>>(mag, angle, x, y, scale);
     }
 
     template <typename T>
@@ -305,9 +305,9 @@ namespace
         const T scale = angleInDegrees ? static_cast<T>(CV_PI / 180.0) : static_cast<T>(1.0);
 
         if (mag.empty())
-            polarToCartDstInterleavedImpl_<T, false> << <grid, block, 0, stream >> >(mag, angle, xy, scale);
+            polarToCartDstInterleavedImpl_<T, false> <<<grid, block, 0, stream >>>(mag, angle, xy, scale);
         else
-            polarToCartDstInterleavedImpl_<T, true> << <grid, block, 0, stream >> >(mag, angle, xy, scale);
+            polarToCartDstInterleavedImpl_<T, true> <<<grid, block, 0, stream >>>(mag, angle, xy, scale);
     }
 
     template <typename T>
@@ -320,7 +320,7 @@ namespace
 
         const T scale = angleInDegrees ? static_cast<T>(CV_PI / 180.0) : static_cast<T>(1.0);
 
-        polarToCartInterleavedImpl_<T> << <grid, block, 0, stream >> >(magAngle, xy, scale);
+        polarToCartInterleavedImpl_<T> <<<grid, block, 0, stream >>>(magAngle, xy, scale);
     }
 }
 

modules/cudaarithm/src/element_operations.cpp

@@ -84,8 +84,13 @@ void cv::cuda::magnitude(InputArray, InputArray, OutputArray, Stream&) { throw_n
 void cv::cuda::magnitudeSqr(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
 void cv::cuda::magnitudeSqr(InputArray, InputArray, OutputArray, Stream&) { throw_no_cuda(); }
 void cv::cuda::phase(InputArray, InputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
+void cv::cuda::phase(InputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
 void cv::cuda::cartToPolar(InputArray, InputArray, OutputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
+void cv::cuda::cartToPolar(InputArray, OutputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
+void cv::cuda::cartToPolar(InputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
 void cv::cuda::polarToCart(InputArray, InputArray, OutputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
+void cv::cuda::polarToCart(InputArray, InputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
+void cv::cuda::polarToCart(InputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
 
 #else
 

modules/cudaarithm/src/reductions.cpp

@@ -69,8 +69,10 @@ void cv::cuda::countNonZero(InputArray, OutputArray, Stream&) { throw_no_cuda();
 
 void cv::cuda::reduce(InputArray, OutputArray, int, int, int, Stream&) { throw_no_cuda(); }
 
-void cv::cuda::meanStdDev(InputArray, Scalar&, Scalar&) { throw_no_cuda(); }
+void cv::cuda::meanStdDev(InputArray, OutputArray, InputArray, Stream&) { throw_no_cuda(); }
 void cv::cuda::meanStdDev(InputArray, OutputArray, Stream&) { throw_no_cuda(); }
+void cv::cuda::meanStdDev(InputArray, Scalar&, Scalar&, InputArray) { throw_no_cuda(); }
+void cv::cuda::meanStdDev(InputArray, Scalar&, Scalar&) { throw_no_cuda(); }
 
 void cv::cuda::rectStdDev(InputArray, InputArray, OutputArray, Rect, Stream&) { throw_no_cuda(); }
 

modules/cudacodec/src/cuda/nv12_to_rgb.cu

@@ -179,9 +179,9 @@ void nv12ToBgra(const GpuMat& decodedFrame, GpuMat& outFrame, int width, int hei
     dim3 block(32, 8);
     dim3 grid(divUp(width, 2 * block.x), divUp(height, block.y));
     if (videoFullRangeFlag)
-        NV12_to_BGRA<true> << <grid, block, 0, stream >> > (decodedFrame.ptr<uchar>(), decodedFrame.step, outFrame.ptr<uint>(), outFrame.step, width, height);
+        NV12_to_BGRA<true> <<<grid, block, 0, stream >>> (decodedFrame.ptr<uchar>(), decodedFrame.step, outFrame.ptr<uint>(), outFrame.step, width, height);
     else
-        NV12_to_BGRA<false> << <grid, block, 0, stream >> > (decodedFrame.ptr<uchar>(), decodedFrame.step, outFrame.ptr<uint>(), outFrame.step, width, height);
+        NV12_to_BGRA<false> <<<grid, block, 0, stream >>> (decodedFrame.ptr<uchar>(), decodedFrame.step, outFrame.ptr<uint>(), outFrame.step, width, height);
     CV_CUDEV_SAFE_CALL(cudaGetLastError());
     if (stream == 0)
         CV_CUDEV_SAFE_CALL(cudaDeviceSynchronize());

modules/cudaimgproc/src/color.cpp

@@ -51,7 +51,7 @@ void cv::cuda::cvtColor(InputArray, OutputArray, int, int, Stream&) { throw_no_c
 
 void cv::cuda::demosaicing(InputArray, OutputArray, int, int, Stream&) { throw_no_cuda(); }
 
-void cv::cuda::swapChannels(InputOutputArray, const int[], Stream&) { throw_no_cuda(); }
+void cv::cuda::swapChannels(InputOutputArray, const int[4], Stream&) { throw_no_cuda(); }
 
 void cv::cuda::gammaCorrection(InputArray, OutputArray, bool, Stream&) { throw_no_cuda(); }
 

modules/cudaimgproc/src/connectedcomponents.cpp

@@ -9,8 +9,8 @@ using namespace cv::cuda;
 
 #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
 
-void cv::cuda::connectedComponents(InputArray img_, OutputArray labels_, int connectivity,
-    int ltype, ConnectedComponentsAlgorithmsTypes ccltype) { throw_no_cuda(); }
+void cv::cuda::connectedComponents(InputArray, OutputArray, int, int, ConnectedComponentsAlgorithmsTypes) { throw_no_cuda(); }
+void cv::cuda::connectedComponents(InputArray, OutputArray, int, int) { throw_no_cuda(); }
 
 #else /* !defined (HAVE_CUDA) */
 

modules/cudaimgproc/src/cuda/canny.cu

@@ -428,7 +428,7 @@ namespace canny
             cudaSafeCall( cudaMemsetAsync(d_counter, 0, sizeof(int), stream) );
 
             const dim3 block(128);
-            const dim3 grid(::min(count, 65535u), divUp(count, 65535), 1);
+            const dim3 grid(std::min(count, 65535), divUp(count, 65535), 1);
 
             edgesHysteresisGlobalKernel<<<grid, block, 0, stream>>>(map, st1, st2, d_counter, count);
             cudaSafeCall( cudaGetLastError() );
@@ -439,7 +439,7 @@ namespace canny
             cudaSafeCall( cudaMemcpyAsync(&count, d_counter, sizeof(int), cudaMemcpyDeviceToHost, stream) );
             cudaSafeCall( cudaStreamSynchronize(stream) );
 
-            count = min(count, map.cols * map.rows);
+            count = std::min(count, map.cols * map.rows);
 
             //std::swap(st1, st2);
             short2* tmp = st1;

modules/cudaimgproc/src/cuda/connectedcomponents.cu

@@ -317,19 +317,19 @@ void BlockBasedKomuraEquivalence(const cv::cuda::GpuMat& img, cv::cuda::GpuMat&
     grid_size = dim3((((img.cols + 1) / 2) - 1) / kblock_cols + 1, (((img.rows + 1) / 2) - 1) / kblock_rows + 1, 1);
     block_size = dim3(kblock_cols, kblock_rows, 1);
 
-    InitLabeling << <grid_size, block_size >> > (img, labels, last_pixel);
+    InitLabeling <<<grid_size, block_size >>> (img, labels, last_pixel);
     cudaSafeCall(cudaGetLastError());
 
-    Compression << <grid_size, block_size >> > (labels);
+    Compression <<<grid_size, block_size >>> (labels);
     cudaSafeCall(cudaGetLastError());
 
-    Merge << <grid_size, block_size >> > (labels, last_pixel);
+    Merge <<<grid_size, block_size >>> (labels, last_pixel);
     cudaSafeCall(cudaGetLastError());
 
-    Compression << <grid_size, block_size >> > (labels);
+    Compression <<<grid_size, block_size >>> (labels);
     cudaSafeCall(cudaGetLastError());
 
-    FinalLabeling << <grid_size, block_size >> > (img, labels);
+    FinalLabeling <<<grid_size, block_size >>> (img, labels);
     cudaSafeCall(cudaGetLastError());
 
     if (last_pixel_allocated) {

modules/cudaimgproc/src/cuda/generalized_hough.cu

@@ -302,7 +302,7 @@ namespace cv { namespace cuda { namespace device
             int totalCount;
             cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
 
-            totalCount = ::min(totalCount, maxSize);
+            totalCount = std::min(totalCount, maxSize);
 
             return totalCount;
         }
@@ -812,7 +812,7 @@ namespace cv { namespace cuda { namespace device
             int totalCount;
             cudaSafeCall( cudaMemcpy(&totalCount, counterPtr, sizeof(int), cudaMemcpyDeviceToHost) );
 
-            totalCount = ::min(totalCount, maxSize);
+            totalCount = std::min(totalCount, maxSize);
 
             return totalCount;
         }