1.x related fixes

modules/gpu/perf/perf_labeling.cpp

@@ -48,7 +48,9 @@ GPU_PERF_TEST(ConnectedComponents, cv::gpu::DeviceInfo, cv::Size)
     cv::gpu::DeviceInfo devInfo = GET_PARAM(0);
     cv::gpu::setDevice(devInfo.deviceID());
 
-    cv::Mat image = readImage("gpu/labeling/label.png", cv::IMREAD_GRAYSCALE);
+    cv::Mat image = readImage("gpu/labeling/aloe-disp.png", cv::IMREAD_GRAYSCALE);
+
+    cv::threshold(image, image, 150, 255, CV_THRESH_BINARY);
 
     cv::gpu::GpuMat mask;
     mask.create(image.rows, image.cols, CV_8UC1);

modules/gpu/perf_cpu/perf_labeling.cpp

@@ -0,0 +1,157 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+// By downloading, copying, installing or using the software you agree to this license.
+// If you do not agree to this license, do not download, install,
+// copy or use the software.
+//
+//
+//                          License Agreement
+//               For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//  * Redistributions of source code must retain the above copyright notice,
+//    this list of conditions and the following disclaimer.
+//
+//  * Redistributions in binary form must reproduce the above copyright notice,
+//    this list of conditions and the following disclaimer in the documentation
+//    and/or other materials provided with the distribution.
+//
+//  * The name of the copyright holders may not be used to endorse or promote products
+//    derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//M*/
+
+#include "perf_precomp.hpp"
+
+#ifdef HAVE_CUDA
+
+namespace {
+
+    struct GreedyLabeling
+    {
+        struct dot
+        {
+            int x;
+            int y;
+
+            static dot make(int i, int j)
+            {
+                dot d; d.x = i; d.y = j;
+                return d;
+            }
+        };
+
+        struct InInterval
+        {
+            InInterval(const int& _lo, const int& _hi) : lo(-_lo), hi(_hi) {};
+            const int lo, hi;
+
+            bool operator() (const unsigned char a, const unsigned char b) const
+            {
+                int d = a - b;
+                return lo <= d && d <= hi;
+            }
+        };
+
+        GreedyLabeling(cv::Mat img)
+        : image(img), _labels(image.size(), CV_32SC1, cv::Scalar::all(-1)) {stack = new dot[image.cols * image.rows];}
+
+        ~GreedyLabeling(){delete[] stack;}
+
+        void operator() (cv::Mat labels) const
+        {
+            InInterval inInt(0, 2);
+            int cc = -1;
+
+            int* dist_labels = (int*)labels.data;
+            int pitch = labels.step1();
+
+            unsigned char* source = (unsigned char*)image.data;
+            int width = image.cols;
+            int height = image.rows;
+
+            for (int j = 0; j < image.rows; ++j)
+                for (int i = 0; i < image.cols; ++i)
+                {
+                    if (dist_labels[j * pitch + i] != -1) continue;
+
+                    dot* top = stack;
+                    dot p = dot::make(i, j);
+                    cc++;
+
+                    dist_labels[j * pitch + i] = cc;
+
+                    while (top >= stack)
+                    {
+                        int*  dl = &dist_labels[p.y * pitch + p.x];
+                        unsigned char* sp = &source[p.y * image.step1() + p.x];
+
+                        dl[0] = cc;
+
+                        //right
+                        if( p.x < (width - 1) && dl[ +1] == -1 && inInt(sp[0], sp[+1]))
+                            *top++ = dot::make(p.x + 1, p.y);
+
+                        //left
+                        if( p.x > 0 && dl[-1] == -1 && inInt(sp[0], sp[-1]))
+                            *top++ = dot::make(p.x - 1, p.y);
+
+                        //bottom
+                        if( p.y < (height - 1) && dl[+pitch] == -1 && inInt(sp[0], sp[+image.step1()]))
+                            *top++ = dot::make(p.x, p.y + 1);
+
+                        //top
+                        if( p.y > 0 && dl[-pitch] == -1 && inInt(sp[0], sp[-image.step1()]))
+                            *top++ = dot::make(p.x, p.y - 1);
+
+                        p = *--top;
+                    }
+                }
+        }
+
+        cv::Mat image;
+        cv::Mat _labels;
+        dot* stack;
+    };
+}
+
+GPU_PERF_TEST(ConnectedComponents, cv::gpu::DeviceInfo, cv::Size)
+{
+    cv::gpu::DeviceInfo devInfo = GET_PARAM(0);
+    cv::gpu::setDevice(devInfo.deviceID());
+
+    cv::Mat image = readImage("gpu/labeling/aloe-disp.png", cv::IMREAD_GRAYSCALE);
+
+    GreedyLabeling host(image);
+
+    host(host._labels);
+
+    declare.time(1.0);
+
+    TEST_CYCLE()
+    {
+        host(host._labels);
+    }
+}
+
+INSTANTIATE_TEST_CASE_P(Labeling, ConnectedComponents, testing::Combine(ALL_DEVICES, testing::Values(cv::Size(261, 262))));
+
+#endif

modules/gpu/src/cuda/ccomponetns.cu

@@ -42,6 +42,7 @@
 #include <opencv2/gpu/device/common.hpp>
 #include <opencv2/gpu/device/vec_traits.hpp>
 #include <opencv2/gpu/device/vec_math.hpp>
+#include <opencv2/gpu/device/emulation.hpp>
 #include <iostream>
 #include <stdio.h>
 
@@ -255,8 +256,7 @@ namespace cv { namespace gpu { namespace device
                     edgesTile[yloc][xloc] = c;
                 }
 
-
-            for (int i = 0; ; ++i)
+            for (int k = 0; ;++k)
             {
                 //1. backup
                 #pragma unroll
@@ -312,11 +312,12 @@ namespace cv { namespace gpu { namespace device
                         if (new_labels[i][j] < old_labels[i][j])
                         {
                             changed = 1;
-                            atomicMin(&labelsTile[0][0] + old_labels[i][j], new_labels[i][j]);
+                            Emulation::smem::atomicMin(&labelsTile[0][0] + old_labels[i][j], new_labels[i][j]);
                         }
                     }
 
-                changed = __syncthreads_or(changed);
+                changed = Emulation::sycthOr(changed);
+
                 if (!changed)
                     break;
 

modules/gpu/src/graphcuts.cpp

@@ -108,6 +108,8 @@ void cv::gpu::connectivityMask(const GpuMat& image, GpuMat& mask, const cv::Scal
 
 void cv::gpu::labelComponents(const GpuMat& mask, GpuMat& components, int flags, Stream& s)
 {
+    if (!TargetArchs::builtWith(SHARED_ATOMICS) || !DeviceInfo().supports(SHARED_ATOMICS))
+        CV_Error(CV_StsNotImplemented, "The device doesn't support shared atomics and communicative synchronization!");
     CV_Assert(!mask.empty() && mask.type() == CV_8U);
 
     if (mask.size() != components.size() || components.type() != CV_32SC1)

modules/gpu/src/opencv2/gpu/device/emulation.hpp

@@ -50,6 +50,17 @@ namespace cv { namespace gpu { namespace device
 {
     struct Emulation
     {
+
+        static __device__ __forceinline__ int sycthOr(int pred)
+        {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
+                // just campilation stab
+                return false;
+#else
+                return __syncthreads_or(pred);
+#endif
+        }
+
         template<int CTA_SIZE>
         static __forceinline__ __device__ int Ballot(int predicate)
         {

modules/gpu/test/test_labeling.cpp

@@ -164,7 +164,7 @@ struct Labeling : testing::TestWithParam<cv::gpu::DeviceInfo>
 
     cv::Mat loat_image()
     {
-        return cv::imread(std::string( cvtest::TS::ptr()->get_data_path() ) + "labeling/label.png");
+        return cv::imread(std::string( cvtest::TS::ptr()->get_data_path() ) + "labeling/IMG_0727.JPG");
     }
 };
 
@@ -173,6 +173,8 @@ TEST_P(Labeling, ConnectedComponents)
     cv::Mat image;
     cvtColor(loat_image(), image, CV_BGR2GRAY);
 
+    cv::threshold(image, image, 150, 255, CV_THRESH_BINARY);
+
     ASSERT_TRUE(image.type() == CV_8UC1);
 
     GreedyLabeling host(image);
@@ -189,6 +191,10 @@ TEST_P(Labeling, ConnectedComponents)
     ASSERT_NO_THROW(cv::gpu::labelComponents(mask, components));
 
     host.checkCorrectness(cv::Mat(components));
+    cv::imshow("test", image);
+    cv::waitKey(0);
+    cv::imshow("test", host._labels);
+    cv::waitKey(0);
 }
 
 INSTANTIATE_TEST_CASE_P(ConnectedComponents, Labeling, ALL_DEVICES);