resize area with block scan

modules/gpu/src/cuda/resize.cu

@@ -620,7 +620,6 @@ namespace cv { namespace gpu { namespace device
         {
             (void)interpolation;
 
-            //TODO: add assert to picture size
             int iscale_x = round(fx);
             int iscale_y = round(fy);
 
@@ -641,8 +640,8 @@ namespace cv { namespace gpu { namespace device
             thred_lines = divUp(src.rows, threads);
             blocks = dst.cols * thred_lines;
 
-            printf("device code executed for Y coordinate with:\nwarps %d, threads %d, thred_lines %d, blocks %d\n",
+            printf("device code executed for Y coordinate with:\nsize %d warps %d, threads %d, thred_lines %d, blocks %d\n",
-                   warps, threads, thred_lines, blocks);
+                   dst.rows, warps, threads, thred_lines, blocks);
 
             resise_scan_fast_y<T, smem_type><<<blocks, threads, warps * 32 * sizeof(smem_type)>>>
                     (buffer, dst, iscale_x, iscale_y, thred_lines);

modules/gpu/src/resize.cpp

@@ -94,6 +94,7 @@ void cv::gpu::resize(const GpuMat& src, GpuMat& dst, GpuMat& buffer, Size dsize,
     CV_Assert(interpolation == INTER_AREA);
     CV_Assert( (fx < 1.0) && (fy < 1.0));
     CV_Assert(!(dsize == Size()) || (fx > 0 && fy > 0));
+    CV_Assert(src.cols >= 128 && src.rows >= 128);
 
     if (dsize == Size())
         dsize = Size(saturate_cast<int>(src.cols * fx), saturate_cast<int>(src.rows * fy));

modules/gpu/test/test_resize.cpp

@@ -201,13 +201,13 @@ TEST_P(ResizeArea, Accuracy)
    cv::Mat gpu;
    dst.download(gpu);
 
-   // std::cout << src
+   std::cout //<< src
    // << std::endl << std::endl
    // << gpu_buff
    // << std::endl << std::endl
-   // << gpu
+   << gpu
-   // << std::endl << std::endl
+   << std::endl << std::endl
-   // << dst_cpu<<  std::endl;
+   << dst_cpu<<  std::endl;
 
 
     EXPECT_MAT_NEAR(dst_cpu, dst, src.depth() == CV_32F ? 1e-2 : 1.0);
@@ -215,9 +215,9 @@ TEST_P(ResizeArea, Accuracy)
 
 INSTANTIATE_TEST_CASE_P(GPU_ImgProc, ResizeArea, testing::Combine(
     ALL_DEVICES,
-    testing::Values(cv::Size(512, 256)),//DIFFERENT_SIZES,
+    testing::Values(cv::Size(640, 10 *  128)),//DIFFERENT_SIZES,
     testing::Values(MatType(CV_8UC1)/*MatType(CV_8UC3), MatType(CV_16UC1), MatType(CV_16UC3), MatType(CV_16UC4), MatType(CV_32FC1), MatType(CV_32FC3), MatType(CV_32FC4)*/),
-    testing::Values(0.5),
+    testing::Values(0.1),
     testing::Values(Interpolation(cv::INTER_AREA)),
     WHOLE_SUBMAT));