fixed #1279

modules/gpu/src/cuda/surf.cu

@@ -566,9 +566,6 @@ namespace cv { namespace gpu { namespace surf
 
                 float* s_sum_row = s_sum + threadIdx.y * 32;
 
-                //reduceSum32(s_sum_row, sumx);
-                //reduceSum32(s_sum_row, sumy);
-
                 warpReduce32(s_sum_row, sumx, threadIdx.x, plus<volatile float>());
                 warpReduce32(s_sum_row, sumy, threadIdx.x, plus<volatile float>());
 

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

@@ -73,7 +73,7 @@ namespace cv {  namespace gpu { namespace device
     }
 
     // warp-synchronous 32 elements reduction
-    template <typename T, typename Op> __device__ __forceinline__ void warpReduce32(volatile T* data, volatile T& partial_reduction, int tid, Op op)
+    template <typename T, typename Op> __device__ __forceinline__ void warpReduce32(volatile T* data, T& partial_reduction, int tid, Op op)
     {
         data[tid] = partial_reduction;
 
@@ -88,7 +88,7 @@ namespace cv {  namespace gpu { namespace device
     }
 
     // warp-synchronous 16 elements reduction
-    template <typename T, typename Op> __device__ __forceinline__ void warpReduce16(volatile T* data, volatile T& partial_reduction, int tid, Op op)
+    template <typename T, typename Op> __device__ __forceinline__ void warpReduce16(volatile T* data, T& partial_reduction, int tid, Op op)
     {
         data[tid] = partial_reduction;
 
@@ -102,7 +102,7 @@ namespace cv {  namespace gpu { namespace device
     }
 
     // warp-synchronous reduction
-    template <int n, typename T, typename Op> __device__ __forceinline__ void warpReduce(volatile T* data, volatile T& partial_reduction, int tid, Op op)
+    template <int n, typename T, typename Op> __device__ __forceinline__ void warpReduce(volatile T* data, T& partial_reduction, int tid, Op op)
     {
         if (tid < n)
             data[tid] = partial_reduction;

modules/gpu/test/test_main.cpp

@@ -109,9 +109,11 @@ int main(int argc, char** argv)
     cvtest::TS::ptr()->init("gpu");
     testing::InitGoogleTest(&argc, argv);
 
-    //cv::CommandLineParser parser(argc, (const char**)argv);
+    const char* keys ="{ nvtest_output_level | nvtest_output_level | none | NVidia test verbosity level }";
 
-    std::string outputLevel = "none";//parser.get<std::string>("nvtest_output_level", "none");
+    cv::CommandLineParser parser(argc, (const char**)argv, keys);
+
+    std::string outputLevel = parser.get<std::string>("nvtest_output_level", "none");
 
     if (outputLevel == "none")
         nvidiaTestOutputLevel = OutputLevelNone;