fixed some bugs in GPU matrix reductions, removed <functional> into precomp.hpp

modules/gpu/src/cuda/matrix_reductions.cu

@@ -328,7 +328,7 @@ namespace cv { namespace gpu { namespace mathfunc
         __shared__ best_type smaxval[nthreads];
         
         uint tid = threadIdx.y * blockDim.x + threadIdx.x;
-        uint idx = min(tid, gridDim.x * gridDim.y - 1);
+        uint idx = min(tid, size - 1);
 
         sminval[tid] = minval[idx];
         smaxval[tid] = maxval[idx];
@@ -671,7 +671,7 @@ namespace cv { namespace gpu { namespace mathfunc
         __shared__ uint smaxloc[nthreads];
 
         uint tid = threadIdx.y * blockDim.x + threadIdx.x;
-        uint idx = min(tid, gridDim.x * gridDim.y - 1);
+        uint idx = min(tid, size - 1);
 
         sminval[tid] = minval[idx];
         smaxval[tid] = maxval[idx];
@@ -1150,7 +1150,7 @@ namespace cv { namespace gpu { namespace mathfunc
 
         const int tid = threadIdx.y * blockDim.x + threadIdx.x;
 
-        DstType res = tid < gridDim.x * gridDim.y ? result[tid] : VecTraits<DstType>::all(0);
+        DstType res = tid < size ? result[tid] : VecTraits<DstType>::all(0);
         smem[tid] = res.x;
         smem[tid + nthreads] = res.y;
         __syncthreads();
@@ -1262,7 +1262,7 @@ namespace cv { namespace gpu { namespace mathfunc
 
         const int tid = threadIdx.y * blockDim.x + threadIdx.x;
 
-        DstType res = tid < gridDim.x * gridDim.y ? result[tid] : VecTraits<DstType>::all(0);
+        DstType res = tid < size ? result[tid] : VecTraits<DstType>::all(0);
         smem[tid] = res.x;
         smem[tid + nthreads] = res.y;
         smem[tid + 2 * nthreads] = res.z;
@@ -1384,7 +1384,7 @@ namespace cv { namespace gpu { namespace mathfunc
 
         const int tid = threadIdx.y * blockDim.x + threadIdx.x;
 
-        DstType res = tid < gridDim.x * gridDim.y ? result[tid] : VecTraits<DstType>::all(0);
+        DstType res = tid < size ? result[tid] : VecTraits<DstType>::all(0);
         smem[tid] = res.x;
         smem[tid + nthreads] = res.y;
         smem[tid + 2 * nthreads] = res.z;

modules/gpu/src/initialization.cpp

@@ -41,7 +41,6 @@
 //M*/
 
 #include "precomp.hpp"
-#include <functional>
 
 using namespace cv;
 using namespace cv::gpu;

modules/gpu/src/matrix_reductions.cpp

@@ -276,11 +276,11 @@ void cv::gpu::minMax(const GpuMat& src, double* minVal, double* maxVal, const Gp
             minMaxMaskCaller<double> };
 
     CV_Assert(src.channels() == 1);
+
     CV_Assert(mask.empty() || (mask.type() == CV_8U && src.size() == mask.size()));
 
-    bool double_ok = hasGreaterOrEqualVersion(1, 3) && 
-                     hasNativeDoubleSupport(getDevice());
-    CV_Assert(src.type() != CV_64F || double_ok);
+    CV_Assert(src.type() != CV_64F || (hasGreaterOrEqualVersion(1, 3) && 
+                                       hasNativeDoubleSupport(getDevice())));
 
     double minVal_; if (!minVal) minVal = &minVal_;
     double maxVal_; if (!maxVal) maxVal = &maxVal_;
@@ -375,11 +375,11 @@ void cv::gpu::minMaxLoc(const GpuMat& src, double* minVal, double* maxVal, Point
             minMaxLocMaskCaller<double> };
 
     CV_Assert(src.channels() == 1);
+
     CV_Assert(mask.empty() || (mask.type() == CV_8U && src.size() == mask.size()));
 
-    bool double_ok = hasGreaterOrEqualVersion(1, 3) && 
-                     hasNativeDoubleSupport(getDevice());
-    CV_Assert(src.type() != CV_64F || double_ok);
+    CV_Assert(src.type() != CV_64F || (hasGreaterOrEqualVersion(1, 3) && 
+                                       hasNativeDoubleSupport(getDevice())));
 
     double minVal_; if (!minVal) minVal = &minVal_;
     double maxVal_; if (!maxVal) maxVal = &maxVal_;
@@ -459,9 +459,8 @@ int cv::gpu::countNonZero(const GpuMat& src, GpuMat& buf)
 
     CV_Assert(src.channels() == 1);
 
-    bool double_ok = hasGreaterOrEqualVersion(1, 3) && 
-                     hasNativeDoubleSupport(getDevice());
-    CV_Assert(src.type() != CV_64F || double_ok);
+    CV_Assert(src.type() != CV_64F || (hasGreaterOrEqualVersion(1, 3) && 
+                                       hasNativeDoubleSupport(getDevice())));
 
     Size buf_size;
     getBufSizeRequired(src.cols, src.rows, buf_size.width, buf_size.height);

modules/gpu/src/precomp.hpp

@@ -57,6 +57,7 @@
 #include <sstream>
 #include <exception>
 #include <iterator>
+#include <functional>
 
 #include "opencv2/gpu/gpu.hpp"
 #include "opencv2/imgproc/imgproc.hpp"

tests/gpu/src/arithm.cpp

@@ -49,7 +49,7 @@ using namespace std;
 using namespace gpu;
 
 #define CHECK(pred, err) if (!(pred)) { \
-    ts->printf(CvTS::LOG, "Fail: \"%s\" at line: %d\n", #pred, __LINE__); \
+    ts->printf(CvTS::CONSOLE, "Fail: \"%s\" at line: %d\n", #pred, __LINE__); \
     ts->set_failed_test_info(err); \
     return; }