1 /*M/////////////////////////////////////////////////////////////////////////////////////// 2 // 3 // IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING. 4 // 5 // By downloading, copying, installing or using the software you agree to this license. 6 // If you do not agree to this license, do not download, install, 7 // copy or use the software. 8 // 9 // 10 // License Agreement 11 // For Open Source Computer Vision Library 12 // 13 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved. 14 // Copyright (C) 2009, Willow Garage Inc., all rights reserved. 15 // Copyright (C) 2013, OpenCV Foundation, all rights reserved. 16 // Third party copyrights are property of their respective owners. 17 // 18 // Redistribution and use in source and binary forms, with or without modification, 19 // are permitted provided that the following conditions are met: 20 // 21 // * Redistribution's of source code must retain the above copyright notice, 22 // this list of conditions and the following disclaimer. 23 // 24 // * Redistribution's in binary form must reproduce the above copyright notice, 25 // this list of conditions and the following disclaimer in the documentation 26 // and/or other materials provided with the distribution. 27 // 28 // * The name of the copyright holders may not be used to endorse or promote products 29 // derived from this software without specific prior written permission. 30 // 31 // This software is provided by the copyright holders and contributors "as is" and 32 // any express or implied warranties, including, but not limited to, the implied 33 // warranties of merchantability and fitness for a particular purpose are disclaimed. 34 // In no event shall the Intel Corporation or contributors be liable for any direct, 35 // indirect, incidental, special, exemplary, or consequential damages 36 // (including, but not limited to, procurement of substitute goods or services; 37 // loss of use, data, or profits; or business interruption) however caused 38 // and on any theory of liability, whether in contract, strict liability, 39 // or tort (including negligence or otherwise) arising in any way out of 40 // the use of this software, even if advised of the possibility of such damage. 41 // 42 //M*/ 43 44 #pragma once 45 46 #ifndef __OPENCV_CUDEV_GRID_MINMAXLOC_DETAIL_HPP__ 47 #define __OPENCV_CUDEV_GRID_MINMAXLOC_DETAIL_HPP__ 48 49 #include "../../common.hpp" 50 #include "../../util/vec_traits.hpp" 51 #include "../../util/type_traits.hpp" 52 #include "../../util/limits.hpp" 53 #include "../../block/reduce.hpp" 54 55 namespace cv { namespace cudev { 56 57 namespace grid_minmaxloc_detail 58 { 59 template <int BLOCK_SIZE, class SrcPtr, typename ResType, class MaskPtr> 60 __global__ void minMaxLoc_pass_1(const SrcPtr src, ResType* minVal, ResType* maxVal, int* minLoc, int* maxLoc, const MaskPtr mask, const int rows, const int cols, const int patch_y, const int patch_x) 61 { 62 __shared__ ResType sMinVal[BLOCK_SIZE]; 63 __shared__ ResType sMaxVal[BLOCK_SIZE]; 64 __shared__ uint sMinLoc[BLOCK_SIZE]; 65 __shared__ uint sMaxLoc[BLOCK_SIZE]; 66 67 const int x0 = blockIdx.x * blockDim.x * patch_x + threadIdx.x; 68 const int y0 = blockIdx.y * blockDim.y * patch_y + threadIdx.y; 69 70 ResType myMin = numeric_limits<ResType>::max(); 71 ResType myMax = -numeric_limits<ResType>::max(); 72 int myMinLoc = -1; 73 int myMaxLoc = -1; 74 75 for (int i = 0, y = y0; i < patch_y && y < rows; ++i, y += blockDim.y) 76 { 77 for (int j = 0, x = x0; j < patch_x && x < cols; ++j, x += blockDim.x) 78 { 79 if (mask(y, x)) 80 { 81 const ResType srcVal = src(y, x); 82 83 if (srcVal < myMin) 84 { 85 myMin = srcVal; 86 myMinLoc = y * cols + x; 87 } 88 89 if (srcVal > myMax) 90 { 91 myMax = srcVal; 92 myMaxLoc = y * cols + x; 93 } 94 } 95 } 96 } 97 98 const int tid = threadIdx.y * blockDim.x + threadIdx.x; 99 100 blockReduceKeyVal<BLOCK_SIZE>(smem_tuple(sMinVal, sMaxVal), tie(myMin, myMax), 101 smem_tuple(sMinLoc, sMaxLoc), tie(myMinLoc, myMaxLoc), 102 tid, 103 make_tuple(less<ResType>(), greater<ResType>())); 104 105 const int bid = blockIdx.y * gridDim.x + blockIdx.x; 106 107 if (tid == 0) 108 { 109 minVal[bid] = myMin; 110 maxVal[bid] = myMax; 111 minLoc[bid] = myMinLoc; 112 maxLoc[bid] = myMaxLoc; 113 } 114 } 115 116 template <int BLOCK_SIZE, typename T> 117 __global__ void minMaxLoc_pass_2(T* minMal, T* maxVal, int* minLoc, int* maxLoc, int count) 118 { 119 __shared__ T sMinVal[BLOCK_SIZE]; 120 __shared__ T sMaxVal[BLOCK_SIZE]; 121 __shared__ int sMinLoc[BLOCK_SIZE]; 122 __shared__ int sMaxLoc[BLOCK_SIZE]; 123 124 const int idx = ::min(threadIdx.x, count - 1); 125 126 T myMin = minMal[idx]; 127 T myMax = maxVal[idx]; 128 int myMinLoc = minLoc[idx]; 129 int myMaxLoc = maxLoc[idx]; 130 131 blockReduceKeyVal<BLOCK_SIZE>(smem_tuple(sMinVal, sMaxVal), tie(myMin, myMax), 132 smem_tuple(sMinLoc, sMaxLoc), tie(myMinLoc, myMaxLoc), 133 threadIdx.x, 134 make_tuple(less<T>(), greater<T>())); 135 136 if (threadIdx.x == 0) 137 { 138 minMal[0] = myMin; 139 maxVal[0] = myMax; 140 minLoc[0] = myMinLoc; 141 maxLoc[0] = myMaxLoc; 142 } 143 } 144 145 template <class Policy> 146 void getLaunchCfg(int rows, int cols, dim3& block, dim3& grid) 147 { 148 block = dim3(Policy::block_size_x, Policy::block_size_y); 149 grid = dim3(divUp(cols, block.x * Policy::patch_size_x), divUp(rows, block.y * Policy::patch_size_y)); 150 151 grid.x = ::min(grid.x, block.x); 152 grid.y = ::min(grid.y, block.y); 153 } 154 155 template <class Policy, class SrcPtr, typename ResType, class MaskPtr> 156 __host__ void minMaxLoc(const SrcPtr& src, ResType* minVal, ResType* maxVal, int* minLoc, int* maxLoc, const MaskPtr& mask, int rows, int cols, cudaStream_t stream) 157 { 158 dim3 block, grid; 159 getLaunchCfg<Policy>(rows, cols, block, grid); 160 161 const int patch_x = divUp(divUp(cols, grid.x), block.x); 162 const int patch_y = divUp(divUp(rows, grid.y), block.y); 163 164 minMaxLoc_pass_1<Policy::block_size_x * Policy::block_size_y><<<grid, block, 0, stream>>>(src, minVal, maxVal, minLoc, maxLoc, mask, rows, cols, patch_y, patch_x); 165 CV_CUDEV_SAFE_CALL( cudaGetLastError() ); 166 167 minMaxLoc_pass_2<Policy::block_size_x * Policy::block_size_y><<<1, Policy::block_size_x * Policy::block_size_y, 0, stream>>>(minVal, maxVal, minLoc, maxLoc, grid.x * grid.y); 168 CV_CUDEV_SAFE_CALL( cudaGetLastError() ); 169 170 if (stream == 0) 171 CV_CUDEV_SAFE_CALL( cudaDeviceSynchronize() ); 172 } 173 } 174 175 }} 176 177 #endif 178
