opencl_kernels_calib3d.cpp
8.51 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
// This file is auto-generated. Do not edit!
#include "opencv2/core.hpp"
#include "cvconfig.h"
#include "opencl_kernels_calib3d.hpp"
#ifdef HAVE_OPENCL
namespace cv
{
namespace ocl
{
namespace calib3d
{
static const char* const moduleName = "calib3d";
struct cv::ocl::internal::ProgramEntry stereobm_oclsrc={moduleName, "stereobm",
"#define MAX_VAL 32767\n"
"#ifndef WSZ\n"
"#define WSZ 2\n"
"#endif\n"
"#define WSZ2 (WSZ / 2)\n"
"#ifdef DEFINE_KERNEL_STEREOBM\n"
"#define DISPARITY_SHIFT 4\n"
"#define FILTERED ((MIN_DISP - 1) << DISPARITY_SHIFT)\n"
"void calcDisp(__local short * cost, __global short * disp, int uniquenessRatio,\n"
"__local int * bestDisp, __local int * bestCost, int d, int x, int y, int cols, int rows)\n"
"{\n"
"int best_disp = *bestDisp, best_cost = *bestCost;\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"short c = cost[0];\n"
"int thresh = best_cost + (best_cost * uniquenessRatio / 100);\n"
"bool notUniq = ( (c <= thresh) && (d < (best_disp - 1) || d > (best_disp + 1) ) );\n"
"if (notUniq)\n"
"*bestCost = FILTERED;\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"if( *bestCost != FILTERED && x < cols - WSZ2 - MIN_DISP && y < rows - WSZ2 && d == best_disp)\n"
"{\n"
"int d_aprox = 0;\n"
"int yp =0, yn = 0;\n"
"if ((0 < best_disp) && (best_disp < NUM_DISP - 1))\n"
"{\n"
"yp = cost[-2 * BLOCK_SIZE_Y];\n"
"yn = cost[2 * BLOCK_SIZE_Y];\n"
"d_aprox = yp + yn - 2 * c + abs(yp - yn);\n"
"}\n"
"disp[0] = (short)(((best_disp + MIN_DISP)*256 + (d_aprox != 0 ? (yp - yn) * 256 / d_aprox : 0) + 15) >> 4);\n"
"}\n"
"}\n"
"short calcCostBorder(__global const uchar * leftptr, __global const uchar * rightptr, int x, int y, int nthread,\n"
"short * costbuf, int *h, int cols, int d, short cost)\n"
"{\n"
"int head = (*h) % WSZ;\n"
"__global const uchar * left, * right;\n"
"int idx = mad24(y + WSZ2 * (2 * nthread - 1), cols, x + WSZ2 * (1 - 2 * nthread));\n"
"left = leftptr + idx;\n"
"right = rightptr + (idx - d);\n"
"short costdiff = 0;\n"
"if (0 == nthread)\n"
"{\n"
"#pragma unroll\n"
"for (int i = 0; i < WSZ; i++)\n"
"{\n"
"costdiff += abs( left[0] - right[0] );\n"
"left += cols;\n"
"right += cols;\n"
"}\n"
"}\n"
"else\n"
"{\n"
"#pragma unroll\n"
"for (int i = 0; i < WSZ; i++)\n"
"{\n"
"costdiff += abs(left[i] - right[i]);\n"
"}\n"
"}\n"
"cost += costdiff - costbuf[head];\n"
"costbuf[head] = costdiff;\n"
"*h = head + 1;\n"
"return cost;\n"
"}\n"
"short calcCostInside(__global const uchar * leftptr, __global const uchar * rightptr, int x, int y,\n"
"int cols, int d, short cost_up_left, short cost_up, short cost_left)\n"
"{\n"
"__global const uchar * left, * right;\n"
"int idx = mad24(y - WSZ2 - 1, cols, x - WSZ2 - 1);\n"
"left = leftptr + idx;\n"
"right = rightptr + (idx - d);\n"
"int idx2 = WSZ*cols;\n"
"uchar corrner1 = abs(left[0] - right[0]),\n"
"corrner2 = abs(left[WSZ] - right[WSZ]),\n"
"corrner3 = abs(left[idx2] - right[idx2]),\n"
"corrner4 = abs(left[idx2 + WSZ] - right[idx2 + WSZ]);\n"
"return cost_up + cost_left - cost_up_left + corrner1 -\n"
"corrner2 - corrner3 + corrner4;\n"
"}\n"
"__kernel void stereoBM(__global const uchar * leftptr,\n"
"__global const uchar * rightptr,\n"
"__global uchar * dispptr, int disp_step, int disp_offset,\n"
"int rows, int cols,\n"
"int textureThreshold, int uniquenessRatio)\n"
"{\n"
"int lz = get_local_id(0);\n"
"int gx = get_global_id(1) * BLOCK_SIZE_X;\n"
"int gy = get_global_id(2) * BLOCK_SIZE_Y;\n"
"int nthread = lz / NUM_DISP;\n"
"int disp_idx = lz % NUM_DISP;\n"
"__global short * disp;\n"
"__global const uchar * left, * right;\n"
"__local short costFunc[2 * BLOCK_SIZE_Y * NUM_DISP];\n"
"__local short * cost;\n"
"__local int best_disp[2];\n"
"__local int best_cost[2];\n"
"best_cost[nthread] = MAX_VAL;\n"
"best_disp[nthread] = -1;\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"short costbuf[WSZ];\n"
"int head = 0;\n"
"int shiftX = WSZ2 + NUM_DISP + MIN_DISP - 1;\n"
"int shiftY = WSZ2;\n"
"int x = gx + shiftX, y = gy + shiftY, lx = 0, ly = 0;\n"
"int costIdx = disp_idx * 2 * BLOCK_SIZE_Y + (BLOCK_SIZE_Y - 1);\n"
"cost = costFunc + costIdx;\n"
"int tempcost = 0;\n"
"if (x < cols - WSZ2 - MIN_DISP && y < rows - WSZ2)\n"
"{\n"
"if (0 == nthread)\n"
"{\n"
"#pragma unroll\n"
"for (int i = 0; i < WSZ; i++)\n"
"{\n"
"int idx = mad24(y - WSZ2, cols, x - WSZ2 + i);\n"
"left = leftptr + idx;\n"
"right = rightptr + (idx - disp_idx);\n"
"short costdiff = 0;\n"
"for(int j = 0; j < WSZ; j++)\n"
"{\n"
"costdiff += abs( left[0] - right[0] );\n"
"left += cols;\n"
"right += cols;\n"
"}\n"
"costbuf[i] = costdiff;\n"
"}\n"
"}\n"
"else\n"
"{\n"
"#pragma unroll\n"
"for (int i = 0; i < WSZ; i++)\n"
"{\n"
"int idx = mad24(y - WSZ2 + i, cols, x - WSZ2);\n"
"left = leftptr + idx;\n"
"right = rightptr + (idx - disp_idx);\n"
"short costdiff = 0;\n"
"for (int j = 0; j < WSZ; j++)\n"
"{\n"
"costdiff += abs( left[j] - right[j]);\n"
"}\n"
"tempcost += costdiff;\n"
"costbuf[i] = costdiff;\n"
"}\n"
"}\n"
"}\n"
"if (nthread == 1)\n"
"{\n"
"cost[0] = tempcost;\n"
"atomic_min(best_cost + 1, tempcost);\n"
"}\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"if (best_cost[1] == tempcost)\n"
"atomic_max(best_disp + 1, disp_idx);\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"int dispIdx = mad24(gy, disp_step, mad24((int)sizeof(short), gx, disp_offset));\n"
"disp = (__global short *)(dispptr + dispIdx);\n"
"calcDisp(cost, disp, uniquenessRatio, best_disp + 1, best_cost + 1, disp_idx, x, y, cols, rows);\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"lx = 1 - nthread;\n"
"ly = nthread;\n"
"for (int i = 0; i < BLOCK_SIZE_Y * BLOCK_SIZE_X / 2; i++)\n"
"{\n"
"x = (lx < BLOCK_SIZE_X) ? gx + shiftX + lx : cols;\n"
"y = (ly < BLOCK_SIZE_Y) ? gy + shiftY + ly : rows;\n"
"best_cost[nthread] = MAX_VAL;\n"
"best_disp[nthread] = -1;\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"costIdx = mad24(2 * BLOCK_SIZE_Y, disp_idx, (BLOCK_SIZE_Y - 1 - ly + lx));\n"
"if (0 > costIdx)\n"
"costIdx = BLOCK_SIZE_Y - 1;\n"
"cost = costFunc + costIdx;\n"
"if (x < cols - WSZ2 - MIN_DISP && y < rows - WSZ2)\n"
"{\n"
"tempcost = (ly * (1 - nthread) + lx * nthread == 0) ?\n"
"calcCostBorder(leftptr, rightptr, x, y, nthread, costbuf, &head, cols, disp_idx, cost[2*nthread-1]) :\n"
"calcCostInside(leftptr, rightptr, x, y, cols, disp_idx, cost[0], cost[1], cost[-1]);\n"
"}\n"
"cost[0] = tempcost;\n"
"atomic_min(best_cost + nthread, tempcost);\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"if (best_cost[nthread] == tempcost)\n"
"atomic_max(best_disp + nthread, disp_idx);\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"dispIdx = mad24(gy + ly, disp_step, mad24((int)sizeof(short), (gx + lx), disp_offset));\n"
"disp = (__global short *)(dispptr + dispIdx);\n"
"calcDisp(cost, disp, uniquenessRatio, best_disp + nthread, best_cost + nthread, disp_idx, x, y, cols, rows);\n"
"barrier(CLK_LOCAL_MEM_FENCE);\n"
"if (lx + nthread - 1 == ly)\n"
"{\n"
"lx = (lx + nthread + 1) * (1 - nthread);\n"
"ly = (ly + 1) * nthread;\n"
"}\n"
"else\n"
"{\n"
"lx += nthread;\n"
"ly = ly - nthread + 1;\n"
"}\n"
"}\n"
"}\n"
"#endif\n"
"__kernel void prefilter_norm(__global unsigned char *input, __global unsigned char *output,\n"
"int rows, int cols, int prefilterCap, int scale_g, int scale_s)\n"
"{\n"
"int x = get_global_id(0);\n"
"int y = get_global_id(1);\n"
"if(x < cols && y < rows)\n"
"{\n"
"int cov1 = input[ max(y-1, 0) * cols + x] * 1 +\n"
"input[y * cols + max(x-1,0)] * 1 + input[ y * cols + x] * 4 + input[y * cols + min(x+1, cols-1)] * 1 +\n"
"input[min(y+1, rows-1) * cols + x] * 1;\n"
"int cov2 = 0;\n"
"for(int i = -WSZ2; i < WSZ2+1; i++)\n"
"for(int j = -WSZ2; j < WSZ2+1; j++)\n"
"cov2 += input[clamp(y+i, 0, rows-1) * cols + clamp(x+j, 0, cols-1)];\n"
"int res = (cov1*scale_g - cov2*scale_s)>>10;\n"
"res = clamp(res, -prefilterCap, prefilterCap) + prefilterCap;\n"
"output[y * cols + x] = res;\n"
"}\n"
"}\n"
"__kernel void prefilter_xsobel(__global unsigned char *input, __global unsigned char *output,\n"
"int rows, int cols, int prefilterCap)\n"
"{\n"
"int x = get_global_id(0);\n"
"int y = get_global_id(1);\n"
"if(x < cols && y < rows)\n"
"{\n"
"if (0 < x && !((y == rows-1) & (rows%2==1) ) )\n"
"{\n"
"int cov = input[ ((y > 0) ? y-1 : y+1) * cols + (x-1)] * (-1) + input[ ((y > 0) ? y-1 : y+1) * cols + ((x<cols-1) ? x+1 : x-1)] * (1) +\n"
"input[ (y) * cols + (x-1)] * (-2) + input[ (y) * cols + ((x<cols-1) ? x+1 : x-1)] * (2) +\n"
"input[((y<rows-1)?(y+1):(y-1))* cols + (x-1)] * (-1) + input[((y<rows-1)?(y+1):(y-1))* cols + ((x<cols-1) ? x+1 : x-1)] * (1);\n"
"cov = clamp(cov, -prefilterCap, prefilterCap) + prefilterCap;\n"
"output[y * cols + x] = cov;\n"
"}\n"
"else\n"
"output[y * cols + x] = prefilterCap;\n"
"}\n"
"}\n"
, "f01ad97e355f1d5b0a7c03f8f324cf43", NULL};
}}}
#endif