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
// This file is part of OpenCV project.
// It is subject to the license terms in the LICENSE file found in the top-level directory
// of this distribution and at http://opencv.org/license.html
#include "precomp.hpp"
namespace cv { namespace hal {
CV_CPU_OPTIMIZATION_NAMESPACE_BEGIN
void split8u(const uchar* src, uchar** dst, int len, int cn);
void split16u(const ushort* src, ushort** dst, int len, int cn);
void split32s(const int* src, int** dst, int len, int cn);
void split64s(const int64* src, int64** dst, int len, int cn);
#ifndef CV_CPU_OPTIMIZATION_DECLARATIONS_ONLY
#if CV_SIMD
// see the comments for vecmerge_ in merge.cpp
template<typename T, typename VecT> static void
vecsplit_( const T* src, T** dst, int len, int cn )
{
const int VECSZ = VecT::nlanes;
int i, i0 = 0;
T* dst0 = dst[0];
T* dst1 = dst[1];
int r0 = (int)((size_t)(void*)dst0 % (VECSZ*sizeof(T)));
int r1 = (int)((size_t)(void*)dst1 % (VECSZ*sizeof(T)));
int r2 = cn > 2 ? (int)((size_t)(void*)dst[2] % (VECSZ*sizeof(T))) : r0;
int r3 = cn > 3 ? (int)((size_t)(void*)dst[3] % (VECSZ*sizeof(T))) : r0;
hal::StoreMode mode = hal::STORE_ALIGNED_NOCACHE;
if( (r0|r1|r2|r3) != 0 )
{
mode = hal::STORE_UNALIGNED;
if( r0 == r1 && r0 == r2 && r0 == r3 && r0 % sizeof(T) == 0 && len > VECSZ*2 )
i0 = VECSZ - (r0 / sizeof(T));
}
if( cn == 2 )
{
for( i = 0; i < len; i += VECSZ )
{
if( i > len - VECSZ )
{
i = len - VECSZ;
mode = hal::STORE_UNALIGNED;
}
VecT a, b;
v_load_deinterleave(src + i*cn, a, b);
v_store(dst0 + i, a, mode);
v_store(dst1 + i, b, mode);
if( i < i0 )
{
i = i0 - VECSZ;
mode = hal::STORE_ALIGNED_NOCACHE;
}
}
}
else if( cn == 3 )
{
T* dst2 = dst[2];
for( i = 0; i < len; i += VECSZ )
{
if( i > len - VECSZ )
{
i = len - VECSZ;
mode = hal::STORE_UNALIGNED;
}
VecT a, b, c;
v_load_deinterleave(src + i*cn, a, b, c);
v_store(dst0 + i, a, mode);
v_store(dst1 + i, b, mode);
v_store(dst2 + i, c, mode);
if( i < i0 )
{
i = i0 - VECSZ;
mode = hal::STORE_ALIGNED_NOCACHE;
}
}
}
else
{
CV_Assert( cn == 4 );
T* dst2 = dst[2];
T* dst3 = dst[3];
for( i = 0; i < len; i += VECSZ )
{
if( i > len - VECSZ )
{
i = len - VECSZ;
mode = hal::STORE_UNALIGNED;
}
VecT a, b, c, d;
v_load_deinterleave(src + i*cn, a, b, c, d);
v_store(dst0 + i, a, mode);
v_store(dst1 + i, b, mode);
v_store(dst2 + i, c, mode);
v_store(dst3 + i, d, mode);
if( i < i0 )
{
i = i0 - VECSZ;
mode = hal::STORE_ALIGNED_NOCACHE;
}
}
}
vx_cleanup();
}
#endif
template<typename T> static void
split_( const T* src, T** dst, int len, int cn )
{
int k = cn % 4 ? cn % 4 : 4;
int i, j;
if( k == 1 )
{
T* dst0 = dst[0];
if(cn == 1)
{
memcpy(dst0, src, len * sizeof(T));
}
else
{
for( i = 0, j = 0 ; i < len; i++, j += cn )
dst0[i] = src[j];
}
}
else if( k == 2 )
{
T *dst0 = dst[0], *dst1 = dst[1];
i = j = 0;
for( ; i < len; i++, j += cn )
{
dst0[i] = src[j];
dst1[i] = src[j+1];
}
}
else if( k == 3 )
{
T *dst0 = dst[0], *dst1 = dst[1], *dst2 = dst[2];
i = j = 0;
for( ; i < len; i++, j += cn )
{
dst0[i] = src[j];
dst1[i] = src[j+1];
dst2[i] = src[j+2];
}
}
else
{
T *dst0 = dst[0], *dst1 = dst[1], *dst2 = dst[2], *dst3 = dst[3];
i = j = 0;
for( ; i < len; i++, j += cn )
{
dst0[i] = src[j]; dst1[i] = src[j+1];
dst2[i] = src[j+2]; dst3[i] = src[j+3];
}
}
for( ; k < cn; k += 4 )
{
T *dst0 = dst[k], *dst1 = dst[k+1], *dst2 = dst[k+2], *dst3 = dst[k+3];
for( i = 0, j = k; i < len; i++, j += cn )
{
dst0[i] = src[j]; dst1[i] = src[j+1];
dst2[i] = src[j+2]; dst3[i] = src[j+3];
}
}
}
void split8u(const uchar* src, uchar** dst, int len, int cn )
{
CV_INSTRUMENT_REGION();
#if CV_SIMD
if( len >= v_uint8::nlanes && 2 <= cn && cn <= 4 )
vecsplit_<uchar, v_uint8>(src, dst, len, cn);
else
#endif
split_(src, dst, len, cn);
}
void split16u(const ushort* src, ushort** dst, int len, int cn )
{
CV_INSTRUMENT_REGION();
#if CV_SIMD
if( len >= v_uint16::nlanes && 2 <= cn && cn <= 4 )
vecsplit_<ushort, v_uint16>(src, dst, len, cn);
else
#endif
split_(src, dst, len, cn);
}
void split32s(const int* src, int** dst, int len, int cn )
{
CV_INSTRUMENT_REGION();
#if CV_SIMD
if( len >= v_uint32::nlanes && 2 <= cn && cn <= 4 )
vecsplit_<int, v_int32>(src, dst, len, cn);
else
#endif
split_(src, dst, len, cn);
}
void split64s(const int64* src, int64** dst, int len, int cn )
{
CV_INSTRUMENT_REGION();
#if CV_SIMD
if( len >= v_int64::nlanes && 2 <= cn && cn <= 4 )
vecsplit_<int64, v_int64>(src, dst, len, cn);
else
#endif
split_(src, dst, len, cn);
}
#endif
CV_CPU_OPTIMIZATION_NAMESPACE_END
}} // namespace