Repository navigation
Expand file tree
/
Copy pathBlock_Copy.cpp
More file actions
212 lines (210 loc) · 4.97 KB
/
Copy pathBlock_Copy.cpp
File metadata and controls
212 lines (210 loc) · 4.97 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
#include "Block_Copy.hxx"
/*!
* Efficiently copy a block of dimension bx*by*bz from volume with dimension nx*ny*nz starting at location x0,y0,z0.
*
* Arguments:
*
* data Pointer to source volume
* x0 .
* y0 Start copy at this location
* z0 .
* nx .
* ny Source volume dimensions
* nz .
* work Pointer to destination block
* bx . Must be multiple of 4.
* by Destination block dimensions.
* bz .
*
*/
void Copy_To_Block(
float* data,
int x0,
int y0,
int z0,
int nx,
int ny,
int nz,
__m128* work,
int bx,
int by,
int bz
)
{
int _mm_bx = bx >> 2;
int nclipx = (x0+bx) - nx;
int clipped_mm_bx = nclipx > 0 ? (bx - nclipx) >> 2 : _mm_bx;
int clipped_bx = nclipx > 0 ? (bx - nclipx) : bx;
int y_stop = y0+by < ny ? y0+by : ny;
int iz;
for (iz = z0; iz < z0+bz && iz < nz; ++iz)
{
int iy = y0;
__m128* src = (__m128*)(data + (long)(iz*ny + iy)*nx + x0);
__m128* dst = work + (long)((iz-z0)*by + (iy-y0))*_mm_bx;
if (nclipx <= 0)
{
if (_mm_bx == 1)
{
for (; iy < y_stop; ++iy)
{
//printf("....copying iz=%d,iy=%d\n",iz,iy);
*(dst++) = _mm_loadu_ps((float*)(src));
src = (__m128*)(((float*)src) + nx);
}
}
else if (_mm_bx == 2)
{
for (; iy < y_stop; ++iy)
{
//printf("....copying iz=%d,iy=%d\n",iz,iy);
*(dst++) = _mm_loadu_ps((float*)(src));
*(dst++) = _mm_loadu_ps((float*)(src+1));
src = (__m128*)(((float*)src) + nx);
}
}
else if (_mm_bx == 4)
{
for (; iy < y_stop; ++iy)
{
//printf("....copying iz=%d,iy=%d\n",iz,iy);
*(dst++) = _mm_loadu_ps((float*)(src));
*(dst++) = _mm_loadu_ps((float*)(src+1));
*(dst++) = _mm_loadu_ps((float*)(src+2));
*(dst++) = _mm_loadu_ps((float*)(src+3));
src = (__m128*)(((float*)src) + nx);
}
}
else
{
for (; iy < y_stop; ++iy)
{
//printf("....copying iz=%d,iy=%d\n",iz,iy);
int ix;
for (ix = 0; ix < clipped_mm_bx; ++ix) *(dst++) = _mm_loadu_ps((float*)(src+ix));
src = (__m128*)(((float*)src) + nx);
}
}
}
else
{
for (; iy < y_stop; ++iy)
{
//printf("....copying iz=%d,iy=%d\n",iz,iy);
int ix;
for (ix = 0; ix < clipped_mm_bx; ++ix) dst[ix] = _mm_loadu_ps((float*)(src+ix));
for (ix=ix*4; ix < clipped_bx; ++ix) ((float*)dst)[ix] = ((float*)src)[ix];
for (; ix < bx; ++ix) ((float*)dst)[ix] = 0.0f;
src = (__m128*)(((float*)src) + nx);
dst += _mm_bx;
}
}
for (; iy < y0+by; ++iy)
{
//printf("....zeroing iz=%d,iy=%d\n",iz,iy);
for (int ix = 0; ix < _mm_bx; ++ix) dst[ix] = _mm_set1_ps(0.0f);
dst += _mm_bx;
}
}
for (; iz < z0+bz; ++iz)
{
//printf("...zeroing iz=%d\n",iz);
__m128* dst = work + by*_mm_bx*(long)(iz-z0);
for (int ixy = 0; ixy < by*_mm_bx; ++ixy) dst[ixy] = _mm_set1_ps(0.0f);
}
}
/*!
* Efficiently copy a block of dimensions bx*by*bz to destination volume starting at location x0,y0,z0.
*
* Arguments:
*
* work Pointer to destination block
* bx . Must be multiple of 4.
* by Destination block dimensions.
* bz .
* data Pointer to source volume
* x0 .
* y0 Start copy at this location
* z0 .
* nx .
* ny Source volume dimensions
* nz .
*
*/
void Copy_From_Block(
__m128* work,
int bx,
int by,
int bz,
float* data,
int x0,
int y0,
int z0,
int nx,
int ny,
int nz
)
{
int _mm_bx = bx >> 2;
int nclipx = (x0+bx) - nx;
int clipped_mm_bx = nclipx > 0 ? (bx - nclipx) >> 2 : bx >> 2;
int clipped_bx = nclipx > 0 ? (bx - nclipx) : bx;
int y_stop = y0+by < ny ? y0+by : ny;
for (int iz = z0; iz < z0+bz && iz < nz; ++iz)
{
int iy = y0;
__m128i* dst = (__m128i*)(data + (long)(iz*ny + iy)*nx + x0);
__m128i* src = (__m128i*)work + (long)((iz-z0)*by + (iy-y0))*_mm_bx;
if (nclipx <= 0)
{
if (_mm_bx == 1)
{
for (; iy < y_stop; ++iy)
{
_mm_storeu_si128(dst, _mm_load_si128(src++));
dst = (__m128i*)(((float*)dst) + nx);
}
}
else if (_mm_bx == 2)
{
for (; iy < y_stop; ++iy)
{
_mm_storeu_si128(dst, _mm_load_si128(src++));
_mm_storeu_si128(dst+1, _mm_load_si128(src++));
dst = (__m128i*)(((float*)dst) + nx);
}
}
else if (_mm_bx == 4)
{
for (; iy < y_stop; ++iy)
{
_mm_storeu_si128(dst, _mm_load_si128(src++));
_mm_storeu_si128(dst+1, _mm_load_si128(src++));
_mm_storeu_si128(dst+2, _mm_load_si128(src++));
_mm_storeu_si128(dst+3, _mm_load_si128(src++));
dst = (__m128i*)(((float*)dst) + nx);
}
}
else
{
for (; iy < y_stop; ++iy)
{
int ix;
for (ix = 0; ix < clipped_mm_bx; ++ix) _mm_storeu_si128(dst+ix, _mm_load_si128(src++));
dst = (__m128i*)(((float*)dst) + nx);
}
}
}
else
{
for (; iy < y_stop; ++iy)
{
int ix;
for (ix = 0; ix < clipped_mm_bx; ++ix) _mm_storeu_si128(dst+ix, _mm_load_si128(src+ix));
for (ix=ix*4; ix < clipped_bx; ++ix) ((float*)dst)[ix] = ((float*)src)[ix];
dst = (__m128i*)(((float*)dst) + nx);
src += _mm_bx;
}
}
}
}