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
261
262
263
|
#ifndef MMX_X64_H_INCLUDED
#define MMX_X64_H_INCLUDED
/* Implementation of x64 MMX substitition functions, before
* pixman is reimplemented not to use __m64 type on Visual C++
*
* Copyright (C)2009 by George Yohng
* Released in public domain.
*/
#include <intrin.h>
#define M64C(a) (*(const __m64 *)(&a))
#define M64U(a) (*(const unsigned long long *)(&a))
__inline __m64
_m_from_int (int a)
{
long long i64 = a;
return M64C (i64);
}
__inline __m64
_mm_setzero_si64 ()
{
long long i64 = 0;
return M64C (i64);
}
__inline __m64
_mm_set_pi32 (int i1, int i0)
{
unsigned long long i64 = ((unsigned)i0) + (((unsigned long long)(unsigned)i1) << 32);
return M64C (i64);
}
__inline void
_m_empty ()
{
}
__inline __m64
_mm_set1_pi16 (short w)
{
unsigned long long i64 = ((unsigned long long)(unsigned short)(w)) * 0x0001000100010001ULL;
return M64C (i64);
}
__inline int
_m_to_int (__m64 m)
{
return m.m64_i32[0];
}
__inline __m64
_mm_movepi64_pi64 (__m128i a)
{
return M64C (a.m128i_i64[0]);
}
__inline __m64
_m_pand (__m64 a, __m64 b)
{
unsigned long long i64 = M64U (a) & M64U (b);
return M64C (i64);
}
__inline __m64
_m_por (__m64 a, __m64 b)
{
unsigned long long i64 = M64U (a) | M64U (b);
return M64C (i64);
}
__inline __m64
_m_pxor (__m64 a, __m64 b)
{
unsigned long long i64 = M64U (a) ^ M64U (b);
return M64C (i64);
}
__inline __m64
_m_pmulhuw (__m64 a, __m64 b) /* unoptimized */
{
unsigned short d[4] =
{
(unsigned short)((((unsigned)a.m64_u16[0]) * b.m64_u16[0]) >> 16),
(unsigned short)((((unsigned)a.m64_u16[1]) * b.m64_u16[1]) >> 16),
(unsigned short)((((unsigned)a.m64_u16[2]) * b.m64_u16[2]) >> 16),
(unsigned short)((((unsigned)a.m64_u16[3]) * b.m64_u16[3]) >> 16)
};
return M64C (d[0]);
}
__inline __m64
_m_pmullw2 (__m64 a, __m64 b) /* unoptimized */
{
unsigned short d[4] =
{
(unsigned short)((((unsigned)a.m64_u16[0]) * b.m64_u16[0])),
(unsigned short)((((unsigned)a.m64_u16[1]) * b.m64_u16[1])),
(unsigned short)((((unsigned)a.m64_u16[2]) * b.m64_u16[2])),
(unsigned short)((((unsigned)a.m64_u16[3]) * b.m64_u16[3]))
};
return M64C (d[0]);
}
__inline __m64
_m_pmullw (__m64 a, __m64 b) /* unoptimized */
{
unsigned long long x =
((unsigned long long)(unsigned short)((((unsigned)a.m64_u16[0]) * b.m64_u16[0]))) +
(((unsigned long long)(unsigned short)((((unsigned)a.m64_u16[1]) * b.m64_u16[1]))) << 16) +
(((unsigned long long)(unsigned short)((((unsigned)a.m64_u16[2]) * b.m64_u16[2]))) << 32) +
(((unsigned long long)(unsigned short)((((unsigned)a.m64_u16[3]) * b.m64_u16[3]))) << 48);
return M64C (x);
}
__inline __m64
_m_paddusb (__m64 a, __m64 b) /* unoptimized */
{
unsigned long long x = (M64U (a) & 0x00FF00FF00FF00FFULL) +
(M64U (b) & 0x00FF00FF00FF00FFULL);
unsigned long long y = ((M64U (a) >> 8) & 0x00FF00FF00FF00FFULL) +
((M64U (b) >> 8) & 0x00FF00FF00FF00FFULL);
x |= ((x & 0xFF00FF00FF00FF00ULL) >> 8) * 0xFF;
y |= ((y & 0xFF00FF00FF00FF00ULL) >> 8) * 0xFF;
x = (x & 0x00FF00FF00FF00FFULL) | ((y & 0x00FF00FF00FF00FFULL) << 8);
return M64C (x);
}
__inline __m64
_m_paddusw (__m64 a, __m64 b) /* unoptimized */
{
unsigned long long x = (M64U (a) & 0x0000FFFF0000FFFFULL) +
(M64U (b) & 0x0000FFFF0000FFFFULL);
unsigned long long y = ((M64U (a) >> 16) & 0x0000FFFF0000FFFFULL) +
((M64U (b) >> 16) & 0x0000FFFF0000FFFFULL);
x |= ((x & 0xFFFF0000FFFF0000) >> 16) * 0xFFFF;
y |= ((y & 0xFFFF0000FFFF0000) >> 16) * 0xFFFF;
x = (x & 0x0000FFFF0000FFFFULL) | ((y & 0x0000FFFF0000FFFFULL) << 16);
return M64C (x);
}
__inline __m64
_m_pshufw (__m64 a, int n) /* unoptimized */
{
unsigned short d[4] =
{
a.m64_u16[n & 3],
a.m64_u16[(n >> 2) & 3],
a.m64_u16[(n >> 4) & 3],
a.m64_u16[(n >> 6) & 3]
};
return M64C (d[0]);
}
__inline unsigned char
sat16 (unsigned short d)
{
if (d > 0xFF) return 0xFF;
else return d & 0xFF;
}
__inline __m64
_m_packuswb (__m64 m1, __m64 m2) /* unoptimized */
{
unsigned char d[8] =
{
sat16 (m1.m64_u16[0]),
sat16 (m1.m64_u16[1]),
sat16 (m1.m64_u16[2]),
sat16 (m1.m64_u16[3]),
sat16 (m2.m64_u16[0]),
sat16 (m2.m64_u16[1]),
sat16 (m2.m64_u16[2]),
sat16 (m2.m64_u16[3])
};
return M64C (d[0]);
}
__inline __m64 _m_punpcklbw (__m64 m1, __m64 m2) /* unoptimized */
{
unsigned char d[8] =
{
m1.m64_u8[0],
m2.m64_u8[0],
m1.m64_u8[1],
m2.m64_u8[1],
m1.m64_u8[2],
m2.m64_u8[2],
m1.m64_u8[3],
m2.m64_u8[3],
};
return M64C (d[0]);
}
__inline __m64 _m_punpckhbw (__m64 m1, __m64 m2) /* unoptimized */
{
unsigned char d[8] =
{
m1.m64_u8[4],
m2.m64_u8[4],
m1.m64_u8[5],
m2.m64_u8[5],
m1.m64_u8[6],
m2.m64_u8[6],
m1.m64_u8[7],
m2.m64_u8[7],
};
return M64C (d[0]);
}
__inline __m64 _m_psrlwi (__m64 a, int n) /* unoptimized */
{
unsigned short d[4] =
{
a.m64_u16[0] >> n,
a.m64_u16[1] >> n,
a.m64_u16[2] >> n,
a.m64_u16[3] >> n
};
return M64C (d[0]);
}
__inline __m64 _m_psrlqi (__m64 m, int n)
{
unsigned long long x = M64U (m) >> n;
return M64C (x);
}
__inline __m64 _m_psllqi (__m64 m, int n)
{
unsigned long long x = M64U (m) << n;
return M64C (x);
}
#endif /* MMX_X64_H_INCLUDED */
|
