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
|
// this code come from a52dec/libao/audio_out_oss.c
// FIXME FIXME FIXME
// a52_resample_init should find the requested converter (from type flags ->
// given number of channels) and set up some function pointers...
// a52_resample() should do the conversion.
#include <inttypes.h>
#include "a52.h"
#include "../config.h"
#ifdef HAVE_MMX
static uint64_t __attribute__((aligned(16))) magicF2W= 0x43c0000043c00000LL;
#endif
static inline int16_t convert (int32_t i)
{
if (i > 0x43c07fff)
return 32767;
else if (i < 0x43bf8000)
return -32768;
else
return i - 0x43c00000;
}
static int chans=2;
static int flags=0;
void a52_resample_init(int _flags,int _chans){
chans=_chans;
flags=_flags;
}
int a52_resample(float * _f, int16_t * s16)
{
int i;
int32_t * f = (int32_t *) _f;
switch (flags) {
case A52_MONO:
for (i = 0; i < 256; i++) {
s16[5*i] = s16[5*i+1] = s16[5*i+2] = s16[5*i+3] = 0;
s16[5*i+4] = convert (f[i]);
}
break;
case A52_CHANNEL:
case A52_STEREO:
case A52_DOLBY:
/* benchmark scores are 0.3% better with SSE but we would need to set bias=0 and premultiply it
#ifdef HAVE_SSE
asm volatile(
"movl $-1024, %%esi \n\t"
"1: \n\t"
"cvtps2pi (%1, %%esi), %%mm0 \n\t"
"cvtps2pi 1024(%1, %%esi), %%mm2\n\t"
"movq %%mm0, %%mm1 \n\t"
"punpcklwd %%mm2, %%mm0 \n\t"
"punpckhwd %%mm2, %%mm1 \n\t"
"movq %%mm0, (%0, %%esi) \n\t"
"movq %%mm1, 8(%0, %%esi) \n\t"
"addl $16, %%esi \n\t"
" jnz 1b \n\t"
"emms \n\t"
:: "r" (s16+512), "r" (f+256)
:"%esi", "memory"
);*/
#ifdef HAVE_MMX
asm volatile(
"movl $-1024, %%esi \n\t"
"movq magicF2W, %%mm7 \n\t"
"1: \n\t"
"movq (%1, %%esi), %%mm0 \n\t"
"movq 8(%1, %%esi), %%mm1 \n\t"
"movq 1024(%1, %%esi), %%mm2 \n\t"
"movq 1032(%1, %%esi), %%mm3 \n\t"
"psubd %%mm7, %%mm0 \n\t"
"psubd %%mm7, %%mm1 \n\t"
"psubd %%mm7, %%mm2 \n\t"
"psubd %%mm7, %%mm3 \n\t"
"packssdw %%mm1, %%mm0 \n\t"
"packssdw %%mm3, %%mm2 \n\t"
"movq %%mm0, %%mm1 \n\t"
"punpcklwd %%mm2, %%mm0 \n\t"
"punpckhwd %%mm2, %%mm1 \n\t"
"movq %%mm0, (%0, %%esi) \n\t"
"movq %%mm1, 8(%0, %%esi) \n\t"
"addl $16, %%esi \n\t"
" jnz 1b \n\t"
"emms \n\t"
:: "r" (s16+512), "r" (f+256)
:"%esi", "memory"
);
#else
for (i = 0; i < 256; i++) {
s16[2*i] = convert (f[i]);
s16[2*i+1] = convert (f[i+256]);
}
#endif
break;
case A52_3F:
for (i = 0; i < 256; i++) {
s16[5*i] = convert (f[i]);
s16[5*i+1] = convert (f[i+512]);
s16[5*i+2] = s16[5*i+3] = 0;
s16[5*i+4] = convert (f[i+256]);
}
break;
case A52_2F2R:
for (i = 0; i < 256; i++) {
s16[4*i] = convert (f[i]);
s16[4*i+1] = convert (f[i+256]);
s16[4*i+2] = convert (f[i+512]);
s16[4*i+3] = convert (f[i+768]);
}
break;
case A52_3F2R:
for (i = 0; i < 256; i++) {
s16[5*i] = convert (f[i]);
s16[5*i+1] = convert (f[i+512]);
s16[5*i+2] = convert (f[i+768]);
s16[5*i+3] = convert (f[i+1024]);
s16[5*i+4] = convert (f[i+256]);
}
break;
case A52_MONO | A52_LFE:
for (i = 0; i < 256; i++) {
s16[6*i] = s16[6*i+1] = s16[6*i+2] = s16[6*i+3] = 0;
s16[6*i+4] = convert (f[i+256]);
s16[6*i+5] = convert (f[i]);
}
break;
case A52_CHANNEL | A52_LFE:
case A52_STEREO | A52_LFE:
case A52_DOLBY | A52_LFE:
for (i = 0; i < 256; i++) {
s16[6*i] = convert (f[i+256]);
s16[6*i+1] = convert (f[i+512]);
s16[6*i+2] = s16[6*i+3] = s16[6*i+4] = 0;
s16[6*i+5] = convert (f[i]);
}
break;
case A52_3F | A52_LFE:
for (i = 0; i < 256; i++) {
s16[6*i] = convert (f[i+256]);
s16[6*i+1] = convert (f[i+768]);
s16[6*i+2] = s16[6*i+3] = 0;
s16[6*i+4] = convert (f[i+512]);
s16[6*i+5] = convert (f[i]);
}
break;
case A52_2F2R | A52_LFE:
for (i = 0; i < 256; i++) {
s16[6*i] = convert (f[i+256]);
s16[6*i+1] = convert (f[i+512]);
s16[6*i+2] = convert (f[i+768]);
s16[6*i+3] = convert (f[i+1024]);
s16[6*i+4] = 0;
s16[6*i+5] = convert (f[i]);
}
break;
case A52_3F2R | A52_LFE:
for (i = 0; i < 256; i++) {
s16[6*i] = convert (f[i+256]);
s16[6*i+1] = convert (f[i+768]);
s16[6*i+2] = convert (f[i+1024]);
s16[6*i+3] = convert (f[i+1280]);
s16[6*i+4] = convert (f[i+512]);
s16[6*i+5] = convert (f[i]);
}
break;
}
return chans*256;
}
|
