aboutsummaryrefslogtreecommitdiff
path: root/src/libmpg123/tabinit.c
blob: 8851672e1edd3b31d71772537c8ef6fa12eadc6e (plain)
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
/*
	tabinit.c: initialize tables...

	copyright ?-2008 by the mpg123 project - free software under the terms of the LGPL 2.1
	see COPYING and AUTHORS files in distribution or http://mpg123.org
	initially written by Michael Hipp
*/

#include "mpg123lib_intern.h"
#include "debug.h"

/* That altivec alignment part here should not hurt generic code, I hope */
#ifdef OPT_ALTIVEC
static ALIGNED(16) real cos64[16];
static ALIGNED(16) real cos32[8];
static ALIGNED(16) real cos16[4];
static ALIGNED(16) real cos8[2];
static ALIGNED(16) real cos4[1];
#else
static real cos64[16],cos32[8],cos16[4],cos8[2],cos4[1];
#endif

real *pnts[] = { cos64,cos32,cos16,cos8,cos4 };


static long intwinbase[] = {
     0,    -1,    -1,    -1,    -1,    -1,    -1,    -2,    -2,    -2,
    -2,    -3,    -3,    -4,    -4,    -5,    -5,    -6,    -7,    -7,
    -8,    -9,   -10,   -11,   -13,   -14,   -16,   -17,   -19,   -21,
   -24,   -26,   -29,   -31,   -35,   -38,   -41,   -45,   -49,   -53,
   -58,   -63,   -68,   -73,   -79,   -85,   -91,   -97,  -104,  -111,
  -117,  -125,  -132,  -139,  -147,  -154,  -161,  -169,  -176,  -183,
  -190,  -196,  -202,  -208,  -213,  -218,  -222,  -225,  -227,  -228,
  -228,  -227,  -224,  -221,  -215,  -208,  -200,  -189,  -177,  -163,
  -146,  -127,  -106,   -83,   -57,   -29,     2,    36,    72,   111,
   153,   197,   244,   294,   347,   401,   459,   519,   581,   645,
   711,   779,   848,   919,   991,  1064,  1137,  1210,  1283,  1356,
  1428,  1498,  1567,  1634,  1698,  1759,  1817,  1870,  1919,  1962,
  2001,  2032,  2057,  2075,  2085,  2087,  2080,  2063,  2037,  2000,
  1952,  1893,  1822,  1739,  1644,  1535,  1414,  1280,  1131,   970,
   794,   605,   402,   185,   -45,  -288,  -545,  -814, -1095, -1388,
 -1692, -2006, -2330, -2663, -3004, -3351, -3705, -4063, -4425, -4788,
 -5153, -5517, -5879, -6237, -6589, -6935, -7271, -7597, -7910, -8209,
 -8491, -8755, -8998, -9219, -9416, -9585, -9727, -9838, -9916, -9959,
 -9966, -9935, -9863, -9750, -9592, -9389, -9139, -8840, -8492, -8092,
 -7640, -7134, -6574, -5959, -5288, -4561, -3776, -2935, -2037, -1082,
   -70,   998,  2122,  3300,  4533,  5818,  7154,  8540,  9975, 11455,
 12980, 14548, 16155, 17799, 19478, 21189, 22929, 24694, 26482, 28289,
 30112, 31947, 33791, 35640, 37489, 39336, 41176, 43006, 44821, 46617,
 48390, 50137, 51853, 53534, 55178, 56778, 58333, 59838, 61289, 62684,
 64019, 65290, 66494, 67629, 68692, 69679, 70590, 71420, 72169, 72835,
 73415, 73908, 74313, 74630, 74856, 74992, 75038 };

void prepare_decode_tables()
{
  int i,k,kr,divv;
  real *costab;

  for(i=0;i<5;i++)
  {
    kr=0x10>>i; divv=0x40>>i;
    costab = pnts[i];
    for(k=0;k<kr;k++)
      costab[k] = DOUBLE_TO_REAL(1.0 / (2.0 * cos(M_PI * ((double) k * 2.0 + 1.0) / (double) divv)));
  }
}

#ifdef OPT_MMXORSSE
#ifndef OPT_X86_64
void make_decode_tables_mmx_asm(long scaleval, float* decwin_mmx, float *decwins);
void make_decode_tables_mmx(mpg123_handle *fr)
{
	debug("MMX decode tables");
	/* Take care: The scale should be like before, when we didn't have float output all around. */
	make_decode_tables_mmx_asm((long)((fr->lastscale < 0 ? fr->p.outscale : fr->lastscale)*SHORT_SCALE), fr->decwin_mmx, fr->decwins);
	debug("MMX decode tables done");
}
#else

/* This mimics round() as defined in C99. We stay C89. */
static int rounded(double f)
{
	return (int)(f>0 ? floor(f+0.5) : ceil(f-0.5));
}

/* x86-64 doesn't use asm version */
void make_decode_tables_mmx(mpg123_handle *fr)
{
	int i,j,val;
	int idx = 0;
	short *ptr = (short *)fr->decwins;
	/* Scale is always based on 1.0 . */
	double scaleval = -0.5*(fr->lastscale < 0 ? fr->p.outscale : fr->lastscale);
	debug1("MMX decode tables with scaleval %g", scaleval);
	for(i=0,j=0;i<256;i++,j++,idx+=32)
	{
		if(idx < 512+16)
		fr->decwin_mmx[idx+16] = fr->decwin_mmx[idx] = DOUBLE_TO_REAL((double) intwinbase[j] * scaleval);
		
		if(i % 32 == 31)
		idx -= 1023;
		if(i % 64 == 63)
		scaleval = - scaleval;
	}
	
	for( /* i=256 */ ;i<512;i++,j--,idx+=32)
	{
		if(idx < 512+16)
		fr->decwin_mmx[idx+16] = fr->decwin_mmx[idx] = DOUBLE_TO_REAL((double) intwinbase[j] * scaleval);
		
		if(i % 32 == 31)
		idx -= 1023;
		if(i % 64 == 63)
		scaleval = - scaleval;
	}
	
	for(i=0; i<512; i++) {
		if(i&1) val = rounded(fr->decwin_mmx[i]*0.5);
		else val = rounded(fr->decwin_mmx[i]*-0.5);
		if(val > 32767) val = 32767;
		else if(val < -32768) val = -32768;
		ptr[i] = val;
	}
	for(i=512; i<512+32; i++) {
		if(i&1) val = rounded(fr->decwin_mmx[i]*0.5);
		else val = 0;
		if(val > 32767) val = 32767;
		else if(val < -32768) val = -32768;
		ptr[i] = val;
	}
	for(i=0; i<512; i++) {
		val = rounded(fr->decwin_mmx[511-i]*-0.5);
		if(val > 32767) val = 32767;
		else if(val < -32768) val = -32768;
		ptr[512+32+i] = val;
	}
	debug("decode tables done");
}
#endif
#endif

void make_decode_tables(mpg123_handle *fr)
{
	int i,j;
	int idx = 0;
	/* Scale is always based on 1.0 . */
	double scaleval = -0.5*(fr->lastscale < 0 ? fr->p.outscale : fr->lastscale);
	debug1("decode tables with scaleval %g", scaleval);
	for(i=0,j=0;i<256;i++,j++,idx+=32)
	{
		if(idx < 512+16)
		fr->decwin[idx+16] = fr->decwin[idx] = DOUBLE_TO_REAL_15((double) intwinbase[j] * scaleval);

		if(i % 32 == 31)
		idx -= 1023;
		if(i % 64 == 63)
		scaleval = - scaleval;
	}

	for( /* i=256 */ ;i<512;i++,j--,idx+=32)
	{
		if(idx < 512+16)
		fr->decwin[idx+16] = fr->decwin[idx] = DOUBLE_TO_REAL_15((double) intwinbase[j] * scaleval);

		if(i % 32 == 31)
		idx -= 1023;
		if(i % 64 == 63)
		scaleval = - scaleval;
	}
#if defined(OPT_X86_64) || defined(OPT_ALTIVEC) || defined(OPT_SSE)
	if(fr->cpu_opts.type == x86_64 || fr->cpu_opts.type == altivec || fr->cpu_opts.type == sse)
	{ /* for float SSE / AltiVec decoder */
		for(i=512; i<512+32; i++)
		{
			fr->decwin[i] = (i&1) ? fr->decwin[i] : 0;
		}
		for(i=0; i<512; i++)
		{
			fr->decwin[512+32+i] = -fr->decwin[511-i];
		}
	}
#endif
	debug("decode tables done");
}

#ifndef NO_8BIT
int make_conv16to8_table(mpg123_handle *fr)
{
  int i;
	int mode = fr->af.encoding;

  /*
   * ????: 8.0 is right but on SB cards '2.0' is a better value ???
   */
  const double mul = 8.0;

  if(!fr->conv16to8_buf){
    fr->conv16to8_buf = (unsigned char *) malloc(8192);
    if(!fr->conv16to8_buf) {
      fr->err = MPG123_ERR_16TO8TABLE;
      if(NOQUIET) error("Can't allocate 16 to 8 converter table!");
      return -1;
    }
    fr->conv16to8 = fr->conv16to8_buf + 4096;
  }

  if(fr->af.encoding == MPG123_ENC_ULAW_8){
    double m=127.0 / log(256.0);
    int c1;

    for(i=-4096;i<4096;i++) {
/* dunno whether this is a valid transformation rule ?!?!? */
      if(i < 0)
        c1 = 127 - (int) (log( 1.0 - 255.0 * (double) i*mul / 32768.0 ) * m);
      else
        c1 = 255 - (int) (log( 1.0 + 255.0 * (double) i*mul / 32768.0 ) * m);
      if((c1 < 0 || c1 > 255) && NOQUIET) error2("Converror %d %d",i,c1);

      if(c1 == 0)
        c1 = 2;
      fr->conv16to8[i] = (unsigned char) c1;
    }
  }
  else if(mode == MPG123_ENC_SIGNED_8) {
    for(i=-4096;i<4096;i++) {
      fr->conv16to8[i] = i>>5;
    }
  }
  else if(mode == MPG123_ENC_UNSIGNED_8) {
    for(i=-4096;i<4096;i++) {
      fr->conv16to8[i] = (i>>5)+128;
    }
  }
  else {
    for(i=-4096;i<4096;i++) {
      fr->conv16to8[i] = 0;
    }
  }
	return 0;
}
#endif