1 /*
2 * MP3 window subband -> subband filtering -> mdct routine
3 *
4 * Copyright (c) 1999 Takehiro TOMINAGA
5 *
6 *
7 * This library is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Library General Public
9 * License as published by the Free Software Foundation; either
10 * version 2 of the License, or (at your option) any later version.
11 *
12 * This library is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Library General Public License for more details.
16 *
17 * You should have received a copy of the GNU Library General Public
18 * License along with this library; if not, write to the
19 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
20 * Boston, MA 02111-1307, USA.
21 */
22
23 /*
24 * Special Thanks to Patrick De Smet for your advices.
25 */
26
27 /* $Id: newmdct.c,v 1.25 2001/01/17 09:54:20 robert Exp $ */
28
29 #ifdef HAVE_CONFIG_H
30 # include <config.h>
31 #endif
32
33 #include "util.h"
34 #include "l3side.h"
35 #include "newmdct.h"
36
37 #ifdef WITH_DMALLOC
38 #include <dmalloc.h>
39 #endif
40
41 #define SCALE (32768.0/ 2.384e-06)
42
43 #ifndef USE_GOGO_SUBBAND
44 static const FLOAT8 enwindow[] =
45 {
46 -4.77e-07*0.740951125354959/2.384e-06, 1.03951e-04*0.740951125354959/2.384e-06, 9.53674e-04*0.740951125354959/2.384e-06, 2.841473e-03*0.740951125354959/2.384e-06,
47 3.5758972e-02*0.740951125354959/2.384e-06, 3.401756e-03*0.740951125354959/2.384e-06, 9.83715e-04*0.740951125354959/2.384e-06, 9.9182e-05*0.740951125354959/2.384e-06, /* 15*/
48 1.2398e-05*0.740951125354959/2.384e-06, 1.91212e-04*0.740951125354959/2.384e-06, 2.283096e-03*0.740951125354959/2.384e-06,1.6994476e-02*0.740951125354959/2.384e-06,
49 -1.8756866e-02*0.740951125354959/2.384e-06,-2.630711e-03*0.740951125354959/2.384e-06, -2.47478e-04*0.740951125354959/2.384e-06, -1.4782e-05*0.740951125354959/2.384e-06,
50 9.063471690191471e-01,
51 1.960342806591213e-01,
52
53
54 -4.77e-07*0.773010453362737/2.384e-06, 1.05858e-04*0.773010453362737/2.384e-06, 9.30786e-04*0.773010453362737/2.384e-06, 2.521515e-03*0.773010453362737/2.384e-06,
55 3.5694122e-02*0.773010453362737/2.384e-06, 3.643036e-03*0.773010453362737/2.384e-06, 9.91821e-04*0.773010453362737/2.384e-06, 9.6321e-05*0.773010453362737/2.384e-06, /* 14*/
56 1.1444e-05*0.773010453362737/2.384e-06, 1.65462e-04*0.773010453362737/2.384e-06, 2.110004e-03*0.773010453362737/2.384e-06,1.6112804e-02*0.773010453362737/2.384e-06,
57 -1.9634247e-02*0.773010453362737/2.384e-06,-2.803326e-03*0.773010453362737/2.384e-06, -2.77042e-04*0.773010453362737/2.384e-06, -1.6689e-05*0.773010453362737/2.384e-06,
58 8.206787908286602e-01,
59 3.901806440322567e-01,
60
61
62 -4.77e-07*0.803207531480645/2.384e-06, 1.07288e-04*0.803207531480645/2.384e-06, 9.02653e-04*0.803207531480645/2.384e-06, 2.174854e-03*0.803207531480645/2.384e-06,
63 3.5586357e-02*0.803207531480645/2.384e-06, 3.858566e-03*0.803207531480645/2.384e-06, 9.95159e-04*0.803207531480645/2.384e-06, 9.3460e-05*0.803207531480645/2.384e-06, /* 13*/
64 1.0014e-05*0.803207531480645/2.384e-06, 1.40190e-04*0.803207531480645/2.384e-06, 1.937389e-03*0.803207531480645/2.384e-06,1.5233517e-02*0.803207531480645/2.384e-06,
65 -2.0506859e-02*0.803207531480645/2.384e-06,-2.974033e-03*0.803207531480645/2.384e-06, -3.07560e-04*0.803207531480645/2.384e-06, -1.8120e-05*0.803207531480645/2.384e-06,
66 7.416505462720353e-01,
67 5.805693545089249e-01,
68
69
70 -4.77e-07*0.831469612302545/2.384e-06, 1.08242e-04*0.831469612302545/2.384e-06, 8.68797e-04*0.831469612302545/2.384e-06, 1.800537e-03*0.831469612302545/2.384e-06,
71 3.5435200e-02*0.831469612302545/2.384e-06, 4.049301e-03*0.831469612302545/2.384e-06, 9.94205e-04*0.831469612302545/2.384e-06, 9.0599e-05*0.831469612302545/2.384e-06, /* 12*/
72 9.060e-06*0.831469612302545/2.384e-06, 1.16348e-04*0.831469612302545/2.384e-06, 1.766682e-03*0.831469612302545/2.384e-06,1.4358521e-02*0.831469612302545/2.384e-06,
73 -2.1372318e-02*0.831469612302545/2.384e-06, -3.14188e-03*0.831469612302545/2.384e-06, -3.39031e-04*0.831469612302545/2.384e-06, -1.9550e-05*0.831469612302545/2.384e-06,
74 6.681786379192989e-01,
75 7.653668647301797e-01,
76
77
78 -4.77e-07*0.857728610000272/2.384e-06, 1.08719e-04*0.857728610000272/2.384e-06, 8.29220e-04*0.857728610000272/2.384e-06, 1.399517e-03*0.857728610000272/2.384e-06,
79 3.5242081e-02*0.857728610000272/2.384e-06, 4.215240e-03*0.857728610000272/2.384e-06, 9.89437e-04*0.857728610000272/2.384e-06, 8.7261e-05*0.857728610000272/2.384e-06, /* 11*/
80 8.106e-06*0.857728610000272/2.384e-06, 9.3937e-05*0.857728610000272/2.384e-06, 1.597881e-03*0.857728610000272/2.384e-06,1.3489246e-02*0.857728610000272/2.384e-06,
81 -2.2228718e-02*0.857728610000272/2.384e-06,-3.306866e-03*0.857728610000272/2.384e-06, -3.71456e-04*0.857728610000272/2.384e-06, -2.1458e-05*0.857728610000272/2.384e-06,
82 5.993769336819237e-01,
83 9.427934736519954e-01,
84
85
86 -4.77e-07*0.881921264348355/2.384e-06, 1.08719e-04*0.881921264348355/2.384e-06, 7.8392e-04*0.881921264348355/2.384e-06, 9.71317e-04*0.881921264348355/2.384e-06,
87 3.5007000e-02*0.881921264348355/2.384e-06, 4.357815e-03*0.881921264348355/2.384e-06, 9.80854e-04*0.881921264348355/2.384e-06, 8.3923e-05*0.881921264348355/2.384e-06, /* 10*/
88 7.629e-06*0.881921264348355/2.384e-06, 7.2956e-05*0.881921264348355/2.384e-06, 1.432419e-03*0.881921264348355/2.384e-06,1.2627602e-02*0.881921264348355/2.384e-06,
89 -2.3074150e-02*0.881921264348355/2.384e-06,-3.467083e-03*0.881921264348355/2.384e-06, -4.04358e-04*0.881921264348355/2.384e-06, -2.3365e-05*0.881921264348355/2.384e-06,
90 5.345111359507916e-01,
91 1.111140466039205e+00,
92
93
94 -9.54e-07*0.903989293123443/2.384e-06, 1.08242e-04*0.903989293123443/2.384e-06, 7.31945e-04*0.903989293123443/2.384e-06, 5.15938e-04*0.903989293123443/2.384e-06,
95 3.4730434e-02*0.903989293123443/2.384e-06, 4.477024e-03*0.903989293123443/2.384e-06, 9.68933e-04*0.903989293123443/2.384e-06, 8.0585e-05*0.903989293123443/2.384e-06, /* 9*/
96 6.676e-06*0.903989293123443/2.384e-06, 5.2929e-05*0.903989293123443/2.384e-06, 1.269817e-03*0.903989293123443/2.384e-06,1.1775017e-02*0.903989293123443/2.384e-06,
97 -2.3907185e-02*0.903989293123443/2.384e-06,-3.622532e-03*0.903989293123443/2.384e-06, -4.38213e-04*0.903989293123443/2.384e-06, -2.5272e-05*0.903989293123443/2.384e-06,
98 4.729647758913199e-01,
99 1.268786568327291e+00,
100
101
102 -9.54e-07*0.92387953251128675613/2.384e-06, 1.06812e-04*0.92387953251128675613/2.384e-06, 6.74248e-04*0.92387953251128675613/2.384e-06, 3.3379e-05*0.92387953251128675613/2.384e-06,
103 3.4412861e-02*0.92387953251128675613/2.384e-06, 4.573822e-03*0.92387953251128675613/2.384e-06, 9.54151e-04*0.92387953251128675613/2.384e-06, 7.6771e-05*0.92387953251128675613/2.384e-06,
104 6.199e-06*0.92387953251128675613/2.384e-06, 3.4332e-05*0.92387953251128675613/2.384e-06, 1.111031e-03*0.92387953251128675613/2.384e-06,1.0933399e-02*0.92387953251128675613/2.384e-06,
105 -2.4725437e-02*0.92387953251128675613/2.384e-06,-3.771782e-03*0.92387953251128675613/2.384e-06, -4.72546e-04*0.92387953251128675613/2.384e-06, -2.7657e-05*0.92387953251128675613/2.384e-06,
106 4.1421356237309504879e-01, /* tan(PI/8) */
107 1.414213562373095e+00,
108
109
110 -9.54e-07*0.941544065183021/2.384e-06, 1.05381e-04*0.941544065183021/2.384e-06, 6.10352e-04*0.941544065183021/2.384e-06, -4.75883e-04*0.941544065183021/2.384e-06,
111 3.4055710e-02*0.941544065183021/2.384e-06, 4.649162e-03*0.941544065183021/2.384e-06, 9.35555e-04*0.941544065183021/2.384e-06, 7.3433e-05*0.941544065183021/2.384e-06, /* 7*/
112 5.245e-06*0.941544065183021/2.384e-06, 1.7166e-05*0.941544065183021/2.384e-06, 9.56535e-04*0.941544065183021/2.384e-06,1.0103703e-02*0.941544065183021/2.384e-06,
113 -2.5527000e-02*0.941544065183021/2.384e-06,-3.914356e-03*0.941544065183021/2.384e-06, -5.07355e-04*0.941544065183021/2.384e-06, -3.0041e-05*0.941544065183021/2.384e-06,
114 3.578057213145241e-01,
115 1.546020906725474e+00,
116
117
118 -9.54e-07*0.956940335732209/2.384e-06, 1.02520e-04*0.956940335732209/2.384e-06, 5.39303e-04*0.956940335732209/2.384e-06,-1.011848e-03*0.956940335732209/2.384e-06,
119 3.3659935e-02*0.956940335732209/2.384e-06, 4.703045e-03*0.956940335732209/2.384e-06, 9.15051e-04*0.956940335732209/2.384e-06, 7.0095e-05*0.956940335732209/2.384e-06, /* 6*/
120 4.768e-06*0.956940335732209/2.384e-06, 9.54e-07*0.956940335732209/2.384e-06, 8.06808e-04*0.956940335732209/2.384e-06, 9.287834e-03*0.956940335732209/2.384e-06,
121 -2.6310921e-02*0.956940335732209/2.384e-06,-4.048824e-03*0.956940335732209/2.384e-06, -5.42164e-04*0.956940335732209/2.384e-06, -3.2425e-05*0.956940335732209/2.384e-06,
122 3.033466836073424e-01,
123 1.662939224605090e+00,
124
125
126 -1.431e-06*0.970031253194544/2.384e-06, 9.9182e-05*0.970031253194544/2.384e-06, 4.62532e-04*0.970031253194544/2.384e-06,-1.573563e-03*0.970031253194544/2.384e-06,
127 3.3225536e-02*0.970031253194544/2.384e-06, 4.737377e-03*0.970031253194544/2.384e-06, 8.91685e-04*0.970031253194544/2.384e-06, 6.6280e-05*0.970031253194544/2.384e-06, /* 5*/
128 4.292e-06*0.970031253194544/2.384e-06, -1.3828e-05*0.970031253194544/2.384e-06, 6.61850e-04*0.970031253194544/2.384e-06, 8.487225e-03*0.970031253194544/2.384e-06,
129 -2.7073860e-02*0.970031253194544/2.384e-06,-4.174709e-03*0.970031253194544/2.384e-06, -5.76973e-04*0.970031253194544/2.384e-06, -3.4809e-05*0.970031253194544/2.384e-06,
130 2.504869601913055e-01,
131 1.763842528696710e+00,
132
133
134 -1.431e-06*0.98078528040323/2.384e-06, 9.5367e-05*0.98078528040323/2.384e-06, 3.78609e-04*0.98078528040323/2.384e-06,-2.161503e-03*0.98078528040323/2.384e-06,
135 3.2754898e-02*0.98078528040323/2.384e-06, 4.752159e-03*0.98078528040323/2.384e-06, 8.66413e-04*0.98078528040323/2.384e-06, 6.2943e-05*0.98078528040323/2.384e-06, /* 4*/
136 3.815e-06*0.98078528040323/2.384e-06, -2.718e-05*0.98078528040323/2.384e-06, 5.22137e-04*0.98078528040323/2.384e-06, 7.703304e-03*0.98078528040323/2.384e-06,
137 -2.7815342e-02*0.98078528040323/2.384e-06,-4.290581e-03*0.98078528040323/2.384e-06, -6.11782e-04*0.98078528040323/2.384e-06, -3.7670e-05*0.98078528040323/2.384e-06,
138 1.989123673796580e-01,
139 1.847759065022573e+00,
140
141
142 -1.907e-06*0.989176509964781/2.384e-06, 9.0122e-05*0.989176509964781/2.384e-06, 2.88486e-04*0.989176509964781/2.384e-06,-2.774239e-03*0.989176509964781/2.384e-06,
143 3.2248020e-02*0.989176509964781/2.384e-06, 4.748821e-03*0.989176509964781/2.384e-06, 8.38757e-04*0.989176509964781/2.384e-06, 5.9605e-05*0.989176509964781/2.384e-06, /* 3*/
144 3.338e-06*0.989176509964781/2.384e-06, -3.9577e-05*0.989176509964781/2.384e-06, 3.88145e-04*0.989176509964781/2.384e-06, 6.937027e-03*0.989176509964781/2.384e-06,
145 -2.8532982e-02*0.989176509964781/2.384e-06,-4.395962e-03*0.989176509964781/2.384e-06, -6.46591e-04*0.989176509964781/2.384e-06, -4.0531e-05*0.989176509964781/2.384e-06,
146 1.483359875383474e-01,
147 1.913880671464418e+00,
148
149
150 -1.907e-06*0.995184726672197/2.384e-06, 8.4400e-05*0.995184726672197/2.384e-06, 1.91689e-04*0.995184726672197/2.384e-06,-3.411293e-03*0.995184726672197/2.384e-06,
151 3.1706810e-02*0.995184726672197/2.384e-06, 4.728317e-03*0.995184726672197/2.384e-06, 8.09669e-04*0.995184726672197/2.384e-06, 5.579e-05*0.995184726672197/2.384e-06,
152 3.338e-06*0.995184726672197/2.384e-06, -5.0545e-05*0.995184726672197/2.384e-06, 2.59876e-04*0.995184726672197/2.384e-06, 6.189346e-03*0.995184726672197/2.384e-06,
153 -2.9224873e-02*0.995184726672197/2.384e-06,-4.489899e-03*0.995184726672197/2.384e-06, -6.80923e-04*0.995184726672197/2.384e-06, -4.3392e-05*0.995184726672197/2.384e-06,
154 9.849140335716425e-02,
155 1.961570560806461e+00,
156
157
158 -2.384e-06*0.998795456205172/2.384e-06, 7.7724e-05*0.998795456205172/2.384e-06, 8.8215e-05*0.998795456205172/2.384e-06,-4.072189e-03*0.998795456205172/2.384e-06,
159 3.1132698e-02*0.998795456205172/2.384e-06, 4.691124e-03*0.998795456205172/2.384e-06, 7.79152e-04*0.998795456205172/2.384e-06, 5.2929e-05*0.998795456205172/2.384e-06,
160 2.861e-06*0.998795456205172/2.384e-06, -6.0558e-05*0.998795456205172/2.384e-06, 1.37329e-04*0.998795456205172/2.384e-06, 5.462170e-03*0.998795456205172/2.384e-06,
161 -2.9890060e-02*0.998795456205172/2.384e-06,-4.570484e-03*0.998795456205172/2.384e-06, -7.14302e-04*0.998795456205172/2.384e-06, -4.6253e-05*0.998795456205172/2.384e-06,
162 4.912684976946725e-02,
163 1.990369453344394e+00,
164
165
166 3.5780907e-02 * SQRT2*0.5/2.384e-06,1.7876148e-02 * SQRT2*0.5/2.384e-06, 3.134727e-03 * SQRT2*0.5/2.384e-06, 2.457142e-03 * SQRT2*0.5/2.384e-06,
167 9.71317e-04 * SQRT2*0.5/2.384e-06, 2.18868e-04 * SQRT2*0.5/2.384e-06, 1.01566e-04 * SQRT2*0.5/2.384e-06, 1.3828e-05 * SQRT2*0.5/2.384e-06,
168
169 3.0526638e-02/2.384e-06, 4.638195e-03/2.384e-06, 7.47204e-04/2.384e-06, 4.9591e-05/2.384e-06,
170 4.756451e-03/2.384e-06, 2.1458e-05/2.384e-06, -6.9618e-05/2.384e-06,/* 2.384e-06/2.384e-06*/
171 };
172 #endif
173
174
175 #define NS 12
176 #define NL 36
177
178 static const FLOAT8 win[4][NL] = {
179 {
180 2.382191739347913e-13,
181 6.423305872147834e-13,
182 9.400849094049688e-13,
183 1.122435026096556e-12,
184 1.183840321267481e-12,
185 1.122435026096556e-12,
186 9.400849094049690e-13,
187 6.423305872147839e-13,
188 2.382191739347918e-13,
189
190 5.456116108943412e-12,
191 4.878985199565852e-12,
192 4.240448995017367e-12,
193 3.559909094758252e-12,
194 2.858043359288075e-12,
195 2.156177623817898e-12,
196 1.475637723558783e-12,
197 8.371015190102974e-13,
198 2.599706096327376e-13,
199
200 -5.456116108943412e-12,
201 -4.878985199565852e-12,
202 -4.240448995017367e-12,
203 -3.559909094758252e-12,
204 -2.858043359288076e-12,
205 -2.156177623817898e-12,
206 -1.475637723558783e-12,
207 -8.371015190102975e-13,
208 -2.599706096327376e-13,
209
210 -2.382191739347923e-13,
211 -6.423305872147843e-13,
212 -9.400849094049696e-13,
213 -1.122435026096556e-12,
214 -1.183840321267481e-12,
215 -1.122435026096556e-12,
216 -9.400849094049694e-13,
217 -6.423305872147840e-13,
218 -2.382191739347918e-13,
219 },
220 {
221 2.382191739347913e-13,
222 6.423305872147834e-13,
223 9.400849094049688e-13,
224 1.122435026096556e-12,
225 1.183840321267481e-12,
226 1.122435026096556e-12,
227 9.400849094049688e-13,
228 6.423305872147841e-13,
229 2.382191739347918e-13,
230
231 5.456116108943413e-12,
232 4.878985199565852e-12,
233 4.240448995017367e-12,
234 3.559909094758253e-12,
235 2.858043359288075e-12,
236 2.156177623817898e-12,
237 1.475637723558782e-12,
238 8.371015190102975e-13,
239 2.599706096327376e-13,
240
241 -5.461314069809755e-12,
242 -4.921085770524055e-12,
243 -4.343405037091838e-12,
244 -3.732668368707687e-12,
245 -3.093523840190885e-12,
246 -2.430835727329465e-12,
247 -1.734679010007751e-12,
248 -9.748253656609281e-13,
249 -2.797435120168326e-13,
250
251 0.000000000000000e+00,
252 0.000000000000000e+00,
253 0.000000000000000e+00,
254 0.000000000000000e+00,
255 0.000000000000000e+00,
256 0.000000000000000e+00,
257 -2.283748241799531e-13,
258 -4.037858874020686e-13,
259 -2.146547464825323e-13,
260 },
261 {
262 1.316524975873958e-01, /* win[SHORT_TYPE] */
263 4.142135623730950e-01,
264 7.673269879789602e-01,
265
266 1.091308501069271e+00, /* tantab_l */
267 1.303225372841206e+00,
268 1.569685577117490e+00,
269 1.920982126971166e+00,
270 2.414213562373094e+00,
271 3.171594802363212e+00,
272 4.510708503662055e+00,
273 7.595754112725146e+00,
274 2.290376554843115e+01,
275
276 0.98480775301220802032, /* cx */
277 0.64278760968653936292,
278 0.34202014332566882393,
279 0.93969262078590842791,
280 -0.17364817766693030343,
281 -0.76604444311897790243,
282 0.86602540378443870761,
283 0.500000000000000e+00,
284
285 -5.144957554275265e-01, /* ca */
286 -4.717319685649723e-01,
287 -3.133774542039019e-01,
288 -1.819131996109812e-01,
289 -9.457419252642064e-02,
290 -4.096558288530405e-02,
291 -1.419856857247115e-02,
292 -3.699974673760037e-03,
293
294 8.574929257125442e-01, /* cs */
295 8.817419973177052e-01,
296 9.496286491027329e-01,
297 9.833145924917901e-01,
298 9.955178160675857e-01,
299 9.991605581781475e-01,
300 9.998991952444470e-01,
301 9.999931550702802e-01,
302 },
303 {
304 0.000000000000000e+00,
305 0.000000000000000e+00,
306 0.000000000000000e+00,
307 0.000000000000000e+00,
308 0.000000000000000e+00,
309 0.000000000000000e+00,
310 2.283748241799531e-13,
311 4.037858874020686e-13,
312 2.146547464825323e-13,
313
314 5.461314069809755e-12,
315 4.921085770524055e-12,
316 4.343405037091838e-12,
317 3.732668368707687e-12,
318 3.093523840190885e-12,
319 2.430835727329466e-12,
320 1.734679010007751e-12,
321 9.748253656609281e-13,
322 2.797435120168326e-13,
323
324 -5.456116108943413e-12,
325 -4.878985199565852e-12,
326 -4.240448995017367e-12,
327 -3.559909094758253e-12,
328 -2.858043359288075e-12,
329 -2.156177623817898e-12,
330 -1.475637723558782e-12,
331 -8.371015190102975e-13,
332 -2.599706096327376e-13,
333
334 -2.382191739347913e-13,
335 -6.423305872147834e-13,
336 -9.400849094049688e-13,
337 -1.122435026096556e-12,
338 -1.183840321267481e-12,
339 -1.122435026096556e-12,
340 -9.400849094049688e-13,
341 -6.423305872147841e-13,
342 -2.382191739347918e-13,
343 }
344 };
345
346 #define tantab_l (win[SHORT_TYPE]+3)
347 #define cx (win[SHORT_TYPE]+12)
348 #define ca (win[SHORT_TYPE]+20)
349 #define cs (win[SHORT_TYPE]+28)
350
351 /************************************************************************
352 *
353 * window_subband()
354 *
355 * PURPOSE: Overlapping window on PCM samples
356 *
357 * SEMANTICS:
358 * 32 16-bit pcm samples are scaled to fractional 2's complement and
359 * concatenated to the end of the window buffer #x#. The updated window
360 * buffer #x# is then windowed by the analysis window #c# to produce the
361 * windowed sample #z#
362 *
363 ************************************************************************/
364
365 /*
366 * new IDCT routine written by Takehiro TOMINAGA
367 */
368 static const int order[] = {
369 0, 1,16,17, 8, 9,24,25, 4, 5,20,21,12,13,28,29,
370 2, 3,18,19,10,11,26,27, 6, 7,22,23,14,15,30,31
371 };
372
373
374 /* returns sum_j=0^31 a[j]*cos(PI*j*(k+1/2)/32), 0<=k<32 */
375 INLINE static void
window_subband(const sample_t * x1,FLOAT8 a[SBLIMIT])376 window_subband(const sample_t *x1, FLOAT8 a[SBLIMIT])
377 {
378 int i;
379 FLOAT8 const *wp = enwindow+10;
380
381 const sample_t *x2 = &x1[238-14-286];
382
383 for (i = -15; i < 0; i++) {
384 FLOAT8 w, s, t;
385
386 w = wp[-10]; s = x2[-224] * w; t = x1[ 224] * w;
387 w = wp[-9]; s += x2[-160] * w; t += x1[ 160] * w;
388 w = wp[-8]; s += x2[- 96] * w; t += x1[ 96] * w;
389 w = wp[-7]; s += x2[- 32] * w; t += x1[ 32] * w;
390 w = wp[-6]; s += x2[ 32] * w; t += x1[- 32] * w;
391 w = wp[-5]; s += x2[ 96] * w; t += x1[- 96] * w;
392 w = wp[-4]; s += x2[ 160] * w; t += x1[-160] * w;
393 w = wp[-3]; s += x2[ 224] * w; t += x1[-224] * w;
394
395 w = wp[-2]; s += x1[-256] * w; t -= x2[ 256] * w;
396 w = wp[-1]; s += x1[-192] * w; t -= x2[ 192] * w;
397 w = wp[ 0]; s += x1[-128] * w; t -= x2[ 128] * w;
398 w = wp[ 1]; s += x1[- 64] * w; t -= x2[ 64] * w;
399 w = wp[ 2]; s += x1[ 0] * w; t -= x2[ 0] * w;
400 w = wp[ 3]; s += x1[ 64] * w; t -= x2[- 64] * w;
401 w = wp[ 4]; s += x1[ 128] * w; t -= x2[-128] * w;
402 w = wp[ 5]; s += x1[ 192] * w; t -= x2[-192] * w;
403
404 /*
405 * this multiplyer could be removed, but it needs more 256 FLOAT data.
406 * thinking about the data cache performance, I think we should not
407 * use such a huge table. tt 2000/Oct/25
408 */
409 s *= wp[6];
410 w = t - s;
411 a[30+i*2] = t + s;
412 a[31+i*2] = wp[7] * w;
413 wp += 18;
414 x1--;
415 x2++;
416 }
417 {
418 FLOAT8 s,t,u,v;
419 t = x1[- 16] * wp[-10]; s = x1[ -32] * wp[-2];
420 t += (x1[- 48] - x1[ 16]) * wp[-9]; s += x1[ -96] * wp[-1];
421 t += (x1[- 80] + x1[ 48]) * wp[-8]; s += x1[-160] * wp[ 0];
422 t += (x1[-112] - x1[ 80]) * wp[-7]; s += x1[-224] * wp[ 1];
423 t += (x1[-144] + x1[112]) * wp[-6]; s -= x1[ 32] * wp[ 2];
424 t += (x1[-176] - x1[144]) * wp[-5]; s -= x1[ 96] * wp[ 3];
425 t += (x1[-208] + x1[176]) * wp[-4]; s -= x1[ 160] * wp[ 4];
426 t += (x1[-240] - x1[208]) * wp[-3]; s -= x1[ 224];
427
428 u = s - t;
429 v = s + t;
430
431 t = a[14];
432 s = a[15] - t;
433
434 a[31] = v + t; // A0
435 a[30] = u + s; // A1
436 a[15] = u - s; // A2
437 a[14] = v - t; // A3
438 }
439 {
440 FLOAT8 xr;
441 xr = a[28] - a[ 0]; a[ 0] += a[28]; a[28] = xr * wp[-2*18+7];
442 xr = a[29] - a[ 1]; a[ 1] += a[29]; a[29] = xr * wp[-2*18+7];
443
444 xr = a[26] - a[ 2]; a[ 2] += a[26]; a[26] = xr * wp[-4*18+7];
445 xr = a[27] - a[ 3]; a[ 3] += a[27]; a[27] = xr * wp[-4*18+7];
446
447 xr = a[24] - a[ 4]; a[ 4] += a[24]; a[24] = xr * wp[-6*18+7];
448 xr = a[25] - a[ 5]; a[ 5] += a[25]; a[25] = xr * wp[-6*18+7];
449
450 xr = a[22] - a[ 6]; a[ 6] += a[22]; a[22] = xr * SQRT2;
451 xr = a[23] - a[ 7]; a[ 7] += a[23]; a[23] = xr * SQRT2 - a[ 7];
452 a[ 7] -= a[ 6];
453 a[22] -= a[ 7];
454 a[23] -= a[22];
455
456 xr = a[ 6]; a[ 6] = a[31] - xr; a[31] = a[31] + xr;
457 xr = a[ 7]; a[ 7] = a[30] - xr; a[30] = a[30] + xr;
458 xr = a[22]; a[22] = a[15] - xr; a[15] = a[15] + xr;
459 xr = a[23]; a[23] = a[14] - xr; a[14] = a[14] + xr;
460
461 xr = a[20] - a[ 8]; a[ 8] += a[20]; a[20] = xr * wp[-10*18+7];
462 xr = a[21] - a[ 9]; a[ 9] += a[21]; a[21] = xr * wp[-10*18+7];
463
464 xr = a[18] - a[10]; a[10] += a[18]; a[18] = xr * wp[-12*18+7];
465 xr = a[19] - a[11]; a[11] += a[19]; a[19] = xr * wp[-12*18+7];
466
467 xr = a[16] - a[12]; a[12] += a[16]; a[16] = xr * wp[-14*18+7];
468 xr = a[17] - a[13]; a[13] += a[17]; a[17] = xr * wp[-14*18+7];
469
470 xr = -a[20] + a[24]; a[20] += a[24]; a[24] = xr * wp[-12*18+7];
471 xr = -a[21] + a[25]; a[21] += a[25]; a[25] = xr * wp[-12*18+7];
472
473 xr = a[ 4] - a[ 8]; a[ 4] += a[ 8]; a[ 8] = xr * wp[-12*18+7];
474 xr = a[ 5] - a[ 9]; a[ 5] += a[ 9]; a[ 9] = xr * wp[-12*18+7];
475
476 xr = a[ 0] - a[12]; a[ 0] += a[12]; a[12] = xr * wp[-4*18+7];
477 xr = a[ 1] - a[13]; a[ 1] += a[13]; a[13] = xr * wp[-4*18+7];
478 xr = a[16] - a[28]; a[16] += a[28]; a[28] = xr * wp[-4*18+7];
479 xr = -a[17] + a[29]; a[17] += a[29]; a[29] = xr * wp[-4*18+7];
480
481 xr = SQRT2 * (a[ 2] - a[10]); a[ 2] += a[10]; a[10] = xr;
482 xr = SQRT2 * (a[ 3] - a[11]); a[ 3] += a[11]; a[11] = xr;
483 xr = SQRT2 * (-a[18] + a[26]); a[18] += a[26]; a[26] = xr - a[18];
484 xr = SQRT2 * (-a[19] + a[27]); a[19] += a[27]; a[27] = xr - a[19];
485
486 xr = a[ 2]; a[19] -= a[ 3]; a[ 3] -= xr; a[ 2] = a[31] - xr; a[31] += xr;
487 xr = a[ 3]; a[11] -= a[19]; a[18] -= xr; a[ 3] = a[30] - xr; a[30] += xr;
488 xr = a[18]; a[27] -= a[11]; a[19] -= xr; a[18] = a[15] - xr; a[15] += xr;
489
490 xr = a[19]; a[10] -= xr; a[19] = a[14] - xr; a[14] += xr;
491 xr = a[10]; a[11] -= xr; a[10] = a[23] - xr; a[23] += xr;
492 xr = a[11]; a[26] -= xr; a[11] = a[22] - xr; a[22] += xr;
493 xr = a[26]; a[27] -= xr; a[26] = a[ 7] - xr; a[ 7] += xr;
494
495 xr = a[27]; a[27] = a[ 6] - xr; a[ 6] += xr;
496
497 xr = SQRT2 * (a[ 0] - a[ 4]); a[ 0] += a[ 4]; a[ 4] = xr;
498 xr = SQRT2 * (a[ 1] - a[ 5]); a[ 1] += a[ 5]; a[ 5] = xr;
499 xr = SQRT2 * (a[16] - a[20]); a[16] += a[20]; a[20] = xr;
500 xr = SQRT2 * (a[17] - a[21]); a[17] += a[21]; a[21] = xr;
501
502 xr = -SQRT2 * (a[ 8] - a[12]); a[ 8] += a[12]; a[12] = xr - a[ 8];
503 xr = -SQRT2 * (a[ 9] - a[13]); a[ 9] += a[13]; a[13] = xr - a[ 9];
504 xr = -SQRT2 * (a[25] - a[29]); a[25] += a[29]; a[29] = xr - a[25];
505 xr = -SQRT2 * (a[24] + a[28]); a[24] -= a[28]; a[28] = xr - a[24];
506
507 xr = a[24] - a[16]; a[24] = xr;
508 xr = a[20] - xr; a[20] = xr;
509 xr = a[28] - xr; a[28] = xr;
510
511 xr = a[25] - a[17]; a[25] = xr;
512 xr = a[21] - xr; a[21] = xr;
513 xr = a[29] - xr; a[29] = xr;
514
515 xr = a[17] - a[ 1]; a[17] = xr;
516 xr = a[ 9] - xr; a[ 9] = xr;
517 xr = a[25] - xr; a[25] = xr;
518 xr = a[ 5] - xr; a[ 5] = xr;
519 xr = a[21] - xr; a[21] = xr;
520 xr = a[13] - xr; a[13] = xr;
521 xr = a[29] - xr; a[29] = xr;
522
523 xr = a[ 1] - a[ 0]; a[ 1] = xr;
524 xr = a[16] - xr; a[16] = xr;
525 xr = a[17] - xr; a[17] = xr;
526 xr = a[ 8] - xr; a[ 8] = xr;
527 xr = a[ 9] - xr; a[ 9] = xr;
528 xr = a[24] - xr; a[24] = xr;
529 xr = a[25] - xr; a[25] = xr;
530 xr = a[ 4] - xr; a[ 4] = xr;
531 xr = a[ 5] - xr; a[ 5] = xr;
532 xr = a[20] - xr; a[20] = xr;
533 xr = a[21] - xr; a[21] = xr;
534 xr = a[12] - xr; a[12] = xr;
535 xr = a[13] - xr; a[13] = xr;
536 xr = a[28] - xr; a[28] = xr;
537 xr = a[29] - xr; a[29] = xr;
538
539 xr = a[ 0]; a[ 0] += a[31]; a[31] -= xr;
540 xr = a[ 1]; a[ 1] += a[30]; a[30] -= xr;
541 xr = a[16]; a[16] += a[15]; a[15] -= xr;
542 xr = a[17]; a[17] += a[14]; a[14] -= xr;
543 xr = a[ 8]; a[ 8] += a[23]; a[23] -= xr;
544 xr = a[ 9]; a[ 9] += a[22]; a[22] -= xr;
545 xr = a[24]; a[24] += a[ 7]; a[ 7] -= xr;
546 xr = a[25]; a[25] += a[ 6]; a[ 6] -= xr;
547 xr = a[ 4]; a[ 4] += a[27]; a[27] -= xr;
548 xr = a[ 5]; a[ 5] += a[26]; a[26] -= xr;
549 xr = a[20]; a[20] += a[11]; a[11] -= xr;
550 xr = a[21]; a[21] += a[10]; a[10] -= xr;
551 xr = a[12]; a[12] += a[19]; a[19] -= xr;
552 xr = a[13]; a[13] += a[18]; a[18] -= xr;
553 xr = a[28]; a[28] += a[ 3]; a[ 3] -= xr;
554 xr = a[29]; a[29] += a[ 2]; a[ 2] -= xr;
555 }
556
557 }
558
559
560 /*-------------------------------------------------------------------*/
561 /* */
562 /* Function: Calculation of the MDCT */
563 /* In the case of long blocks (type 0,1,3) there are */
564 /* 36 coefficents in the time domain and 18 in the frequency */
565 /* domain. */
566 /* In the case of short blocks (type 2) there are 3 */
567 /* transformations with short length. This leads to 12 coefficents */
568 /* in the time and 6 in the frequency domain. In this case the */
569 /* results are stored side by side in the vector out[]. */
570 /* */
571 /* New layer3 */
572 /* */
573 /*-------------------------------------------------------------------*/
574
mdct_short(FLOAT8 * inout)575 inline static void mdct_short(FLOAT8 *inout)
576 {
577 int l;
578 for ( l = 0; l < 3; l++ ) {
579 FLOAT8 tc0,tc1,tc2,ts0,ts1,ts2;
580
581 ts0 = inout[2*3] * win[SHORT_TYPE][0] - inout[5*3];
582 tc0 = inout[0*3] * win[SHORT_TYPE][2] - inout[3*3];
583 tc1 = ts0 + tc0;
584 tc2 = ts0 - tc0;
585
586 ts0 = inout[5*3] * win[SHORT_TYPE][0] + inout[2*3];
587 tc0 = inout[3*3] * win[SHORT_TYPE][2] + inout[0*3];
588 ts1 = ts0 + tc0;
589 ts2 = -ts0 + tc0;
590
591 tc0 = (inout[1*3] * win[SHORT_TYPE][1] - inout[4*3]) * 2.069978111953089e-11; /* tritab_s[1] */
592 ts0 = (inout[4*3] * win[SHORT_TYPE][1] + inout[1*3]) * 2.069978111953089e-11; /* tritab_s[1] */
593
594 inout[3*0] = tc1 * 1.907525191737280e-11 /* tritab_s[2] */ + tc0;
595 inout[3*5] = -ts1 * 1.907525191737280e-11 /* tritab_s[0] */ + ts0;
596
597 tc2 = tc2 * 0.86602540378443870761 * 1.907525191737281e-11 /* tritab_s[2] */;
598 ts1 = ts1 * 0.5 * 1.907525191737281e-11 + ts0;
599 inout[3*1] = tc2-ts1;
600 inout[3*2] = tc2+ts1;
601
602 tc1 = tc1 * 0.5 * 1.907525191737281e-11 - tc0;
603 ts2 = ts2 * 0.86602540378443870761 * 1.907525191737281e-11 /* tritab_s[0] */;
604 inout[3*3] = tc1+ts2;
605 inout[3*4] = tc1-ts2;
606
607 inout++;
608 }
609 }
610
mdct_long(FLOAT8 * out,FLOAT8 * in)611 inline static void mdct_long(FLOAT8 *out, FLOAT8 *in)
612 {
613 FLOAT8 ct,st;
614 {
615 FLOAT8 tc1, tc2, tc3, tc4, ts5, ts6, ts7, ts8;
616 // 1,2, 5,6, 9,10, 13,14, 17
617 tc1 = in[17]-in[ 9];
618 tc3 = in[15]-in[11];
619 tc4 = in[14]-in[12];
620 ts5 = in[ 0]+in[ 8];
621 ts6 = in[ 1]+in[ 7];
622 ts7 = in[ 2]+in[ 6];
623 ts8 = in[ 3]+in[ 5];
624
625 out[17] = (ts5+ts7-ts8)-(ts6-in[4]);
626 st = (ts5+ts7-ts8)*cx[7]+(ts6-in[4]);
627 ct = (tc1-tc3-tc4)*cx[6];
628 out[5] = ct+st;
629 out[6] = ct-st;
630
631 tc2 = (in[16]-in[10])*cx[6];
632 ts6 = ts6*cx[7] + in[4];
633 ct = tc1*cx[0] + tc2 + tc3*cx[1] + tc4*cx[2];
634 st = -ts5*cx[4] + ts6 - ts7*cx[5] + ts8*cx[3];
635 out[1] = ct+st;
636 out[2] = ct-st;
637
638 ct = tc1*cx[1] - tc2 - tc3*cx[2] + tc4*cx[0];
639 st = -ts5*cx[5] + ts6 - ts7*cx[3] + ts8*cx[4];
640 out[ 9] = ct+st;
641 out[10] = ct-st;
642
643 ct = tc1*cx[2] - tc2 + tc3*cx[0] - tc4*cx[1];
644 st = ts5*cx[3] - ts6 + ts7*cx[4] - ts8*cx[5];
645 out[13] = ct+st;
646 out[14] = ct-st;
647 }
648 {
649 FLOAT8 ts1, ts2, ts3, ts4, tc5, tc6, tc7, tc8;
650
651 ts1 = in[ 8]-in[ 0];
652 ts3 = in[ 6]-in[ 2];
653 ts4 = in[ 5]-in[ 3];
654 tc5 = in[17]+in[ 9];
655 tc6 = in[16]+in[10];
656 tc7 = in[15]+in[11];
657 tc8 = in[14]+in[12];
658
659 out[0] = (tc5+tc7+tc8)+(tc6+in[13]);
660 ct = (tc5+tc7+tc8)*cx[7]-(tc6+in[13]);
661 st = (ts1-ts3+ts4)*cx[6];
662 out[11] = ct+st;
663 out[12] = ct-st;
664
665 ts2 = (in[7]-in[1])*cx[6];
666 tc6 = in[13] - tc6*cx[7];
667 ct = tc5*cx[3] - tc6 + tc7*cx[4] + tc8*cx[5];
668 st = ts1*cx[2] + ts2 + ts3*cx[0] + ts4*cx[1];
669 out[3] = ct+st;
670 out[4] = ct-st;
671
672 ct = -tc5*cx[5] + tc6 - tc7*cx[3] - tc8*cx[4];
673 st = ts1*cx[1] + ts2 - ts3*cx[2] - ts4*cx[0];
674 out[7] = ct+st;
675 out[8] = ct-st;
676
677 ct = -tc5*cx[4] + tc6 - tc7*cx[5] - tc8*cx[3];
678 st = ts1*cx[0] - ts2 + ts3*cx[1] - ts4*cx[2];
679 out[15] = ct+st;
680 out[16] = ct-st;
681 }
682 }
683
684
mdct_sub48(lame_internal_flags * gfc,const sample_t * w0,const sample_t * w1,FLOAT8 mdct_freq[2][2][576])685 void mdct_sub48( lame_internal_flags *gfc, const sample_t *w0, const sample_t *w1,
686 FLOAT8 mdct_freq[2][2][576] )
687 {
688 int gr, k, ch;
689 const sample_t *wk;
690
691 wk = w0 + 286;
692 /* thinking cache performance, ch->gr loop is better than gr->ch loop */
693 for (ch = 0; ch < gfc->channels_out; ch++) {
694 for (gr = 0; gr < gfc->mode_gr; gr++) {
695 int band;
696 FLOAT8 *mdct_enc = &mdct_freq[gr][ch][0];
697 gr_info *gi = &(gfc->l3_side.gr[gr].ch[ch].tt);
698 FLOAT8 *samp = gfc->sb_sample[ch][1 - gr][0];
699
700 for (k = 0; k < 18 / 2; k++) {
701 window_subband(wk, samp);
702 window_subband(wk + 32, samp + 32);
703 samp += 64;
704 wk += 64;
705 /*
706 * Compensate for inversion in the analysis filter
707 */
708 for (band = 1; band < 32; band+=2) {
709 samp[band-32] *= -1;
710 }
711 }
712
713
714 /* apply filters on the polyphase filterbank outputs */
715 /* bands <= gfc->highpass_band will be zeroed out below */
716 /* bands >= gfc->lowpass_band will be zeroed out below */
717 if (gfc->filter_type==0) {
718 for (band=gfc->highpass_start_band; band <= gfc->highpass_end_band; band++) {
719 for (k=0; k<18; k++)
720 gfc->sb_sample[ch][1-gr][k][order[band]]*=gfc->amp_highpass[band];
721 }
722 for (band=gfc->lowpass_start_band; band <= gfc->lowpass_end_band; band++) {
723 for (k=0; k<18; k++)
724 gfc->sb_sample[ch][1-gr][k][order[band]]*=gfc->amp_lowpass[band];
725 }
726 }
727
728
729
730 /*
731 * Perform imdct of 18 previous subband samples
732 * + 18 current subband samples
733 */
734 for (band = 0; band < 32; band++, mdct_enc += 18) {
735 int type = gi->block_type;
736 FLOAT8 *band0, *band1;
737 band0 = gfc->sb_sample[ch][ gr][0] + order[band];
738 band1 = gfc->sb_sample[ch][1-gr][0] + order[band];
739 if (gi->mixed_block_flag && band < 2)
740 type = 0;
741 if (band >= gfc->lowpass_band || band <= gfc->highpass_band) {
742 memset((char *)mdct_enc,0,18*sizeof(FLOAT8));
743 } else {
744 if (type == SHORT_TYPE) {
745 for (k = -NS/4; k < 0; k++) {
746 FLOAT8 w = win[SHORT_TYPE][k+3];
747 mdct_enc[k*3+ 9] = band0[( 9+k)*32] * w - band0[( 8-k)*32];
748 mdct_enc[k*3+18] = band0[(14-k)*32] * w + band0[(15+k)*32];
749 mdct_enc[k*3+10] = band0[(15+k)*32] * w - band0[(14-k)*32];
750 mdct_enc[k*3+19] = band1[( 2-k)*32] * w + band1[( 3+k)*32];
751 mdct_enc[k*3+11] = band1[( 3+k)*32] * w - band1[( 2-k)*32];
752 mdct_enc[k*3+20] = band1[( 8-k)*32] * w + band1[( 9+k)*32];
753 }
754 mdct_short(mdct_enc);
755 } else {
756 FLOAT8 work[18];
757 for (k = -NL/4; k < 0; k++) {
758 FLOAT8 a, b;
759 a = win[type][k+27] * band1[(k+9)*32]
760 + win[type][k+36] * band1[(8-k)*32];
761 b = win[type][k+ 9] * band0[(k+9)*32]
762 - win[type][k+18] * band0[(8-k)*32];
763 work[k+ 9] = a - b*tantab_l[k+9];
764 work[k+18] = a*tantab_l[k+9] + b;
765 }
766
767 mdct_long(mdct_enc, work);
768 }
769 }
770 /*
771 * Perform aliasing reduction butterfly
772 */
773 if (type != SHORT_TYPE) {
774 if (band == 0)
775 continue;
776 for (k = 7; k >= 0; --k) {
777 FLOAT8 bu,bd;
778 bu = mdct_enc[k] * ca[k] + mdct_enc[-1-k] * cs[k];
779 bd = mdct_enc[k] * cs[k] - mdct_enc[-1-k] * ca[k];
780
781 mdct_enc[-1-k] = bu;
782 mdct_enc[k] = bd;
783 }
784 }
785 }
786 }
787 wk = w1 + 286;
788 if (gfc->mode_gr == 1) {
789 memcpy(gfc->sb_sample[ch][0], gfc->sb_sample[ch][1], 576 * sizeof(FLOAT8));
790 }
791 }
792 }
793