libm: add fallbacks for various single-precision functions
[libav.git] / libavcodec / aacps_tablegen.h
1 /*
2 * Header file for hardcoded Parametric Stereo tables
3 *
4 * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
5 *
6 * This file is part of Libav.
7 *
8 * Libav is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * Libav is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with Libav; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 #ifndef AACPS_TABLEGEN_H
24 #define AACPS_TABLEGEN_H
25
26 #include <math.h>
27 #include <stdint.h>
28
29 #if CONFIG_HARDCODED_TABLES
30 #define ps_tableinit()
31 #include "libavcodec/aacps_tables.h"
32 #else
33 #include "libavutil/common.h"
34 #include "libavutil/libm.h"
35 #include "libavutil/mathematics.h"
36 #include "libavutil/mem.h"
37 #define NR_ALLPASS_BANDS20 30
38 #define NR_ALLPASS_BANDS34 50
39 #define PS_AP_LINKS 3
40 static float pd_re_smooth[8*8*8];
41 static float pd_im_smooth[8*8*8];
42 static float HA[46][8][4];
43 static float HB[46][8][4];
44 static DECLARE_ALIGNED(16, float, f20_0_8) [ 8][8][2];
45 static DECLARE_ALIGNED(16, float, f34_0_12)[12][8][2];
46 static DECLARE_ALIGNED(16, float, f34_1_8) [ 8][8][2];
47 static DECLARE_ALIGNED(16, float, f34_2_4) [ 4][8][2];
48 static DECLARE_ALIGNED(16, float, Q_fract_allpass)[2][50][3][2];
49 static DECLARE_ALIGNED(16, float, phi_fract)[2][50][2];
50
51 static const float g0_Q8[] = {
52 0.00746082949812f, 0.02270420949825f, 0.04546865930473f, 0.07266113929591f,
53 0.09885108575264f, 0.11793710567217f, 0.125f
54 };
55
56 static const float g0_Q12[] = {
57 0.04081179924692f, 0.03812810994926f, 0.05144908135699f, 0.06399831151592f,
58 0.07428313801106f, 0.08100347892914f, 0.08333333333333f
59 };
60
61 static const float g1_Q8[] = {
62 0.01565675600122f, 0.03752716391991f, 0.05417891378782f, 0.08417044116767f,
63 0.10307344158036f, 0.12222452249753f, 0.125f
64 };
65
66 static const float g2_Q4[] = {
67 -0.05908211155639f, -0.04871498374946f, 0.0f, 0.07778723915851f,
68 0.16486303567403f, 0.23279856662996f, 0.25f
69 };
70
71 static void make_filters_from_proto(float (*filter)[8][2], const float *proto, int bands)
72 {
73 int q, n;
74 for (q = 0; q < bands; q++) {
75 for (n = 0; n < 7; n++) {
76 double theta = 2 * M_PI * (q + 0.5) * (n - 6) / bands;
77 filter[q][n][0] = proto[n] * cos(theta);
78 filter[q][n][1] = proto[n] * -sin(theta);
79 }
80 }
81 }
82
83 static void ps_tableinit(void)
84 {
85 static const float ipdopd_sin[] = { 0, M_SQRT1_2, 1, M_SQRT1_2, 0, -M_SQRT1_2, -1, -M_SQRT1_2 };
86 static const float ipdopd_cos[] = { 1, M_SQRT1_2, 0, -M_SQRT1_2, -1, -M_SQRT1_2, 0, M_SQRT1_2 };
87 int pd0, pd1, pd2;
88
89 static const float iid_par_dequant[] = {
90 //iid_par_dequant_default
91 0.05623413251903, 0.12589254117942, 0.19952623149689, 0.31622776601684,
92 0.44668359215096, 0.63095734448019, 0.79432823472428, 1,
93 1.25892541179417, 1.58489319246111, 2.23872113856834, 3.16227766016838,
94 5.01187233627272, 7.94328234724282, 17.7827941003892,
95 //iid_par_dequant_fine
96 0.00316227766017, 0.00562341325190, 0.01, 0.01778279410039,
97 0.03162277660168, 0.05623413251903, 0.07943282347243, 0.11220184543020,
98 0.15848931924611, 0.22387211385683, 0.31622776601684, 0.39810717055350,
99 0.50118723362727, 0.63095734448019, 0.79432823472428, 1,
100 1.25892541179417, 1.58489319246111, 1.99526231496888, 2.51188643150958,
101 3.16227766016838, 4.46683592150963, 6.30957344480193, 8.91250938133745,
102 12.5892541179417, 17.7827941003892, 31.6227766016838, 56.2341325190349,
103 100, 177.827941003892, 316.227766016837,
104 };
105 static const float icc_invq[] = {
106 1, 0.937, 0.84118, 0.60092, 0.36764, 0, -0.589, -1
107 };
108 static const float acos_icc_invq[] = {
109 0, 0.35685527, 0.57133466, 0.92614472, 1.1943263, M_PI/2, 2.2006171, M_PI
110 };
111 int iid, icc;
112
113 int k, m;
114 static const int8_t f_center_20[] = {
115 -3, -1, 1, 3, 5, 7, 10, 14, 18, 22,
116 };
117 static const int8_t f_center_34[] = {
118 2, 6, 10, 14, 18, 22, 26, 30,
119 34,-10, -6, -2, 51, 57, 15, 21,
120 27, 33, 39, 45, 54, 66, 78, 42,
121 102, 66, 78, 90,102,114,126, 90,
122 };
123 static const float fractional_delay_links[] = { 0.43f, 0.75f, 0.347f };
124 const float fractional_delay_gain = 0.39f;
125
126 for (pd0 = 0; pd0 < 8; pd0++) {
127 float pd0_re = ipdopd_cos[pd0];
128 float pd0_im = ipdopd_sin[pd0];
129 for (pd1 = 0; pd1 < 8; pd1++) {
130 float pd1_re = ipdopd_cos[pd1];
131 float pd1_im = ipdopd_sin[pd1];
132 for (pd2 = 0; pd2 < 8; pd2++) {
133 float pd2_re = ipdopd_cos[pd2];
134 float pd2_im = ipdopd_sin[pd2];
135 float re_smooth = 0.25f * pd0_re + 0.5f * pd1_re + pd2_re;
136 float im_smooth = 0.25f * pd0_im + 0.5f * pd1_im + pd2_im;
137 float pd_mag = 1 / sqrt(im_smooth * im_smooth + re_smooth * re_smooth);
138 pd_re_smooth[pd0*64+pd1*8+pd2] = re_smooth * pd_mag;
139 pd_im_smooth[pd0*64+pd1*8+pd2] = im_smooth * pd_mag;
140 }
141 }
142 }
143
144 for (iid = 0; iid < 46; iid++) {
145 float c = iid_par_dequant[iid]; ///< Linear Inter-channel Intensity Difference
146 float c1 = (float)M_SQRT2 / sqrtf(1.0f + c*c);
147 float c2 = c * c1;
148 for (icc = 0; icc < 8; icc++) {
149 /*if (PS_BASELINE || ps->icc_mode < 3)*/ {
150 float alpha = 0.5f * acos_icc_invq[icc];
151 float beta = alpha * (c1 - c2) * (float)M_SQRT1_2;
152 HA[iid][icc][0] = c2 * cosf(beta + alpha);
153 HA[iid][icc][1] = c1 * cosf(beta - alpha);
154 HA[iid][icc][2] = c2 * sinf(beta + alpha);
155 HA[iid][icc][3] = c1 * sinf(beta - alpha);
156 } /* else */ {
157 float alpha, gamma, mu, rho;
158 float alpha_c, alpha_s, gamma_c, gamma_s;
159 rho = FFMAX(icc_invq[icc], 0.05f);
160 alpha = 0.5f * atan2f(2.0f * c * rho, c*c - 1.0f);
161 mu = c + 1.0f / c;
162 mu = sqrtf(1 + (4 * rho * rho - 4)/(mu * mu));
163 gamma = atanf(sqrtf((1.0f - mu)/(1.0f + mu)));
164 if (alpha < 0) alpha += M_PI/2;
165 alpha_c = cosf(alpha);
166 alpha_s = sinf(alpha);
167 gamma_c = cosf(gamma);
168 gamma_s = sinf(gamma);
169 HB[iid][icc][0] = M_SQRT2 * alpha_c * gamma_c;
170 HB[iid][icc][1] = M_SQRT2 * alpha_s * gamma_c;
171 HB[iid][icc][2] = -M_SQRT2 * alpha_s * gamma_s;
172 HB[iid][icc][3] = M_SQRT2 * alpha_c * gamma_s;
173 }
174 }
175 }
176
177 for (k = 0; k < NR_ALLPASS_BANDS20; k++) {
178 double f_center, theta;
179 if (k < FF_ARRAY_ELEMS(f_center_20))
180 f_center = f_center_20[k] * 0.125;
181 else
182 f_center = k - 6.5f;
183 for (m = 0; m < PS_AP_LINKS; m++) {
184 theta = -M_PI * fractional_delay_links[m] * f_center;
185 Q_fract_allpass[0][k][m][0] = cos(theta);
186 Q_fract_allpass[0][k][m][1] = sin(theta);
187 }
188 theta = -M_PI*fractional_delay_gain*f_center;
189 phi_fract[0][k][0] = cos(theta);
190 phi_fract[0][k][1] = sin(theta);
191 }
192 for (k = 0; k < NR_ALLPASS_BANDS34; k++) {
193 double f_center, theta;
194 if (k < FF_ARRAY_ELEMS(f_center_34))
195 f_center = f_center_34[k] / 24.;
196 else
197 f_center = k - 26.5f;
198 for (m = 0; m < PS_AP_LINKS; m++) {
199 theta = -M_PI * fractional_delay_links[m] * f_center;
200 Q_fract_allpass[1][k][m][0] = cos(theta);
201 Q_fract_allpass[1][k][m][1] = sin(theta);
202 }
203 theta = -M_PI*fractional_delay_gain*f_center;
204 phi_fract[1][k][0] = cos(theta);
205 phi_fract[1][k][1] = sin(theta);
206 }
207
208 make_filters_from_proto(f20_0_8, g0_Q8, 8);
209 make_filters_from_proto(f34_0_12, g0_Q12, 12);
210 make_filters_from_proto(f34_1_8, g1_Q8, 8);
211 make_filters_from_proto(f34_2_4, g2_Q4, 4);
212 }
213 #endif /* CONFIG_HARDCODED_TABLES */
214
215 #endif /* AACPS_TABLEGEN_H */