1/*
2 * (I)DCT Transforms
3 * Copyright (c) 2009 Peter Ross <pross@xvid.org>
4 * Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
5 * Copyright (c) 2010 Vitor Sessak
6 *
7 * This file is part of Libav.
8 *
9 * Libav is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
13 *
14 * Libav is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
17 * Lesser General Public License for more details.
18 *
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with Libav; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
22 */
23
24/**
25 * @file
26 * (Inverse) Discrete Cosine Transforms. These are also known as the
27 * type II and type III DCTs respectively.
28 */
29
30#include <math.h>
31
32#include "libavutil/mathematics.h"
33#include "dct.h"
34#include "dct32.h"
35
36/* sin((M_PI * x / (2 * n)) */
37#define SIN(s, n, x) (s->costab[(n) - (x)])
38
39/* cos((M_PI * x / (2 * n)) */
40#define COS(s, n, x) (s->costab[x])
41
42static void ff_dst_calc_I_c(DCTContext *ctx, FFTSample *data)
43{
44    int n = 1 << ctx->nbits;
45    int i;
46
47    data[0] = 0;
48    for (i = 1; i < n / 2; i++) {
49        float tmp1   = data[i    ];
50        float tmp2   = data[n - i];
51        float s      = SIN(ctx, n, 2 * i);
52
53        s           *= tmp1 + tmp2;
54        tmp1         = (tmp1 - tmp2) * 0.5f;
55        data[i]      = s + tmp1;
56        data[n - i]  = s - tmp1;
57    }
58
59    data[n / 2] *= 2;
60    ctx->rdft.rdft_calc(&ctx->rdft, data);
61
62    data[0] *= 0.5f;
63
64    for (i = 1; i < n - 2; i += 2) {
65        data[i + 1] +=  data[i - 1];
66        data[i]      = -data[i + 2];
67    }
68
69    data[n - 1] = 0;
70}
71
72static void ff_dct_calc_I_c(DCTContext *ctx, FFTSample *data)
73{
74    int n = 1 << ctx->nbits;
75    int i;
76    float next = -0.5f * (data[0] - data[n]);
77
78    for (i = 0; i < n / 2; i++) {
79        float tmp1 = data[i];
80        float tmp2 = data[n - i];
81        float s    = SIN(ctx, n, 2 * i);
82        float c    = COS(ctx, n, 2 * i);
83
84        c *= tmp1 - tmp2;
85        s *= tmp1 - tmp2;
86
87        next += c;
88
89        tmp1        = (tmp1 + tmp2) * 0.5f;
90        data[i]     = tmp1 - s;
91        data[n - i] = tmp1 + s;
92    }
93
94    ctx->rdft.rdft_calc(&ctx->rdft, data);
95    data[n] = data[1];
96    data[1] = next;
97
98    for (i = 3; i <= n; i += 2)
99        data[i] = data[i - 2] - data[i];
100}
101
102static void ff_dct_calc_III_c(DCTContext *ctx, FFTSample *data)
103{
104    int n = 1 << ctx->nbits;
105    int i;
106
107    float next  = data[n - 1];
108    float inv_n = 1.0f / n;
109
110    for (i = n - 2; i >= 2; i -= 2) {
111        float val1 = data[i];
112        float val2 = data[i - 1] - data[i + 1];
113        float c    = COS(ctx, n, i);
114        float s    = SIN(ctx, n, i);
115
116        data[i]     = c * val1 + s * val2;
117        data[i + 1] = s * val1 - c * val2;
118    }
119
120    data[1] = 2 * next;
121
122    ctx->rdft.rdft_calc(&ctx->rdft, data);
123
124    for (i = 0; i < n / 2; i++) {
125        float tmp1 = data[i]         * inv_n;
126        float tmp2 = data[n - i - 1] * inv_n;
127        float csc  = ctx->csc2[i] * (tmp1 - tmp2);
128
129        tmp1            += tmp2;
130        data[i]          = tmp1 + csc;
131        data[n - i - 1]  = tmp1 - csc;
132    }
133}
134
135static void ff_dct_calc_II_c(DCTContext *ctx, FFTSample *data)
136{
137    int n = 1 << ctx->nbits;
138    int i;
139    float next;
140
141    for (i = 0; i < n / 2; i++) {
142        float tmp1 = data[i];
143        float tmp2 = data[n - i - 1];
144        float s    = SIN(ctx, n, 2 * i + 1);
145
146        s    *= tmp1 - tmp2;
147        tmp1  = (tmp1 + tmp2) * 0.5f;
148
149        data[i]     = tmp1 + s;
150        data[n-i-1] = tmp1 - s;
151    }
152
153    ctx->rdft.rdft_calc(&ctx->rdft, data);
154
155    next     = data[1] * 0.5;
156    data[1] *= -1;
157
158    for (i = n - 2; i >= 0; i -= 2) {
159        float inr = data[i    ];
160        float ini = data[i + 1];
161        float c   = COS(ctx, n, i);
162        float s   = SIN(ctx, n, i);
163
164        data[i]     = c * inr + s * ini;
165        data[i + 1] = next;
166
167        next += s * inr - c * ini;
168    }
169}
170
171static void dct32_func(DCTContext *ctx, FFTSample *data)
172{
173    ctx->dct32(data, data);
174}
175
176av_cold int ff_dct_init(DCTContext *s, int nbits, enum DCTTransformType inverse)
177{
178    int n = 1 << nbits;
179    int i;
180
181    memset(s, 0, sizeof(*s));
182
183    s->nbits   = nbits;
184    s->inverse = inverse;
185
186    if (inverse == DCT_II && nbits == 5) {
187        s->dct_calc = dct32_func;
188    } else {
189        ff_init_ff_cos_tabs(nbits + 2);
190
191        s->costab = ff_cos_tabs[nbits + 2];
192        s->csc2   = av_malloc(n / 2 * sizeof(FFTSample));
193
194        if (ff_rdft_init(&s->rdft, nbits, inverse == DCT_III) < 0) {
195            av_free(s->csc2);
196            return -1;
197        }
198
199        for (i = 0; i < n / 2; i++)
200            s->csc2[i] = 0.5 / sin((M_PI / (2 * n) * (2 * i + 1)));
201
202        switch (inverse) {
203        case DCT_I  : s->dct_calc = ff_dct_calc_I_c;   break;
204        case DCT_II : s->dct_calc = ff_dct_calc_II_c;  break;
205        case DCT_III: s->dct_calc = ff_dct_calc_III_c; break;
206        case DST_I  : s->dct_calc = ff_dst_calc_I_c;   break;
207        }
208    }
209
210    s->dct32 = ff_dct32_float;
211    if (HAVE_MMX)
212        ff_dct_init_mmx(s);
213
214    return 0;
215}
216
217av_cold void ff_dct_end(DCTContext *s)
218{
219    ff_rdft_end(&s->rdft);
220    av_free(s->csc2);
221}
222