1/*
2 * Simple IDCT
3 *
4 * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at>
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg 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 * FFmpeg 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 FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23/**
24 * @file
25 * simpleidct in C.
26 */
27
28#include "libavutil/intreadwrite.h"
29#include "avcodec.h"
30#include "mathops.h"
31#include "simple_idct.h"
32
33#define BIT_DEPTH 8
34#include "simple_idct_template.c"
35#undef BIT_DEPTH
36
37#define BIT_DEPTH 10
38#include "simple_idct_template.c"
39#undef BIT_DEPTH
40
41#define BIT_DEPTH 12
42#include "simple_idct_template.c"
43#undef BIT_DEPTH
44
45/* 2x4x8 idct */
46
47#define CN_SHIFT 12
48#define C_FIX(x) ((int)((x) * (1 << CN_SHIFT) + 0.5))
49#define C1 C_FIX(0.6532814824)
50#define C2 C_FIX(0.2705980501)
51
52/* row idct is multiple by 16 * sqrt(2.0), col idct4 is normalized,
53   and the butterfly must be multiplied by 0.5 * sqrt(2.0) */
54#define C_SHIFT (4+1+12)
55
56static inline void idct4col_put(uint8_t *dest, int line_size, const int16_t *col)
57{
58    int c0, c1, c2, c3, a0, a1, a2, a3;
59
60    a0 = col[8*0];
61    a1 = col[8*2];
62    a2 = col[8*4];
63    a3 = col[8*6];
64    c0 = ((a0 + a2) << (CN_SHIFT - 1)) + (1 << (C_SHIFT - 1));
65    c2 = ((a0 - a2) << (CN_SHIFT - 1)) + (1 << (C_SHIFT - 1));
66    c1 = a1 * C1 + a3 * C2;
67    c3 = a1 * C2 - a3 * C1;
68    dest[0] = av_clip_uint8((c0 + c1) >> C_SHIFT);
69    dest += line_size;
70    dest[0] = av_clip_uint8((c2 + c3) >> C_SHIFT);
71    dest += line_size;
72    dest[0] = av_clip_uint8((c2 - c3) >> C_SHIFT);
73    dest += line_size;
74    dest[0] = av_clip_uint8((c0 - c1) >> C_SHIFT);
75}
76
77#define BF(k) \
78{\
79    int a0, a1;\
80    a0 = ptr[k];\
81    a1 = ptr[8 + k];\
82    ptr[k] = a0 + a1;\
83    ptr[8 + k] = a0 - a1;\
84}
85
86/* only used by DV codec. The input must be interlaced. 128 is added
87   to the pixels before clamping to avoid systematic error
88   (1024*sqrt(2)) offset would be needed otherwise. */
89/* XXX: I think a 1.0/sqrt(2) normalization should be needed to
90   compensate the extra butterfly stage - I don't have the full DV
91   specification */
92void ff_simple_idct248_put(uint8_t *dest, int line_size, int16_t *block)
93{
94    int i;
95    int16_t *ptr;
96
97    /* butterfly */
98    ptr = block;
99    for(i=0;i<4;i++) {
100        BF(0);
101        BF(1);
102        BF(2);
103        BF(3);
104        BF(4);
105        BF(5);
106        BF(6);
107        BF(7);
108        ptr += 2 * 8;
109    }
110
111    /* IDCT8 on each line */
112    for(i=0; i<8; i++) {
113        idctRowCondDC_8(block + i*8, 0);
114    }
115
116    /* IDCT4 and store */
117    for(i=0;i<8;i++) {
118        idct4col_put(dest + i, 2 * line_size, block + i);
119        idct4col_put(dest + line_size + i, 2 * line_size, block + 8 + i);
120    }
121}
122
123/* 8x4 & 4x8 WMV2 IDCT */
124#undef CN_SHIFT
125#undef C_SHIFT
126#undef C_FIX
127#undef C1
128#undef C2
129#define CN_SHIFT 12
130#define C_FIX(x) ((int)((x) * 1.414213562 * (1 << CN_SHIFT) + 0.5))
131#define C1 C_FIX(0.6532814824)
132#define C2 C_FIX(0.2705980501)
133#define C3 C_FIX(0.5)
134#define C_SHIFT (4+1+12)
135static inline void idct4col_add(uint8_t *dest, int line_size, const int16_t *col)
136{
137    int c0, c1, c2, c3, a0, a1, a2, a3;
138
139    a0 = col[8*0];
140    a1 = col[8*1];
141    a2 = col[8*2];
142    a3 = col[8*3];
143    c0 = (a0 + a2)*C3 + (1 << (C_SHIFT - 1));
144    c2 = (a0 - a2)*C3 + (1 << (C_SHIFT - 1));
145    c1 = a1 * C1 + a3 * C2;
146    c3 = a1 * C2 - a3 * C1;
147    dest[0] = av_clip_uint8(dest[0] + ((c0 + c1) >> C_SHIFT));
148    dest += line_size;
149    dest[0] = av_clip_uint8(dest[0] + ((c2 + c3) >> C_SHIFT));
150    dest += line_size;
151    dest[0] = av_clip_uint8(dest[0] + ((c2 - c3) >> C_SHIFT));
152    dest += line_size;
153    dest[0] = av_clip_uint8(dest[0] + ((c0 - c1) >> C_SHIFT));
154}
155
156#define RN_SHIFT 15
157#define R_FIX(x) ((int)((x) * 1.414213562 * (1 << RN_SHIFT) + 0.5))
158#define R1 R_FIX(0.6532814824)
159#define R2 R_FIX(0.2705980501)
160#define R3 R_FIX(0.5)
161#define R_SHIFT 11
162static inline void idct4row(int16_t *row)
163{
164    int c0, c1, c2, c3, a0, a1, a2, a3;
165
166    a0 = row[0];
167    a1 = row[1];
168    a2 = row[2];
169    a3 = row[3];
170    c0 = (a0 + a2)*R3 + (1 << (R_SHIFT - 1));
171    c2 = (a0 - a2)*R3 + (1 << (R_SHIFT - 1));
172    c1 = a1 * R1 + a3 * R2;
173    c3 = a1 * R2 - a3 * R1;
174    row[0]= (c0 + c1) >> R_SHIFT;
175    row[1]= (c2 + c3) >> R_SHIFT;
176    row[2]= (c2 - c3) >> R_SHIFT;
177    row[3]= (c0 - c1) >> R_SHIFT;
178}
179
180void ff_simple_idct84_add(uint8_t *dest, int line_size, int16_t *block)
181{
182    int i;
183
184    /* IDCT8 on each line */
185    for(i=0; i<4; i++) {
186        idctRowCondDC_8(block + i*8, 0);
187    }
188
189    /* IDCT4 and store */
190    for(i=0;i<8;i++) {
191        idct4col_add(dest + i, line_size, block + i);
192    }
193}
194
195void ff_simple_idct48_add(uint8_t *dest, int line_size, int16_t *block)
196{
197    int i;
198
199    /* IDCT4 on each line */
200    for(i=0; i<8; i++) {
201        idct4row(block + i*8);
202    }
203
204    /* IDCT8 and store */
205    for(i=0; i<4; i++){
206        idctSparseColAdd_8(dest + i, line_size, block + i);
207    }
208}
209
210void ff_simple_idct44_add(uint8_t *dest, int line_size, int16_t *block)
211{
212    int i;
213
214    /* IDCT4 on each line */
215    for(i=0; i<4; i++) {
216        idct4row(block + i*8);
217    }
218
219    /* IDCT4 and store */
220    for(i=0; i<4; i++){
221        idct4col_add(dest + i, line_size, block + i);
222    }
223}
224
225void ff_prores_idct(int16_t *block, const int16_t *qmat)
226{
227    int i;
228
229    for (i = 0; i < 64; i++)
230        block[i] *= qmat[i];
231
232    for (i = 0; i < 8; i++)
233        idctRowCondDC_10(block + i*8, 2);
234
235    for (i = 0; i < 8; i++) {
236        block[i] += 8192;
237        idctSparseCol_10(block + i);
238    }
239}
240