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 "mathops.h"
30#include "simple_idct.h"
31
32#define IN_IDCT_DEPTH 16
33
34#define BIT_DEPTH 8
35#include "simple_idct_template.c"
36#undef BIT_DEPTH
37
38#define BIT_DEPTH 10
39#include "simple_idct_template.c"
40
41#define EXTRA_SHIFT  2
42#include "simple_idct_template.c"
43
44#undef EXTRA_SHIFT
45#undef BIT_DEPTH
46
47#define BIT_DEPTH 12
48#include "simple_idct_template.c"
49#undef BIT_DEPTH
50#undef IN_IDCT_DEPTH
51
52#define IN_IDCT_DEPTH 32
53#define BIT_DEPTH 10
54#include "simple_idct_template.c"
55#undef BIT_DEPTH
56#undef IN_IDCT_DEPTH
57
58/* 2x4x8 idct */
59
60#define CN_SHIFT 12
61#define C_FIX(x) ((int)((x) * (1 << CN_SHIFT) + 0.5))
62#define C1 C_FIX(0.6532814824)
63#define C2 C_FIX(0.2705980501)
64
65/* row idct is multiple by 16 * sqrt(2.0), col idct4 is normalized,
66   and the butterfly must be multiplied by 0.5 * sqrt(2.0) */
67#define C_SHIFT (4+1+12)
68
69static inline void idct4col_put(uint8_t *dest, ptrdiff_t line_size, const int16_t *col)
70{
71    int c0, c1, c2, c3, a0, a1, a2, a3;
72
73    a0 = col[8*0];
74    a1 = col[8*2];
75    a2 = col[8*4];
76    a3 = col[8*6];
77    c0 = ((a0 + a2) * (1 << CN_SHIFT - 1)) + (1 << (C_SHIFT - 1));
78    c2 = ((a0 - a2) * (1 << CN_SHIFT - 1)) + (1 << (C_SHIFT - 1));
79    c1 = a1 * C1 + a3 * C2;
80    c3 = a1 * C2 - a3 * C1;
81    dest[0] = av_clip_uint8((c0 + c1) >> C_SHIFT);
82    dest += line_size;
83    dest[0] = av_clip_uint8((c2 + c3) >> C_SHIFT);
84    dest += line_size;
85    dest[0] = av_clip_uint8((c2 - c3) >> C_SHIFT);
86    dest += line_size;
87    dest[0] = av_clip_uint8((c0 - c1) >> C_SHIFT);
88}
89
90#define BF(k) \
91{\
92    int a0, a1;\
93    a0 = ptr[k];\
94    a1 = ptr[8 + k];\
95    ptr[k] = a0 + a1;\
96    ptr[8 + k] = a0 - a1;\
97}
98
99/* only used by DV codec. The input must be interlaced. 128 is added
100   to the pixels before clamping to avoid systematic error
101   (1024*sqrt(2)) offset would be needed otherwise. */
102/* XXX: I think a 1.0/sqrt(2) normalization should be needed to
103   compensate the extra butterfly stage - I don't have the full DV
104   specification */
105void ff_simple_idct248_put(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
106{
107    int i;
108    int16_t *ptr;
109
110    /* butterfly */
111    ptr = block;
112    for(i=0;i<4;i++) {
113        BF(0);
114        BF(1);
115        BF(2);
116        BF(3);
117        BF(4);
118        BF(5);
119        BF(6);
120        BF(7);
121        ptr += 2 * 8;
122    }
123
124    /* IDCT8 on each line */
125    for(i=0; i<8; i++) {
126        idctRowCondDC_int16_8bit(block + i*8, 0);
127    }
128
129    /* IDCT4 and store */
130    for(i=0;i<8;i++) {
131        idct4col_put(dest + i, 2 * line_size, block + i);
132        idct4col_put(dest + line_size + i, 2 * line_size, block + 8 + i);
133    }
134}
135
136/* 8x4 & 4x8 WMV2 IDCT */
137#undef CN_SHIFT
138#undef C_SHIFT
139#undef C_FIX
140#undef C1
141#undef C2
142#define CN_SHIFT 12
143#define C_FIX(x) ((int)((x) * M_SQRT2 * (1 << CN_SHIFT) + 0.5))
144#define C1 C_FIX(0.6532814824)
145#define C2 C_FIX(0.2705980501)
146#define C3 C_FIX(0.5)
147#define C_SHIFT (4+1+12)
148static inline void idct4col_add(uint8_t *dest, ptrdiff_t line_size, const int16_t *col)
149{
150    int c0, c1, c2, c3, a0, a1, a2, a3;
151
152    a0 = col[8*0];
153    a1 = col[8*1];
154    a2 = col[8*2];
155    a3 = col[8*3];
156    c0 = (a0 + a2)*C3 + (1 << (C_SHIFT - 1));
157    c2 = (a0 - a2)*C3 + (1 << (C_SHIFT - 1));
158    c1 = a1 * C1 + a3 * C2;
159    c3 = a1 * C2 - a3 * C1;
160    dest[0] = av_clip_uint8(dest[0] + ((c0 + c1) >> C_SHIFT));
161    dest += line_size;
162    dest[0] = av_clip_uint8(dest[0] + ((c2 + c3) >> C_SHIFT));
163    dest += line_size;
164    dest[0] = av_clip_uint8(dest[0] + ((c2 - c3) >> C_SHIFT));
165    dest += line_size;
166    dest[0] = av_clip_uint8(dest[0] + ((c0 - c1) >> C_SHIFT));
167}
168
169#define RN_SHIFT 15
170#define R_FIX(x) ((int)((x) * M_SQRT2 * (1 << RN_SHIFT) + 0.5))
171#define R1 R_FIX(0.6532814824)
172#define R2 R_FIX(0.2705980501)
173#define R3 R_FIX(0.5)
174#define R_SHIFT 11
175static inline void idct4row(int16_t *row)
176{
177    unsigned c0, c1, c2, c3;
178    int a0, a1, a2, a3;
179
180    a0 = row[0];
181    a1 = row[1];
182    a2 = row[2];
183    a3 = row[3];
184    c0 = (a0 + a2)*R3 + (1 << (R_SHIFT - 1));
185    c2 = (a0 - a2)*R3 + (1 << (R_SHIFT - 1));
186    c1 = a1 * R1 + a3 * R2;
187    c3 = a1 * R2 - a3 * R1;
188    row[0]= (c0 + c1) >> R_SHIFT;
189    row[1]= (c2 + c3) >> R_SHIFT;
190    row[2]= (c2 - c3) >> R_SHIFT;
191    row[3]= (c0 - c1) >> R_SHIFT;
192}
193
194void ff_simple_idct84_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
195{
196    int i;
197
198    /* IDCT8 on each line */
199    for(i=0; i<4; i++) {
200        idctRowCondDC_int16_8bit(block + i*8, 0);
201    }
202
203    /* IDCT4 and store */
204    for(i=0;i<8;i++) {
205        idct4col_add(dest + i, line_size, block + i);
206    }
207}
208
209void ff_simple_idct48_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
210{
211    int i;
212
213    /* IDCT4 on each line */
214    for(i=0; i<8; i++) {
215        idct4row(block + i*8);
216    }
217
218    /* IDCT8 and store */
219    for(i=0; i<4; i++){
220        idctSparseColAdd_int16_8bit(dest + i, line_size, block + i);
221    }
222}
223
224void ff_simple_idct44_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
225{
226    int i;
227
228    /* IDCT4 on each line */
229    for(i=0; i<4; i++) {
230        idct4row(block + i*8);
231    }
232
233    /* IDCT4 and store */
234    for(i=0; i<4; i++){
235        idct4col_add(dest + i, line_size, block + i);
236    }
237}
238
239void ff_prores_idct_10(int16_t *block, const int16_t *qmat)
240{
241    int i;
242
243    for (i = 0; i < 64; i++)
244        block[i] *= qmat[i];
245
246    for (i = 0; i < 8; i++)
247        idctRowCondDC_extrashift_10(block + i*8, 2);
248
249    for (i = 0; i < 8; i++) {
250        block[i] += 8192;
251        idctSparseCol_extrashift_10(block + i);
252    }
253}
254
255void ff_prores_idct_12(int16_t *block, const int16_t *qmat)
256{
257    int i;
258
259    for (i = 0; i < 64; i++)
260        block[i] *= qmat[i];
261
262    for (i = 0; i < 8; i++)
263        idctRowCondDC_int16_12bit(block + i*8, 0);
264
265    for (i = 0; i < 8; i++) {
266        block[i] += 8192;
267        idctSparseCol_int16_12bit(block + i);
268    }
269}
270