simple_idct: make repeated code a macro
[libav.git] / libavcodec / simple_idct_template.c
CommitLineData
e7a972e1
MR
1/*
2 * Simple IDCT
3 *
4 * Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at>
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/**
24 * @file
25 * simpleidct in C.
26 */
27
28/*
29 based upon some outcommented c code from mpeg2dec (idct_mmx.c
30 written by Aaron Holtzman <aholtzma@ess.engr.uvic.ca>)
31 */
32
33#include "bit_depth_template.c"
34
35#undef W1
36#undef W2
37#undef W3
38#undef W4
39#undef W5
40#undef W6
41#undef W7
42#undef ROW_SHIFT
43#undef COL_SHIFT
44#undef DC_SHIFT
45#undef MUL
46#undef MAC
47
48#if BIT_DEPTH == 8
49
50#define W1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
51#define W2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
52#define W3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
53#define W4 16383 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
54#define W5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
55#define W6 8867 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
56#define W7 4520 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
57
58#define ROW_SHIFT 11
59#define COL_SHIFT 20
60#define DC_SHIFT 3
61
62#define MUL(a, b) MUL16(a, b)
63#define MAC(a, b, c) MAC16(a, b, c)
64
65#elif BIT_DEPTH == 10
66
67#define W1 90901
68#define W2 85627
69#define W3 77062
70#define W4 65535
71#define W5 51491
72#define W6 35468
73#define W7 18081
74
75#define ROW_SHIFT 15
76#define COL_SHIFT 20
77#define DC_SHIFT 1
78
79#define MUL(a, b) ((a) * (b))
80#define MAC(a, b, c) ((a) += (b) * (c))
81
82#else
83
84#error "Unsupported bitdepth"
85
86#endif
87
88static inline void FUNC(idctRowCondDC)(DCTELEM *row)
89{
90 int a0, a1, a2, a3, b0, b1, b2, b3;
91
92#if HAVE_FAST_64BIT
93#define ROW0_MASK (0xffffLL << 48 * HAVE_BIGENDIAN)
94 if (((((uint64_t *)row)[0] & ~ROW0_MASK) | ((uint64_t *)row)[1]) == 0) {
95 uint64_t temp = (row[0] << DC_SHIFT) & 0xffff;
96 temp += temp << 16;
97 temp += temp << 32;
98 ((uint64_t *)row)[0] = temp;
99 ((uint64_t *)row)[1] = temp;
100 return;
101 }
102#else
103 if (!(((uint32_t*)row)[1] |
104 ((uint32_t*)row)[2] |
105 ((uint32_t*)row)[3] |
106 row[1])) {
107 uint32_t temp = (row[0] << DC_SHIFT) & 0xffff;
108 temp += temp << 16;
109 ((uint32_t*)row)[0]=((uint32_t*)row)[1] =
110 ((uint32_t*)row)[2]=((uint32_t*)row)[3] = temp;
111 return;
112 }
113#endif
114
115 a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1));
116 a1 = a0;
117 a2 = a0;
118 a3 = a0;
119
120 /* no need to optimize : gcc does it */
121 a0 += W2 * row[2];
122 a1 += W6 * row[2];
123 a2 -= W6 * row[2];
124 a3 -= W2 * row[2];
125
126 b0 = MUL(W1, row[1]);
127 MAC(b0, W3, row[3]);
128 b1 = MUL(W3, row[1]);
129 MAC(b1, -W7, row[3]);
130 b2 = MUL(W5, row[1]);
131 MAC(b2, -W1, row[3]);
132 b3 = MUL(W7, row[1]);
133 MAC(b3, -W5, row[3]);
134
135 if (AV_RN64A(row + 4)) {
136 a0 += W4*row[4] + W6*row[6];
137 a1 += - W4*row[4] - W2*row[6];
138 a2 += - W4*row[4] + W2*row[6];
139 a3 += W4*row[4] - W6*row[6];
140
141 MAC(b0, W5, row[5]);
142 MAC(b0, W7, row[7]);
143
144 MAC(b1, -W1, row[5]);
145 MAC(b1, -W5, row[7]);
146
147 MAC(b2, W7, row[5]);
148 MAC(b2, W3, row[7]);
149
150 MAC(b3, W3, row[5]);
151 MAC(b3, -W1, row[7]);
152 }
153
154 row[0] = (a0 + b0) >> ROW_SHIFT;
155 row[7] = (a0 - b0) >> ROW_SHIFT;
156 row[1] = (a1 + b1) >> ROW_SHIFT;
157 row[6] = (a1 - b1) >> ROW_SHIFT;
158 row[2] = (a2 + b2) >> ROW_SHIFT;
159 row[5] = (a2 - b2) >> ROW_SHIFT;
160 row[3] = (a3 + b3) >> ROW_SHIFT;
161 row[4] = (a3 - b3) >> ROW_SHIFT;
162}
163
a402f109
MR
164#define IDCT_COLS do { \
165 a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4)); \
166 a1 = a0; \
167 a2 = a0; \
168 a3 = a0; \
169 \
170 a0 += W2*col[8*2]; \
171 a1 += W6*col[8*2]; \
172 a2 += -W6*col[8*2]; \
173 a3 += -W2*col[8*2]; \
174 \
175 b0 = MUL(W1, col[8*1]); \
176 b1 = MUL(W3, col[8*1]); \
177 b2 = MUL(W5, col[8*1]); \
178 b3 = MUL(W7, col[8*1]); \
179 \
180 MAC(b0, W3, col[8*3]); \
181 MAC(b1, -W7, col[8*3]); \
182 MAC(b2, -W1, col[8*3]); \
183 MAC(b3, -W5, col[8*3]); \
184 \
185 if (col[8*4]) { \
186 a0 += W4*col[8*4]; \
187 a1 += -W4*col[8*4]; \
188 a2 += -W4*col[8*4]; \
189 a3 += W4*col[8*4]; \
190 } \
191 \
192 if (col[8*5]) { \
193 MAC(b0, W5, col[8*5]); \
194 MAC(b1, -W1, col[8*5]); \
195 MAC(b2, W7, col[8*5]); \
196 MAC(b3, W3, col[8*5]); \
197 } \
198 \
199 if (col[8*6]) { \
200 a0 += W6*col[8*6]; \
201 a1 += -W2*col[8*6]; \
202 a2 += W2*col[8*6]; \
203 a3 += -W6*col[8*6]; \
204 } \
205 \
206 if (col[8*7]) { \
207 MAC(b0, W7, col[8*7]); \
208 MAC(b1, -W5, col[8*7]); \
209 MAC(b2, W3, col[8*7]); \
210 MAC(b3, -W1, col[8*7]); \
211 } \
212 } while (0)
213
e7a972e1
MR
214static inline void FUNC(idctSparseColPut)(pixel *dest, int line_size,
215 DCTELEM *col)
216{
217 int a0, a1, a2, a3, b0, b1, b2, b3;
218 INIT_CLIP;
219
a402f109 220 IDCT_COLS;
e7a972e1
MR
221
222 dest[0] = CLIP((a0 + b0) >> COL_SHIFT);
223 dest += line_size;
224 dest[0] = CLIP((a1 + b1) >> COL_SHIFT);
225 dest += line_size;
226 dest[0] = CLIP((a2 + b2) >> COL_SHIFT);
227 dest += line_size;
228 dest[0] = CLIP((a3 + b3) >> COL_SHIFT);
229 dest += line_size;
230 dest[0] = CLIP((a3 - b3) >> COL_SHIFT);
231 dest += line_size;
232 dest[0] = CLIP((a2 - b2) >> COL_SHIFT);
233 dest += line_size;
234 dest[0] = CLIP((a1 - b1) >> COL_SHIFT);
235 dest += line_size;
236 dest[0] = CLIP((a0 - b0) >> COL_SHIFT);
237}
238
239static inline void FUNC(idctSparseColAdd)(pixel *dest, int line_size,
240 DCTELEM *col)
241{
242 int a0, a1, a2, a3, b0, b1, b2, b3;
243 INIT_CLIP;
244
a402f109 245 IDCT_COLS;
e7a972e1
MR
246
247 dest[0] = CLIP(dest[0] + ((a0 + b0) >> COL_SHIFT));
248 dest += line_size;
249 dest[0] = CLIP(dest[0] + ((a1 + b1) >> COL_SHIFT));
250 dest += line_size;
251 dest[0] = CLIP(dest[0] + ((a2 + b2) >> COL_SHIFT));
252 dest += line_size;
253 dest[0] = CLIP(dest[0] + ((a3 + b3) >> COL_SHIFT));
254 dest += line_size;
255 dest[0] = CLIP(dest[0] + ((a3 - b3) >> COL_SHIFT));
256 dest += line_size;
257 dest[0] = CLIP(dest[0] + ((a2 - b2) >> COL_SHIFT));
258 dest += line_size;
259 dest[0] = CLIP(dest[0] + ((a1 - b1) >> COL_SHIFT));
260 dest += line_size;
261 dest[0] = CLIP(dest[0] + ((a0 - b0) >> COL_SHIFT));
262}
263
264static inline void FUNC(idctSparseCol)(DCTELEM *col)
265{
266 int a0, a1, a2, a3, b0, b1, b2, b3;
267
a402f109 268 IDCT_COLS;
e7a972e1
MR
269
270 col[0 ] = ((a0 + b0) >> COL_SHIFT);
271 col[8 ] = ((a1 + b1) >> COL_SHIFT);
272 col[16] = ((a2 + b2) >> COL_SHIFT);
273 col[24] = ((a3 + b3) >> COL_SHIFT);
274 col[32] = ((a3 - b3) >> COL_SHIFT);
275 col[40] = ((a2 - b2) >> COL_SHIFT);
276 col[48] = ((a1 - b1) >> COL_SHIFT);
277 col[56] = ((a0 - b0) >> COL_SHIFT);
278}
279
280void FUNC(ff_simple_idct_put)(uint8_t *dest_, int line_size, DCTELEM *block)
281{
282 pixel *dest = (pixel *)dest_;
283 int i;
e7bcc5ba
MR
284
285 line_size /= sizeof(pixel);
286
e7a972e1
MR
287 for(i=0; i<8; i++)
288 FUNC(idctRowCondDC)(block + i*8);
289
290 for(i=0; i<8; i++)
291 FUNC(idctSparseColPut)(dest + i, line_size, block + i);
292}
293
294void FUNC(ff_simple_idct_add)(uint8_t *dest_, int line_size, DCTELEM *block)
295{
296 pixel *dest = (pixel *)dest_;
297 int i;
e7bcc5ba
MR
298
299 line_size /= sizeof(pixel);
300
e7a972e1
MR
301 for(i=0; i<8; i++)
302 FUNC(idctRowCondDC)(block + i*8);
303
304 for(i=0; i<8; i++)
305 FUNC(idctSparseColAdd)(dest + i, line_size, block + i);
306}
307
308void FUNC(ff_simple_idct)(DCTELEM *block)
309{
310 int i;
311 for(i=0; i<8; i++)
312 FUNC(idctRowCondDC)(block + i*8);
313
314 for(i=0; i<8; i++)
315 FUNC(idctSparseCol)(block + i);
316}