31de23da464a051ffcc22dd3d310f5a75e998127
[libav.git] / libavcodec / vp3dsp.c
1 /*
2 * Copyright (C) 2004 the ffmpeg project
3 *
4 * This file is part of Libav.
5 *
6 * Libav is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * Libav is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 /**
22 * @file
23 * Standard C DSP-oriented functions cribbed from the original VP3
24 * source code.
25 */
26
27 #include "libavutil/attributes.h"
28 #include "libavutil/common.h"
29 #include "libavutil/intreadwrite.h"
30
31 #include "avcodec.h"
32 #include "dsputil.h"
33 #include "rnd_avg.h"
34 #include "vp3dsp.h"
35
36 #define IdctAdjustBeforeShift 8
37 #define xC1S7 64277
38 #define xC2S6 60547
39 #define xC3S5 54491
40 #define xC4S4 46341
41 #define xC5S3 36410
42 #define xC6S2 25080
43 #define xC7S1 12785
44
45 #define M(a, b) (((a) * (b)) >> 16)
46
47 static av_always_inline void idct(uint8_t *dst, int stride,
48 int16_t *input, int type)
49 {
50 int16_t *ip = input;
51
52 int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
53 int Ed, Gd, Add, Bdd, Fd, Hd;
54
55 int i;
56
57 /* Inverse DCT on the rows now */
58 for (i = 0; i < 8; i++) {
59 /* Check for non-zero values */
60 if (ip[0 * 8] | ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
61 ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8]) {
62 A = M(xC1S7, ip[1 * 8]) + M(xC7S1, ip[7 * 8]);
63 B = M(xC7S1, ip[1 * 8]) - M(xC1S7, ip[7 * 8]);
64 C = M(xC3S5, ip[3 * 8]) + M(xC5S3, ip[5 * 8]);
65 D = M(xC3S5, ip[5 * 8]) - M(xC5S3, ip[3 * 8]);
66
67 Ad = M(xC4S4, (A - C));
68 Bd = M(xC4S4, (B - D));
69
70 Cd = A + C;
71 Dd = B + D;
72
73 E = M(xC4S4, (ip[0 * 8] + ip[4 * 8]));
74 F = M(xC4S4, (ip[0 * 8] - ip[4 * 8]));
75
76 G = M(xC2S6, ip[2 * 8]) + M(xC6S2, ip[6 * 8]);
77 H = M(xC6S2, ip[2 * 8]) - M(xC2S6, ip[6 * 8]);
78
79 Ed = E - G;
80 Gd = E + G;
81
82 Add = F + Ad;
83 Bdd = Bd - H;
84
85 Fd = F - Ad;
86 Hd = Bd + H;
87
88 /* Final sequence of operations over-write original inputs. */
89 ip[0 * 8] = Gd + Cd;
90 ip[7 * 8] = Gd - Cd;
91
92 ip[1 * 8] = Add + Hd;
93 ip[2 * 8] = Add - Hd;
94
95 ip[3 * 8] = Ed + Dd;
96 ip[4 * 8] = Ed - Dd;
97
98 ip[5 * 8] = Fd + Bdd;
99 ip[6 * 8] = Fd - Bdd;
100 }
101
102 ip += 1; /* next row */
103 }
104
105 ip = input;
106
107 for (i = 0; i < 8; i++) {
108 /* Check for non-zero values (bitwise or faster than ||) */
109 if (ip[1] | ip[2] | ip[3] |
110 ip[4] | ip[5] | ip[6] | ip[7]) {
111 A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]);
112 B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]);
113 C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]);
114 D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]);
115
116 Ad = M(xC4S4, (A - C));
117 Bd = M(xC4S4, (B - D));
118
119 Cd = A + C;
120 Dd = B + D;
121
122 E = M(xC4S4, (ip[0] + ip[4])) + 8;
123 F = M(xC4S4, (ip[0] - ip[4])) + 8;
124
125 if (type == 1) { // HACK
126 E += 16 * 128;
127 F += 16 * 128;
128 }
129
130 G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]);
131 H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]);
132
133 Ed = E - G;
134 Gd = E + G;
135
136 Add = F + Ad;
137 Bdd = Bd - H;
138
139 Fd = F - Ad;
140 Hd = Bd + H;
141
142 /* Final sequence of operations over-write original inputs. */
143 if (type == 1) {
144 dst[0 * stride] = av_clip_uint8((Gd + Cd) >> 4);
145 dst[7 * stride] = av_clip_uint8((Gd - Cd) >> 4);
146
147 dst[1 * stride] = av_clip_uint8((Add + Hd) >> 4);
148 dst[2 * stride] = av_clip_uint8((Add - Hd) >> 4);
149
150 dst[3 * stride] = av_clip_uint8((Ed + Dd) >> 4);
151 dst[4 * stride] = av_clip_uint8((Ed - Dd) >> 4);
152
153 dst[5 * stride] = av_clip_uint8((Fd + Bdd) >> 4);
154 dst[6 * stride] = av_clip_uint8((Fd - Bdd) >> 4);
155 } else {
156 dst[0 * stride] = av_clip_uint8(dst[0 * stride] + ((Gd + Cd) >> 4));
157 dst[7 * stride] = av_clip_uint8(dst[7 * stride] + ((Gd - Cd) >> 4));
158
159 dst[1 * stride] = av_clip_uint8(dst[1 * stride] + ((Add + Hd) >> 4));
160 dst[2 * stride] = av_clip_uint8(dst[2 * stride] + ((Add - Hd) >> 4));
161
162 dst[3 * stride] = av_clip_uint8(dst[3 * stride] + ((Ed + Dd) >> 4));
163 dst[4 * stride] = av_clip_uint8(dst[4 * stride] + ((Ed - Dd) >> 4));
164
165 dst[5 * stride] = av_clip_uint8(dst[5 * stride] + ((Fd + Bdd) >> 4));
166 dst[6 * stride] = av_clip_uint8(dst[6 * stride] + ((Fd - Bdd) >> 4));
167 }
168 } else {
169 if (type == 1) {
170 dst[0*stride] =
171 dst[1*stride] =
172 dst[2*stride] =
173 dst[3*stride] =
174 dst[4*stride] =
175 dst[5*stride] =
176 dst[6*stride] =
177 dst[7*stride] = av_clip_uint8(128 + ((xC4S4 * ip[0] + (IdctAdjustBeforeShift << 16)) >> 20));
178 } else {
179 if (ip[0]) {
180 int v = (xC4S4 * ip[0] + (IdctAdjustBeforeShift << 16)) >> 20;
181 dst[0 * stride] = av_clip_uint8(dst[0 * stride] + v);
182 dst[1 * stride] = av_clip_uint8(dst[1 * stride] + v);
183 dst[2 * stride] = av_clip_uint8(dst[2 * stride] + v);
184 dst[3 * stride] = av_clip_uint8(dst[3 * stride] + v);
185 dst[4 * stride] = av_clip_uint8(dst[4 * stride] + v);
186 dst[5 * stride] = av_clip_uint8(dst[5 * stride] + v);
187 dst[6 * stride] = av_clip_uint8(dst[6 * stride] + v);
188 dst[7 * stride] = av_clip_uint8(dst[7 * stride] + v);
189 }
190 }
191 }
192
193 ip += 8; /* next column */
194 dst++;
195 }
196 }
197
198 static void vp3_idct_put_c(uint8_t *dest /* align 8 */, int line_size,
199 int16_t *block /* align 16 */)
200 {
201 idct(dest, line_size, block, 1);
202 memset(block, 0, sizeof(*block) * 64);
203 }
204
205 static void vp3_idct_add_c(uint8_t *dest /* align 8 */, int line_size,
206 int16_t *block /* align 16 */)
207 {
208 idct(dest, line_size, block, 2);
209 memset(block, 0, sizeof(*block) * 64);
210 }
211
212 static void vp3_idct_dc_add_c(uint8_t *dest /* align 8 */, int line_size,
213 int16_t *block /* align 16 */)
214 {
215 int i, dc = (block[0] + 15) >> 5;
216
217 for (i = 0; i < 8; i++) {
218 dest[0] = av_clip_uint8(dest[0] + dc);
219 dest[1] = av_clip_uint8(dest[1] + dc);
220 dest[2] = av_clip_uint8(dest[2] + dc);
221 dest[3] = av_clip_uint8(dest[3] + dc);
222 dest[4] = av_clip_uint8(dest[4] + dc);
223 dest[5] = av_clip_uint8(dest[5] + dc);
224 dest[6] = av_clip_uint8(dest[6] + dc);
225 dest[7] = av_clip_uint8(dest[7] + dc);
226 dest += line_size;
227 }
228 block[0] = 0;
229 }
230
231 static void vp3_v_loop_filter_c(uint8_t *first_pixel, int stride,
232 int *bounding_values)
233 {
234 unsigned char *end;
235 int filter_value;
236 const int nstride = -stride;
237
238 for (end = first_pixel + 8; first_pixel < end; first_pixel++) {
239 filter_value = (first_pixel[2 * nstride] - first_pixel[stride]) +
240 (first_pixel[0] - first_pixel[nstride]) * 3;
241 filter_value = bounding_values[(filter_value + 4) >> 3];
242
243 first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
244 first_pixel[0] = av_clip_uint8(first_pixel[0] - filter_value);
245 }
246 }
247
248 static void vp3_h_loop_filter_c(uint8_t *first_pixel, int stride,
249 int *bounding_values)
250 {
251 unsigned char *end;
252 int filter_value;
253
254 for (end = first_pixel + 8 * stride; first_pixel != end; first_pixel += stride) {
255 filter_value = (first_pixel[-2] - first_pixel[1]) +
256 (first_pixel[ 0] - first_pixel[-1]) * 3;
257 filter_value = bounding_values[(filter_value + 4) >> 3];
258
259 first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
260 first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
261 }
262 }
263
264 static void put_no_rnd_pixels_l2(uint8_t *dst, const uint8_t *src1,
265 const uint8_t *src2, ptrdiff_t stride, int h)
266 {
267 int i;
268
269 for (i = 0; i < h; i++) {
270 uint32_t a, b;
271
272 a = AV_RN32(&src1[i * stride]);
273 b = AV_RN32(&src2[i * stride]);
274 AV_WN32A(&dst[i * stride], no_rnd_avg32(a, b));
275 a = AV_RN32(&src1[i * stride + 4]);
276 b = AV_RN32(&src2[i * stride + 4]);
277 AV_WN32A(&dst[i * stride + 4], no_rnd_avg32(a, b));
278 }
279 }
280
281 av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags)
282 {
283 c->put_no_rnd_pixels_l2 = put_no_rnd_pixels_l2;
284
285 c->idct_put = vp3_idct_put_c;
286 c->idct_add = vp3_idct_add_c;
287 c->idct_dc_add = vp3_idct_dc_add_c;
288 c->v_loop_filter = vp3_v_loop_filter_c;
289 c->h_loop_filter = vp3_h_loop_filter_c;
290
291 if (ARCH_ARM)
292 ff_vp3dsp_init_arm(c, flags);
293 if (ARCH_BFIN)
294 ff_vp3dsp_init_bfin(c, flags);
295 if (ARCH_PPC)
296 ff_vp3dsp_init_ppc(c, flags);
297 if (ARCH_X86)
298 ff_vp3dsp_init_x86(c, flags);
299 }