1/*
2 * Copyright (C) 2004 the ffmpeg project
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg 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 * FFmpeg 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 FFmpeg; 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 "rnd_avg.h"
33#include "vp3dsp.h"
34
35#define IdctAdjustBeforeShift 8
36#define xC1S7 64277
37#define xC2S6 60547
38#define xC3S5 54491
39#define xC4S4 46341
40#define xC5S3 36410
41#define xC6S2 25080
42#define xC7S1 12785
43
44#define M(a, b) (((a) * (b)) >> 16)
45
46static av_always_inline void idct(uint8_t *dst, int stride,
47                                  int16_t *input, int type)
48{
49    int16_t *ip = input;
50
51    int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
52    int Ed, Gd, Add, Bdd, Fd, Hd;
53
54    int i;
55
56    /* Inverse DCT on the rows now */
57    for (i = 0; i < 8; i++) {
58        /* Check for non-zero values */
59        if (ip[0 * 8] | ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
60            ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8]) {
61            A = M(xC1S7, ip[1 * 8]) + M(xC7S1, ip[7 * 8]);
62            B = M(xC7S1, ip[1 * 8]) - M(xC1S7, ip[7 * 8]);
63            C = M(xC3S5, ip[3 * 8]) + M(xC5S3, ip[5 * 8]);
64            D = M(xC3S5, ip[5 * 8]) - M(xC5S3, ip[3 * 8]);
65
66            Ad = M(xC4S4, (A - C));
67            Bd = M(xC4S4, (B - D));
68
69            Cd = A + C;
70            Dd = B + D;
71
72            E = M(xC4S4, (ip[0 * 8] + ip[4 * 8]));
73            F = M(xC4S4, (ip[0 * 8] - ip[4 * 8]));
74
75            G = M(xC2S6, ip[2 * 8]) + M(xC6S2, ip[6 * 8]);
76            H = M(xC6S2, ip[2 * 8]) - M(xC2S6, ip[6 * 8]);
77
78            Ed = E - G;
79            Gd = E + G;
80
81            Add = F + Ad;
82            Bdd = Bd - H;
83
84            Fd = F - Ad;
85            Hd = Bd + H;
86
87            /*  Final sequence of operations over-write original inputs. */
88            ip[0 * 8] = Gd + Cd;
89            ip[7 * 8] = Gd - Cd;
90
91            ip[1 * 8] = Add + Hd;
92            ip[2 * 8] = Add - Hd;
93
94            ip[3 * 8] = Ed + Dd;
95            ip[4 * 8] = Ed - Dd;
96
97            ip[5 * 8] = Fd + Bdd;
98            ip[6 * 8] = Fd - Bdd;
99        }
100
101        ip += 1;            /* next row */
102    }
103
104    ip = input;
105
106    for (i = 0; i < 8; i++) {
107        /* Check for non-zero values (bitwise or faster than ||) */
108        if (ip[1] | ip[2] | ip[3] |
109            ip[4] | ip[5] | ip[6] | ip[7]) {
110            A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]);
111            B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]);
112            C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]);
113            D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]);
114
115            Ad = M(xC4S4, (A - C));
116            Bd = M(xC4S4, (B - D));
117
118            Cd = A + C;
119            Dd = B + D;
120
121            E = M(xC4S4, (ip[0] + ip[4])) + 8;
122            F = M(xC4S4, (ip[0] - ip[4])) + 8;
123
124            if (type == 1) { // HACK
125                E += 16 * 128;
126                F += 16 * 128;
127            }
128
129            G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]);
130            H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]);
131
132            Ed = E - G;
133            Gd = E + G;
134
135            Add = F + Ad;
136            Bdd = Bd - H;
137
138            Fd = F - Ad;
139            Hd = Bd + H;
140
141            /* Final sequence of operations over-write original inputs. */
142            if (type == 1) {
143                dst[0 * stride] = av_clip_uint8((Gd + Cd) >> 4);
144                dst[7 * stride] = av_clip_uint8((Gd - Cd) >> 4);
145
146                dst[1 * stride] = av_clip_uint8((Add + Hd) >> 4);
147                dst[2 * stride] = av_clip_uint8((Add - Hd) >> 4);
148
149                dst[3 * stride] = av_clip_uint8((Ed + Dd) >> 4);
150                dst[4 * stride] = av_clip_uint8((Ed - Dd) >> 4);
151
152                dst[5 * stride] = av_clip_uint8((Fd + Bdd) >> 4);
153                dst[6 * stride] = av_clip_uint8((Fd - Bdd) >> 4);
154            } else {
155                dst[0 * stride] = av_clip_uint8(dst[0 * stride] + ((Gd + Cd) >> 4));
156                dst[7 * stride] = av_clip_uint8(dst[7 * stride] + ((Gd - Cd) >> 4));
157
158                dst[1 * stride] = av_clip_uint8(dst[1 * stride] + ((Add + Hd) >> 4));
159                dst[2 * stride] = av_clip_uint8(dst[2 * stride] + ((Add - Hd) >> 4));
160
161                dst[3 * stride] = av_clip_uint8(dst[3 * stride] + ((Ed + Dd) >> 4));
162                dst[4 * stride] = av_clip_uint8(dst[4 * stride] + ((Ed - Dd) >> 4));
163
164                dst[5 * stride] = av_clip_uint8(dst[5 * stride] + ((Fd + Bdd) >> 4));
165                dst[6 * stride] = av_clip_uint8(dst[6 * stride] + ((Fd - Bdd) >> 4));
166            }
167        } else {
168            if (type == 1) {
169                dst[0*stride] =
170                dst[1*stride] =
171                dst[2*stride] =
172                dst[3*stride] =
173                dst[4*stride] =
174                dst[5*stride] =
175                dst[6*stride] =
176                dst[7*stride] = av_clip_uint8(128 + ((xC4S4 * ip[0] + (IdctAdjustBeforeShift << 16)) >> 20));
177            } else {
178                if (ip[0]) {
179                    int v = (xC4S4 * ip[0] + (IdctAdjustBeforeShift << 16)) >> 20;
180                    dst[0 * stride] = av_clip_uint8(dst[0 * stride] + v);
181                    dst[1 * stride] = av_clip_uint8(dst[1 * stride] + v);
182                    dst[2 * stride] = av_clip_uint8(dst[2 * stride] + v);
183                    dst[3 * stride] = av_clip_uint8(dst[3 * stride] + v);
184                    dst[4 * stride] = av_clip_uint8(dst[4 * stride] + v);
185                    dst[5 * stride] = av_clip_uint8(dst[5 * stride] + v);
186                    dst[6 * stride] = av_clip_uint8(dst[6 * stride] + v);
187                    dst[7 * stride] = av_clip_uint8(dst[7 * stride] + v);
188                }
189            }
190        }
191
192        ip += 8;            /* next column */
193        dst++;
194    }
195}
196
197static void vp3_idct_put_c(uint8_t *dest /* align 8 */, int line_size,
198                           int16_t *block /* align 16 */)
199{
200    idct(dest, line_size, block, 1);
201    memset(block, 0, sizeof(*block) * 64);
202}
203
204static void vp3_idct_add_c(uint8_t *dest /* align 8 */, int line_size,
205                           int16_t *block /* align 16 */)
206{
207    idct(dest, line_size, block, 2);
208    memset(block, 0, sizeof(*block) * 64);
209}
210
211static void vp3_idct_dc_add_c(uint8_t *dest /* align 8 */, int line_size,
212                              int16_t *block /* align 16 */)
213{
214    int i, dc = (block[0] + 15) >> 5;
215
216    for (i = 0; i < 8; i++) {
217        dest[0] = av_clip_uint8(dest[0] + dc);
218        dest[1] = av_clip_uint8(dest[1] + dc);
219        dest[2] = av_clip_uint8(dest[2] + dc);
220        dest[3] = av_clip_uint8(dest[3] + dc);
221        dest[4] = av_clip_uint8(dest[4] + dc);
222        dest[5] = av_clip_uint8(dest[5] + dc);
223        dest[6] = av_clip_uint8(dest[6] + dc);
224        dest[7] = av_clip_uint8(dest[7] + dc);
225        dest   += line_size;
226    }
227    block[0] = 0;
228}
229
230static void vp3_v_loop_filter_c(uint8_t *first_pixel, int stride,
231                                int *bounding_values)
232{
233    unsigned char *end;
234    int filter_value;
235    const int nstride = -stride;
236
237    for (end = first_pixel + 8; first_pixel < end; first_pixel++) {
238        filter_value = (first_pixel[2 * nstride] - first_pixel[stride]) +
239                       (first_pixel[0] - first_pixel[nstride]) * 3;
240        filter_value = bounding_values[(filter_value + 4) >> 3];
241
242        first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
243        first_pixel[0]       = av_clip_uint8(first_pixel[0] - filter_value);
244    }
245}
246
247static void vp3_h_loop_filter_c(uint8_t *first_pixel, int stride,
248                                int *bounding_values)
249{
250    unsigned char *end;
251    int filter_value;
252
253    for (end = first_pixel + 8 * stride; first_pixel != end; first_pixel += stride) {
254        filter_value = (first_pixel[-2] - first_pixel[1]) +
255                       (first_pixel[ 0] - first_pixel[-1]) * 3;
256        filter_value = bounding_values[(filter_value + 4) >> 3];
257
258        first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
259        first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
260    }
261}
262
263static void put_no_rnd_pixels_l2(uint8_t *dst, const uint8_t *src1,
264                                 const uint8_t *src2, ptrdiff_t stride, int h)
265{
266    int i;
267
268    for (i = 0; i < h; i++) {
269        uint32_t a, b;
270
271        a = AV_RN32(&src1[i * stride]);
272        b = AV_RN32(&src2[i * stride]);
273        AV_WN32A(&dst[i * stride], no_rnd_avg32(a, b));
274        a = AV_RN32(&src1[i * stride + 4]);
275        b = AV_RN32(&src2[i * stride + 4]);
276        AV_WN32A(&dst[i * stride + 4], no_rnd_avg32(a, b));
277    }
278}
279
280av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags)
281{
282    c->put_no_rnd_pixels_l2 = put_no_rnd_pixels_l2;
283
284    c->idct_put      = vp3_idct_put_c;
285    c->idct_add      = vp3_idct_add_c;
286    c->idct_dc_add   = vp3_idct_dc_add_c;
287    c->v_loop_filter = vp3_v_loop_filter_c;
288    c->h_loop_filter = vp3_h_loop_filter_c;
289
290    if (ARCH_ARM)
291        ff_vp3dsp_init_arm(c, flags);
292    if (ARCH_PPC)
293        ff_vp3dsp_init_ppc(c, flags);
294    if (ARCH_X86)
295        ff_vp3dsp_init_x86(c, flags);
296}
297