GCC Code Coverage Report
Directory: ../../../ffmpeg/ Exec Total Coverage
File: src/libavcodec/cavs.c Lines: 470 487 96.5 %
Date: 2021-04-20 04:37:23 Branches: 210 231 90.9 %

Line Branch Exec Source
1
/*
2
 * Chinese AVS video (AVS1-P2, JiZhun profile) decoder.
3
 * Copyright (c) 2006  Stefan Gehrer <stefan.gehrer@gmx.de>
4
 *
5
 * This file is part of FFmpeg.
6
 *
7
 * FFmpeg is free software; you can redistribute it and/or
8
 * modify it under the terms of the GNU Lesser General Public
9
 * License as published by the Free Software Foundation; either
10
 * version 2.1 of the License, or (at your option) any later version.
11
 *
12
 * FFmpeg is distributed in the hope that it will be useful,
13
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15
 * Lesser General Public License for more details.
16
 *
17
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with FFmpeg; if not, write to the Free Software
19
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20
 */
21
22
/**
23
 * @file
24
 * Chinese AVS video (AVS1-P2, JiZhun profile) decoder
25
 * @author Stefan Gehrer <stefan.gehrer@gmx.de>
26
 */
27
28
#include "libavutil/mem_internal.h"
29
30
#include "avcodec.h"
31
#include "get_bits.h"
32
#include "golomb.h"
33
#include "h264chroma.h"
34
#include "idctdsp.h"
35
#include "internal.h"
36
#include "mathops.h"
37
#include "qpeldsp.h"
38
#include "cavs.h"
39
40
static const uint8_t alpha_tab[64] = {
41
     0,  0,  0,  0,  0,  0,  1,  1,  1,  1,  1,  2,  2,  2,  3,  3,
42
     4,  4,  5,  5,  6,  7,  8,  9, 10, 11, 12, 13, 15, 16, 18, 20,
43
    22, 24, 26, 28, 30, 33, 33, 35, 35, 36, 37, 37, 39, 39, 42, 44,
44
    46, 48, 50, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64
45
};
46
47
static const uint8_t beta_tab[64] = {
48
     0,  0,  0,  0,  0,  0,  1,  1,  1,  1,  1,  1,  1,  2,  2,  2,
49
     2,  2,  3,  3,  3,  3,  4,  4,  4,  4,  5,  5,  5,  5,  6,  6,
50
     6,  7,  7,  7,  8,  8,  8,  9,  9, 10, 10, 11, 11, 12, 13, 14,
51
    15, 16, 17, 18, 19, 20, 21, 22, 23, 23, 24, 24, 25, 25, 26, 27
52
};
53
54
static const uint8_t tc_tab[64] = {
55
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
56
    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2,
57
    2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4,
58
    5, 5, 5, 6, 6, 6, 7, 7, 7, 7, 8, 8, 8, 9, 9, 9
59
};
60
61
/** mark block as unavailable, i.e. out of picture
62
 *  or not yet decoded */
63
static const cavs_vector un_mv = { 0, 0, 1, NOT_AVAIL };
64
65
static const int8_t left_modifier_l[8] = {  0, -1,  6, -1, -1, 7, 6, 7 };
66
static const int8_t top_modifier_l[8]  = { -1,  1,  5, -1, -1, 5, 7, 7 };
67
static const int8_t left_modifier_c[7] = {  5, -1,  2, -1,  6, 5, 6 };
68
static const int8_t top_modifier_c[7]  = {  4,  1, -1, -1,  4, 6, 6 };
69
70
/*****************************************************************************
71
 *
72
 * in-loop deblocking filter
73
 *
74
 ****************************************************************************/
75
76
1793782
static inline int get_bs(cavs_vector *mvP, cavs_vector *mvQ, int b)
77
{
78

1793782
    if ((mvP->ref == REF_INTRA) || (mvQ->ref == REF_INTRA))
79
15674
        return 2;
80
1778108
    if((abs(mvP->x - mvQ->x) >= 4) ||
81
1673007
       (abs(mvP->y - mvQ->y) >= 4) ||
82
1611712
       (mvP->ref != mvQ->ref))
83
303544
        return 1;
84
1474564
    if (b) {
85
1217805
        mvP += MV_BWD_OFFS;
86
1217805
        mvQ += MV_BWD_OFFS;
87
1217805
        if((abs(mvP->x - mvQ->x) >= 4) ||
88
1203525
           (abs(mvP->y - mvQ->y) >= 4) ||
89
1195077
           (mvP->ref != mvQ->ref))
90
67614
            return 1;
91
    }
92
1406950
    return 0;
93
}
94
95
#define SET_PARAMS                                                \
96
    alpha = alpha_tab[av_clip_uintp2(qp_avg + h->alpha_offset, 6)];  \
97
    beta  =  beta_tab[av_clip_uintp2(qp_avg + h->beta_offset,  6)];  \
98
    tc    =    tc_tab[av_clip_uintp2(qp_avg + h->alpha_offset, 6)];
99
100
/**
101
 * in-loop deblocking filter for a single macroblock
102
 *
103
 * boundary strength (bs) mapping:
104
 *
105
 * --4---5--
106
 * 0   2   |
107
 * | 6 | 7 |
108
 * 1   3   |
109
 * ---------
110
 */
111
297745
void ff_cavs_filter(AVSContext *h, enum cavs_mb mb_type)
112
{
113
    uint8_t bs[8];
114
    int qp_avg, alpha, beta, tc;
115
    int i;
116
117
    /* save un-deblocked lines */
118
297745
    h->topleft_border_y = h->top_border_y[h->mbx * 16 + 15];
119
297745
    h->topleft_border_u = h->top_border_u[h->mbx * 10 + 8];
120
297745
    h->topleft_border_v = h->top_border_v[h->mbx * 10 + 8];
121
297745
    memcpy(&h->top_border_y[h->mbx * 16],     h->cy + 15 * h->l_stride, 16);
122
297745
    memcpy(&h->top_border_u[h->mbx * 10 + 1], h->cu +  7 * h->c_stride, 8);
123
297745
    memcpy(&h->top_border_v[h->mbx * 10 + 1], h->cv +  7 * h->c_stride, 8);
124
2679705
    for (i = 0; i < 8; i++) {
125
2381960
        h->left_border_y[i * 2 + 1] = *(h->cy + 15 + (i * 2 + 0) * h->l_stride);
126
2381960
        h->left_border_y[i * 2 + 2] = *(h->cy + 15 + (i * 2 + 1) * h->l_stride);
127
2381960
        h->left_border_u[i + 1]     = *(h->cu + 7  +  i          * h->c_stride);
128
2381960
        h->left_border_v[i + 1]     = *(h->cv + 7  +  i          * h->c_stride);
129
    }
130
297745
    if (!h->loop_filter_disable) {
131
        /* determine bs */
132
297745
        if (mb_type == I_8X8)
133
25344
            memset(bs, 2, 8);
134
        else {
135
272401
            memset(bs, 0, 8);
136
272401
            if (ff_cavs_partition_flags[mb_type] & SPLITV) {
137
175404
                bs[2] = get_bs(&h->mv[MV_FWD_X0], &h->mv[MV_FWD_X1], mb_type > P_8X8);
138
175404
                bs[3] = get_bs(&h->mv[MV_FWD_X2], &h->mv[MV_FWD_X3], mb_type > P_8X8);
139
            }
140
272401
            if (ff_cavs_partition_flags[mb_type] & SPLITH) {
141
176685
                bs[6] = get_bs(&h->mv[MV_FWD_X0], &h->mv[MV_FWD_X2], mb_type > P_8X8);
142
176685
                bs[7] = get_bs(&h->mv[MV_FWD_X1], &h->mv[MV_FWD_X3], mb_type > P_8X8);
143
            }
144
272401
            bs[0] = get_bs(&h->mv[MV_FWD_A1], &h->mv[MV_FWD_X0], mb_type > P_8X8);
145
272401
            bs[1] = get_bs(&h->mv[MV_FWD_A3], &h->mv[MV_FWD_X2], mb_type > P_8X8);
146
272401
            bs[4] = get_bs(&h->mv[MV_FWD_B2], &h->mv[MV_FWD_X0], mb_type > P_8X8);
147
272401
            bs[5] = get_bs(&h->mv[MV_FWD_B3], &h->mv[MV_FWD_X1], mb_type > P_8X8);
148
        }
149
297745
        if (AV_RN64(bs)) {
150
134358
            if (h->flags & A_AVAIL) {
151
127741
                qp_avg = (h->qp + h->left_qp + 1) >> 1;
152
127741
                SET_PARAMS;
153
127741
                h->cdsp.cavs_filter_lv(h->cy, h->l_stride, alpha, beta, tc, bs[0], bs[1]);
154
127741
                qp_avg = (ff_cavs_chroma_qp[h->qp] + ff_cavs_chroma_qp[h->left_qp] + 1) >> 1;
155
127741
                SET_PARAMS;
156
127741
                h->cdsp.cavs_filter_cv(h->cu, h->c_stride, alpha, beta, tc, bs[0], bs[1]);
157
127741
                h->cdsp.cavs_filter_cv(h->cv, h->c_stride, alpha, beta, tc, bs[0], bs[1]);
158
            }
159
134358
            qp_avg = h->qp;
160
134358
            SET_PARAMS;
161
134358
            h->cdsp.cavs_filter_lv(h->cy + 8,               h->l_stride, alpha, beta, tc, bs[2], bs[3]);
162
134358
            h->cdsp.cavs_filter_lh(h->cy + 8 * h->l_stride, h->l_stride, alpha, beta, tc, bs[6], bs[7]);
163
164
134358
            if (h->flags & B_AVAIL) {
165
126078
                qp_avg = (h->qp + h->top_qp[h->mbx] + 1) >> 1;
166
126078
                SET_PARAMS;
167
126078
                h->cdsp.cavs_filter_lh(h->cy, h->l_stride, alpha, beta, tc, bs[4], bs[5]);
168
126078
                qp_avg = (ff_cavs_chroma_qp[h->qp] + ff_cavs_chroma_qp[h->top_qp[h->mbx]] + 1) >> 1;
169
126078
                SET_PARAMS;
170
126078
                h->cdsp.cavs_filter_ch(h->cu, h->c_stride, alpha, beta, tc, bs[4], bs[5]);
171
126078
                h->cdsp.cavs_filter_ch(h->cv, h->c_stride, alpha, beta, tc, bs[4], bs[5]);
172
            }
173
        }
174
    }
175
297745
    h->left_qp        = h->qp;
176
297745
    h->top_qp[h->mbx] = h->qp;
177
297745
}
178
179
#undef SET_PARAMS
180
181
/*****************************************************************************
182
 *
183
 * spatial intra prediction
184
 *
185
 ****************************************************************************/
186
187
101376
void ff_cavs_load_intra_pred_luma(AVSContext *h, uint8_t *top,
188
                                  uint8_t **left, int block)
189
{
190
    int i;
191
192

101376
    switch (block) {
193
25344
    case 0:
194
25344
        *left               = h->left_border_y;
195
25344
        h->left_border_y[0] = h->left_border_y[1];
196
25344
        memset(&h->left_border_y[17], h->left_border_y[16], 9);
197
25344
        memcpy(&top[1], &h->top_border_y[h->mbx * 16], 16);
198
25344
        top[17] = top[16];
199
25344
        top[0]  = top[1];
200

25344
        if ((h->flags & A_AVAIL) && (h->flags & B_AVAIL))
201
24410
            h->left_border_y[0] = top[0] = h->topleft_border_y;
202
25344
        break;
203
25344
    case 1:
204
25344
        *left = h->intern_border_y;
205
228096
        for (i = 0; i < 8; i++)
206
202752
            h->intern_border_y[i + 1] = *(h->cy + 7 + i * h->l_stride);
207
25344
        memset(&h->intern_border_y[9], h->intern_border_y[8], 9);
208
25344
        h->intern_border_y[0] = h->intern_border_y[1];
209
25344
        memcpy(&top[1], &h->top_border_y[h->mbx * 16 + 8], 8);
210
25344
        if (h->flags & C_AVAIL)
211
24175
            memcpy(&top[9], &h->top_border_y[(h->mbx + 1) * 16], 8);
212
        else
213
1169
            memset(&top[9], top[8], 9);
214
25344
        top[17] = top[16];
215
25344
        top[0]  = top[1];
216
25344
        if (h->flags & B_AVAIL)
217
24808
            h->intern_border_y[0] = top[0] = h->top_border_y[h->mbx * 16 + 7];
218
25344
        break;
219
25344
    case 2:
220
25344
        *left = &h->left_border_y[8];
221
25344
        memcpy(&top[1], h->cy + 7 * h->l_stride, 16);
222
25344
        top[17] = top[16];
223
25344
        top[0]  = top[1];
224
25344
        if (h->flags & A_AVAIL)
225
24937
            top[0] = h->left_border_y[8];
226
25344
        break;
227
25344
    case 3:
228
25344
        *left = &h->intern_border_y[8];
229
228096
        for (i = 0; i < 8; i++)
230
202752
            h->intern_border_y[i + 9] = *(h->cy + 7 + (i + 8) * h->l_stride);
231
25344
        memset(&h->intern_border_y[17], h->intern_border_y[16], 9);
232
25344
        memcpy(&top[0], h->cy + 7 + 7 * h->l_stride, 9);
233
25344
        memset(&top[9], top[8], 9);
234
25344
        break;
235
    }
236
101376
}
237
238
25344
void ff_cavs_load_intra_pred_chroma(AVSContext *h)
239
{
240
    /* extend borders by one pixel */
241
25344
    h->left_border_u[9]              = h->left_border_u[8];
242
25344
    h->left_border_v[9]              = h->left_border_v[8];
243
25344
    if(h->flags & C_AVAIL) {
244
24175
        h->top_border_u[h->mbx*10 + 9] = h->top_border_u[h->mbx*10 + 11];
245
24175
        h->top_border_v[h->mbx*10 + 9] = h->top_border_v[h->mbx*10 + 11];
246
    } else {
247
1169
        h->top_border_u[h->mbx * 10 + 9] = h->top_border_u[h->mbx * 10 + 8];
248
1169
        h->top_border_v[h->mbx * 10 + 9] = h->top_border_v[h->mbx * 10 + 8];
249
    }
250

25344
    if((h->flags & A_AVAIL) && (h->flags & B_AVAIL)) {
251
24410
        h->top_border_u[h->mbx * 10] = h->left_border_u[0] = h->topleft_border_u;
252
24410
        h->top_border_v[h->mbx * 10] = h->left_border_v[0] = h->topleft_border_v;
253
    } else {
254
934
        h->left_border_u[0]          = h->left_border_u[1];
255
934
        h->left_border_v[0]          = h->left_border_v[1];
256
934
        h->top_border_u[h->mbx * 10] = h->top_border_u[h->mbx * 10 + 1];
257
934
        h->top_border_v[h->mbx * 10] = h->top_border_v[h->mbx * 10 + 1];
258
    }
259
25344
}
260
261
45576
static void intra_pred_vert(uint8_t *d, uint8_t *top, uint8_t *left, ptrdiff_t stride)
262
{
263
    int y;
264
45576
    uint64_t a = AV_RN64(&top[1]);
265
410184
    for (y = 0; y < 8; y++)
266
364608
        *((uint64_t *)(d + y * stride)) = a;
267
45576
}
268
269
33618
static void intra_pred_horiz(uint8_t *d, uint8_t *top, uint8_t *left, ptrdiff_t stride)
270
{
271
    int y;
272
    uint64_t a;
273
302562
    for (y = 0; y < 8; y++) {
274
268944
        a = left[y + 1] * 0x0101010101010101ULL;
275
268944
        *((uint64_t *)(d + y * stride)) = a;
276
    }
277
33618
}
278
279
27
static void intra_pred_dc_128(uint8_t *d, uint8_t *top, uint8_t *left, ptrdiff_t stride)
280
{
281
    int y;
282
27
    uint64_t a = 0x8080808080808080ULL;
283
243
    for (y = 0; y < 8; y++)
284
216
        *((uint64_t *)(d + y * stride)) = a;
285
27
}
286
287
2756
static void intra_pred_plane(uint8_t *d, uint8_t *top, uint8_t *left, ptrdiff_t stride)
288
{
289
    int x, y, ia;
290
2756
    int ih = 0;
291
2756
    int iv = 0;
292
2756
    const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
293
294
13780
    for (x = 0; x < 4; x++) {
295
11024
        ih += (x + 1) *  (top[5 + x] -  top[3 - x]);
296
11024
        iv += (x + 1) * (left[5 + x] - left[3 - x]);
297
    }
298
2756
    ia = (top[8] + left[8]) << 4;
299
2756
    ih = (17 * ih + 16) >> 5;
300
2756
    iv = (17 * iv + 16) >> 5;
301
24804
    for (y = 0; y < 8; y++)
302
198432
        for (x = 0; x < 8; x++)
303
176384
            d[y * stride + x] = cm[(ia + (x - 3) * ih + (y - 3) * iv + 16) >> 5];
304
2756
}
305
306
#define LOWPASS(ARRAY, INDEX)                                           \
307
    ((ARRAY[(INDEX) - 1] + 2 * ARRAY[(INDEX)] + ARRAY[(INDEX) + 1] + 2) >> 2)
308
309
61656
static void intra_pred_lp(uint8_t *d, uint8_t *top, uint8_t *left, ptrdiff_t stride)
310
{
311
    int x, y;
312
554904
    for (y = 0; y < 8; y++)
313
4439232
        for (x = 0; x < 8; x++)
314
3945984
            d[y * stride + x] = (LOWPASS(top, x + 1) + LOWPASS(left, y + 1)) >> 1;
315
61656
}
316
317
2887
static void intra_pred_down_left(uint8_t *d, uint8_t *top, uint8_t *left, ptrdiff_t stride)
318
{
319
    int x, y;
320
25983
    for (y = 0; y < 8; y++)
321
207864
        for (x = 0; x < 8; x++)
322
184768
            d[y * stride + x] = (LOWPASS(top, x + y + 2) + LOWPASS(left, x + y + 2)) >> 1;
323
2887
}
324
325
1826
static void intra_pred_down_right(uint8_t *d, uint8_t *top, uint8_t *left, ptrdiff_t stride)
326
{
327
    int x, y;
328
16434
    for (y = 0; y < 8; y++)
329
131472
        for (x = 0; x < 8; x++)
330
116864
            if (x == y)
331
14608
                d[y * stride + x] = (left[1] + 2 * top[0] + top[1] + 2) >> 2;
332
102256
            else if (x > y)
333
51128
                d[y * stride + x] = LOWPASS(top, x - y);
334
            else
335
51128
                d[y * stride + x] = LOWPASS(left, y - x);
336
1826
}
337
338
2117
static void intra_pred_lp_left(uint8_t *d, uint8_t *top, uint8_t *left, ptrdiff_t stride)
339
{
340
    int x, y;
341
19053
    for (y = 0; y < 8; y++)
342
152424
        for (x = 0; x < 8; x++)
343
135488
            d[y * stride + x] = LOWPASS(left, y + 1);
344
2117
}
345
346
1601
static void intra_pred_lp_top(uint8_t *d, uint8_t *top, uint8_t *left, ptrdiff_t stride)
347
{
348
    int x, y;
349
14409
    for (y = 0; y < 8; y++)
350
115272
        for (x = 0; x < 8; x++)
351
102464
            d[y * stride + x] = LOWPASS(top, x + 1);
352
1601
}
353
354
#undef LOWPASS
355
356
2829
static inline void modify_pred(const int8_t *mod_table, int *mode)
357
{
358
2829
    *mode = mod_table[*mode];
359
2829
    if (*mode < 0) {
360
        av_log(NULL, AV_LOG_ERROR, "Illegal intra prediction mode\n");
361
        *mode = 0;
362
    }
363
2829
}
364
365
25344
void ff_cavs_modify_mb_i(AVSContext *h, int *pred_mode_uv)
366
{
367
    /* save pred modes before they get modified */
368
25344
    h->pred_mode_Y[3]             = h->pred_mode_Y[5];
369
25344
    h->pred_mode_Y[6]             = h->pred_mode_Y[8];
370
25344
    h->top_pred_Y[h->mbx * 2 + 0] = h->pred_mode_Y[7];
371
25344
    h->top_pred_Y[h->mbx * 2 + 1] = h->pred_mode_Y[8];
372
373
    /* modify pred modes according to availability of neighbour samples */
374
25344
    if (!(h->flags & A_AVAIL)) {
375
407
        modify_pred(left_modifier_l, &h->pred_mode_Y[4]);
376
407
        modify_pred(left_modifier_l, &h->pred_mode_Y[7]);
377
407
        modify_pred(left_modifier_c, pred_mode_uv);
378
    }
379
25344
    if (!(h->flags & B_AVAIL)) {
380
536
        modify_pred(top_modifier_l, &h->pred_mode_Y[4]);
381
536
        modify_pred(top_modifier_l, &h->pred_mode_Y[5]);
382
536
        modify_pred(top_modifier_c, pred_mode_uv);
383
    }
384
25344
}
385
386
/*****************************************************************************
387
 *
388
 * motion compensation
389
 *
390
 ****************************************************************************/
391
392
1448997
static inline void mc_dir_part(AVSContext *h, AVFrame *pic, int chroma_height,
393
                               int delta, int list, uint8_t *dest_y,
394
                               uint8_t *dest_cb, uint8_t *dest_cr,
395
                               int src_x_offset, int src_y_offset,
396
                               qpel_mc_func *qpix_op,
397
                               h264_chroma_mc_func chroma_op, cavs_vector *mv)
398
{
399
1448997
    const int mx         = mv->x + src_x_offset * 8;
400
1448997
    const int my         = mv->y + src_y_offset * 8;
401
1448997
    const int luma_xy    = (mx & 3) + ((my & 3) << 2);
402
1448997
    uint8_t *src_y       = pic->data[0] + (mx >> 2) + (my >> 2) * h->l_stride;
403
1448997
    uint8_t *src_cb      = pic->data[1] + (mx >> 3) + (my >> 3) * h->c_stride;
404
1448997
    uint8_t *src_cr      = pic->data[2] + (mx >> 3) + (my >> 3) * h->c_stride;
405
1448997
    int extra_width      = 0;
406
1448997
    int extra_height     = extra_width;
407
1448997
    const int full_mx    = mx >> 2;
408
1448997
    const int full_my    = my >> 2;
409
1448997
    const int pic_width  = 16 * h->mb_width;
410
1448997
    const int pic_height = 16 * h->mb_height;
411
1448997
    int emu = 0;
412
413
1448997
    if (!pic->data[0])
414
        return;
415
1448997
    if (mx & 7)
416
120702
        extra_width  -= 3;
417
1448997
    if (my & 7)
418
125744
        extra_height -= 3;
419
420
1448997
    if (full_mx < 0 - extra_width ||
421
1448061
        full_my < 0 - extra_height ||
422
1446581
        full_mx + 16 /* FIXME */ > pic_width + extra_width ||
423
1431388
        full_my + 16 /* FIXME */ > pic_height + extra_height) {
424
37368
        h->vdsp.emulated_edge_mc(h->edge_emu_buffer,
425
37368
                                 src_y - 2 - 2 * h->l_stride,
426
                                 h->l_stride, h->l_stride,
427
                                 16 + 5, 16 + 5 /* FIXME */,
428
                                 full_mx - 2, full_my - 2,
429
                                 pic_width, pic_height);
430
37368
        src_y = h->edge_emu_buffer + 2 + 2 * h->l_stride;
431
37368
        emu   = 1;
432
    }
433
434
    // FIXME try variable height perhaps?
435
1448997
    qpix_op[luma_xy](dest_y, src_y, h->l_stride);
436
437
1448997
    if (emu) {
438
37368
        h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cb,
439
                                 h->c_stride, h->c_stride,
440
                                 9, 9 /* FIXME */,
441
                                 mx >> 3, my >> 3,
442
                                 pic_width >> 1, pic_height >> 1);
443
37368
        src_cb = h->edge_emu_buffer;
444
    }
445
1448997
    chroma_op(dest_cb, src_cb, h->c_stride, chroma_height, mx & 7, my & 7);
446
447
1448997
    if (emu) {
448
37368
        h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cr,
449
                                 h->c_stride, h->c_stride,
450
                                 9, 9 /* FIXME */,
451
                                 mx >> 3, my >> 3,
452
                                 pic_width >> 1, pic_height >> 1);
453
37368
        src_cr = h->edge_emu_buffer;
454
    }
455
1448997
    chroma_op(dest_cr, src_cr, h->c_stride, chroma_height, mx & 7, my & 7);
456
}
457
458
832240
static inline void mc_part_std(AVSContext *h, int chroma_height, int delta,
459
                               uint8_t *dest_y,
460
                               uint8_t *dest_cb,
461
                               uint8_t *dest_cr,
462
                               int x_offset, int y_offset,
463
                               qpel_mc_func *qpix_put,
464
                               h264_chroma_mc_func chroma_put,
465
                               qpel_mc_func *qpix_avg,
466
                               h264_chroma_mc_func chroma_avg,
467
                               cavs_vector *mv)
468
{
469
832240
    qpel_mc_func *qpix_op =  qpix_put;
470
832240
    h264_chroma_mc_func chroma_op = chroma_put;
471
472
832240
    dest_y   += x_offset * 2 + y_offset * h->l_stride * 2;
473
832240
    dest_cb  += x_offset     + y_offset * h->c_stride;
474
832240
    dest_cr  += x_offset     + y_offset * h->c_stride;
475
832240
    x_offset += 8 * h->mbx;
476
832240
    y_offset += 8 * h->mby;
477
478
832240
    if (mv->ref >= 0) {
479
788976
        AVFrame *ref = h->DPB[mv->ref].f;
480
788976
        mc_dir_part(h, ref, chroma_height, delta, 0,
481
                    dest_y, dest_cb, dest_cr, x_offset, y_offset,
482
                    qpix_op, chroma_op, mv);
483
484
788976
        qpix_op   = qpix_avg;
485
788976
        chroma_op = chroma_avg;
486
    }
487
488
832240
    if ((mv + MV_BWD_OFFS)->ref >= 0) {
489
660021
        AVFrame *ref = h->DPB[0].f;
490
660021
        mc_dir_part(h, ref, chroma_height, delta, 1,
491
                    dest_y, dest_cb, dest_cr, x_offset, y_offset,
492
                    qpix_op, chroma_op, mv + MV_BWD_OFFS);
493
    }
494
832240
}
495
496
272401
void ff_cavs_inter(AVSContext *h, enum cavs_mb mb_type)
497
{
498
272401
    if (ff_cavs_partition_flags[mb_type] == 0) { // 16x16
499
85788
        mc_part_std(h, 8, 0, h->cy, h->cu, h->cv, 0, 0,
500
85788
                    h->cdsp.put_cavs_qpel_pixels_tab[0],
501
                    h->h264chroma.put_h264_chroma_pixels_tab[0],
502
85788
                    h->cdsp.avg_cavs_qpel_pixels_tab[0],
503
                    h->h264chroma.avg_h264_chroma_pixels_tab[0],
504
                    &h->mv[MV_FWD_X0]);
505
    } else {
506
186613
        mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 0, 0,
507
186613
                    h->cdsp.put_cavs_qpel_pixels_tab[1],
508
                    h->h264chroma.put_h264_chroma_pixels_tab[1],
509
186613
                    h->cdsp.avg_cavs_qpel_pixels_tab[1],
510
                    h->h264chroma.avg_h264_chroma_pixels_tab[1],
511
                    &h->mv[MV_FWD_X0]);
512
186613
        mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 4, 0,
513
186613
                    h->cdsp.put_cavs_qpel_pixels_tab[1],
514
                    h->h264chroma.put_h264_chroma_pixels_tab[1],
515
186613
                    h->cdsp.avg_cavs_qpel_pixels_tab[1],
516
                    h->h264chroma.avg_h264_chroma_pixels_tab[1],
517
                    &h->mv[MV_FWD_X1]);
518
186613
        mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 0, 4,
519
186613
                    h->cdsp.put_cavs_qpel_pixels_tab[1],
520
                    h->h264chroma.put_h264_chroma_pixels_tab[1],
521
186613
                    h->cdsp.avg_cavs_qpel_pixels_tab[1],
522
                    h->h264chroma.avg_h264_chroma_pixels_tab[1],
523
                    &h->mv[MV_FWD_X2]);
524
186613
        mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 4, 4,
525
186613
                    h->cdsp.put_cavs_qpel_pixels_tab[1],
526
                    h->h264chroma.put_h264_chroma_pixels_tab[1],
527
186613
                    h->cdsp.avg_cavs_qpel_pixels_tab[1],
528
                    h->h264chroma.avg_h264_chroma_pixels_tab[1],
529
                    &h->mv[MV_FWD_X3]);
530
    }
531
272401
}
532
533
/*****************************************************************************
534
 *
535
 * motion vector prediction
536
 *
537
 ****************************************************************************/
538
539
510849
static inline void scale_mv(AVSContext *h, int *d_x, int *d_y,
540
                            cavs_vector *src, int distp)
541
{
542
510849
    int64_t den = h->scale_den[FFMAX(src->ref, 0)];
543
510849
    *d_x = (src->x * distp * den + 256 + FF_SIGNBIT(src->x)) >> 9;
544
510849
    *d_y = (src->y * distp * den + 256 + FF_SIGNBIT(src->y)) >> 9;
545
510849
}
546
547
170283
static inline void mv_pred_median(AVSContext *h,
548
                                  cavs_vector *mvP,
549
                                  cavs_vector *mvA,
550
                                  cavs_vector *mvB,
551
                                  cavs_vector *mvC)
552
{
553
    int ax, ay, bx, by, cx, cy;
554
    int len_ab, len_bc, len_ca, len_mid;
555
556
    /* scale candidates according to their temporal span */
557
170283
    scale_mv(h, &ax, &ay, mvA, mvP->dist);
558
170283
    scale_mv(h, &bx, &by, mvB, mvP->dist);
559
170283
    scale_mv(h, &cx, &cy, mvC, mvP->dist);
560
    /* find the geometrical median of the three candidates */
561
170283
    len_ab  = abs(ax - bx) + abs(ay - by);
562
170283
    len_bc  = abs(bx - cx) + abs(by - cy);
563
170283
    len_ca  = abs(cx - ax) + abs(cy - ay);
564
170283
    len_mid = mid_pred(len_ab, len_bc, len_ca);
565
170283
    if (len_mid == len_ab) {
566
137907
        mvP->x = cx;
567
137907
        mvP->y = cy;
568
32376
    } else if (len_mid == len_bc) {
569
24630
        mvP->x = ax;
570
24630
        mvP->y = ay;
571
    } else {
572
7746
        mvP->x = bx;
573
7746
        mvP->y = by;
574
    }
575
170283
}
576
577
238918
void ff_cavs_mv(AVSContext *h, enum cavs_mv_loc nP, enum cavs_mv_loc nC,
578
                enum cavs_mv_pred mode, enum cavs_block size, int ref)
579
{
580
238918
    cavs_vector *mvP = &h->mv[nP];
581
238918
    cavs_vector *mvA = &h->mv[nP-1];
582
238918
    cavs_vector *mvB = &h->mv[nP-4];
583
238918
    cavs_vector *mvC = &h->mv[nC];
584
238918
    const cavs_vector *mvP2 = NULL;
585
586
238918
    mvP->ref  = ref;
587
238918
    mvP->dist = h->dist[mvP->ref];
588

238918
    if (mvC->ref == NOT_AVAIL || (nP == MV_FWD_X3) || (nP == MV_BWD_X3 ))
589
28240
        mvC = &h->mv[nP - 5];  // set to top-left (mvD)
590
238918
    if (mode == MV_PRED_PSKIP &&
591
20801
        (mvA->ref == NOT_AVAIL ||
592
20081
         mvB->ref == NOT_AVAIL ||
593
19308
         (mvA->x | mvA->y | mvA->ref) == 0 ||
594
850
         (mvB->x | mvB->y | mvB->ref) == 0)) {
595
20434
        mvP2 = &un_mv;
596
    /* if there is only one suitable candidate, take it */
597

218484
    } else if (mvA->ref >= 0 && mvB->ref < 0  && mvC->ref < 0) {
598
11224
        mvP2 = mvA;
599

207260
    } else if (mvA->ref < 0  && mvB->ref >= 0 && mvC->ref < 0) {
600
4577
        mvP2 = mvB;
601

202683
    } else if (mvA->ref < 0  && mvB->ref < 0  && mvC->ref >= 0) {
602
3525
        mvP2 = mvC;
603

199158
    } else if (mode == MV_PRED_LEFT     && mvA->ref == ref) {
604
14269
        mvP2 = mvA;
605

184889
    } else if (mode == MV_PRED_TOP      && mvB->ref == ref) {
606
7898
        mvP2 = mvB;
607

176991
    } else if (mode == MV_PRED_TOPRIGHT && mvC->ref == ref) {
608
6708
        mvP2 = mvC;
609
    }
610
238918
    if (mvP2) {
611
68635
        mvP->x = mvP2->x;
612
68635
        mvP->y = mvP2->y;
613
    } else
614
170283
        mv_pred_median(h, mvP, mvA, mvB, mvC);
615
616
238918
    if (mode < MV_PRED_PSKIP) {
617
189273
        int mx = get_se_golomb(&h->gb) + (unsigned)mvP->x;
618
189273
        int my = get_se_golomb(&h->gb) + (unsigned)mvP->y;
619
620

189273
        if (mx != (int16_t)mx || my != (int16_t)my) {
621
            av_log(h->avctx, AV_LOG_ERROR, "MV %d %d out of supported range\n", mx, my);
622
        } else {
623
189273
            mvP->x = mx;
624
189273
            mvP->y = my;
625
        }
626
    }
627
238918
    set_mvs(mvP, size);
628
238918
}
629
630
/*****************************************************************************
631
 *
632
 * macroblock level
633
 *
634
 ****************************************************************************/
635
636
/**
637
 * initialise predictors for motion vectors and intra prediction
638
 */
639
297745
void ff_cavs_init_mb(AVSContext *h)
640
{
641
    int i;
642
643
    /* copy predictors from top line (MB B and C) into cache */
644
1190980
    for (i = 0; i < 3; i++) {
645
893235
        h->mv[MV_FWD_B2 + i] = h->top_mv[0][h->mbx * 2 + i];
646
893235
        h->mv[MV_BWD_B2 + i] = h->top_mv[1][h->mbx * 2 + i];
647
    }
648
297745
    h->pred_mode_Y[1] = h->top_pred_Y[h->mbx * 2 + 0];
649
297745
    h->pred_mode_Y[2] = h->top_pred_Y[h->mbx * 2 + 1];
650
    /* clear top predictors if MB B is not available */
651
297745
    if (!(h->flags & B_AVAIL)) {
652
8280
        h->mv[MV_FWD_B2]  = un_mv;
653
8280
        h->mv[MV_FWD_B3]  = un_mv;
654
8280
        h->mv[MV_BWD_B2]  = un_mv;
655
8280
        h->mv[MV_BWD_B3]  = un_mv;
656
8280
        h->pred_mode_Y[1] = h->pred_mode_Y[2] = NOT_AVAIL;
657
8280
        h->flags         &= ~(C_AVAIL | D_AVAIL);
658
289465
    } else if (h->mbx) {
659
283032
        h->flags |= D_AVAIL;
660
    }
661
297745
    if (h->mbx == h->mb_width - 1) // MB C not available
662
6616
        h->flags &= ~C_AVAIL;
663
    /* clear top-right predictors if MB C is not available */
664
297745
    if (!(h->flags & C_AVAIL)) {
665
14712
        h->mv[MV_FWD_C2] = un_mv;
666
14712
        h->mv[MV_BWD_C2] = un_mv;
667
    }
668
    /* clear top-left predictors if MB D is not available */
669
297745
    if (!(h->flags & D_AVAIL)) {
670
14713
        h->mv[MV_FWD_D3] = un_mv;
671
14713
        h->mv[MV_BWD_D3] = un_mv;
672
    }
673
297745
}
674
675
/**
676
 * save predictors for later macroblocks and increase
677
 * macroblock address
678
 * @return 0 if end of frame is reached, 1 otherwise
679
 */
680
297745
int ff_cavs_next_mb(AVSContext *h)
681
{
682
    int i;
683
684
297745
    h->flags |= A_AVAIL;
685
297745
    h->cy    += 16;
686
297745
    h->cu    += 8;
687
297745
    h->cv    += 8;
688
    /* copy mvs as predictors to the left */
689
2084215
    for (i = 0; i <= 20; i += 4)
690
1786470
        h->mv[i] = h->mv[i + 2];
691
    /* copy bottom mvs from cache to top line */
692
297745
    h->top_mv[0][h->mbx * 2 + 0] = h->mv[MV_FWD_X2];
693
297745
    h->top_mv[0][h->mbx * 2 + 1] = h->mv[MV_FWD_X3];
694
297745
    h->top_mv[1][h->mbx * 2 + 0] = h->mv[MV_BWD_X2];
695
297745
    h->top_mv[1][h->mbx * 2 + 1] = h->mv[MV_BWD_X3];
696
    /* next MB address */
697
297745
    h->mbidx++;
698
297745
    h->mbx++;
699
297745
    if (h->mbx == h->mb_width) { // New mb line
700
6616
        h->flags = B_AVAIL | C_AVAIL;
701
        /* clear left pred_modes */
702
6616
        h->pred_mode_Y[3] = h->pred_mode_Y[6] = NOT_AVAIL;
703
        /* clear left mv predictors */
704
46312
        for (i = 0; i <= 20; i += 4)
705
39696
            h->mv[i] = un_mv;
706
6616
        h->mbx = 0;
707
6616
        h->mby++;
708
        /* re-calculate sample pointers */
709
6616
        h->cy = h->cur.f->data[0] + h->mby * 16 * h->l_stride;
710
6616
        h->cu = h->cur.f->data[1] + h->mby * 8 * h->c_stride;
711
6616
        h->cv = h->cur.f->data[2] + h->mby * 8 * h->c_stride;
712
6616
        if (h->mby == h->mb_height) { // Frame end
713
183
            return 0;
714
        }
715
    }
716
297562
    return 1;
717
}
718
719
/*****************************************************************************
720
 *
721
 * frame level
722
 *
723
 ****************************************************************************/
724
725
184
int ff_cavs_init_pic(AVSContext *h)
726
{
727
    int i;
728
729
    /* clear some predictors */
730
1288
    for (i = 0; i <= 20; i += 4)
731
1104
        h->mv[i] = un_mv;
732
184
    h->mv[MV_BWD_X0] = ff_cavs_dir_mv;
733
184
    set_mvs(&h->mv[MV_BWD_X0], BLK_16X16);
734
184
    h->mv[MV_FWD_X0] = ff_cavs_dir_mv;
735
184
    set_mvs(&h->mv[MV_FWD_X0], BLK_16X16);
736
184
    h->pred_mode_Y[3] = h->pred_mode_Y[6] = NOT_AVAIL;
737
184
    h->cy             = h->cur.f->data[0];
738
184
    h->cu             = h->cur.f->data[1];
739
184
    h->cv             = h->cur.f->data[2];
740
184
    h->l_stride       = h->cur.f->linesize[0];
741
184
    h->c_stride       = h->cur.f->linesize[1];
742
184
    h->luma_scan[2]   = 8 * h->l_stride;
743
184
    h->luma_scan[3]   = 8 * h->l_stride + 8;
744
184
    h->mbx            = h->mby = h->mbidx = 0;
745
184
    h->flags          = 0;
746
747
184
    return 0;
748
}
749
750
/*****************************************************************************
751
 *
752
 * headers and interface
753
 *
754
 ****************************************************************************/
755
756
/**
757
 * some predictions require data from the top-neighbouring macroblock.
758
 * this data has to be stored for one complete row of macroblocks
759
 * and this storage space is allocated here
760
 */
761
4
int ff_cavs_init_top_lines(AVSContext *h)
762
{
763
    /* alloc top line of predictors */
764
4
    h->top_qp       = av_mallocz(h->mb_width);
765
4
    h->top_mv[0]    = av_mallocz_array(h->mb_width * 2 + 1,  sizeof(cavs_vector));
766
4
    h->top_mv[1]    = av_mallocz_array(h->mb_width * 2 + 1,  sizeof(cavs_vector));
767
4
    h->top_pred_Y   = av_mallocz_array(h->mb_width * 2,  sizeof(*h->top_pred_Y));
768
4
    h->top_border_y = av_mallocz_array(h->mb_width + 1,  16);
769
4
    h->top_border_u = av_mallocz_array(h->mb_width,  10);
770
4
    h->top_border_v = av_mallocz_array(h->mb_width,  10);
771
772
    /* alloc space for co-located MVs and types */
773
4
    h->col_mv        = av_mallocz_array(h->mb_width * h->mb_height,
774
                                        4 * sizeof(cavs_vector));
775
4
    h->col_type_base = av_mallocz(h->mb_width * h->mb_height);
776
4
    h->block         = av_mallocz(64 * sizeof(int16_t));
777
778


4
    if (!h->top_qp || !h->top_mv[0] || !h->top_mv[1] || !h->top_pred_Y ||
779

4
        !h->top_border_y || !h->top_border_u || !h->top_border_v ||
780

4
        !h->col_mv || !h->col_type_base || !h->block) {
781
        av_freep(&h->top_qp);
782
        av_freep(&h->top_mv[0]);
783
        av_freep(&h->top_mv[1]);
784
        av_freep(&h->top_pred_Y);
785
        av_freep(&h->top_border_y);
786
        av_freep(&h->top_border_u);
787
        av_freep(&h->top_border_v);
788
        av_freep(&h->col_mv);
789
        av_freep(&h->col_type_base);
790
        av_freep(&h->block);
791
        return AVERROR(ENOMEM);
792
    }
793
4
    return 0;
794
}
795
796
4
av_cold int ff_cavs_init(AVCodecContext *avctx)
797
{
798
4
    AVSContext *h = avctx->priv_data;
799
800
4
    ff_blockdsp_init(&h->bdsp, avctx);
801
4
    ff_h264chroma_init(&h->h264chroma, 8);
802
4
    ff_idctdsp_init(&h->idsp, avctx);
803
4
    ff_videodsp_init(&h->vdsp, 8);
804
4
    ff_cavsdsp_init(&h->cdsp, avctx);
805
4
    ff_init_scantable_permutation(h->idsp.idct_permutation,
806
4
                                  h->cdsp.idct_perm);
807
4
    ff_init_scantable(h->idsp.idct_permutation, &h->scantable, ff_zigzag_direct);
808
809
4
    h->avctx       = avctx;
810
4
    avctx->pix_fmt = AV_PIX_FMT_YUV420P;
811
812
4
    h->cur.f    = av_frame_alloc();
813
4
    h->DPB[0].f = av_frame_alloc();
814
4
    h->DPB[1].f = av_frame_alloc();
815

4
    if (!h->cur.f || !h->DPB[0].f || !h->DPB[1].f) {
816
        ff_cavs_end(avctx);
817
        return AVERROR(ENOMEM);
818
    }
819
820
4
    h->luma_scan[0]                     = 0;
821
4
    h->luma_scan[1]                     = 8;
822
4
    h->intra_pred_l[INTRA_L_VERT]       = intra_pred_vert;
823
4
    h->intra_pred_l[INTRA_L_HORIZ]      = intra_pred_horiz;
824
4
    h->intra_pred_l[INTRA_L_LP]         = intra_pred_lp;
825
4
    h->intra_pred_l[INTRA_L_DOWN_LEFT]  = intra_pred_down_left;
826
4
    h->intra_pred_l[INTRA_L_DOWN_RIGHT] = intra_pred_down_right;
827
4
    h->intra_pred_l[INTRA_L_LP_LEFT]    = intra_pred_lp_left;
828
4
    h->intra_pred_l[INTRA_L_LP_TOP]     = intra_pred_lp_top;
829
4
    h->intra_pred_l[INTRA_L_DC_128]     = intra_pred_dc_128;
830
4
    h->intra_pred_c[INTRA_C_LP]         = intra_pred_lp;
831
4
    h->intra_pred_c[INTRA_C_HORIZ]      = intra_pred_horiz;
832
4
    h->intra_pred_c[INTRA_C_VERT]       = intra_pred_vert;
833
4
    h->intra_pred_c[INTRA_C_PLANE]      = intra_pred_plane;
834
4
    h->intra_pred_c[INTRA_C_LP_LEFT]    = intra_pred_lp_left;
835
4
    h->intra_pred_c[INTRA_C_LP_TOP]     = intra_pred_lp_top;
836
4
    h->intra_pred_c[INTRA_C_DC_128]     = intra_pred_dc_128;
837
4
    h->mv[7]                            = un_mv;
838
4
    h->mv[19]                           = un_mv;
839
4
    return 0;
840
}
841
842
4
av_cold int ff_cavs_end(AVCodecContext *avctx)
843
{
844
4
    AVSContext *h = avctx->priv_data;
845
846
4
    av_frame_free(&h->cur.f);
847
4
    av_frame_free(&h->DPB[0].f);
848
4
    av_frame_free(&h->DPB[1].f);
849
850
4
    av_freep(&h->top_qp);
851
4
    av_freep(&h->top_mv[0]);
852
4
    av_freep(&h->top_mv[1]);
853
4
    av_freep(&h->top_pred_Y);
854
4
    av_freep(&h->top_border_y);
855
4
    av_freep(&h->top_border_u);
856
4
    av_freep(&h->top_border_v);
857
4
    av_freep(&h->col_mv);
858
4
    av_freep(&h->col_type_base);
859
4
    av_freep(&h->block);
860
4
    av_freep(&h->edge_emu_buffer);
861
4
    return 0;
862
}