h264_mb_template.c 16.7 KB
Newer Older
1 2 3 4
/*
 * H.26L/H.264/AVC/JVT/14496-10/... decoder
 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
 *
5
 * This file is part of FFmpeg.
6
 *
7
 * FFmpeg is free software; you can redistribute it and/or
8 9 10 11
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
12
 * FFmpeg is distributed in the hope that it will be useful,
13 14 15 16 17
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * 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 20 21
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

22 23
#include "svq3.h"

24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42
#undef FUNC
#undef PIXEL_SHIFT

#if SIMPLE
#   define FUNC(n) AV_JOIN(n ## _simple_, BITS)
#   define PIXEL_SHIFT (BITS >> 4)
#else
#   define FUNC(n) n ## _complex
#   define PIXEL_SHIFT h->pixel_shift
#endif

#undef  CHROMA_IDC
#define CHROMA_IDC 1
#include "h264_mc_template.c"

#undef  CHROMA_IDC
#define CHROMA_IDC 2
#include "h264_mc_template.c"

43
static av_noinline void FUNC(hl_decode_mb)(const H264Context *h, H264SliceContext *sl)
44
{
45 46
    const int mb_x    = sl->mb_x;
    const int mb_y    = sl->mb_y;
47
    const int mb_xy   = sl->mb_xy;
48
    const int mb_type = h->cur_pic.mb_type[mb_xy];
49 50 51
    uint8_t *dest_y, *dest_cb, *dest_cr;
    int linesize, uvlinesize /*dct_offset*/;
    int i, j;
52
    const int *block_offset = &h->block_offset[0];
53
    const int transform_bypass = !SIMPLE && (sl->qscale == 0 && h->sps.transform_bypass);
54
    /* is_h264 should always be true if SVQ3 is disabled. */
55
    const int is_h264 = !CONFIG_SVQ3_DECODER || SIMPLE || h->avctx->codec_id == AV_CODEC_ID_H264;
Diego Biurrun's avatar
Diego Biurrun committed
56
    void (*idct_add)(uint8_t *dst, int16_t *block, int stride);
57
    const int block_h   = 16 >> h->chroma_y_shift;
58
    const int chroma422 = CHROMA422(h);
59

60 61 62
    dest_y  = h->cur_pic.f->data[0] + ((mb_x << PIXEL_SHIFT)     + mb_y * sl->linesize)  * 16;
    dest_cb = h->cur_pic.f->data[1] +  (mb_x << PIXEL_SHIFT) * 8 + mb_y * sl->uvlinesize * block_h;
    dest_cr = h->cur_pic.f->data[2] +  (mb_x << PIXEL_SHIFT) * 8 + mb_y * sl->uvlinesize * block_h;
63

64 65
    h->vdsp.prefetch(dest_y  + (sl->mb_x & 3) * 4 * sl->linesize   + (64 << PIXEL_SHIFT), sl->linesize,       4);
    h->vdsp.prefetch(dest_cb + (sl->mb_x & 7)     * sl->uvlinesize + (64 << PIXEL_SHIFT), dest_cr - dest_cb, 2);
66

67
    h->list_counts[mb_xy] = sl->list_count;
68

69
    if (!SIMPLE && MB_FIELD(sl)) {
70 71
        linesize     = sl->mb_linesize = sl->linesize * 2;
        uvlinesize   = sl->mb_uvlinesize = sl->uvlinesize * 2;
72 73
        block_offset = &h->block_offset[48];
        if (mb_y & 1) { // FIXME move out of this function?
74 75 76
            dest_y  -= sl->linesize * 15;
            dest_cb -= sl->uvlinesize * (block_h - 1);
            dest_cr -= sl->uvlinesize * (block_h - 1);
77
        }
78
        if (FRAME_MBAFF(h)) {
79
            int list;
80
            for (list = 0; list < sl->list_count; list++) {
81 82 83
                if (!USES_LIST(mb_type, list))
                    continue;
                if (IS_16X16(mb_type)) {
84
                    int8_t *ref = &sl->ref_cache[list][scan8[0]];
85
                    fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (sl->mb_y & 1), 1);
86 87
                } else {
                    for (i = 0; i < 16; i += 4) {
88
                        int ref = sl->ref_cache[list][scan8[i]];
89
                        if (ref >= 0)
90
                            fill_rectangle(&sl->ref_cache[list][scan8[i]], 2, 2,
91
                                           8, (16 + ref) ^ (sl->mb_y & 1), 1);
92 93 94 95 96
                    }
                }
            }
        }
    } else {
97 98
        linesize   = sl->mb_linesize   = sl->linesize;
        uvlinesize = sl->mb_uvlinesize = sl->uvlinesize;
99 100 101 102
        // dct_offset = s->linesize * 16;
    }

    if (!SIMPLE && IS_INTRA_PCM(mb_type)) {
103
        const int bit_depth = h->sps.bit_depth_luma;
104 105 106
        if (PIXEL_SHIFT) {
            int j;
            GetBitContext gb;
107
            init_get_bits(&gb, sl->intra_pcm_ptr,
108 109 110 111 112 113 114
                          ff_h264_mb_sizes[h->sps.chroma_format_idc] * bit_depth);

            for (i = 0; i < 16; i++) {
                uint16_t *tmp_y = (uint16_t *)(dest_y + i * linesize);
                for (j = 0; j < 16; j++)
                    tmp_y[j] = get_bits(&gb, bit_depth);
            }
115
            if (SIMPLE || !CONFIG_GRAY || !(h->flags & AV_CODEC_FLAG_GRAY)) {
116 117 118 119
                if (!h->sps.chroma_format_idc) {
                    for (i = 0; i < block_h; i++) {
                        uint16_t *tmp_cb = (uint16_t *)(dest_cb + i * uvlinesize);
                        uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize);
120 121 122
                        for (j = 0; j < 8; j++) {
                            tmp_cb[j] = tmp_cr[j] = 1 << (bit_depth - 1);
                        }
123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
                    }
                } else {
                    for (i = 0; i < block_h; i++) {
                        uint16_t *tmp_cb = (uint16_t *)(dest_cb + i * uvlinesize);
                        for (j = 0; j < 8; j++)
                            tmp_cb[j] = get_bits(&gb, bit_depth);
                    }
                    for (i = 0; i < block_h; i++) {
                        uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize);
                        for (j = 0; j < 8; j++)
                            tmp_cr[j] = get_bits(&gb, bit_depth);
                    }
                }
            }
        } else {
            for (i = 0; i < 16; i++)
139
                memcpy(dest_y + i * linesize, sl->intra_pcm_ptr + i * 16, 16);
140
            if (SIMPLE || !CONFIG_GRAY || !(h->flags & AV_CODEC_FLAG_GRAY)) {
141
                if (!h->sps.chroma_format_idc) {
142
                    for (i = 0; i < 8; i++) {
143 144
                        memset(dest_cb + i * uvlinesize, 1 << (bit_depth - 1), 8);
                        memset(dest_cr + i * uvlinesize, 1 << (bit_depth - 1), 8);
145 146
                    }
                } else {
147 148
                    const uint8_t *src_cb = sl->intra_pcm_ptr + 256;
                    const uint8_t *src_cr = sl->intra_pcm_ptr + 256 + block_h * 8;
149 150 151 152 153 154 155 156 157
                    for (i = 0; i < block_h; i++) {
                        memcpy(dest_cb + i * uvlinesize, src_cb + i * 8, 8);
                        memcpy(dest_cr + i * uvlinesize, src_cr + i * 8, 8);
                    }
                }
            }
        }
    } else {
        if (IS_INTRA(mb_type)) {
158
            if (sl->deblocking_filter)
159
                xchg_mb_border(h, sl, dest_y, dest_cb, dest_cr, linesize,
160 161
                               uvlinesize, 1, 0, SIMPLE, PIXEL_SHIFT);

162
            if (SIMPLE || !CONFIG_GRAY || !(h->flags & AV_CODEC_FLAG_GRAY)) {
163 164
                h->hpc.pred8x8[sl->chroma_pred_mode](dest_cb, uvlinesize);
                h->hpc.pred8x8[sl->chroma_pred_mode](dest_cr, uvlinesize);
165 166
            }

167
            hl_decode_mb_predict_luma(h, sl, mb_type, is_h264, SIMPLE,
168 169 170
                                      transform_bypass, PIXEL_SHIFT,
                                      block_offset, linesize, dest_y, 0);

171
            if (sl->deblocking_filter)
172
                xchg_mb_border(h, sl, dest_y, dest_cb, dest_cr, linesize,
173 174 175
                               uvlinesize, 0, 0, SIMPLE, PIXEL_SHIFT);
        } else if (is_h264) {
            if (chroma422) {
176
                FUNC(hl_motion_422)(h, sl, dest_y, dest_cb, dest_cr,
177 178
                              h->qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
                              h->qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
179 180 181
                              h->h264dsp.weight_h264_pixels_tab,
                              h->h264dsp.biweight_h264_pixels_tab);
            } else {
182
                FUNC(hl_motion_420)(h, sl, dest_y, dest_cb, dest_cr,
183 184
                              h->qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
                              h->qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
185 186 187 188 189
                              h->h264dsp.weight_h264_pixels_tab,
                              h->h264dsp.biweight_h264_pixels_tab);
            }
        }

190
        hl_decode_mb_idct_luma(h, sl, mb_type, is_h264, SIMPLE, transform_bypass,
191 192
                               PIXEL_SHIFT, block_offset, linesize, dest_y, 0);

193
        if ((SIMPLE || !CONFIG_GRAY || !(h->flags & AV_CODEC_FLAG_GRAY)) &&
194
            (sl->cbp & 0x30)) {
195 196 197
            uint8_t *dest[2] = { dest_cb, dest_cr };
            if (transform_bypass) {
                if (IS_INTRA(mb_type) && h->sps.profile_idc == 244 &&
198 199 200
                    (sl->chroma_pred_mode == VERT_PRED8x8 ||
                     sl->chroma_pred_mode == HOR_PRED8x8)) {
                    h->hpc.pred8x8_add[sl->chroma_pred_mode](dest[0],
201
                                                            block_offset + 16,
202
                                                            sl->mb + (16 * 16 * 1 << PIXEL_SHIFT),
203
                                                            uvlinesize);
204
                    h->hpc.pred8x8_add[sl->chroma_pred_mode](dest[1],
205
                                                            block_offset + 32,
206
                                                            sl->mb + (16 * 16 * 2 << PIXEL_SHIFT),
207 208
                                                            uvlinesize);
                } else {
209
                    idct_add = h->h264dsp.h264_add_pixels4_clear;
210 211
                    for (j = 1; j < 3; j++) {
                        for (i = j * 16; i < j * 16 + 4; i++)
212
                            if (sl->non_zero_count_cache[scan8[i]] ||
213
                                dctcoef_get(sl->mb, PIXEL_SHIFT, i * 16))
214
                                idct_add(dest[j - 1] + block_offset[i],
215
                                         sl->mb + (i * 16 << PIXEL_SHIFT),
216 217 218
                                         uvlinesize);
                        if (chroma422) {
                            for (i = j * 16 + 4; i < j * 16 + 8; i++)
219
                                if (sl->non_zero_count_cache[scan8[i + 4]] ||
220
                                    dctcoef_get(sl->mb, PIXEL_SHIFT, i * 16))
221
                                    idct_add(dest[j - 1] + block_offset[i + 4],
222
                                             sl->mb + (i * 16 << PIXEL_SHIFT),
223 224 225 226 227 228 229 230
                                             uvlinesize);
                        }
                    }
                }
            } else {
                if (is_h264) {
                    int qp[2];
                    if (chroma422) {
231 232
                        qp[0] = sl->chroma_qp[0] + 3;
                        qp[1] = sl->chroma_qp[1] + 3;
233
                    } else {
234 235
                        qp[0] = sl->chroma_qp[0];
                        qp[1] = sl->chroma_qp[1];
236
                    }
237
                    if (sl->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 0]])
238
                        h->h264dsp.h264_chroma_dc_dequant_idct(sl->mb + (16 * 16 * 1 << PIXEL_SHIFT),
239
                                                               h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][qp[0]][0]);
240
                    if (sl->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 1]])
241
                        h->h264dsp.h264_chroma_dc_dequant_idct(sl->mb + (16 * 16 * 2 << PIXEL_SHIFT),
242 243
                                                               h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][qp[1]][0]);
                    h->h264dsp.h264_idct_add8(dest, block_offset,
244
                                              sl->mb, uvlinesize,
245
                                              sl->non_zero_count_cache);
246
                } else if (CONFIG_SVQ3_DECODER) {
247
                    h->h264dsp.h264_chroma_dc_dequant_idct(sl->mb + 16 * 16 * 1,
248
                                                           h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][sl->chroma_qp[0]][0]);
249
                    h->h264dsp.h264_chroma_dc_dequant_idct(sl->mb + 16 * 16 * 2,
250
                                                           h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][sl->chroma_qp[1]][0]);
251 252
                    for (j = 1; j < 3; j++) {
                        for (i = j * 16; i < j * 16 + 4; i++)
253
                            if (sl->non_zero_count_cache[scan8[i]] || sl->mb[i * 16]) {
254
                                uint8_t *const ptr = dest[j - 1] + block_offset[i];
255
                                ff_svq3_add_idct_c(ptr, sl->mb + i * 16,
256
                                                   uvlinesize,
257
                                                   ff_h264_chroma_qp[0][sl->qscale + 12] - 12, 2);
258 259 260 261 262 263 264 265 266 267 268 269 270 271
                            }
                    }
                }
            }
        }
    }
}

#if !SIMPLE || BITS == 8

#undef  CHROMA_IDC
#define CHROMA_IDC 3
#include "h264_mc_template.c"

272
static av_noinline void FUNC(hl_decode_mb_444)(const H264Context *h, H264SliceContext *sl)
273
{
274 275
    const int mb_x    = sl->mb_x;
    const int mb_y    = sl->mb_y;
276
    const int mb_xy   = sl->mb_xy;
277
    const int mb_type = h->cur_pic.mb_type[mb_xy];
278 279 280
    uint8_t *dest[3];
    int linesize;
    int i, j, p;
281
    const int *block_offset = &h->block_offset[0];
282
    const int transform_bypass = !SIMPLE && (sl->qscale == 0 && h->sps.transform_bypass);
283
    const int plane_count      = (SIMPLE || !CONFIG_GRAY || !(h->flags & AV_CODEC_FLAG_GRAY)) ? 3 : 1;
284 285

    for (p = 0; p < plane_count; p++) {
286
        dest[p] = h->cur_pic.f->data[p] +
287 288 289
                  ((mb_x << PIXEL_SHIFT) + mb_y * sl->linesize) * 16;
        h->vdsp.prefetch(dest[p] + (sl->mb_x & 3) * 4 * sl->linesize + (64 << PIXEL_SHIFT),
                         sl->linesize, 4);
290 291
    }

292
    h->list_counts[mb_xy] = sl->list_count;
293

294
    if (!SIMPLE && MB_FIELD(sl)) {
295
        linesize     = sl->mb_linesize = sl->mb_uvlinesize = sl->linesize * 2;
296 297 298
        block_offset = &h->block_offset[48];
        if (mb_y & 1) // FIXME move out of this function?
            for (p = 0; p < 3; p++)
299
                dest[p] -= sl->linesize * 15;
300
        if (FRAME_MBAFF(h)) {
301
            int list;
302
            for (list = 0; list < sl->list_count; list++) {
303 304 305
                if (!USES_LIST(mb_type, list))
                    continue;
                if (IS_16X16(mb_type)) {
306
                    int8_t *ref = &sl->ref_cache[list][scan8[0]];
307
                    fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (sl->mb_y & 1), 1);
308 309
                } else {
                    for (i = 0; i < 16; i += 4) {
310
                        int ref = sl->ref_cache[list][scan8[i]];
311
                        if (ref >= 0)
312
                            fill_rectangle(&sl->ref_cache[list][scan8[i]], 2, 2,
313
                                           8, (16 + ref) ^ (sl->mb_y & 1), 1);
314 315 316 317 318
                    }
                }
            }
        }
    } else {
319
        linesize = sl->mb_linesize = sl->mb_uvlinesize = sl->linesize;
320 321 322 323 324 325
    }

    if (!SIMPLE && IS_INTRA_PCM(mb_type)) {
        if (PIXEL_SHIFT) {
            const int bit_depth = h->sps.bit_depth_luma;
            GetBitContext gb;
326
            init_get_bits(&gb, sl->intra_pcm_ptr, 768 * bit_depth);
327 328 329 330 331 332 333 334 335 336 337

            for (p = 0; p < plane_count; p++)
                for (i = 0; i < 16; i++) {
                    uint16_t *tmp = (uint16_t *)(dest[p] + i * linesize);
                    for (j = 0; j < 16; j++)
                        tmp[j] = get_bits(&gb, bit_depth);
                }
        } else {
            for (p = 0; p < plane_count; p++)
                for (i = 0; i < 16; i++)
                    memcpy(dest[p] + i * linesize,
338
                           sl->intra_pcm_ptr + p * 256 + i * 16, 16);
339 340 341
        }
    } else {
        if (IS_INTRA(mb_type)) {
342
            if (sl->deblocking_filter)
343
                xchg_mb_border(h, sl, dest[0], dest[1], dest[2], linesize,
344 345 346
                               linesize, 1, 1, SIMPLE, PIXEL_SHIFT);

            for (p = 0; p < plane_count; p++)
347
                hl_decode_mb_predict_luma(h, sl, mb_type, 1, SIMPLE,
348 349 350
                                          transform_bypass, PIXEL_SHIFT,
                                          block_offset, linesize, dest[p], p);

351
            if (sl->deblocking_filter)
352
                xchg_mb_border(h, sl, dest[0], dest[1], dest[2], linesize,
353 354
                               linesize, 0, 1, SIMPLE, PIXEL_SHIFT);
        } else {
355
            FUNC(hl_motion_444)(h, sl, dest[0], dest[1], dest[2],
356 357
                      h->qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
                      h->qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
358 359 360 361 362
                      h->h264dsp.weight_h264_pixels_tab,
                      h->h264dsp.biweight_h264_pixels_tab);
        }

        for (p = 0; p < plane_count; p++)
363
            hl_decode_mb_idct_luma(h, sl, mb_type, 1, SIMPLE, transform_bypass,
364 365 366 367 368 369
                                   PIXEL_SHIFT, block_offset, linesize,
                                   dest[p], p);
    }
}

#endif