Go to most recent revision | Details | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 4349 | Serge | 1 | /* |
| 2 | * H.26L/H.264/AVC/JVT/14496-10/... decoder |
||
| 3 | * Copyright (c) 2003 Michael Niedermayer |
||
| 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 | #undef FUNC |
||
| 23 | #undef PIXEL_SHIFT |
||
| 24 | |||
| 25 | #if SIMPLE |
||
| 26 | # define FUNC(n) AV_JOIN(n ## _simple_, BITS) |
||
| 27 | # define PIXEL_SHIFT (BITS >> 4) |
||
| 28 | #else |
||
| 29 | # define FUNC(n) n ## _complex |
||
| 30 | # define PIXEL_SHIFT h->pixel_shift |
||
| 31 | #endif |
||
| 32 | |||
| 33 | #undef CHROMA_IDC |
||
| 34 | #define CHROMA_IDC 1 |
||
| 35 | #include "h264_mc_template.c" |
||
| 36 | |||
| 37 | #undef CHROMA_IDC |
||
| 38 | #define CHROMA_IDC 2 |
||
| 39 | #include "h264_mc_template.c" |
||
| 40 | |||
| 41 | static av_noinline void FUNC(hl_decode_mb)(H264Context *h) |
||
| 42 | { |
||
| 43 | const int mb_x = h->mb_x; |
||
| 44 | const int mb_y = h->mb_y; |
||
| 45 | const int mb_xy = h->mb_xy; |
||
| 46 | const int mb_type = h->cur_pic.mb_type[mb_xy]; |
||
| 47 | uint8_t *dest_y, *dest_cb, *dest_cr; |
||
| 48 | int linesize, uvlinesize /*dct_offset*/; |
||
| 49 | int i, j; |
||
| 50 | int *block_offset = &h->block_offset[0]; |
||
| 51 | const int transform_bypass = !SIMPLE && (h->qscale == 0 && h->sps.transform_bypass); |
||
| 52 | /* is_h264 should always be true if SVQ3 is disabled. */ |
||
| 53 | const int is_h264 = !CONFIG_SVQ3_DECODER || SIMPLE || h->avctx->codec_id == AV_CODEC_ID_H264; |
||
| 54 | void (*idct_add)(uint8_t *dst, int16_t *block, int stride); |
||
| 55 | const int block_h = 16 >> h->chroma_y_shift; |
||
| 56 | const int chroma422 = CHROMA422(h); |
||
| 57 | |||
| 58 | dest_y = h->cur_pic.f.data[0] + ((mb_x << PIXEL_SHIFT) + mb_y * h->linesize) * 16; |
||
| 59 | dest_cb = h->cur_pic.f.data[1] + (mb_x << PIXEL_SHIFT) * 8 + mb_y * h->uvlinesize * block_h; |
||
| 60 | dest_cr = h->cur_pic.f.data[2] + (mb_x << PIXEL_SHIFT) * 8 + mb_y * h->uvlinesize * block_h; |
||
| 61 | |||
| 62 | h->vdsp.prefetch(dest_y + (h->mb_x & 3) * 4 * h->linesize + (64 << PIXEL_SHIFT), h->linesize, 4); |
||
| 63 | h->vdsp.prefetch(dest_cb + (h->mb_x & 7) * h->uvlinesize + (64 << PIXEL_SHIFT), dest_cr - dest_cb, 2); |
||
| 64 | |||
| 65 | h->list_counts[mb_xy] = h->list_count; |
||
| 66 | |||
| 67 | if (!SIMPLE && MB_FIELD(h)) { |
||
| 68 | linesize = h->mb_linesize = h->linesize * 2; |
||
| 69 | uvlinesize = h->mb_uvlinesize = h->uvlinesize * 2; |
||
| 70 | block_offset = &h->block_offset[48]; |
||
| 71 | if (mb_y & 1) { // FIXME move out of this function? |
||
| 72 | dest_y -= h->linesize * 15; |
||
| 73 | dest_cb -= h->uvlinesize * (block_h - 1); |
||
| 74 | dest_cr -= h->uvlinesize * (block_h - 1); |
||
| 75 | } |
||
| 76 | if (FRAME_MBAFF(h)) { |
||
| 77 | int list; |
||
| 78 | for (list = 0; list < h->list_count; list++) { |
||
| 79 | if (!USES_LIST(mb_type, list)) |
||
| 80 | continue; |
||
| 81 | if (IS_16X16(mb_type)) { |
||
| 82 | int8_t *ref = &h->ref_cache[list][scan8[0]]; |
||
| 83 | fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (h->mb_y & 1), 1); |
||
| 84 | } else { |
||
| 85 | for (i = 0; i < 16; i += 4) { |
||
| 86 | int ref = h->ref_cache[list][scan8[i]]; |
||
| 87 | if (ref >= 0) |
||
| 88 | fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2, |
||
| 89 | 8, (16 + ref) ^ (h->mb_y & 1), 1); |
||
| 90 | } |
||
| 91 | } |
||
| 92 | } |
||
| 93 | } |
||
| 94 | } else { |
||
| 95 | linesize = h->mb_linesize = h->linesize; |
||
| 96 | uvlinesize = h->mb_uvlinesize = h->uvlinesize; |
||
| 97 | // dct_offset = s->linesize * 16; |
||
| 98 | } |
||
| 99 | |||
| 100 | if (!SIMPLE && IS_INTRA_PCM(mb_type)) { |
||
| 101 | const int bit_depth = h->sps.bit_depth_luma; |
||
| 102 | if (PIXEL_SHIFT) { |
||
| 103 | int j; |
||
| 104 | GetBitContext gb; |
||
| 105 | init_get_bits(&gb, h->intra_pcm_ptr, |
||
| 106 | ff_h264_mb_sizes[h->sps.chroma_format_idc] * bit_depth); |
||
| 107 | |||
| 108 | for (i = 0; i < 16; i++) { |
||
| 109 | uint16_t *tmp_y = (uint16_t *)(dest_y + i * linesize); |
||
| 110 | for (j = 0; j < 16; j++) |
||
| 111 | tmp_y[j] = get_bits(&gb, bit_depth); |
||
| 112 | } |
||
| 113 | if (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) { |
||
| 114 | if (!h->sps.chroma_format_idc) { |
||
| 115 | for (i = 0; i < block_h; i++) { |
||
| 116 | uint16_t *tmp_cb = (uint16_t *)(dest_cb + i * uvlinesize); |
||
| 117 | uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize); |
||
| 118 | for (j = 0; j < 8; j++) { |
||
| 119 | tmp_cb[j] = tmp_cr[j] = 1 << (bit_depth - 1); |
||
| 120 | } |
||
| 121 | } |
||
| 122 | } else { |
||
| 123 | for (i = 0; i < block_h; i++) { |
||
| 124 | uint16_t *tmp_cb = (uint16_t *)(dest_cb + i * uvlinesize); |
||
| 125 | for (j = 0; j < 8; j++) |
||
| 126 | tmp_cb[j] = get_bits(&gb, bit_depth); |
||
| 127 | } |
||
| 128 | for (i = 0; i < block_h; i++) { |
||
| 129 | uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize); |
||
| 130 | for (j = 0; j < 8; j++) |
||
| 131 | tmp_cr[j] = get_bits(&gb, bit_depth); |
||
| 132 | } |
||
| 133 | } |
||
| 134 | } |
||
| 135 | } else { |
||
| 136 | for (i = 0; i < 16; i++) |
||
| 137 | memcpy(dest_y + i * linesize, h->intra_pcm_ptr + i * 16, 16); |
||
| 138 | if (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) { |
||
| 139 | if (!h->sps.chroma_format_idc) { |
||
| 140 | for (i = 0; i < 8; i++) { |
||
| 141 | memset(dest_cb + i*uvlinesize, 1 << (bit_depth - 1), 8); |
||
| 142 | memset(dest_cr + i*uvlinesize, 1 << (bit_depth - 1), 8); |
||
| 143 | } |
||
| 144 | } else { |
||
| 145 | const uint8_t *src_cb = h->intra_pcm_ptr + 256; |
||
| 146 | const uint8_t *src_cr = h->intra_pcm_ptr + 256 + block_h * 8; |
||
| 147 | for (i = 0; i < block_h; i++) { |
||
| 148 | memcpy(dest_cb + i * uvlinesize, src_cb + i * 8, 8); |
||
| 149 | memcpy(dest_cr + i * uvlinesize, src_cr + i * 8, 8); |
||
| 150 | } |
||
| 151 | } |
||
| 152 | } |
||
| 153 | } |
||
| 154 | } else { |
||
| 155 | if (IS_INTRA(mb_type)) { |
||
| 156 | if (h->deblocking_filter) |
||
| 157 | xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, |
||
| 158 | uvlinesize, 1, 0, SIMPLE, PIXEL_SHIFT); |
||
| 159 | |||
| 160 | if (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) { |
||
| 161 | h->hpc.pred8x8[h->chroma_pred_mode](dest_cb, uvlinesize); |
||
| 162 | h->hpc.pred8x8[h->chroma_pred_mode](dest_cr, uvlinesize); |
||
| 163 | } |
||
| 164 | |||
| 165 | hl_decode_mb_predict_luma(h, mb_type, is_h264, SIMPLE, |
||
| 166 | transform_bypass, PIXEL_SHIFT, |
||
| 167 | block_offset, linesize, dest_y, 0); |
||
| 168 | |||
| 169 | if (h->deblocking_filter) |
||
| 170 | xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, |
||
| 171 | uvlinesize, 0, 0, SIMPLE, PIXEL_SHIFT); |
||
| 172 | } else if (is_h264) { |
||
| 173 | if (chroma422) { |
||
| 174 | FUNC(hl_motion_422)(h, dest_y, dest_cb, dest_cr, |
||
| 175 | h->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab, |
||
| 176 | h->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab, |
||
| 177 | h->h264dsp.weight_h264_pixels_tab, |
||
| 178 | h->h264dsp.biweight_h264_pixels_tab); |
||
| 179 | } else { |
||
| 180 | FUNC(hl_motion_420)(h, dest_y, dest_cb, dest_cr, |
||
| 181 | h->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab, |
||
| 182 | h->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab, |
||
| 183 | h->h264dsp.weight_h264_pixels_tab, |
||
| 184 | h->h264dsp.biweight_h264_pixels_tab); |
||
| 185 | } |
||
| 186 | } |
||
| 187 | |||
| 188 | hl_decode_mb_idct_luma(h, mb_type, is_h264, SIMPLE, transform_bypass, |
||
| 189 | PIXEL_SHIFT, block_offset, linesize, dest_y, 0); |
||
| 190 | |||
| 191 | if ((SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) && |
||
| 192 | (h->cbp & 0x30)) { |
||
| 193 | uint8_t *dest[2] = { dest_cb, dest_cr }; |
||
| 194 | if (transform_bypass) { |
||
| 195 | if (IS_INTRA(mb_type) && h->sps.profile_idc == 244 && |
||
| 196 | (h->chroma_pred_mode == VERT_PRED8x8 || |
||
| 197 | h->chroma_pred_mode == HOR_PRED8x8)) { |
||
| 198 | h->hpc.pred8x8_add[h->chroma_pred_mode](dest[0], |
||
| 199 | block_offset + 16, |
||
| 200 | h->mb + (16 * 16 * 1 << PIXEL_SHIFT), |
||
| 201 | uvlinesize); |
||
| 202 | h->hpc.pred8x8_add[h->chroma_pred_mode](dest[1], |
||
| 203 | block_offset + 32, |
||
| 204 | h->mb + (16 * 16 * 2 << PIXEL_SHIFT), |
||
| 205 | uvlinesize); |
||
| 206 | } else { |
||
| 207 | idct_add = h->h264dsp.h264_add_pixels4_clear; |
||
| 208 | for (j = 1; j < 3; j++) { |
||
| 209 | for (i = j * 16; i < j * 16 + 4; i++) |
||
| 210 | if (h->non_zero_count_cache[scan8[i]] || |
||
| 211 | dctcoef_get(h->mb, PIXEL_SHIFT, i * 16)) |
||
| 212 | idct_add(dest[j - 1] + block_offset[i], |
||
| 213 | h->mb + (i * 16 << PIXEL_SHIFT), |
||
| 214 | uvlinesize); |
||
| 215 | if (chroma422) { |
||
| 216 | for (i = j * 16 + 4; i < j * 16 + 8; i++) |
||
| 217 | if (h->non_zero_count_cache[scan8[i + 4]] || |
||
| 218 | dctcoef_get(h->mb, PIXEL_SHIFT, i * 16)) |
||
| 219 | idct_add(dest[j - 1] + block_offset[i + 4], |
||
| 220 | h->mb + (i * 16 << PIXEL_SHIFT), |
||
| 221 | uvlinesize); |
||
| 222 | } |
||
| 223 | } |
||
| 224 | } |
||
| 225 | } else { |
||
| 226 | if (is_h264) { |
||
| 227 | int qp[2]; |
||
| 228 | if (chroma422) { |
||
| 229 | qp[0] = h->chroma_qp[0] + 3; |
||
| 230 | qp[1] = h->chroma_qp[1] + 3; |
||
| 231 | } else { |
||
| 232 | qp[0] = h->chroma_qp[0]; |
||
| 233 | qp[1] = h->chroma_qp[1]; |
||
| 234 | } |
||
| 235 | if (h->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 0]]) |
||
| 236 | h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 1 << PIXEL_SHIFT), |
||
| 237 | h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][qp[0]][0]); |
||
| 238 | if (h->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 1]]) |
||
| 239 | h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 2 << PIXEL_SHIFT), |
||
| 240 | h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][qp[1]][0]); |
||
| 241 | h->h264dsp.h264_idct_add8(dest, block_offset, |
||
| 242 | h->mb, uvlinesize, |
||
| 243 | h->non_zero_count_cache); |
||
| 244 | } else if (CONFIG_SVQ3_DECODER) { |
||
| 245 | h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16 * 16 * 1, |
||
| 246 | h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][h->chroma_qp[0]][0]); |
||
| 247 | h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16 * 16 * 2, |
||
| 248 | h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][h->chroma_qp[1]][0]); |
||
| 249 | for (j = 1; j < 3; j++) { |
||
| 250 | for (i = j * 16; i < j * 16 + 4; i++) |
||
| 251 | if (h->non_zero_count_cache[scan8[i]] || h->mb[i * 16]) { |
||
| 252 | uint8_t *const ptr = dest[j - 1] + block_offset[i]; |
||
| 253 | ff_svq3_add_idct_c(ptr, h->mb + i * 16, |
||
| 254 | uvlinesize, |
||
| 255 | ff_h264_chroma_qp[0][h->qscale + 12] - 12, 2); |
||
| 256 | } |
||
| 257 | } |
||
| 258 | } |
||
| 259 | } |
||
| 260 | } |
||
| 261 | } |
||
| 262 | } |
||
| 263 | |||
| 264 | #if !SIMPLE || BITS == 8 |
||
| 265 | |||
| 266 | #undef CHROMA_IDC |
||
| 267 | #define CHROMA_IDC 3 |
||
| 268 | #include "h264_mc_template.c" |
||
| 269 | |||
| 270 | static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h) |
||
| 271 | { |
||
| 272 | const int mb_x = h->mb_x; |
||
| 273 | const int mb_y = h->mb_y; |
||
| 274 | const int mb_xy = h->mb_xy; |
||
| 275 | const int mb_type = h->cur_pic.mb_type[mb_xy]; |
||
| 276 | uint8_t *dest[3]; |
||
| 277 | int linesize; |
||
| 278 | int i, j, p; |
||
| 279 | int *block_offset = &h->block_offset[0]; |
||
| 280 | const int transform_bypass = !SIMPLE && (h->qscale == 0 && h->sps.transform_bypass); |
||
| 281 | const int plane_count = (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) ? 3 : 1; |
||
| 282 | |||
| 283 | for (p = 0; p < plane_count; p++) { |
||
| 284 | dest[p] = h->cur_pic.f.data[p] + |
||
| 285 | ((mb_x << PIXEL_SHIFT) + mb_y * h->linesize) * 16; |
||
| 286 | h->vdsp.prefetch(dest[p] + (h->mb_x & 3) * 4 * h->linesize + (64 << PIXEL_SHIFT), |
||
| 287 | h->linesize, 4); |
||
| 288 | } |
||
| 289 | |||
| 290 | h->list_counts[mb_xy] = h->list_count; |
||
| 291 | |||
| 292 | if (!SIMPLE && MB_FIELD(h)) { |
||
| 293 | linesize = h->mb_linesize = h->mb_uvlinesize = h->linesize * 2; |
||
| 294 | block_offset = &h->block_offset[48]; |
||
| 295 | if (mb_y & 1) // FIXME move out of this function? |
||
| 296 | for (p = 0; p < 3; p++) |
||
| 297 | dest[p] -= h->linesize * 15; |
||
| 298 | if (FRAME_MBAFF(h)) { |
||
| 299 | int list; |
||
| 300 | for (list = 0; list < h->list_count; list++) { |
||
| 301 | if (!USES_LIST(mb_type, list)) |
||
| 302 | continue; |
||
| 303 | if (IS_16X16(mb_type)) { |
||
| 304 | int8_t *ref = &h->ref_cache[list][scan8[0]]; |
||
| 305 | fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (h->mb_y & 1), 1); |
||
| 306 | } else { |
||
| 307 | for (i = 0; i < 16; i += 4) { |
||
| 308 | int ref = h->ref_cache[list][scan8[i]]; |
||
| 309 | if (ref >= 0) |
||
| 310 | fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2, |
||
| 311 | 8, (16 + ref) ^ (h->mb_y & 1), 1); |
||
| 312 | } |
||
| 313 | } |
||
| 314 | } |
||
| 315 | } |
||
| 316 | } else { |
||
| 317 | linesize = h->mb_linesize = h->mb_uvlinesize = h->linesize; |
||
| 318 | } |
||
| 319 | |||
| 320 | if (!SIMPLE && IS_INTRA_PCM(mb_type)) { |
||
| 321 | if (PIXEL_SHIFT) { |
||
| 322 | const int bit_depth = h->sps.bit_depth_luma; |
||
| 323 | GetBitContext gb; |
||
| 324 | init_get_bits(&gb, h->intra_pcm_ptr, 768 * bit_depth); |
||
| 325 | |||
| 326 | for (p = 0; p < plane_count; p++) |
||
| 327 | for (i = 0; i < 16; i++) { |
||
| 328 | uint16_t *tmp = (uint16_t *)(dest[p] + i * linesize); |
||
| 329 | for (j = 0; j < 16; j++) |
||
| 330 | tmp[j] = get_bits(&gb, bit_depth); |
||
| 331 | } |
||
| 332 | } else { |
||
| 333 | for (p = 0; p < plane_count; p++) |
||
| 334 | for (i = 0; i < 16; i++) |
||
| 335 | memcpy(dest[p] + i * linesize, |
||
| 336 | h->intra_pcm_ptr + p * 256 + i * 16, 16); |
||
| 337 | } |
||
| 338 | } else { |
||
| 339 | if (IS_INTRA(mb_type)) { |
||
| 340 | if (h->deblocking_filter) |
||
| 341 | xchg_mb_border(h, dest[0], dest[1], dest[2], linesize, |
||
| 342 | linesize, 1, 1, SIMPLE, PIXEL_SHIFT); |
||
| 343 | |||
| 344 | for (p = 0; p < plane_count; p++) |
||
| 345 | hl_decode_mb_predict_luma(h, mb_type, 1, SIMPLE, |
||
| 346 | transform_bypass, PIXEL_SHIFT, |
||
| 347 | block_offset, linesize, dest[p], p); |
||
| 348 | |||
| 349 | if (h->deblocking_filter) |
||
| 350 | xchg_mb_border(h, dest[0], dest[1], dest[2], linesize, |
||
| 351 | linesize, 0, 1, SIMPLE, PIXEL_SHIFT); |
||
| 352 | } else { |
||
| 353 | FUNC(hl_motion_444)(h, dest[0], dest[1], dest[2], |
||
| 354 | h->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab, |
||
| 355 | h->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab, |
||
| 356 | h->h264dsp.weight_h264_pixels_tab, |
||
| 357 | h->h264dsp.biweight_h264_pixels_tab); |
||
| 358 | } |
||
| 359 | |||
| 360 | for (p = 0; p < plane_count; p++) |
||
| 361 | hl_decode_mb_idct_luma(h, mb_type, 1, SIMPLE, transform_bypass, |
||
| 362 | PIXEL_SHIFT, block_offset, linesize, |
||
| 363 | dest[p], p); |
||
| 364 | } |
||
| 365 | } |
||
| 366 | |||
| 367 | #endif |