2 * H.26L/H.264/AVC/JVT/14496-10/... decoder
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
5 * This file is part of Libav.
7 * Libav 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.
12 * Libav 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.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
26 # define FUNC(n) AV_JOIN(n ## _simple_, BITS)
27 # define PIXEL_SHIFT (BITS >> 4)
29 # define FUNC(n) n ## _complex
30 # define PIXEL_SHIFT h->pixel_shift
35 #include "h264_mc_template.c"
39 #include "h264_mc_template.c"
41 static av_noinline void FUNC(hl_decode_mb)(H264Context *h)
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*/;
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);
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;
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);
65 h->list_counts[mb_xy] = h->list_count;
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);
78 for (list = 0; list < h->list_count; list++) {
79 if (!USES_LIST(mb_type, list))
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);
85 for (i = 0; i < 16; i += 4) {
86 int ref = h->ref_cache[list][scan8[i]];
88 fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2,
89 8, (16 + ref) ^ (h->mb_y & 1), 1);
95 linesize = h->mb_linesize = h->linesize;
96 uvlinesize = h->mb_uvlinesize = h->uvlinesize;
97 // dct_offset = s->linesize * 16;
100 if (!SIMPLE && IS_INTRA_PCM(mb_type)) {
102 const int bit_depth = h->sps.bit_depth_luma;
105 init_get_bits(&gb, h->intra_pcm_ptr,
106 ff_h264_mb_sizes[h->sps.chroma_format_idc] * bit_depth);
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);
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 for (j = 0; j < 8; j++)
118 tmp_cb[j] = 1 << (bit_depth - 1);
120 for (i = 0; i < block_h; i++) {
121 uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize);
122 for (j = 0; j < 8; j++)
123 tmp_cr[j] = 1 << (bit_depth - 1);
126 for (i = 0; i < block_h; i++) {
127 uint16_t *tmp_cb = (uint16_t *)(dest_cb + i * uvlinesize);
128 for (j = 0; j < 8; j++)
129 tmp_cb[j] = get_bits(&gb, bit_depth);
131 for (i = 0; i < block_h; i++) {
132 uint16_t *tmp_cr = (uint16_t *)(dest_cr + i * uvlinesize);
133 for (j = 0; j < 8; j++)
134 tmp_cr[j] = get_bits(&gb, bit_depth);
139 for (i = 0; i < 16; i++)
140 memcpy(dest_y + i * linesize, h->intra_pcm_ptr + i * 16, 16);
141 if (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) {
142 if (!h->sps.chroma_format_idc) {
143 for (i = 0; i < block_h; i++) {
144 memset(dest_cb + i * uvlinesize, 128, 8);
145 memset(dest_cr + i * uvlinesize, 128, 8);
148 const uint8_t *src_cb = h->intra_pcm_ptr + 256;
149 const uint8_t *src_cr = h->intra_pcm_ptr + 256 + block_h * 8;
150 for (i = 0; i < block_h; i++) {
151 memcpy(dest_cb + i * uvlinesize, src_cb + i * 8, 8);
152 memcpy(dest_cr + i * uvlinesize, src_cr + i * 8, 8);
158 if (IS_INTRA(mb_type)) {
159 if (h->deblocking_filter)
160 xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize,
161 uvlinesize, 1, 0, SIMPLE, PIXEL_SHIFT);
163 if (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) {
164 h->hpc.pred8x8[h->chroma_pred_mode](dest_cb, uvlinesize);
165 h->hpc.pred8x8[h->chroma_pred_mode](dest_cr, uvlinesize);
168 hl_decode_mb_predict_luma(h, mb_type, is_h264, SIMPLE,
169 transform_bypass, PIXEL_SHIFT,
170 block_offset, linesize, dest_y, 0);
172 if (h->deblocking_filter)
173 xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize,
174 uvlinesize, 0, 0, SIMPLE, PIXEL_SHIFT);
175 } else if (is_h264) {
177 FUNC(hl_motion_422)(h, dest_y, dest_cb, dest_cr,
178 h->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
179 h->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
180 h->h264dsp.weight_h264_pixels_tab,
181 h->h264dsp.biweight_h264_pixels_tab);
183 FUNC(hl_motion_420)(h, dest_y, dest_cb, dest_cr,
184 h->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
185 h->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
186 h->h264dsp.weight_h264_pixels_tab,
187 h->h264dsp.biweight_h264_pixels_tab);
191 hl_decode_mb_idct_luma(h, mb_type, is_h264, SIMPLE, transform_bypass,
192 PIXEL_SHIFT, block_offset, linesize, dest_y, 0);
194 if ((SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) &&
196 uint8_t *dest[2] = { dest_cb, dest_cr };
197 if (transform_bypass) {
198 if (IS_INTRA(mb_type) && h->sps.profile_idc == 244 &&
199 (h->chroma_pred_mode == VERT_PRED8x8 ||
200 h->chroma_pred_mode == HOR_PRED8x8)) {
201 h->hpc.pred8x8_add[h->chroma_pred_mode](dest[0],
203 h->mb + (16 * 16 * 1 << PIXEL_SHIFT),
205 h->hpc.pred8x8_add[h->chroma_pred_mode](dest[1],
207 h->mb + (16 * 16 * 2 << PIXEL_SHIFT),
210 idct_add = h->h264dsp.h264_add_pixels4_clear;
211 for (j = 1; j < 3; j++) {
212 for (i = j * 16; i < j * 16 + 4; i++)
213 if (h->non_zero_count_cache[scan8[i]] ||
214 dctcoef_get(h->mb, PIXEL_SHIFT, i * 16))
215 idct_add(dest[j - 1] + block_offset[i],
216 h->mb + (i * 16 << PIXEL_SHIFT),
219 for (i = j * 16 + 4; i < j * 16 + 8; i++)
220 if (h->non_zero_count_cache[scan8[i + 4]] ||
221 dctcoef_get(h->mb, PIXEL_SHIFT, i * 16))
222 idct_add(dest[j - 1] + block_offset[i + 4],
223 h->mb + (i * 16 << PIXEL_SHIFT),
232 qp[0] = h->chroma_qp[0] + 3;
233 qp[1] = h->chroma_qp[1] + 3;
235 qp[0] = h->chroma_qp[0];
236 qp[1] = h->chroma_qp[1];
238 if (h->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 0]])
239 h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 1 << PIXEL_SHIFT),
240 h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][qp[0]][0]);
241 if (h->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 1]])
242 h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 2 << PIXEL_SHIFT),
243 h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][qp[1]][0]);
244 h->h264dsp.h264_idct_add8(dest, block_offset,
246 h->non_zero_count_cache);
247 } else if (CONFIG_SVQ3_DECODER) {
248 h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16 * 16 * 1,
249 h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][h->chroma_qp[0]][0]);
250 h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16 * 16 * 2,
251 h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][h->chroma_qp[1]][0]);
252 for (j = 1; j < 3; j++) {
253 for (i = j * 16; i < j * 16 + 4; i++)
254 if (h->non_zero_count_cache[scan8[i]] || h->mb[i * 16]) {
255 uint8_t *const ptr = dest[j - 1] + block_offset[i];
256 ff_svq3_add_idct_c(ptr, h->mb + i * 16,
258 ff_h264_chroma_qp[0][h->qscale + 12] - 12, 2);
267 #if !SIMPLE || BITS == 8
271 #include "h264_mc_template.c"
273 static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h)
275 const int mb_x = h->mb_x;
276 const int mb_y = h->mb_y;
277 const int mb_xy = h->mb_xy;
278 const int mb_type = h->cur_pic.mb_type[mb_xy];
282 int *block_offset = &h->block_offset[0];
283 const int transform_bypass = !SIMPLE && (h->qscale == 0 && h->sps.transform_bypass);
284 const int plane_count = (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) ? 3 : 1;
286 for (p = 0; p < plane_count; p++) {
287 dest[p] = h->cur_pic.f.data[p] +
288 ((mb_x << PIXEL_SHIFT) + mb_y * h->linesize) * 16;
289 h->vdsp.prefetch(dest[p] + (h->mb_x & 3) * 4 * h->linesize + (64 << PIXEL_SHIFT),
293 h->list_counts[mb_xy] = h->list_count;
295 if (!SIMPLE && MB_FIELD(h)) {
296 linesize = h->mb_linesize = h->mb_uvlinesize = h->linesize * 2;
297 block_offset = &h->block_offset[48];
298 if (mb_y & 1) // FIXME move out of this function?
299 for (p = 0; p < 3; p++)
300 dest[p] -= h->linesize * 15;
301 if (FRAME_MBAFF(h)) {
303 for (list = 0; list < h->list_count; list++) {
304 if (!USES_LIST(mb_type, list))
306 if (IS_16X16(mb_type)) {
307 int8_t *ref = &h->ref_cache[list][scan8[0]];
308 fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (h->mb_y & 1), 1);
310 for (i = 0; i < 16; i += 4) {
311 int ref = h->ref_cache[list][scan8[i]];
313 fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2,
314 8, (16 + ref) ^ (h->mb_y & 1), 1);
320 linesize = h->mb_linesize = h->mb_uvlinesize = h->linesize;
323 if (!SIMPLE && IS_INTRA_PCM(mb_type)) {
325 const int bit_depth = h->sps.bit_depth_luma;
327 init_get_bits(&gb, h->intra_pcm_ptr, 768 * bit_depth);
329 for (p = 0; p < plane_count; p++)
330 for (i = 0; i < 16; i++) {
331 uint16_t *tmp = (uint16_t *)(dest[p] + i * linesize);
332 for (j = 0; j < 16; j++)
333 tmp[j] = get_bits(&gb, bit_depth);
336 for (p = 0; p < plane_count; p++)
337 for (i = 0; i < 16; i++)
338 memcpy(dest[p] + i * linesize,
339 h->intra_pcm_ptr + p * 256 + i * 16, 16);
342 if (IS_INTRA(mb_type)) {
343 if (h->deblocking_filter)
344 xchg_mb_border(h, dest[0], dest[1], dest[2], linesize,
345 linesize, 1, 1, SIMPLE, PIXEL_SHIFT);
347 for (p = 0; p < plane_count; p++)
348 hl_decode_mb_predict_luma(h, mb_type, 1, SIMPLE,
349 transform_bypass, PIXEL_SHIFT,
350 block_offset, linesize, dest[p], p);
352 if (h->deblocking_filter)
353 xchg_mb_border(h, dest[0], dest[1], dest[2], linesize,
354 linesize, 0, 1, SIMPLE, PIXEL_SHIFT);
356 FUNC(hl_motion_444)(h, dest[0], dest[1], dest[2],
357 h->me.qpel_put, h->h264chroma.put_h264_chroma_pixels_tab,
358 h->me.qpel_avg, h->h264chroma.avg_h264_chroma_pixels_tab,
359 h->h264dsp.weight_h264_pixels_tab,
360 h->h264dsp.biweight_h264_pixels_tab);
363 for (p = 0; p < plane_count; p++)
364 hl_decode_mb_idct_luma(h, mb_type, 1, SIMPLE, transform_bypass,
365 PIXEL_SHIFT, block_offset, linesize,