From f51d0f39c0fb3470d6b186974109428da91f1d1b Mon Sep 17 00:00:00 2001 From: Luca Barbato Date: Wed, 19 Mar 2014 16:52:47 +0100 Subject: h264: Split h264 slice decoding from nal decoding --- libavcodec/h264.c | 2541 +---------------------------------------------------- 1 file changed, 24 insertions(+), 2517 deletions(-) (limited to 'libavcodec/h264.c') diff --git a/libavcodec/h264.c b/libavcodec/h264.c index e039dbb8fc..9a96937381 100644 --- a/libavcodec/h264.c +++ b/libavcodec/h264.c @@ -50,139 +50,6 @@ const uint16_t ff_h264_mb_sizes[4] = { 256, 384, 512, 768 }; -static const uint8_t rem6[QP_MAX_NUM + 1] = { - 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, - 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, - 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, -}; - -static const uint8_t div6[QP_MAX_NUM + 1] = { - 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, - 3, 3, 3, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, - 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 10, 10, 10, 10, -}; - -static const uint8_t field_scan[16] = { - 0 + 0 * 4, 0 + 1 * 4, 1 + 0 * 4, 0 + 2 * 4, - 0 + 3 * 4, 1 + 1 * 4, 1 + 2 * 4, 1 + 3 * 4, - 2 + 0 * 4, 2 + 1 * 4, 2 + 2 * 4, 2 + 3 * 4, - 3 + 0 * 4, 3 + 1 * 4, 3 + 2 * 4, 3 + 3 * 4, -}; - -static const uint8_t field_scan8x8[64] = { - 0 + 0 * 8, 0 + 1 * 8, 0 + 2 * 8, 1 + 0 * 8, - 1 + 1 * 8, 0 + 3 * 8, 0 + 4 * 8, 1 + 2 * 8, - 2 + 0 * 8, 1 + 3 * 8, 0 + 5 * 8, 0 + 6 * 8, - 0 + 7 * 8, 1 + 4 * 8, 2 + 1 * 8, 3 + 0 * 8, - 2 + 2 * 8, 1 + 5 * 8, 1 + 6 * 8, 1 + 7 * 8, - 2 + 3 * 8, 3 + 1 * 8, 4 + 0 * 8, 3 + 2 * 8, - 2 + 4 * 8, 2 + 5 * 8, 2 + 6 * 8, 2 + 7 * 8, - 3 + 3 * 8, 4 + 1 * 8, 5 + 0 * 8, 4 + 2 * 8, - 3 + 4 * 8, 3 + 5 * 8, 3 + 6 * 8, 3 + 7 * 8, - 4 + 3 * 8, 5 + 1 * 8, 6 + 0 * 8, 5 + 2 * 8, - 4 + 4 * 8, 4 + 5 * 8, 4 + 6 * 8, 4 + 7 * 8, - 5 + 3 * 8, 6 + 1 * 8, 6 + 2 * 8, 5 + 4 * 8, - 5 + 5 * 8, 5 + 6 * 8, 5 + 7 * 8, 6 + 3 * 8, - 7 + 0 * 8, 7 + 1 * 8, 6 + 4 * 8, 6 + 5 * 8, - 6 + 6 * 8, 6 + 7 * 8, 7 + 2 * 8, 7 + 3 * 8, - 7 + 4 * 8, 7 + 5 * 8, 7 + 6 * 8, 7 + 7 * 8, -}; - -static const uint8_t field_scan8x8_cavlc[64] = { - 0 + 0 * 8, 1 + 1 * 8, 2 + 0 * 8, 0 + 7 * 8, - 2 + 2 * 8, 2 + 3 * 8, 2 + 4 * 8, 3 + 3 * 8, - 3 + 4 * 8, 4 + 3 * 8, 4 + 4 * 8, 5 + 3 * 8, - 5 + 5 * 8, 7 + 0 * 8, 6 + 6 * 8, 7 + 4 * 8, - 0 + 1 * 8, 0 + 3 * 8, 1 + 3 * 8, 1 + 4 * 8, - 1 + 5 * 8, 3 + 1 * 8, 2 + 5 * 8, 4 + 1 * 8, - 3 + 5 * 8, 5 + 1 * 8, 4 + 5 * 8, 6 + 1 * 8, - 5 + 6 * 8, 7 + 1 * 8, 6 + 7 * 8, 7 + 5 * 8, - 0 + 2 * 8, 0 + 4 * 8, 0 + 5 * 8, 2 + 1 * 8, - 1 + 6 * 8, 4 + 0 * 8, 2 + 6 * 8, 5 + 0 * 8, - 3 + 6 * 8, 6 + 0 * 8, 4 + 6 * 8, 6 + 2 * 8, - 5 + 7 * 8, 6 + 4 * 8, 7 + 2 * 8, 7 + 6 * 8, - 1 + 0 * 8, 1 + 2 * 8, 0 + 6 * 8, 3 + 0 * 8, - 1 + 7 * 8, 3 + 2 * 8, 2 + 7 * 8, 4 + 2 * 8, - 3 + 7 * 8, 5 + 2 * 8, 4 + 7 * 8, 5 + 4 * 8, - 6 + 3 * 8, 6 + 5 * 8, 7 + 3 * 8, 7 + 7 * 8, -}; - -// zigzag_scan8x8_cavlc[i] = zigzag_scan8x8[(i/4) + 16*(i%4)] -static const uint8_t zigzag_scan8x8_cavlc[64] = { - 0 + 0 * 8, 1 + 1 * 8, 1 + 2 * 8, 2 + 2 * 8, - 4 + 1 * 8, 0 + 5 * 8, 3 + 3 * 8, 7 + 0 * 8, - 3 + 4 * 8, 1 + 7 * 8, 5 + 3 * 8, 6 + 3 * 8, - 2 + 7 * 8, 6 + 4 * 8, 5 + 6 * 8, 7 + 5 * 8, - 1 + 0 * 8, 2 + 0 * 8, 0 + 3 * 8, 3 + 1 * 8, - 3 + 2 * 8, 0 + 6 * 8, 4 + 2 * 8, 6 + 1 * 8, - 2 + 5 * 8, 2 + 6 * 8, 6 + 2 * 8, 5 + 4 * 8, - 3 + 7 * 8, 7 + 3 * 8, 4 + 7 * 8, 7 + 6 * 8, - 0 + 1 * 8, 3 + 0 * 8, 0 + 4 * 8, 4 + 0 * 8, - 2 + 3 * 8, 1 + 5 * 8, 5 + 1 * 8, 5 + 2 * 8, - 1 + 6 * 8, 3 + 5 * 8, 7 + 1 * 8, 4 + 5 * 8, - 4 + 6 * 8, 7 + 4 * 8, 5 + 7 * 8, 6 + 7 * 8, - 0 + 2 * 8, 2 + 1 * 8, 1 + 3 * 8, 5 + 0 * 8, - 1 + 4 * 8, 2 + 4 * 8, 6 + 0 * 8, 4 + 3 * 8, - 0 + 7 * 8, 4 + 4 * 8, 7 + 2 * 8, 3 + 6 * 8, - 5 + 5 * 8, 6 + 5 * 8, 6 + 6 * 8, 7 + 7 * 8, -}; - -static const uint8_t dequant4_coeff_init[6][3] = { - { 10, 13, 16 }, - { 11, 14, 18 }, - { 13, 16, 20 }, - { 14, 18, 23 }, - { 16, 20, 25 }, - { 18, 23, 29 }, -}; - -static const uint8_t dequant8_coeff_init_scan[16] = { - 0, 3, 4, 3, 3, 1, 5, 1, 4, 5, 2, 5, 3, 1, 5, 1 -}; - -static const uint8_t dequant8_coeff_init[6][6] = { - { 20, 18, 32, 19, 25, 24 }, - { 22, 19, 35, 21, 28, 26 }, - { 26, 23, 42, 24, 33, 31 }, - { 28, 25, 45, 26, 35, 33 }, - { 32, 28, 51, 30, 40, 38 }, - { 36, 32, 58, 34, 46, 43 }, -}; - -static const enum AVPixelFormat h264_hwaccel_pixfmt_list_420[] = { -#if CONFIG_H264_DXVA2_HWACCEL - AV_PIX_FMT_DXVA2_VLD, -#endif -#if CONFIG_H264_VAAPI_HWACCEL - AV_PIX_FMT_VAAPI_VLD, -#endif -#if CONFIG_H264_VDA_HWACCEL - AV_PIX_FMT_VDA_VLD, -#endif -#if CONFIG_H264_VDPAU_HWACCEL - AV_PIX_FMT_VDPAU, -#endif - AV_PIX_FMT_YUV420P, - AV_PIX_FMT_NONE -}; - -static const enum AVPixelFormat h264_hwaccel_pixfmt_list_jpeg_420[] = { -#if CONFIG_H264_DXVA2_HWACCEL - AV_PIX_FMT_DXVA2_VLD, -#endif -#if CONFIG_H264_VAAPI_HWACCEL - AV_PIX_FMT_VAAPI_VLD, -#endif -#if CONFIG_H264_VDA_HWACCEL - AV_PIX_FMT_VDA_VLD, -#endif -#if CONFIG_H264_VDPAU_HWACCEL - AV_PIX_FMT_VDPAU, -#endif - AV_PIX_FMT_YUVJ420P, - AV_PIX_FMT_NONE -}; - static void h264_er_decode_mb(void *opaque, int ref, int mv_dir, int mv_type, int (*mv)[2][4][2], int mb_x, int mb_y, int mb_intra, int mb_skipped) @@ -253,231 +120,6 @@ void ff_h264_draw_horiz_band(H264Context *h, int y, int height) } } -static void unref_picture(H264Context *h, H264Picture *pic) -{ - int off = offsetof(H264Picture, tf) + sizeof(pic->tf); - int i; - - if (!pic->f.buf[0]) - return; - - ff_thread_release_buffer(h->avctx, &pic->tf); - av_buffer_unref(&pic->hwaccel_priv_buf); - - av_buffer_unref(&pic->qscale_table_buf); - av_buffer_unref(&pic->mb_type_buf); - for (i = 0; i < 2; i++) { - av_buffer_unref(&pic->motion_val_buf[i]); - av_buffer_unref(&pic->ref_index_buf[i]); - } - - memset((uint8_t*)pic + off, 0, sizeof(*pic) - off); -} - -static void release_unused_pictures(H264Context *h, int remove_current) -{ - int i; - - /* release non reference frames */ - for (i = 0; i < H264_MAX_PICTURE_COUNT; i++) { - if (h->DPB[i].f.buf[0] && !h->DPB[i].reference && - (remove_current || &h->DPB[i] != h->cur_pic_ptr)) { - unref_picture(h, &h->DPB[i]); - } - } -} - -static int ref_picture(H264Context *h, H264Picture *dst, H264Picture *src) -{ - int ret, i; - - av_assert0(!dst->f.buf[0]); - av_assert0(src->f.buf[0]); - - src->tf.f = &src->f; - dst->tf.f = &dst->f; - ret = ff_thread_ref_frame(&dst->tf, &src->tf); - if (ret < 0) - goto fail; - - dst->qscale_table_buf = av_buffer_ref(src->qscale_table_buf); - dst->mb_type_buf = av_buffer_ref(src->mb_type_buf); - if (!dst->qscale_table_buf || !dst->mb_type_buf) - goto fail; - dst->qscale_table = src->qscale_table; - dst->mb_type = src->mb_type; - - for (i = 0; i < 2; i++) { - dst->motion_val_buf[i] = av_buffer_ref(src->motion_val_buf[i]); - dst->ref_index_buf[i] = av_buffer_ref(src->ref_index_buf[i]); - if (!dst->motion_val_buf[i] || !dst->ref_index_buf[i]) - goto fail; - dst->motion_val[i] = src->motion_val[i]; - dst->ref_index[i] = src->ref_index[i]; - } - - if (src->hwaccel_picture_private) { - dst->hwaccel_priv_buf = av_buffer_ref(src->hwaccel_priv_buf); - if (!dst->hwaccel_priv_buf) - goto fail; - dst->hwaccel_picture_private = dst->hwaccel_priv_buf->data; - } - - for (i = 0; i < 2; i++) - dst->field_poc[i] = src->field_poc[i]; - - memcpy(dst->ref_poc, src->ref_poc, sizeof(src->ref_poc)); - memcpy(dst->ref_count, src->ref_count, sizeof(src->ref_count)); - - dst->poc = src->poc; - dst->frame_num = src->frame_num; - dst->mmco_reset = src->mmco_reset; - dst->pic_id = src->pic_id; - dst->long_ref = src->long_ref; - dst->mbaff = src->mbaff; - dst->field_picture = src->field_picture; - dst->needs_realloc = src->needs_realloc; - dst->reference = src->reference; - dst->recovered = src->recovered; - - return 0; -fail: - unref_picture(h, dst); - return ret; -} - -static int alloc_scratch_buffers(H264Context *h, int linesize) -{ - int alloc_size = FFALIGN(FFABS(linesize) + 32, 32); - - if (h->bipred_scratchpad) - return 0; - - h->bipred_scratchpad = av_malloc(16 * 6 * alloc_size); - // edge emu needs blocksize + filter length - 1 - // (= 21x21 for h264) - h->edge_emu_buffer = av_mallocz(alloc_size * 2 * 21); - - if (!h->bipred_scratchpad || !h->edge_emu_buffer) { - av_freep(&h->bipred_scratchpad); - av_freep(&h->edge_emu_buffer); - return AVERROR(ENOMEM); - } - - return 0; -} - -static int init_table_pools(H264Context *h) -{ - const int big_mb_num = h->mb_stride * (h->mb_height + 1) + 1; - const int mb_array_size = h->mb_stride * h->mb_height; - const int b4_stride = h->mb_width * 4 + 1; - const int b4_array_size = b4_stride * h->mb_height * 4; - - h->qscale_table_pool = av_buffer_pool_init(big_mb_num + h->mb_stride, - av_buffer_allocz); - h->mb_type_pool = av_buffer_pool_init((big_mb_num + h->mb_stride) * - sizeof(uint32_t), av_buffer_allocz); - h->motion_val_pool = av_buffer_pool_init(2 * (b4_array_size + 4) * - sizeof(int16_t), av_buffer_allocz); - h->ref_index_pool = av_buffer_pool_init(4 * mb_array_size, av_buffer_allocz); - - if (!h->qscale_table_pool || !h->mb_type_pool || !h->motion_val_pool || - !h->ref_index_pool) { - av_buffer_pool_uninit(&h->qscale_table_pool); - av_buffer_pool_uninit(&h->mb_type_pool); - av_buffer_pool_uninit(&h->motion_val_pool); - av_buffer_pool_uninit(&h->ref_index_pool); - return AVERROR(ENOMEM); - } - - return 0; -} - -static int alloc_picture(H264Context *h, H264Picture *pic) -{ - int i, ret = 0; - - av_assert0(!pic->f.data[0]); - - pic->tf.f = &pic->f; - ret = ff_thread_get_buffer(h->avctx, &pic->tf, pic->reference ? - AV_GET_BUFFER_FLAG_REF : 0); - if (ret < 0) - goto fail; - - h->linesize = pic->f.linesize[0]; - h->uvlinesize = pic->f.linesize[1]; - - if (h->avctx->hwaccel) { - const AVHWAccel *hwaccel = h->avctx->hwaccel; - av_assert0(!pic->hwaccel_picture_private); - if (hwaccel->priv_data_size) { - pic->hwaccel_priv_buf = av_buffer_allocz(hwaccel->priv_data_size); - if (!pic->hwaccel_priv_buf) - return AVERROR(ENOMEM); - pic->hwaccel_picture_private = pic->hwaccel_priv_buf->data; - } - } - - if (!h->qscale_table_pool) { - ret = init_table_pools(h); - if (ret < 0) - goto fail; - } - - pic->qscale_table_buf = av_buffer_pool_get(h->qscale_table_pool); - pic->mb_type_buf = av_buffer_pool_get(h->mb_type_pool); - if (!pic->qscale_table_buf || !pic->mb_type_buf) - goto fail; - - pic->mb_type = (uint32_t*)pic->mb_type_buf->data + 2 * h->mb_stride + 1; - pic->qscale_table = pic->qscale_table_buf->data + 2 * h->mb_stride + 1; - - for (i = 0; i < 2; i++) { - pic->motion_val_buf[i] = av_buffer_pool_get(h->motion_val_pool); - pic->ref_index_buf[i] = av_buffer_pool_get(h->ref_index_pool); - if (!pic->motion_val_buf[i] || !pic->ref_index_buf[i]) - goto fail; - - pic->motion_val[i] = (int16_t (*)[2])pic->motion_val_buf[i]->data + 4; - pic->ref_index[i] = pic->ref_index_buf[i]->data; - } - - return 0; -fail: - unref_picture(h, pic); - return (ret < 0) ? ret : AVERROR(ENOMEM); -} - -static inline int pic_is_unused(H264Context *h, H264Picture *pic) -{ - if (!pic->f.buf[0]) - return 1; - if (pic->needs_realloc && !(pic->reference & DELAYED_PIC_REF)) - return 1; - return 0; -} - -static int find_unused_picture(H264Context *h) -{ - int i; - - for (i = 0; i < H264_MAX_PICTURE_COUNT; i++) { - if (pic_is_unused(h, &h->DPB[i])) - break; - } - if (i == H264_MAX_PICTURE_COUNT) - return AVERROR_INVALIDDATA; - - if (h->DPB[i].needs_realloc) { - h->DPB[i].needs_realloc = 0; - unref_picture(h, &h->DPB[i]); - } - - return i; -} - /** * Check if the top & left blocks are available if needed and * change the dc mode so it only uses the available blocks. @@ -697,7 +339,7 @@ static int decode_rbsp_trailing(H264Context *h, const uint8_t *src) return 0; } -static void free_tables(H264Context *h, int free_rbsp) +void ff_h264_free_tables(H264Context *h, int free_rbsp) { int i; H264Context *hx; @@ -723,7 +365,7 @@ static void free_tables(H264Context *h, int free_rbsp) if (free_rbsp && h->DPB) { for (i = 0; i < H264_MAX_PICTURE_COUNT; i++) - unref_picture(h, &h->DPB[i]); + ff_h264_unref_picture(h, &h->DPB[i]); av_freep(&h->DPB); } else if (h->DPB) { for (i = 0; i < H264_MAX_PICTURE_COUNT; i++) @@ -758,76 +400,6 @@ static void free_tables(H264Context *h, int free_rbsp) } } -static void init_dequant8_coeff_table(H264Context *h) -{ - int i, j, q, x; - const int max_qp = 51 + 6 * (h->sps.bit_depth_luma - 8); - - for (i = 0; i < 6; i++) { - h->dequant8_coeff[i] = h->dequant8_buffer[i]; - for (j = 0; j < i; j++) - if (!memcmp(h->pps.scaling_matrix8[j], h->pps.scaling_matrix8[i], - 64 * sizeof(uint8_t))) { - h->dequant8_coeff[i] = h->dequant8_buffer[j]; - break; - } - if (j < i) - continue; - - for (q = 0; q < max_qp + 1; q++) { - int shift = div6[q]; - int idx = rem6[q]; - for (x = 0; x < 64; x++) - h->dequant8_coeff[i][q][(x >> 3) | ((x & 7) << 3)] = - ((uint32_t)dequant8_coeff_init[idx][dequant8_coeff_init_scan[((x >> 1) & 12) | (x & 3)]] * - h->pps.scaling_matrix8[i][x]) << shift; - } - } -} - -static void init_dequant4_coeff_table(H264Context *h) -{ - int i, j, q, x; - const int max_qp = 51 + 6 * (h->sps.bit_depth_luma - 8); - for (i = 0; i < 6; i++) { - h->dequant4_coeff[i] = h->dequant4_buffer[i]; - for (j = 0; j < i; j++) - if (!memcmp(h->pps.scaling_matrix4[j], h->pps.scaling_matrix4[i], - 16 * sizeof(uint8_t))) { - h->dequant4_coeff[i] = h->dequant4_buffer[j]; - break; - } - if (j < i) - continue; - - for (q = 0; q < max_qp + 1; q++) { - int shift = div6[q] + 2; - int idx = rem6[q]; - for (x = 0; x < 16; x++) - h->dequant4_coeff[i][q][(x >> 2) | ((x << 2) & 0xF)] = - ((uint32_t)dequant4_coeff_init[idx][(x & 1) + ((x >> 2) & 1)] * - h->pps.scaling_matrix4[i][x]) << shift; - } - } -} - -static void init_dequant_tables(H264Context *h) -{ - int i, x; - init_dequant4_coeff_table(h); - if (h->pps.transform_8x8_mode) - init_dequant8_coeff_table(h); - if (h->sps.transform_bypass) { - for (i = 0; i < 6; i++) - for (x = 0; x < 16; x++) - h->dequant4_coeff[i][0][x] = 1 << 6; - if (h->pps.transform_8x8_mode) - for (i = 0; i < 6; i++) - for (x = 0; x < 64; x++) - h->dequant8_coeff[i][0][x] = 1 << 6; - } -} - int ff_h264_alloc_tables(H264Context *h) { const int big_mb_num = h->mb_stride * (h->mb_height + 1); @@ -871,7 +443,7 @@ int ff_h264_alloc_tables(H264Context *h) } if (!h->dequant4_coeff[0]) - init_dequant_tables(h); + h264_init_dequant_tables(h); if (!h->DPB) { h->DPB = av_mallocz_array(H264_MAX_PICTURE_COUNT, sizeof(*h->DPB)); @@ -885,40 +457,15 @@ int ff_h264_alloc_tables(H264Context *h) return 0; fail: - free_tables(h, 1); + ff_h264_free_tables(h, 1); return AVERROR(ENOMEM); } -/** - * Mimic alloc_tables(), but for every context thread. - */ -static void clone_tables(H264Context *dst, H264Context *src, int i) -{ - dst->intra4x4_pred_mode = src->intra4x4_pred_mode + i * 8 * 2 * src->mb_stride; - dst->non_zero_count = src->non_zero_count; - dst->slice_table = src->slice_table; - dst->cbp_table = src->cbp_table; - dst->mb2b_xy = src->mb2b_xy; - dst->mb2br_xy = src->mb2br_xy; - dst->chroma_pred_mode_table = src->chroma_pred_mode_table; - dst->mvd_table[0] = src->mvd_table[0] + i * 8 * 2 * src->mb_stride; - dst->mvd_table[1] = src->mvd_table[1] + i * 8 * 2 * src->mb_stride; - dst->direct_table = src->direct_table; - dst->list_counts = src->list_counts; - dst->DPB = src->DPB; - dst->cur_pic_ptr = src->cur_pic_ptr; - dst->cur_pic = src->cur_pic; - dst->bipred_scratchpad = NULL; - dst->edge_emu_buffer = NULL; - ff_h264_pred_init(&dst->hpc, src->avctx->codec_id, src->sps.bit_depth_luma, - src->sps.chroma_format_idc); -} - /** * Init context * Allocate buffers which are not shared amongst multiple threads. */ -static int context_init(H264Context *h) +int ff_h264_context_init(H264Context *h) { ERContext *er = &h->er; int mb_array_size = h->mb_height * h->mb_stride; @@ -984,7 +531,7 @@ static int context_init(H264Context *h) return 0; fail: - return AVERROR(ENOMEM); // free_tables will clean up for us + return AVERROR(ENOMEM); // ff_h264_free_tables will clean up for us } static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size, @@ -1126,48 +673,6 @@ av_cold int ff_h264_decode_init(AVCodecContext *avctx) return 0; } -#define IN_RANGE(a, b, size) (((a) >= (b)) && ((a) < ((b) + (size)))) -#undef REBASE_PICTURE -#define REBASE_PICTURE(pic, new_ctx, old_ctx) \ - ((pic && pic >= old_ctx->DPB && \ - pic < old_ctx->DPB + H264_MAX_PICTURE_COUNT) ? \ - &new_ctx->DPB[pic - old_ctx->DPB] : NULL) - -static void copy_picture_range(H264Picture **to, H264Picture **from, int count, - H264Context *new_base, - H264Context *old_base) -{ - int i; - - for (i = 0; i < count; i++) { - assert((IN_RANGE(from[i], old_base, sizeof(*old_base)) || - IN_RANGE(from[i], old_base->DPB, - sizeof(H264Picture) * H264_MAX_PICTURE_COUNT) || - !from[i])); - to[i] = REBASE_PICTURE(from[i], new_base, old_base); - } -} - -static int copy_parameter_set(void **to, void **from, int count, int size) -{ - int i; - - for (i = 0; i < count; i++) { - if (to[i] && !from[i]) { - av_freep(&to[i]); - } else if (from[i] && !to[i]) { - to[i] = av_malloc(size); - if (!to[i]) - return AVERROR(ENOMEM); - } - - if (from[i]) - memcpy(to[i], from[i], size); - } - - return 0; -} - static int decode_init_thread_copy(AVCodecContext *avctx) { H264Context *h = avctx->priv_data; @@ -1182,294 +687,6 @@ static int decode_init_thread_copy(AVCodecContext *avctx) return 0; } -#define copy_fields(to, from, start_field, end_field) \ - memcpy(&to->start_field, &from->start_field, \ - (char *)&to->end_field - (char *)&to->start_field) - -static int h264_slice_header_init(H264Context *, int); - -static int h264_set_parameter_from_sps(H264Context *h); - -static int decode_update_thread_context(AVCodecContext *dst, - const AVCodecContext *src) -{ - H264Context *h = dst->priv_data, *h1 = src->priv_data; - int inited = h->context_initialized, err = 0; - int context_reinitialized = 0; - int i, ret; - - if (dst == src || !h1->context_initialized) - return 0; - - if (inited && - (h->width != h1->width || - h->height != h1->height || - h->mb_width != h1->mb_width || - h->mb_height != h1->mb_height || - h->sps.bit_depth_luma != h1->sps.bit_depth_luma || - h->sps.chroma_format_idc != h1->sps.chroma_format_idc || - h->sps.colorspace != h1->sps.colorspace)) { - - /* set bits_per_raw_sample to the previous value. the check for changed - * bit depth in h264_set_parameter_from_sps() uses it and sets it to - * the current value */ - h->avctx->bits_per_raw_sample = h->sps.bit_depth_luma; - - av_freep(&h->bipred_scratchpad); - - h->width = h1->width; - h->height = h1->height; - h->mb_height = h1->mb_height; - h->mb_width = h1->mb_width; - h->mb_num = h1->mb_num; - h->mb_stride = h1->mb_stride; - h->b_stride = h1->b_stride; - - if ((err = h264_slice_header_init(h, 1)) < 0) { - av_log(h->avctx, AV_LOG_ERROR, "h264_slice_header_init() failed"); - return err; - } - context_reinitialized = 1; - - /* update linesize on resize. The decoder doesn't - * necessarily call h264_frame_start in the new thread */ - h->linesize = h1->linesize; - h->uvlinesize = h1->uvlinesize; - - /* copy block_offset since frame_start may not be called */ - memcpy(h->block_offset, h1->block_offset, sizeof(h->block_offset)); - } - - if (!inited) { - for (i = 0; i < MAX_SPS_COUNT; i++) - av_freep(h->sps_buffers + i); - - for (i = 0; i < MAX_PPS_COUNT; i++) - av_freep(h->pps_buffers + i); - - memcpy(h, h1, sizeof(*h1)); - memset(h->sps_buffers, 0, sizeof(h->sps_buffers)); - memset(h->pps_buffers, 0, sizeof(h->pps_buffers)); - memset(&h->er, 0, sizeof(h->er)); - memset(&h->mb, 0, sizeof(h->mb)); - memset(&h->mb_luma_dc, 0, sizeof(h->mb_luma_dc)); - memset(&h->mb_padding, 0, sizeof(h->mb_padding)); - h->context_initialized = 0; - - memset(&h->cur_pic, 0, sizeof(h->cur_pic)); - av_frame_unref(&h->cur_pic.f); - h->cur_pic.tf.f = &h->cur_pic.f; - - h->avctx = dst; - h->DPB = NULL; - h->qscale_table_pool = NULL; - h->mb_type_pool = NULL; - h->ref_index_pool = NULL; - h->motion_val_pool = NULL; - - ret = ff_h264_alloc_tables(h); - if (ret < 0) { - av_log(dst, AV_LOG_ERROR, "Could not allocate memory\n"); - return ret; - } - ret = context_init(h); - if (ret < 0) { - av_log(dst, AV_LOG_ERROR, "context_init() failed.\n"); - return ret; - } - - for (i = 0; i < 2; i++) { - h->rbsp_buffer[i] = NULL; - h->rbsp_buffer_size[i] = 0; - } - h->bipred_scratchpad = NULL; - h->edge_emu_buffer = NULL; - - h->thread_context[0] = h; - - h->context_initialized = 1; - } - - h->avctx->coded_height = h1->avctx->coded_height; - h->avctx->coded_width = h1->avctx->coded_width; - h->avctx->width = h1->avctx->width; - h->avctx->height = h1->avctx->height; - h->coded_picture_number = h1->coded_picture_number; - h->first_field = h1->first_field; - h->picture_structure = h1->picture_structure; - h->qscale = h1->qscale; - h->droppable = h1->droppable; - h->low_delay = h1->low_delay; - - for (i = 0; i < H264_MAX_PICTURE_COUNT; i++) { - unref_picture(h, &h->DPB[i]); - if (h1->DPB[i].f.buf[0] && - (ret = ref_picture(h, &h->DPB[i], &h1->DPB[i])) < 0) - return ret; - } - - h->cur_pic_ptr = REBASE_PICTURE(h1->cur_pic_ptr, h, h1); - unref_picture(h, &h->cur_pic); - if ((ret = ref_picture(h, &h->cur_pic, &h1->cur_pic)) < 0) - return ret; - - h->workaround_bugs = h1->workaround_bugs; - h->low_delay = h1->low_delay; - h->droppable = h1->droppable; - - /* frame_start may not be called for the next thread (if it's decoding - * a bottom field) so this has to be allocated here */ - err = alloc_scratch_buffers(h, h1->linesize); - if (err < 0) - return err; - - // extradata/NAL handling - h->is_avc = h1->is_avc; - - // SPS/PPS - if ((ret = copy_parameter_set((void **)h->sps_buffers, - (void **)h1->sps_buffers, - MAX_SPS_COUNT, sizeof(SPS))) < 0) - return ret; - h->sps = h1->sps; - if ((ret = copy_parameter_set((void **)h->pps_buffers, - (void **)h1->pps_buffers, - MAX_PPS_COUNT, sizeof(PPS))) < 0) - return ret; - h->pps = h1->pps; - - // Dequantization matrices - // FIXME these are big - can they be only copied when PPS changes? - copy_fields(h, h1, dequant4_buffer, dequant4_coeff); - - for (i = 0; i < 6; i++) - h->dequant4_coeff[i] = h->dequant4_buffer[0] + - (h1->dequant4_coeff[i] - h1->dequant4_buffer[0]); - - for (i = 0; i < 6; i++) - h->dequant8_coeff[i] = h->dequant8_buffer[0] + - (h1->dequant8_coeff[i] - h1->dequant8_buffer[0]); - - h->dequant_coeff_pps = h1->dequant_coeff_pps; - - // POC timing - copy_fields(h, h1, poc_lsb, redundant_pic_count); - - // reference lists - copy_fields(h, h1, short_ref, cabac_init_idc); - - copy_picture_range(h->short_ref, h1->short_ref, 32, h, h1); - copy_picture_range(h->long_ref, h1->long_ref, 32, h, h1); - copy_picture_range(h->delayed_pic, h1->delayed_pic, - MAX_DELAYED_PIC_COUNT + 2, h, h1); - - h->last_slice_type = h1->last_slice_type; - - if (context_reinitialized) - h264_set_parameter_from_sps(h); - - if (!h->cur_pic_ptr) - return 0; - - if (!h->droppable) { - err = ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index); - h->prev_poc_msb = h->poc_msb; - h->prev_poc_lsb = h->poc_lsb; - } - h->prev_frame_num_offset = h->frame_num_offset; - h->prev_frame_num = h->frame_num; - h->outputed_poc = h->next_outputed_poc; - - h->recovery_frame = h1->recovery_frame; - h->frame_recovered = h1->frame_recovered; - - return err; -} - -static int h264_frame_start(H264Context *h) -{ - H264Picture *pic; - int i, ret; - const int pixel_shift = h->pixel_shift; - - release_unused_pictures(h, 1); - h->cur_pic_ptr = NULL; - - i = find_unused_picture(h); - if (i < 0) { - av_log(h->avctx, AV_LOG_ERROR, "no frame buffer available\n"); - return i; - } - pic = &h->DPB[i]; - - pic->reference = h->droppable ? 0 : h->picture_structure; - pic->f.coded_picture_number = h->coded_picture_number++; - pic->field_picture = h->picture_structure != PICT_FRAME; - /* - * Zero key_frame here; IDR markings per slice in frame or fields are ORed - * in later. - * See decode_nal_units(). - */ - pic->f.key_frame = 0; - pic->mmco_reset = 0; - pic->recovered = 0; - - if ((ret = alloc_picture(h, pic)) < 0) - return ret; - - h->cur_pic_ptr = pic; - unref_picture(h, &h->cur_pic); - if ((ret = ref_picture(h, &h->cur_pic, h->cur_pic_ptr)) < 0) - return ret; - - if (CONFIG_ERROR_RESILIENCE) - ff_er_frame_start(&h->er); - - assert(h->linesize && h->uvlinesize); - - for (i = 0; i < 16; i++) { - h->block_offset[i] = (4 * ((scan8[i] - scan8[0]) & 7) << pixel_shift) + 4 * h->linesize * ((scan8[i] - scan8[0]) >> 3); - h->block_offset[48 + i] = (4 * ((scan8[i] - scan8[0]) & 7) << pixel_shift) + 8 * h->linesize * ((scan8[i] - scan8[0]) >> 3); - } - for (i = 0; i < 16; i++) { - h->block_offset[16 + i] = - h->block_offset[32 + i] = (4 * ((scan8[i] - scan8[0]) & 7) << pixel_shift) + 4 * h->uvlinesize * ((scan8[i] - scan8[0]) >> 3); - h->block_offset[48 + 16 + i] = - h->block_offset[48 + 32 + i] = (4 * ((scan8[i] - scan8[0]) & 7) << pixel_shift) + 8 * h->uvlinesize * ((scan8[i] - scan8[0]) >> 3); - } - - /* can't be in alloc_tables because linesize isn't known there. - * FIXME: redo bipred weight to not require extra buffer? */ - for (i = 0; i < h->slice_context_count; i++) - if (h->thread_context[i]) { - ret = alloc_scratch_buffers(h->thread_context[i], h->linesize); - if (ret < 0) - return ret; - } - - /* Some macroblocks can be accessed before they're available in case - * of lost slices, MBAFF or threading. */ - memset(h->slice_table, -1, - (h->mb_height * h->mb_stride - 1) * sizeof(*h->slice_table)); - - // s->decode = (s->flags & CODEC_FLAG_PSNR) || !s->encoding || - // s->current_picture.f.reference /* || h->contains_intra */ || 1; - - /* We mark the current picture as non-reference after allocating it, so - * that if we break out due to an error it can be released automatically - * in the next ff_MPV_frame_start(). - */ - h->cur_pic_ptr->reference = 0; - - h->cur_pic_ptr->field_poc[0] = h->cur_pic_ptr->field_poc[1] = INT_MAX; - - h->next_output_pic = NULL; - - assert(h->cur_pic_ptr->long_ref == 0); - - return 0; -} - /** * Run setup operations that must be run after slice header decoding. * This includes finding the next displayed frame. @@ -1737,106 +954,10 @@ static void decode_postinit(H264Context *h, int setup_finished) ff_thread_finish_setup(h->avctx); } -static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, - uint8_t *src_cb, uint8_t *src_cr, - int linesize, int uvlinesize, - int simple) +int ff_pred_weight_table(H264Context *h) { - uint8_t *top_border; - int top_idx = 1; - const int pixel_shift = h->pixel_shift; - int chroma444 = CHROMA444(h); - int chroma422 = CHROMA422(h); - - src_y -= linesize; - src_cb -= uvlinesize; - src_cr -= uvlinesize; - - if (!simple && FRAME_MBAFF(h)) { - if (h->mb_y & 1) { - if (!MB_MBAFF(h)) { - top_border = h->top_borders[0][h->mb_x]; - AV_COPY128(top_border, src_y + 15 * linesize); - if (pixel_shift) - AV_COPY128(top_border + 16, src_y + 15 * linesize + 16); - if (simple || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) { - if (chroma444) { - if (pixel_shift) { - AV_COPY128(top_border + 32, src_cb + 15 * uvlinesize); - AV_COPY128(top_border + 48, src_cb + 15 * uvlinesize + 16); - AV_COPY128(top_border + 64, src_cr + 15 * uvlinesize); - AV_COPY128(top_border + 80, src_cr + 15 * uvlinesize + 16); - } else { - AV_COPY128(top_border + 16, src_cb + 15 * uvlinesize); - AV_COPY128(top_border + 32, src_cr + 15 * uvlinesize); - } - } else if (chroma422) { - if (pixel_shift) { - AV_COPY128(top_border + 32, src_cb + 15 * uvlinesize); - AV_COPY128(top_border + 48, src_cr + 15 * uvlinesize); - } else { - AV_COPY64(top_border + 16, src_cb + 15 * uvlinesize); - AV_COPY64(top_border + 24, src_cr + 15 * uvlinesize); - } - } else { - if (pixel_shift) { - AV_COPY128(top_border + 32, src_cb + 7 * uvlinesize); - AV_COPY128(top_border + 48, src_cr + 7 * uvlinesize); - } else { - AV_COPY64(top_border + 16, src_cb + 7 * uvlinesize); - AV_COPY64(top_border + 24, src_cr + 7 * uvlinesize); - } - } - } - } - } else if (MB_MBAFF(h)) { - top_idx = 0; - } else - return; - } - - top_border = h->top_borders[top_idx][h->mb_x]; - /* There are two lines saved, the line above the top macroblock - * of a pair, and the line above the bottom macroblock. */ - AV_COPY128(top_border, src_y + 16 * linesize); - if (pixel_shift) - AV_COPY128(top_border + 16, src_y + 16 * linesize + 16); - - if (simple || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) { - if (chroma444) { - if (pixel_shift) { - AV_COPY128(top_border + 32, src_cb + 16 * linesize); - AV_COPY128(top_border + 48, src_cb + 16 * linesize + 16); - AV_COPY128(top_border + 64, src_cr + 16 * linesize); - AV_COPY128(top_border + 80, src_cr + 16 * linesize + 16); - } else { - AV_COPY128(top_border + 16, src_cb + 16 * linesize); - AV_COPY128(top_border + 32, src_cr + 16 * linesize); - } - } else if (chroma422) { - if (pixel_shift) { - AV_COPY128(top_border + 32, src_cb + 16 * uvlinesize); - AV_COPY128(top_border + 48, src_cr + 16 * uvlinesize); - } else { - AV_COPY64(top_border + 16, src_cb + 16 * uvlinesize); - AV_COPY64(top_border + 24, src_cr + 16 * uvlinesize); - } - } else { - if (pixel_shift) { - AV_COPY128(top_border + 32, src_cb + 8 * uvlinesize); - AV_COPY128(top_border + 48, src_cr + 8 * uvlinesize); - } else { - AV_COPY64(top_border + 16, src_cb + 8 * uvlinesize); - AV_COPY64(top_border + 24, src_cr + 8 * uvlinesize); - } - } - } -} - -int ff_pred_weight_table(H264Context *h) -{ - int list, i; - int luma_def, chroma_def; + int list, i; + int luma_def, chroma_def; h->use_weight = 0; h->use_weight_chroma = 0; @@ -1895,72 +1016,6 @@ int ff_pred_weight_table(H264Context *h) return 0; } -/** - * Initialize implicit_weight table. - * @param field 0/1 initialize the weight for interlaced MBAFF - * -1 initializes the rest - */ -static void implicit_weight_table(H264Context *h, int field) -{ - int ref0, ref1, i, cur_poc, ref_start, ref_count0, ref_count1; - - for (i = 0; i < 2; i++) { - h->luma_weight_flag[i] = 0; - h->chroma_weight_flag[i] = 0; - } - - if (field < 0) { - if (h->picture_structure == PICT_FRAME) { - cur_poc = h->cur_pic_ptr->poc; - } else { - cur_poc = h->cur_pic_ptr->field_poc[h->picture_structure - 1]; - } - if (h->ref_count[0] == 1 && h->ref_count[1] == 1 && !FRAME_MBAFF(h) && - h->ref_list[0][0].poc + h->ref_list[1][0].poc == 2 * cur_poc) { - h->use_weight = 0; - h->use_weight_chroma = 0; - return; - } - ref_start = 0; - ref_count0 = h->ref_count[0]; - ref_count1 = h->ref_count[1]; - } else { - cur_poc = h->cur_pic_ptr->field_poc[field]; - ref_start = 16; - ref_count0 = 16 + 2 * h->ref_count[0]; - ref_count1 = 16 + 2 * h->ref_count[1]; - } - - h->use_weight = 2; - h->use_weight_chroma = 2; - h->luma_log2_weight_denom = 5; - h->chroma_log2_weight_denom = 5; - - for (ref0 = ref_start; ref0 < ref_count0; ref0++) { - int poc0 = h->ref_list[0][ref0].poc; - for (ref1 = ref_start; ref1 < ref_count1; ref1++) { - int w = 32; - if (!h->ref_list[0][ref0].long_ref && !h->ref_list[1][ref1].long_ref) { - int poc1 = h->ref_list[1][ref1].poc; - int td = av_clip(poc1 - poc0, -128, 127); - if (td) { - int tb = av_clip(cur_poc - poc0, -128, 127); - int tx = (16384 + (FFABS(td) >> 1)) / td; - int dist_scale_factor = (tb * tx + 32) >> 8; - if (dist_scale_factor >= -64 && dist_scale_factor <= 128) - w = 64 - dist_scale_factor; - } - } - if (field < 0) { - h->implicit_weight[ref0][ref1][0] = - h->implicit_weight[ref0][ref1][1] = w; - } else { - h->implicit_weight[ref0][ref1][field] = w; - } - } - } -} - /** * instantaneous decoder refresh. */ @@ -1974,7 +1029,7 @@ static void idr(H264Context *h) } /* forget old pics after a seek */ -static void flush_change(H264Context *h) +void ff_h264_flush_change(H264Context *h) { int i; for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++) @@ -2006,13 +1061,13 @@ static void flush_dpb(AVCodecContext *avctx) h->delayed_pic[i] = NULL; } - flush_change(h); + ff_h264_flush_change(h); if (h->DPB) for (i = 0; i < H264_MAX_PICTURE_COUNT; i++) - unref_picture(h, &h->DPB[i]); + ff_h264_unref_picture(h, &h->DPB[i]); h->cur_pic_ptr = NULL; - unref_picture(h, &h->cur_pic); + ff_h264_unref_picture(h, &h->cur_pic); h->mb_x = h->mb_y = 0; @@ -2023,7 +1078,7 @@ static void flush_dpb(AVCodecContext *avctx) h->parse_context.index = 0; h->parse_context.last_index = 0; - free_tables(h, 1); + ff_h264_free_tables(h, 1); h->context_initialized = 0; } @@ -2105,146 +1160,6 @@ int ff_init_poc(H264Context *h, int pic_field_poc[2], int *pic_poc) return 0; } -/** - * initialize scan tables - */ -static void init_scan_tables(H264Context *h) -{ - int i; - for (i = 0; i < 16; i++) { -#define TRANSPOSE(x) (x >> 2) | ((x << 2) & 0xF) - h->zigzag_scan[i] = TRANSPOSE(zigzag_scan[i]); - h->field_scan[i] = TRANSPOSE(field_scan[i]); -#undef TRANSPOSE - } - for (i = 0; i < 64; i++) { -#define TRANSPOSE(x) (x >> 3) | ((x & 7) << 3) - h->zigzag_scan8x8[i] = TRANSPOSE(ff_zigzag_direct[i]); - h->zigzag_scan8x8_cavlc[i] = TRANSPOSE(zigzag_scan8x8_cavlc[i]); - h->field_scan8x8[i] = TRANSPOSE(field_scan8x8[i]); - h->field_scan8x8_cavlc[i] = TRANSPOSE(field_scan8x8_cavlc[i]); -#undef TRANSPOSE - } - if (h->sps.transform_bypass) { // FIXME same ugly - h->zigzag_scan_q0 = zigzag_scan; - h->zigzag_scan8x8_q0 = ff_zigzag_direct; - h->zigzag_scan8x8_cavlc_q0 = zigzag_scan8x8_cavlc; - h->field_scan_q0 = field_scan; - h->field_scan8x8_q0 = field_scan8x8; - h->field_scan8x8_cavlc_q0 = field_scan8x8_cavlc; - } else { - h->zigzag_scan_q0 = h->zigzag_scan; - h->zigzag_scan8x8_q0 = h->zigzag_scan8x8; - h->zigzag_scan8x8_cavlc_q0 = h->zigzag_scan8x8_cavlc; - h->field_scan_q0 = h->field_scan; - h->field_scan8x8_q0 = h->field_scan8x8; - h->field_scan8x8_cavlc_q0 = h->field_scan8x8_cavlc; - } -} - -#if CONFIG_ERROR_RESILIENCE -static void h264_set_erpic(ERPicture *dst, H264Picture *src) -{ - int i; - - if (!src) - return; - - dst->f = &src->f; - dst->tf = &src->tf; - - for (i = 0; i < 2; i++) { - dst->motion_val[i] = src->motion_val[i]; - dst->ref_index[i] = src->ref_index[i]; - } - - dst->mb_type = src->mb_type; - dst->field_picture = src->field_picture; -} -#endif /* CONFIG_ERROR_RESILIENCE */ - -static int field_end(H264Context *h, int in_setup) -{ - AVCodecContext *const avctx = h->avctx; - int err = 0; - h->mb_y = 0; - - if (!in_setup && !h->droppable) - ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, - h->picture_structure == PICT_BOTTOM_FIELD); - - if (in_setup || !(avctx->active_thread_type & FF_THREAD_FRAME)) { - if (!h->droppable) { - err = ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index); - h->prev_poc_msb = h->poc_msb; - h->prev_poc_lsb = h->poc_lsb; - } - h->prev_frame_num_offset = h->frame_num_offset; - h->prev_frame_num = h->frame_num; - h->outputed_poc = h->next_outputed_poc; - } - - if (avctx->hwaccel) { - if (avctx->hwaccel->end_frame(avctx) < 0) - av_log(avctx, AV_LOG_ERROR, - "hardware accelerator failed to decode picture\n"); - } - - /* - * FIXME: Error handling code does not seem to support interlaced - * when slices span multiple rows - * The ff_er_add_slice calls don't work right for bottom - * fields; they cause massive erroneous error concealing - * Error marking covers both fields (top and bottom). - * This causes a mismatched s->error_count - * and a bad error table. Further, the error count goes to - * INT_MAX when called for bottom field, because mb_y is - * past end by one (callers fault) and resync_mb_y != 0 - * causes problems for the first MB line, too. - */ - if (CONFIG_ERROR_RESILIENCE && !FIELD_PICTURE(h)) { - h264_set_erpic(&h->er.cur_pic, h->cur_pic_ptr); - h264_set_erpic(&h->er.last_pic, - h->ref_count[0] ? &h->ref_list[0][0] : NULL); - h264_set_erpic(&h->er.next_pic, - h->ref_count[1] ? &h->ref_list[1][0] : NULL); - ff_er_frame_end(&h->er); - } - emms_c(); - - h->current_slice = 0; - - return err; -} - -/** - * Replicate H264 "master" context to thread contexts. - */ -static int clone_slice(H264Context *dst, H264Context *src) -{ - memcpy(dst->block_offset, src->block_offset, sizeof(dst->block_offset)); - dst->cur_pic_ptr = src->cur_pic_ptr; - dst->cur_pic = src->cur_pic; - dst->linesize = src->linesize; - dst->uvlinesize = src->uvlinesize; - dst->first_field = src->first_field; - - dst->prev_poc_msb = src->prev_poc_msb; - dst->prev_poc_lsb = src->prev_poc_lsb; - dst->prev_frame_num_offset = src->prev_frame_num_offset; - dst->prev_frame_num = src->prev_frame_num; - dst->short_ref_count = src->short_ref_count; - - memcpy(dst->short_ref, src->short_ref, sizeof(dst->short_ref)); - memcpy(dst->long_ref, src->long_ref, sizeof(dst->long_ref)); - memcpy(dst->default_ref_list, src->default_ref_list, sizeof(dst->default_ref_list)); - - memcpy(dst->dequant4_coeff, src->dequant4_coeff, sizeof(src->dequant4_coeff)); - memcpy(dst->dequant8_coeff, src->dequant8_coeff, sizeof(src->dequant8_coeff)); - - return 0; -} - /** * Compute profile from profile_idc and constraint_set?_flags. * @@ -2272,7 +1187,7 @@ int ff_h264_get_profile(SPS *sps) return profile; } -static int h264_set_parameter_from_sps(H264Context *h) +int ff_h264_set_parameter_from_sps(H264Context *h) { if (h->flags & CODEC_FLAG_LOW_DELAY || (h->sps.bitstream_restriction_flag && @@ -2312,194 +1227,6 @@ static int h264_set_parameter_from_sps(H264Context *h) return 0; } -static enum AVPixelFormat get_pixel_format(H264Context *h) -{ - switch (h->sps.bit_depth_luma) { - case 9: - if (CHROMA444(h)) { - if (h->avctx->colorspace == AVCOL_SPC_RGB) { - return AV_PIX_FMT_GBRP9; - } else - return AV_PIX_FMT_YUV444P9; - } else if (CHROMA422(h)) - return AV_PIX_FMT_YUV422P9; - else - return AV_PIX_FMT_YUV420P9; - break; - case 10: - if (CHROMA444(h)) { - if (h->avctx->colorspace == AVCOL_SPC_RGB) { - return AV_PIX_FMT_GBRP10; - } else - return AV_PIX_FMT_YUV444P10; - } else if (CHROMA422(h)) - return AV_PIX_FMT_YUV422P10; - else - return AV_PIX_FMT_YUV420P10; - break; - case 8: - if (CHROMA444(h)) { - if (h->avctx->colorspace == AVCOL_SPC_RGB) { - return AV_PIX_FMT_GBRP; - } else - return h->avctx->color_range == AVCOL_RANGE_JPEG ? AV_PIX_FMT_YUVJ444P - : AV_PIX_FMT_YUV444P; - } else if (CHROMA422(h)) { - return h->avctx->color_range == AVCOL_RANGE_JPEG ? AV_PIX_FMT_YUVJ422P - : AV_PIX_FMT_YUV422P; - } else { - return h->avctx->get_format(h->avctx, h->avctx->codec->pix_fmts ? - h->avctx->codec->pix_fmts : - h->avctx->color_range == AVCOL_RANGE_JPEG ? - h264_hwaccel_pixfmt_list_jpeg_420 : - h264_hwaccel_pixfmt_list_420); - } - break; - default: - av_log(h->avctx, AV_LOG_ERROR, - "Unsupported bit depth %d\n", h->sps.bit_depth_luma); - return AVERROR_INVALIDDATA; - } -} - -/* export coded and cropped frame dimensions to AVCodecContext */ -static int init_dimensions(H264Context *h) -{ - int width = h->width - (h->sps.crop_right + h->sps.crop_left); - int height = h->height - (h->sps.crop_top + h->sps.crop_bottom); - - /* handle container cropping */ - if (!h->sps.crop && - FFALIGN(h->avctx->width, 16) == h->width && - FFALIGN(h->avctx->height, 16) == h->height) { - width = h->avctx->width; - height = h->avctx->height; - } - - if (width <= 0 || height <= 0) { - av_log(h->avctx, AV_LOG_ERROR, "Invalid cropped dimensions: %dx%d.\n", - width, height); - if (h->avctx->err_recognition & AV_EF_EXPLODE) - return AVERROR_INVALIDDATA; - - av_log(h->avctx, AV_LOG_WARNING, "Ignoring cropping information.\n"); - h->sps.crop_bottom = h->sps.crop_top = h->sps.crop_right = h->sps.crop_left = 0; - h->sps.crop = 0; - - width = h->width; - height = h->height; - } - - h->avctx->coded_width = h->width; - h->avctx->coded_height = h->height; - h->avctx->width = width; - h->avctx->height = height; - - return 0; -} - -static int h264_slice_header_init(H264Context *h, int reinit) -{ - int nb_slices = (HAVE_THREADS && - h->avctx->active_thread_type & FF_THREAD_SLICE) ? - h->avctx->thread_count : 1; - int i, ret; - - h->avctx->sample_aspect_ratio = h->sps.sar; - av_assert0(h->avctx->sample_aspect_ratio.den); - av_pix_fmt_get_chroma_sub_sample(h->avctx->pix_fmt, - &h->chroma_x_shift, &h->chroma_y_shift); - - if (h->sps.timing_info_present_flag) { - int64_t den = h->sps.time_scale; - if (h->x264_build < 44U) - den *= 2; - av_reduce(&h->avctx->time_base.num, &h->avctx->time_base.den, - h->sps.num_units_in_tick, den, 1 << 30); - } - - h->avctx->hwaccel = ff_find_hwaccel(h->avctx); - - if (reinit) - free_tables(h, 0); - h->first_field = 0; - h->prev_interlaced_frame = 1; - - init_scan_tables(h); - ret = ff_h264_alloc_tables(h); - if (ret < 0) { - av_log(h->avctx, AV_LOG_ERROR, "Could not allocate memory\n"); - return ret; - } - - if (nb_slices > H264_MAX_THREADS || (nb_slices > h->mb_height && h->mb_height)) { - int max_slices; - if (h->mb_height) - max_slices = FFMIN(H264_MAX_THREADS, h->mb_height); - else - max_slices = H264_MAX_THREADS; - av_log(h->avctx, AV_LOG_WARNING, "too many threads/slices %d," - " reducing to %d\n", nb_slices, max_slices); - nb_slices = max_slices; - } - h->slice_context_count = nb_slices; - - if (!HAVE_THREADS || !(h->avctx->active_thread_type & FF_THREAD_SLICE)) { - ret = context_init(h); - if (ret < 0) { - av_log(h->avctx, AV_LOG_ERROR, "context_init() failed.\n"); - return ret; - } - } else { - for (i = 1; i < h->slice_context_count; i++) { - H264Context *c; - c = h->thread_context[i] = av_mallocz(sizeof(H264Context)); - if (!c) - return AVERROR(ENOMEM); - c->avctx = h->avctx; - c->dsp = h->dsp; - c->vdsp = h->vdsp; - c->h264dsp = h->h264dsp; - c->h264qpel = h->h264qpel; - c->h264chroma = h->h264chroma; - c->sps = h->sps; - c->pps = h->pps; - c->pixel_shift = h->pixel_shift; - c->width = h->width; - c->height = h->height; - c->linesize = h->linesize; - c->uvlinesize = h->uvlinesize; - c->chroma_x_shift = h->chroma_x_shift; - c->chroma_y_shift = h->chroma_y_shift; - c->qscale = h->qscale; - c->droppable = h->droppable; - c->data_partitioning = h->data_partitioning; - c->low_delay = h->low_delay; - c->mb_width = h->mb_width; - c->mb_height = h->mb_height; - c->mb_stride = h->mb_stride; - c->mb_num = h->mb_num; - c->flags = h->flags; - c->workaround_bugs = h->workaround_bugs; - c->pict_type = h->pict_type; - - init_scan_tables(c); - clone_tables(c, h, i); - c->context_initialized = 1; - } - - for (i = 0; i < h->slice_context_count; i++) - if ((ret = context_init(h->thread_context[i])) < 0) { - av_log(h->avctx, AV_LOG_ERROR, "context_init() failed.\n"); - return ret; - } - } - - h->context_initialized = 1; - - return 0; -} - int ff_set_ref_count(H264Context *h) { int ref_count[2], list_count; @@ -2554,1226 +1281,6 @@ int ff_set_ref_count(H264Context *h) return 0; } -/** - * Decode a slice header. - * This will (re)intialize the decoder and call h264_frame_start() as needed. - * - * @param h h264context - * @param h0 h264 master context (differs from 'h' when doing sliced based - * parallel decoding) - * - * @return 0 if okay, <0 if an error occurred, 1 if decoding must not be multithreaded - */ -static int decode_slice_header(H264Context *h, H264Context *h0) -{ - unsigned int first_mb_in_slice; - unsigned int pps_id; - int ret; - unsigned int slice_type, tmp, i, j; - int default_ref_list_done = 0; - int last_pic_structure, last_pic_droppable; - int needs_reinit = 0; - int field_pic_flag, bottom_field_flag; - - h->qpel_put = h->h264qpel.put_h264_qpel_pixels_tab; - h->qpel_avg = h->h264qpel.avg_h264_qpel_pixels_tab; - - first_mb_in_slice = get_ue_golomb(&h->gb); - - if (first_mb_in_slice == 0) { // FIXME better field boundary detection - if (h0->current_slice && h->cur_pic_ptr && FIELD_PICTURE(h)) { - field_end(h, 1); - } - - h0->current_slice = 0; - if (!h0->first_field) { - if (h->cur_pic_ptr && !h->droppable) { - ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, - h->picture_structure == PICT_BOTTOM_FIELD); - } - h->cur_pic_ptr = NULL; - } - } - - slice_type = get_ue_golomb_31(&h->gb); - if (slice_type > 9) { - av_log(h->avctx, AV_LOG_ERROR, - "slice type %d too large at %d %d\n", - slice_type, h->mb_x, h->mb_y); - return AVERROR_INVALIDDATA; - } - if (slice_type > 4) { - slice_type -= 5; - h->slice_type_fixed = 1; - } else - h->slice_type_fixed = 0; - - slice_type = golomb_to_pict_type[slice_type]; - if (slice_type == AV_PICTURE_TYPE_I || - (h0->current_slice != 0 && slice_type == h0->last_slice_type)) { - default_ref_list_done = 1; - } - h->slice_type = slice_type; - h->slice_type_nos = slice_type & 3; - - if (h->nal_unit_type == NAL_IDR_SLICE && - h->slice_type_nos != AV_PICTURE_TYPE_I) { - av_log(h->avctx, AV_LOG_ERROR, "A non-intra slice in an IDR NAL unit.\n"); - return AVERROR_INVALIDDATA; - } - - // to make a few old functions happy, it's wrong though - h->pict_type = h->slice_type; - - pps_id = get_ue_golomb(&h->gb); - if (pps_id >= MAX_PPS_COUNT) { - av_log(h->avctx, AV_LOG_ERROR, "pps_id %u out of range\n", pps_id); - return AVERROR_INVALIDDATA; - } - if (!h0->pps_buffers[pps_id]) { - av_log(h->avctx, AV_LOG_ERROR, - "non-existing PPS %u referenced\n", - pps_id); - return AVERROR_INVALIDDATA; - } - h->pps = *h0->pps_buffers[pps_id]; - - if (!h0->sps_buffers[h->pps.sps_id]) { - av_log(h->avctx, AV_LOG_ERROR, - "non-existing SPS %u referenced\n", - h->pps.sps_id); - return AVERROR_INVALIDDATA; - } - - if (h->pps.sps_id != h->sps.sps_id || - h0->sps_buffers[h->pps.sps_id]->new) { - h0->sps_buffers[h->pps.sps_id]->new = 0; - - h->sps = *h0->sps_buffers[h->pps.sps_id]; - - if (h->bit_depth_luma != h->sps.bit_depth_luma || - h->chroma_format_idc != h->sps.chroma_format_idc) { - h->bit_depth_luma = h->sps.bit_depth_luma; - h->chroma_format_idc = h->sps.chroma_format_idc; - needs_reinit = 1; - } - if ((ret = h264_set_parameter_from_sps(h)) < 0) - return ret; - } - - h->avctx->profile = ff_h264_get_profile(&h->sps); - h->avctx->level = h->sps.level_idc; - h->avctx->refs = h->sps.ref_frame_count; - - if (h->mb_width != h->sps.mb_width || - h->mb_height != h->sps.mb_height * (2 - h->sps.frame_mbs_only_flag)) - needs_reinit = 1; - - h->mb_width = h->sps.mb_width; - h->mb_height = h->sps.mb_height * (2 - h->sps.frame_mbs_only_flag); - h->mb_num = h->mb_width * h->mb_height; - h->mb_stride = h->mb_width + 1; - - h->b_stride = h->mb_width * 4; - - h->chroma_y_shift = h->sps.chroma_format_idc <= 1; // 400 uses yuv420p - - h->width = 16 * h->mb_width; - h->height = 16 * h->mb_height; - - ret = init_dimensions(h); - if (ret < 0) - return ret; - - if (h->sps.video_signal_type_present_flag) { - h->avctx->color_range = h->sps.full_range ? AVCOL_RANGE_JPEG - : AVCOL_RANGE_MPEG; - if (h->sps.colour_description_present_flag) { - if (h->avctx->colorspace != h->sps.colorspace) - needs_reinit = 1; - h->avctx->color_primaries = h->sps.color_primaries; - h->avctx->color_trc = h->sps.color_trc; - h->avctx->colorspace = h->sps.colorspace; - } - } - - if (h->context_initialized && - (h->width != h->avctx->coded_width || - h->height != h->avctx->coded_height || - needs_reinit)) { - if (h != h0) { - av_log(h->avctx, AV_LOG_ERROR, - "changing width %d -> %d / height %d -> %d on " - "slice %d\n", - h->width, h->avctx->coded_width, - h->height, h->avctx->coded_height, - h0->current_slice + 1); - return AVERROR_INVALIDDATA; - } - - flush_change(h); - - if ((ret = get_pixel_format(h)) < 0) - return ret; - h->avctx->pix_fmt = ret; - - av_log(h->avctx, AV_LOG_INFO, "Reinit context to %dx%d, " - "pix_fmt: %d\n", h->width, h->height, h->avctx->pix_fmt); - - if ((ret = h264_slice_header_init(h, 1)) < 0) { - av_log(h->avctx, AV_LOG_ERROR, - "h264_slice_header_init() failed\n"); - return ret; - } - } - if (!h->context_initialized) { - if (h != h0) { - av_log(h->avctx, AV_LOG_ERROR, - "Cannot (re-)initialize context during parallel decoding.\n"); - return AVERROR_PATCHWELCOME; - } - - if ((ret = get_pixel_format(h)) < 0) - return ret; - h->avctx->pix_fmt = ret; - - if ((ret = h264_slice_header_init(h, 0)) < 0) { - av_log(h->avctx, AV_LOG_ERROR, - "h264_slice_header_init() failed\n"); - return ret; - } - } - - if (h == h0 && h->dequant_coeff_pps != pps_id) { - h->dequant_coeff_pps = pps_id; - init_dequant_tables(h); - } - - h->frame_num = get_bits(&h->gb, h->sps.log2_max_frame_num); - - h->mb_mbaff = 0; - h->mb_aff_frame = 0; - last_pic_structure = h0->picture_structure; - last_pic_droppable = h0->droppable; - h->droppable = h->nal_ref_idc == 0; - if (h->sps.frame_mbs_only_flag) { - h->picture_structure = PICT_FRAME; - } else { - field_pic_flag = get_bits1(&h->gb); - if (field_pic_flag) { - bottom_field_flag = get_bits1(&h->gb); - h->picture_structure = PICT_TOP_FIELD + bottom_field_flag; - } else { - h->picture_structure = PICT_FRAME; - h->mb_aff_frame = h->sps.mb_aff; - } - } - h->mb_field_decoding_flag = h->picture_structure != PICT_FRAME; - - if (h0->current_slice != 0) { - if (last_pic_structure != h->picture_structure || - last_pic_droppable != h->droppable) { - av_log(h->avctx, AV_LOG_ERROR, - "Changing field mode (%d -> %d) between slices is not allowed\n", - last_pic_structure, h->picture_structure); - h->picture_structure = last_pic_structure; - h->droppable = last_pic_droppable; - return AVERROR_INVALIDDATA; - } else if (!h0->cur_pic_ptr) { - av_log(h->avctx, AV_LOG_ERROR, - "unset cur_pic_ptr on slice %d\n", - h0->current_slice + 1); - return AVERROR_INVALIDDATA; - } - } else { - /* Shorten frame num gaps so we don't have to allocate reference - * frames just to throw them away */ - if (h->frame_num != h->prev_frame_num) { - int unwrap_prev_frame_num = h->prev_frame_num; - int max_frame_num = 1 << h->sps.log2_max_frame_num; - - if (unwrap_prev_frame_num > h->frame_num) - unwrap_prev_frame_num -= max_frame_num; - - if ((h->frame_num - unwrap_prev_frame_num) > h->sps.ref_frame_count) { - unwrap_prev_frame_num = (h->frame_num - h->sps.ref_frame_count) - 1; - if (unwrap_prev_frame_num < 0) - unwrap_prev_frame_num += max_frame_num; - - h->prev_frame_num = unwrap_prev_frame_num; - } - } - - /* See if we have a decoded first field looking for a pair... - * Here, we're using that to see if we should mark previously - * decode frames as "finished". - * We have to do that before the "dummy" in-between frame allocation, - * since that can modify s->current_picture_ptr. */ - if (h0->first_field) { - assert(h0->cur_pic_ptr); - assert(h0->cur_pic_ptr->f.buf[0]); - assert(h0->cur_pic_ptr->reference != DELAYED_PIC_REF); - - /* figure out if we have a complementary field pair */ - if (!FIELD_PICTURE(h) || h->picture_structure == last_pic_structure) { - /* Previous field is unmatched. Don't display it, but let it - * remain for reference if marked as such. */ - if (!last_pic_droppable && last_pic_structure != PICT_FRAME) { - ff_thread_report_progress(&h0->cur_pic_ptr->tf, INT_MAX, - last_pic_structure == PICT_TOP_FIELD); - } - } else { - if (h0->cur_pic_ptr->frame_num != h->frame_num) { - /* This and previous field were reference, but had - * different frame_nums. Consider this field first in - * pair. Throw away previous field except for reference - * purposes. */ - if (!last_pic_droppable && last_pic_structure != PICT_FRAME) { - ff_thread_report_progress(&h0->cur_pic_ptr->tf, INT_MAX, - last_pic_structure == PICT_TOP_FIELD); - } - } else { - /* Second field in complementary pair */ - if (!((last_pic_structure == PICT_TOP_FIELD && - h->picture_structure == PICT_BOTTOM_FIELD) || - (last_pic_structure == PICT_BOTTOM_FIELD && - h->picture_structure == PICT_TOP_FIELD))) { - av_log(h->avctx, AV_LOG_ERROR, - "Invalid field mode combination %d/%d\n", - last_pic_structure, h->picture_structure); - h->picture_structure = last_pic_structure; - h->droppable = last_pic_droppable; - return AVERROR_INVALIDDATA; - } else if (last_pic_droppable != h->droppable) { - avpriv_request_sample(h->avctx, - "Found reference and non-reference fields in the same frame, which"); - h->picture_structure = last_pic_structure; - h->droppable = last_pic_droppable; - return AVERROR_PATCHWELCOME; - } - } - } - } - - while (h->frame_num != h->prev_frame_num && - h->frame_num != (h->prev_frame_num + 1) % (1 << h->sps.log2_max_frame_num)) { - H264Picture *prev = h->short_ref_count ? h->short_ref[0] : NULL; - av_log(h->avctx, AV_LOG_DEBUG, "Frame num gap %d %d\n", - h->frame_num, h->prev_frame_num); - ret = h264_frame_start(h); - if (ret < 0) { - h0->first_field = 0; - return ret; - } - - h->prev_frame_num++; - h->prev_frame_num %= 1 << h->sps.log2_max_frame_num; - h->cur_pic_ptr->frame_num = h->prev_frame_num; - ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, 0); - ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, 1); - ret = ff_generate_sliding_window_mmcos(h, 1); - if (ret < 0 && (h->avctx->err_recognition & AV_EF_EXPLODE)) - return ret; - ret = ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index); - if (ret < 0 && (h->avctx->err_recognition & AV_EF_EXPLODE)) - return ret; - /* Error concealment: If a ref is missing, copy the previous ref - * in its place. - * FIXME: Avoiding a memcpy would be nice, but ref handling makes - * many assumptions about there being no actual duplicates. - * FIXME: This does not copy padding for out-of-frame motion - * vectors. Given we are concealing a lost frame, this probably - * is not noticeable by comparison, but it should be fixed. */ - if (h->short_ref_count) { - if (prev) { - av_image_copy(h->short_ref[0]->f.data, - h->short_ref[0]->f.linesize, - (const uint8_t **)prev->f.data, - prev->f.linesize, - h->avctx->pix_fmt, - h->mb_width * 16, - h->mb_height * 16); - h->short_ref[0]->poc = prev->poc + 2; - } - h->short_ref[0]->frame_num = h->prev_frame_num; - } - } - - /* See if we have a decoded first field looking for a pair... - * We're using that to see whether to continue decoding in that - * frame, or to allocate a new one. */ - if (h0->first_field) { - assert(h0->cur_pic_ptr); - assert(h0->cur_pic_ptr->f.buf[0]); - assert(h0->cur_pic_ptr->reference != DELAYED_PIC_REF); - - /* figure out if we have a complementary field pair */ - if (!FIELD_PICTURE(h) || h->picture_structure == last_pic_structure) { - /* Previous field is unmatched. Don't display it, but let it - * remain for reference if marked as such. */ - h0->cur_pic_ptr = NULL; - h0->first_field = FIELD_PICTURE(h); - } else { - if (h0->cur_pic_ptr->frame_num != h->frame_num) { - /* This and the previous field had different frame_nums. - * Consider this field first in pair. Throw away previous - * one except for reference purposes. */ - h0->first_field = 1; - h0->cur_pic_ptr = NULL; - } else { - /* Second field in complementary pair */ - h0->first_field = 0; - } - } - } else { - /* Frame or first field in a potentially complementary pair */ - h0->first_field = FIELD_PICTURE(h); - } - - if (!FIELD_PICTURE(h) || h0->first_field) { - if (h264_frame_start(h) < 0) { - h0->first_field = 0; - return AVERROR_INVALIDDATA; - } - } else { - release_unused_pictures(h, 0); - } - } - if (h != h0 && (ret = clone_slice(h, h0)) < 0) - return ret; - - h->cur_pic_ptr->frame_num = h->frame_num; // FIXME frame_num cleanup - - assert(h->mb_num == h->mb_width * h->mb_height); - if (first_mb_in_slice << FIELD_OR_MBAFF_PICTURE(h) >= h->mb_num || - first_mb_in_slice >= h->mb_num) { - av_log(h->avctx, AV_LOG_ERROR, "first_mb_in_slice overflow\n"); - return AVERROR_INVALIDDATA; - } - h->resync_mb_x = h->mb_x = first_mb_in_slice % h->mb_width; - h->resync_mb_y = h->mb_y = (first_mb_in_slice / h->mb_width) << - FIELD_OR_MBAFF_PICTURE(h); - if (h->picture_structure == PICT_BOTTOM_FIELD) - h->resync_mb_y = h->mb_y = h->mb_y + 1; - assert(h->mb_y < h->mb_height); - - if (h->picture_structure == PICT_FRAME) { - h->curr_pic_num = h->frame_num; - h->max_pic_num = 1 << h->sps.log2_max_frame_num; - } else { - h->curr_pic_num = 2 * h->frame_num + 1; - h->max_pic_num = 1 << (h->sps.log2_max_frame_num + 1); - } - - if (h->nal_unit_type == NAL_IDR_SLICE) - get_ue_golomb(&h->gb); /* idr_pic_id */ - - if (h->sps.poc_type == 0) { - h->poc_lsb = get_bits(&h->gb, h->sps.log2_max_poc_lsb); - - if (h->pps.pic_order_present == 1 && h->picture_structure == PICT_FRAME) - h->delta_poc_bottom = get_se_golomb(&h->gb); - } - - if (h->sps.poc_type == 1 && !h->sps.delta_pic_order_always_zero_flag) { - h->delta_poc[0] = get_se_golomb(&h->gb); - - if (h->pps.pic_order_present == 1 && h->picture_structure == PICT_FRAME) - h->delta_poc[1] = get_se_golomb(&h->gb); - } - - ff_init_poc(h, h->cur_pic_ptr->field_poc, &h->cur_pic_ptr->poc); - - if (h->pps.redundant_pic_cnt_present) - h->redundant_pic_count = get_ue_golomb(&h->gb); - - ret = ff_set_ref_count(h); - if (ret < 0) - return ret; - else if (ret == 1) - default_ref_list_done = 0; - - if (!default_ref_list_done) - ff_h264_fill_default_ref_list(h); - - if (h->slice_type_nos != AV_PICTURE_TYPE_I) { - ret = ff_h264_decode_ref_pic_list_reordering(h); - if (ret < 0) { - h->ref_count[1] = h->ref_count[0] = 0; - return ret; - } - } - - if ((h->pps.weighted_pred && h->slice_type_nos == AV_PICTURE_TYPE_P) || - (h->pps.weighted_bipred_idc == 1 && - h->slice_type_nos == AV_PICTURE_TYPE_B)) - ff_pred_weight_table(h); - else if (h->pps.weighted_bipred_idc == 2 && - h->slice_type_nos == AV_PICTURE_TYPE_B) { - implicit_weight_table(h, -1); - } else { - h->use_weight = 0; - for (i = 0; i < 2; i++) { - h->luma_weight_flag[i] = 0; - h->chroma_weight_flag[i] = 0; - } - } - - // If frame-mt is enabled, only update mmco tables for the first slice - // in a field. Subsequent slices can temporarily clobber h->mmco_index - // or h->mmco, which will cause ref list mix-ups and decoding errors - // further down the line. This may break decoding if the first slice is - // corrupt, thus we only do this if frame-mt is enabled. - if (h->nal_ref_idc) { - ret = ff_h264_decode_ref_pic_marking(h0, &h->gb, - !(h->avctx->active_thread_type & FF_THREAD_FRAME) || - h0->current_slice == 0); - if (ret < 0 && (h->avctx->err_recognition & AV_EF_EXPLODE)) - return AVERROR_INVALIDDATA; - } - - if (FRAME_MBAFF(h)) { - ff_h264_fill_mbaff_ref_list(h); - - if (h->pps.weighted_bipred_idc == 2 && h->slice_type_nos == AV_PICTURE_TYPE_B) { - implicit_weight_table(h, 0); - implicit_weight_table(h, 1); - } - } - - if (h->slice_type_nos == AV_PICTURE_TYPE_B && !h->direct_spatial_mv_pred) - ff_h264_direct_dist_scale_factor(h); - ff_h264_direct_ref_list_init(h); - - if (h->slice_type_nos != AV_PICTURE_TYPE_I && h->pps.cabac) { - tmp = get_ue_golomb_31(&h->gb); - if (tmp > 2) { - av_log(h->avctx, AV_LOG_ERROR, "cabac_init_idc %u overflow\n", tmp); - return AVERROR_INVALIDDATA; - } - h->cabac_init_idc = tmp; - } - - h->last_qscale_diff = 0; - tmp = h->pps.init_qp + get_se_golomb(&h->gb); - if (tmp > 51 + 6 * (h->sps.bit_depth_luma - 8)) { - av_log(h->avctx, AV_LOG_ERROR, "QP %u out of range\n", tmp); - return AVERROR_INVALIDDATA; - } - h->qscale = tmp; - h->chroma_qp[0] = get_chroma_qp(h, 0, h->qscale); - h->chroma_qp[1] = get_chroma_qp(h, 1, h->qscale); - // FIXME qscale / qp ... stuff - if (h->slice_type == AV_PICTURE_TYPE_SP) - get_bits1(&h->gb); /* sp_for_switch_flag */ - if (h->slice_type == AV_PICTURE_TYPE_SP || - h->slice_type == AV_PICTURE_TYPE_SI) - get_se_golomb(&h->gb); /* slice_qs_delta */ - - h->deblocking_filter = 1; - h->slice_alpha_c0_offset = 0; - h->slice_beta_offset = 0; - if (h->pps.deblocking_filter_parameters_present) { - tmp = get_ue_golomb_31(&h->gb); - if (tmp > 2) { - av_log(h->avctx, AV_LOG_ERROR, - "deblocking_filter_idc %u out of range\n", tmp); - return AVERROR_INVALIDDATA; - } - h->deblocking_filter = tmp; - if (h->deblocking_filter < 2) - h->deblocking_filter ^= 1; // 1<->0 - - if (h->deblocking_filter) { - h->slice_alpha_c0_offset = get_se_golomb(&h->gb) * 2; - h->slice_beta_offset = get_se_golomb(&h->gb) * 2; - if (h->slice_alpha_c0_offset > 12 || - h->slice_alpha_c0_offset < -12 || - h->slice_beta_offset > 12 || - h->slice_beta_offset < -12) { - av_log(h->avctx, AV_LOG_ERROR, - "deblocking filter parameters %d %d out of range\n", - h->slice_alpha_c0_offset, h->slice_beta_offset); - return AVERROR_INVALIDDATA; - } - } - } - - if (h->avctx->skip_loop_filter >= AVDISCARD_ALL || - (h->avctx->skip_loop_filter >= AVDISCARD_NONKEY && - h->slice_type_nos != AV_PICTURE_TYPE_I) || - (h->avctx->skip_loop_filter >= AVDISCARD_BIDIR && - h->slice_type_nos == AV_PICTURE_TYPE_B) || - (h->avctx->skip_loop_filter >= AVDISCARD_NONREF && - h->nal_ref_idc == 0)) - h->deblocking_filter = 0; - - if (h->deblocking_filter == 1 && h0->max_contexts > 1) { - if (h->avctx->flags2 & CODEC_FLAG2_FAST) { - /* Cheat slightly for speed: - * Do not bother to deblock across slices. */ - h->deblocking_filter = 2; - } else { - h0->max_contexts = 1; - if (!h0->single_decode_warning) { - av_log(h->avctx, AV_LOG_INFO, - "Cannot parallelize deblocking type 1, decoding such frames in sequential order\n"); - h0->single_decode_warning = 1; - } - if (h != h0) { - av_log(h->avctx, AV_LOG_ERROR, - "Deblocking switched inside frame.\n"); - return 1; - } - } - } - h->qp_thresh = 15 - - FFMIN(h->slice_alpha_c0_offset, h->slice_beta_offset) - - FFMAX3(0, - h->pps.chroma_qp_index_offset[0], - h->pps.chroma_qp_index_offset[1]) + - 6 * (h->sps.bit_depth_luma - 8); - - h0->last_slice_type = slice_type; - h->slice_num = ++h0->current_slice; - if (h->slice_num >= MAX_SLICES) { - av_log(h->avctx, AV_LOG_ERROR, - "Too many slices, increase MAX_SLICES and recompile\n"); - } - - for (j = 0; j < 2; j++) { - int id_list[16]; - int *ref2frm = h->ref2frm[h->slice_num & (MAX_SLICES - 1)][j]; - for (i = 0; i < 16; i++) { - id_list[i] = 60; - if (j < h->list_count && i < h->ref_count[j] && - h->ref_list[j][i].f.buf[0]) { - int k; - AVBuffer *buf = h->ref_list[j][i].f.buf[0]->buffer; - for (k = 0; k < h->short_ref_count; k++) - if (h->short_ref[k]->f.buf[0]->buffer == buf) { - id_list[i] = k; - break; - } - for (k = 0; k < h->long_ref_count; k++) - if (h->long_ref[k] && h->long_ref[k]->f.buf[0]->buffer == buf) { - id_list[i] = h->short_ref_count + k; - break; - } - } - } - - ref2frm[0] = - ref2frm[1] = -1; - for (i = 0; i < 16; i++) - ref2frm[i + 2] = 4 * id_list[i] + (h->ref_list[j][i].reference & 3); - ref2frm[18 + 0] = - ref2frm[18 + 1] = -1; - for (i = 16; i < 48; i++) - ref2frm[i + 4] = 4 * id_list[(i - 16) >> 1] + - (h->ref_list[j][i].reference & 3); - } - - if (h->avctx->debug & FF_DEBUG_PICT_INFO) { - av_log(h->avctx, AV_LOG_DEBUG, - "slice:%d %s mb:%d %c%s%s pps:%u frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d:%d:%d weight:%d%s %s\n", - h->slice_num, - (h->picture_structure == PICT_FRAME ? "F" : h->picture_structure == PICT_TOP_FIELD ? "T" : "B"), - first_mb_in_slice, - av_get_picture_type_char(h->slice_type), - h->slice_type_fixed ? " fix" : "", - h->nal_unit_type == NAL_IDR_SLICE ? " IDR" : "", - pps_id, h->frame_num, - h->cur_pic_ptr->field_poc[0], - h->cur_pic_ptr->field_poc[1], - h->ref_count[0], h->ref_count[1], - h->qscale, - h->deblocking_filter, - h->slice_alpha_c0_offset, h->slice_beta_offset, - h->use_weight, - h->use_weight == 1 && h->use_weight_chroma ? "c" : "", - h->slice_type == AV_PICTURE_TYPE_B ? (h->direct_spatial_mv_pred ? "SPAT" : "TEMP") : ""); - } - - return 0; -} - -int ff_h264_get_slice_type(const H264Context *h) -{ - switch (h->slice_type) { - case AV_PICTURE_TYPE_P: - return 0; - case AV_PICTURE_TYPE_B: - return 1; - case AV_PICTURE_TYPE_I: - return 2; - case AV_PICTURE_TYPE_SP: - return 3; - case AV_PICTURE_TYPE_SI: - return 4; - default: - return AVERROR_INVALIDDATA; - } -} - -static av_always_inline void fill_filter_caches_inter(H264Context *h, - int mb_type, int top_xy, - int left_xy[LEFT_MBS], - int top_type, - int left_type[LEFT_MBS], - int mb_xy, int list) -{ - int b_stride = h->b_stride; - int16_t(*mv_dst)[2] = &h->mv_cache[list][scan8[0]]; - int8_t *ref_cache = &h->ref_cache[list][scan8[0]]; - if (IS_INTER(mb_type) || IS_DIRECT(mb_type)) { - if (USES_LIST(top_type, list)) { - const int b_xy = h->mb2b_xy[top_xy] + 3 * b_stride; - const int b8_xy = 4 * top_xy + 2; - int (*ref2frm)[64] = h->ref2frm[h->slice_table[top_xy] & (MAX_SLICES - 1)][0] + (MB_MBAFF(h) ? 20 : 2); - AV_COPY128(mv_dst - 1 * 8, h->cur_pic.motion_val[list][b_xy + 0]); - ref_cache[0 - 1 * 8] = - ref_cache[1 - 1 * 8] = ref2frm[list][h->cur_pic.ref_index[list][b8_xy + 0]]; - ref_cache[2 - 1 * 8] = - ref_cache[3 - 1 * 8] = ref2frm[list][h->cur_pic.ref_index[list][b8_xy + 1]]; - } else { - AV_ZERO128(mv_dst - 1 * 8); - AV_WN32A(&ref_cache[0 - 1 * 8], ((LIST_NOT_USED) & 0xFF) * 0x01010101u); - } - - if (!IS_INTERLACED(mb_type ^ left_type[LTOP])) { - if (USES_LIST(left_type[LTOP], list)) { - const int b_xy = h->mb2b_xy[left_xy[LTOP]] + 3; - const int b8_xy = 4 * left_xy[LTOP] + 1; - int (*ref2frm)[64] = h->ref2frm[h->slice_table[left_xy[LTOP]] & (MAX_SLICES - 1)][0] + (MB_MBAFF(h) ? 20 : 2); - AV_COPY32(mv_dst - 1 + 0, h->cur_pic.motion_val[list][b_xy + b_stride * 0]); - AV_COPY32(mv_dst - 1 + 8, h->cur_pic.motion_val[list][b_xy + b_stride * 1]); - AV_COPY32(mv_dst - 1 + 16, h->cur_pic.motion_val[list][b_xy + b_stride * 2]); - AV_COPY32(mv_dst - 1 + 24, h->cur_pic.motion_val[list][b_xy + b_stride * 3]); - ref_cache[-1 + 0] = - ref_cache[-1 + 8] = ref2frm[list][h->cur_pic.ref_index[list][b8_xy + 2 * 0]]; - ref_cache[-1 + 16] = - ref_cache[-1 + 24] = ref2frm[list][h->cur_pic.ref_index[list][b8_xy + 2 * 1]]; - } else { - AV_ZERO32(mv_dst - 1 + 0); - AV_ZERO32(mv_dst - 1 + 8); - AV_ZERO32(mv_dst - 1 + 16); - AV_ZERO32(mv_dst - 1 + 24); - ref_cache[-1 + 0] = - ref_cache[-1 + 8] = - ref_cache[-1 + 16] = - ref_cache[-1 + 24] = LIST_NOT_USED; - } - } - } - - if (!USES_LIST(mb_type, list)) { - fill_rectangle(mv_dst, 4, 4, 8, pack16to32(0, 0), 4); - AV_WN32A(&ref_cache[0 * 8], ((LIST_NOT_USED) & 0xFF) * 0x01010101u); - AV_WN32A(&ref_cache[1 * 8], ((LIST_NOT_USED) & 0xFF) * 0x01010101u); - AV_WN32A(&ref_cache[2 * 8], ((LIST_NOT_USED) & 0xFF) * 0x01010101u); - AV_WN32A(&ref_cache[3 * 8], ((LIST_NOT_USED) & 0xFF) * 0x01010101u); - return; - } - - { - int8_t *ref = &h->cur_pic.ref_index[list][4 * mb_xy]; - int (*ref2frm)[64] = h->ref2frm[h->slice_num & (MAX_SLICES - 1)][0] + (MB_MBAFF(h) ? 20 : 2); - uint32_t ref01 = (pack16to32(ref2frm[list][ref[0]], ref2frm[list][ref[1]]) & 0x00FF00FF) * 0x0101; - uint32_t ref23 = (pack16to32(ref2frm[list][ref[2]], ref2frm[list][ref[3]]) & 0x00FF00FF) * 0x0101; - AV_WN32A(&ref_cache[0 * 8], ref01); - AV_WN32A(&ref_cache[1 * 8], ref01); - AV_WN32A(&ref_cache[2 * 8], ref23); - AV_WN32A(&ref_cache[3 * 8], ref23); - } - - { - int16_t(*mv_src)[2] = &h->cur_pic.motion_val[list][4 * h->mb_x + 4 * h->mb_y * b_stride]; - AV_COPY128(mv_dst + 8 * 0, mv_src + 0 * b_stride); - AV_COPY128(mv_dst + 8 * 1, mv_src + 1 * b_stride); - AV_COPY128(mv_dst + 8 * 2, mv_src + 2 * b_stride); - AV_COPY128(mv_dst + 8 * 3, mv_src + 3 * b_stride); - } -} - -/** - * - * @return non zero if the loop filter can be skipped - */ -static int fill_filter_caches(H264Context *h, int mb_type) -{ - const int mb_xy = h->mb_xy; - int top_xy, left_xy[LEFT_MBS]; - int top_type, left_type[LEFT_MBS]; - uint8_t *nnz; - uint8_t *nnz_cache; - - top_xy = mb_xy - (h->mb_stride << MB_FIELD(h)); - - /* Wow, what a mess, why didn't they simplify the interlacing & intra - * stuff, I can't imagine that these complex rules are worth it. */ - - left_xy[LBOT] = left_xy[LTOP] = mb_xy - 1; - if (FRAME_MBAFF(h)) { - const int left_mb_field_flag = IS_INTERLACED(h->cur_pic.mb_type[mb_xy - 1]); - const int curr_mb_field_flag = IS_INTERLACED(mb_type); - if (h->mb_y & 1) { - if (left_mb_field_flag != curr_mb_field_flag) - left_xy[LTOP] -= h->mb_stride; - } else { - if (curr_mb_field_flag) - top_xy += h->mb_stride & - (((h->cur_pic.mb_type[top_xy] >> 7) & 1) - 1); - if (left_mb_field_flag != curr_mb_field_flag) - left_xy[LBOT] += h->mb_stride; - } - } - - h->top_mb_xy = top_xy; - h->left_mb_xy[LTOP] = left_xy[LTOP]; - h->left_mb_xy[LBOT] = left_xy[LBOT]; - { - /* For sufficiently low qp, filtering wouldn't do anything. - * This is a conservative estimate: could also check beta_offset - * and more accurate chroma_qp. */ - int qp_thresh = h->qp_thresh; // FIXME strictly we should store qp_thresh for each mb of a slice - int qp = h->cur_pic.qscale_table[mb_xy]; - if (qp <= qp_thresh && - (left_xy[LTOP] < 0 || - ((qp + h->cur_pic.qscale_table[left_xy[LTOP]] + 1) >> 1) <= qp_thresh) && - (top_xy < 0 || - ((qp + h->cur_pic.qscale_table[top_xy] + 1) >> 1) <= qp_thresh)) { - if (!FRAME_MBAFF(h)) - return 1; - if ((left_xy[LTOP] < 0 || - ((qp + h->cur_pic.qscale_table[left_xy[LBOT]] + 1) >> 1) <= qp_thresh) && - (top_xy < h->mb_stride || - ((qp + h->cur_pic.qscale_table[top_xy - h->mb_stride] + 1) >> 1) <= qp_thresh)) - return 1; - } - } - - top_type = h->cur_pic.mb_type[top_xy]; - left_type[LTOP] = h->cur_pic.mb_type[left_xy[LTOP]]; - left_type[LBOT] = h->cur_pic.mb_type[left_xy[LBOT]]; - if (h->deblocking_filter == 2) { - if (h->slice_table[top_xy] != h->slice_num) - top_type = 0; - if (h->slice_table[left_xy[LBOT]] != h->slice_num) - left_type[LTOP] = left_type[LBOT] = 0; - } else { - if (h->slice_table[top_xy] == 0xFFFF) - top_type = 0; - if (h->slice_table[left_xy[LBOT]] == 0xFFFF) - left_type[LTOP] = left_type[LBOT] = 0; - } - h->top_type = top_type; - h->left_type[LTOP] = left_type[LTOP]; - h->left_type[LBOT] = left_type[LBOT]; - - if (IS_INTRA(mb_type)) - return 0; - - fill_filter_caches_inter(h, mb_type, top_xy, left_xy, - top_type, left_type, mb_xy, 0); - if (h->list_count == 2) - fill_filter_caches_inter(h, mb_type, top_xy, left_xy, - top_type, left_type, mb_xy, 1); - - nnz = h->non_zero_count[mb_xy]; - nnz_cache = h->non_zero_count_cache; - AV_COPY32(&nnz_cache[4 + 8 * 1], &nnz[0]); - AV_COPY32(&nnz_cache[4 + 8 * 2], &nnz[4]); - AV_COPY32(&nnz_cache[4 + 8 * 3], &nnz[8]); - AV_COPY32(&nnz_cache[4 + 8 * 4], &nnz[12]); - h->cbp = h->cbp_table[mb_xy]; - - if (top_type) { - nnz = h->non_zero_count[top_xy]; - AV_COPY32(&nnz_cache[4 + 8 * 0], &nnz[3 * 4]); - } - - if (left_type[LTOP]) { - nnz = h->non_zero_count[left_xy[LTOP]]; - nnz_cache[3 + 8 * 1] = nnz[3 + 0 * 4]; - nnz_cache[3 + 8 * 2] = nnz[3 + 1 * 4]; - nnz_cache[3 + 8 * 3] = nnz[3 + 2 * 4]; - nnz_cache[3 + 8 * 4] = nnz[3 + 3 * 4]; - } - - /* CAVLC 8x8dct requires NNZ values for residual decoding that differ - * from what the loop filter needs */ - if (!CABAC(h) && h->pps.transform_8x8_mode) { - if (IS_8x8DCT(top_type)) { - nnz_cache[4 + 8 * 0] = - nnz_cache[5 + 8 * 0] = (h->cbp_table[top_xy] & 0x4000) >> 12; - nnz_cache[6 + 8 * 0] = - nnz_cache[7 + 8 * 0] = (h->cbp_table[top_xy] & 0x8000) >> 12; - } - if (IS_8x8DCT(left_type[LTOP])) { - nnz_cache[3 + 8 * 1] = - nnz_cache[3 + 8 * 2] = (h->cbp_table[left_xy[LTOP]] & 0x2000) >> 12; // FIXME check MBAFF - } - if (IS_8x8DCT(left_type[LBOT])) { - nnz_cache[3 + 8 * 3] = - nnz_cache[3 + 8 * 4] = (h->cbp_table[left_xy[LBOT]] & 0x8000) >> 12; // FIXME check MBAFF - } - - if (IS_8x8DCT(mb_type)) { - nnz_cache[scan8[0]] = - nnz_cache[scan8[1]] = - nnz_cache[scan8[2]] = - nnz_cache[scan8[3]] = (h->cbp & 0x1000) >> 12; - - nnz_cache[scan8[0 + 4]] = - nnz_cache[scan8[1 + 4]] = - nnz_cache[scan8[2 + 4]] = - nnz_cache[scan8[3 + 4]] = (h->cbp & 0x2000) >> 12; - - nnz_cache[scan8[0 + 8]] = - nnz_cache[scan8[1 + 8]] = - nnz_cache[scan8[2 + 8]] = - nnz_cache[scan8[3 + 8]] = (h->cbp & 0x4000) >> 12; - - nnz_cache[scan8[0 + 12]] = - nnz_cache[scan8[1 + 12]] = - nnz_cache[scan8[2 + 12]] = - nnz_cache[scan8[3 + 12]] = (h->cbp & 0x8000) >> 12; - } - } - - return 0; -} - -static void loop_filter(H264Context *h, int start_x, int end_x) -{ - uint8_t *dest_y, *dest_cb, *dest_cr; - int linesize, uvlinesize, mb_x, mb_y; - const int end_mb_y = h->mb_y + FRAME_MBAFF(h); - const int old_slice_type = h->slice_type; - const int pixel_shift = h->pixel_shift; - const int block_h = 16 >> h->chroma_y_shift; - - if (h->deblocking_filter) { - for (mb_x = start_x; mb_x < end_x; mb_x++) - for (mb_y = end_mb_y - FRAME_MBAFF(h); mb_y <= end_mb_y; mb_y++) { - int mb_xy, mb_type; - mb_xy = h->mb_xy = mb_x + mb_y * h->mb_stride; - h->slice_num = h->slice_table[mb_xy]; - mb_type = h->cur_pic.mb_type[mb_xy]; - h->list_count = h->list_counts[mb_xy]; - - if (FRAME_MBAFF(h)) - h->mb_mbaff = - h->mb_field_decoding_flag = !!IS_INTERLACED(mb_type); - - h->mb_x = mb_x; - h->mb_y = mb_y; - dest_y = h->cur_pic.f.data[0] + - ((mb_x << pixel_shift) + mb_y * h->linesize) * 16; - dest_cb = h->cur_pic.f.data[1] + - (mb_x << pixel_shift) * (8 << CHROMA444(h)) + - mb_y * h->uvlinesize * block_h; - dest_cr = h->cur_pic.f.data[2] + - (mb_x << pixel_shift) * (8 << CHROMA444(h)) + - mb_y * h->uvlinesize * block_h; - // FIXME simplify above - - if (MB_FIELD(h)) { - linesize = h->mb_linesize = h->linesize * 2; - uvlinesize = h->mb_uvlinesize = h->uvlinesize * 2; - if (mb_y & 1) { // FIXME move out of this function? - dest_y -= h->linesize * 15; - dest_cb -= h->uvlinesize * (block_h - 1); - dest_cr -= h->uvlinesize * (block_h - 1); - } - } else { - linesize = h->mb_linesize = h->linesize; - uvlinesize = h->mb_uvlinesize = h->uvlinesize; - } - backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, - uvlinesize, 0); - if (fill_filter_caches(h, mb_type)) - continue; - h->chroma_qp[0] = get_chroma_qp(h, 0, h->cur_pic.qscale_table[mb_xy]); - h->chroma_qp[1] = get_chroma_qp(h, 1, h->cur_pic.qscale_table[mb_xy]); - - if (FRAME_MBAFF(h)) { - ff_h264_filter_mb(h, mb_x, mb_y, dest_y, dest_cb, dest_cr, - linesize, uvlinesize); - } else { - ff_h264_filter_mb_fast(h, mb_x, mb_y, dest_y, dest_cb, - dest_cr, linesize, uvlinesize); - } - } - } - h->slice_type = old_slice_type; - h->mb_x = end_x; - h->mb_y = end_mb_y - FRAME_MBAFF(h); - h->chroma_qp[0] = get_chroma_qp(h, 0, h->qscale); - h->chroma_qp[1] = get_chroma_qp(h, 1, h->qscale); -} - -static void predict_field_decoding_flag(H264Context *h) -{ - const int mb_xy = h->mb_x + h->mb_y * h->mb_stride; - int mb_type = (h->slice_table[mb_xy - 1] == h->slice_num) ? - h->cur_pic.mb_type[mb_xy - 1] : - (h->slice_table[mb_xy - h->mb_stride] == h->slice_num) ? - h->cur_pic.mb_type[mb_xy - h->mb_stride] : 0; - h->mb_mbaff = h->mb_field_decoding_flag = IS_INTERLACED(mb_type) ? 1 : 0; -} - -/** - * Draw edges and report progress for the last MB row. - */ -static void decode_finish_row(H264Context *h) -{ - int top = 16 * (h->mb_y >> FIELD_PICTURE(h)); - int pic_height = 16 * h->mb_height >> FIELD_PICTURE(h); - int height = 16 << FRAME_MBAFF(h); - int deblock_border = (16 + 4) << FRAME_MBAFF(h); - - if (h->deblocking_filter) { - if ((top + height) >= pic_height) - height += deblock_border; - top -= deblock_border; - } - - if (top >= pic_height || (top + height) < 0) - return; - - height = FFMIN(height, pic_height - top); - if (top < 0) { - height = top + height; - top = 0; - } - - ff_h264_draw_horiz_band(h, top, height); - - if (h->droppable) - return; - - ff_thread_report_progress(&h->cur_pic_ptr->tf, top + height - 1, - h->picture_structure == PICT_BOTTOM_FIELD); -} - -static void er_add_slice(H264Context *h, int startx, int starty, - int endx, int endy, int status) -{ -#if CONFIG_ERROR_RESILIENCE - ERContext *er = &h->er; - - er->ref_count = h->ref_count[0]; - ff_er_add_slice(er, startx, starty, endx, endy, status); -#endif -} - -static int decode_slice(struct AVCodecContext *avctx, void *arg) -{ - H264Context *h = *(void **)arg; - int lf_x_start = h->mb_x; - - h->mb_skip_run = -1; - - h->is_complex = FRAME_MBAFF(h) || h->picture_structure != PICT_FRAME || - avctx->codec_id != AV_CODEC_ID_H264 || - (CONFIG_GRAY && (h->flags & CODEC_FLAG_GRAY)); - - if (h->pps.cabac) { - /* realign */ - align_get_bits(&h->gb); - - /* init cabac */ - ff_init_cabac_decoder(&h->cabac, - h->gb.buffer + get_bits_count(&h->gb) / 8, - (get_bits_left(&h->gb) + 7) / 8); - - ff_h264_init_cabac_states(h); - - for (;;) { - // START_TIMER - int ret = ff_h264_decode_mb_cabac(h); - int eos; - // STOP_TIMER("decode_mb_cabac") - - if (ret >= 0) - ff_h264_hl_decode_mb(h); - - // FIXME optimal? or let mb_decode decode 16x32 ? - if (ret >= 0 && FRAME_MBAFF(h)) { - h->mb_y++; - - ret = ff_h264_decode_mb_cabac(h); - - if (ret >= 0) - ff_h264_hl_decode_mb(h); - h->mb_y--; - } - eos = get_cabac_terminate(&h->cabac); - - if ((h->workaround_bugs & FF_BUG_TRUNCATED) && - h->cabac.bytestream > h->cabac.bytestream_end + 2) { - er_add_slice(h, h->resync_mb_x, h->resync_mb_y, h->mb_x - 1, - h->mb_y, ER_MB_END); - if (h->mb_x >= lf_x_start) - loop_filter(h, lf_x_start, h->mb_x + 1); - return 0; - } - if (ret < 0 || h->cabac.bytestream > h->cabac.bytestream_end + 2) { - av_log(h->avctx, AV_LOG_ERROR, - "error while decoding MB %d %d, bytestream %td\n", - h->mb_x, h->mb_y, - h->cabac.bytestream_end - h->cabac.bytestream); - er_add_slice(h, h->resync_mb_x, h->resync_mb_y, h->mb_x, - h->mb_y, ER_MB_ERROR); - return AVERROR_INVALIDDATA; - } - - if (++h->mb_x >= h->mb_width) { - loop_filter(h, lf_x_start, h->mb_x); - h->mb_x = lf_x_start = 0; - decode_finish_row(h); - ++h->mb_y; - if (FIELD_OR_MBAFF_PICTURE(h)) { - ++h->mb_y; - if (FRAME_MBAFF(h) && h->mb_y < h->mb_height) - predict_field_decoding_flag(h); - } - } - - if (eos || h->mb_y >= h->mb_height) { - tprintf(h->avctx, "slice end %d %d\n", - get_bits_count(&h->gb), h->gb.size_in_bits); - er_add_slice(h, h->resync_mb_x, h->resync_mb_y, h->mb_x - 1, - h->mb_y, ER_MB_END); - if (h->mb_x > lf_x_start) - loop_filter(h, lf_x_start, h->mb_x); - return 0; - } - } - } else { - for (;;) { - int ret = ff_h264_decode_mb_cavlc(h); - - if (ret >= 0) - ff_h264_hl_decode_mb(h); - - // FIXME optimal? or let mb_decode decode 16x32 ? - if (ret >= 0 && FRAME_MBAFF(h)) { - h->mb_y++; - ret = ff_h264_decode_mb_cavlc(h); - - if (ret >= 0) - ff_h264_hl_decode_mb(h); - h->mb_y--; - } - - if (ret < 0) { - av_log(h->avctx, AV_LOG_ERROR, - "error while decoding MB %d %d\n", h->mb_x, h->mb_y); - er_add_slice(h, h->resync_mb_x, h->resync_mb_y, h->mb_x, - h->mb_y, ER_MB_ERROR); - return ret; - } - - if (++h->mb_x >= h->mb_width) { - loop_filter(h, lf_x_start, h->mb_x); - h->mb_x = lf_x_start = 0; - decode_finish_row(h); - ++h->mb_y; - if (FIELD_OR_MBAFF_PICTURE(h)) { - ++h->mb_y; - if (FRAME_MBAFF(h) && h->mb_y < h->mb_height) - predict_field_decoding_flag(h); - } - if (h->mb_y >= h->mb_height) { - tprintf(h->avctx, "slice end %d %d\n", - get_bits_count(&h->gb), h->gb.size_in_bits); - - if (get_bits_left(&h->gb) == 0) { - er_add_slice(h, h->resync_mb_x, h->resync_mb_y, - h->mb_x - 1, h->mb_y, - ER_MB_END); - - return 0; - } else { - er_add_slice(h, h->resync_mb_x, h->resync_mb_y, - h->mb_x - 1, h->mb_y, - ER_MB_END); - - return AVERROR_INVALIDDATA; - } - } - } - - if (get_bits_left(&h->gb) <= 0 && h->mb_skip_run <= 0) { - tprintf(h->avctx, "slice end %d %d\n", - get_bits_count(&h->gb), h->gb.size_in_bits); - - if (get_bits_left(&h->gb) == 0) { - er_add_slice(h, h->resync_mb_x, h->resync_mb_y, - h->mb_x - 1, h->mb_y, - ER_MB_END); - if (h->mb_x > lf_x_start) - loop_filter(h, lf_x_start, h->mb_x); - - return 0; - } else { - er_add_slice(h, h->resync_mb_x, h->resync_mb_y, h->mb_x, - h->mb_y, ER_MB_ERROR); - - return AVERROR_INVALIDDATA; - } - } - } - } -} - -/** - * Call decode_slice() for each context. - * - * @param h h264 master context - * @param context_count number of contexts to execute - */ -static int execute_decode_slices(H264Context *h, unsigned context_count) -{ - AVCodecContext *const avctx = h->avctx; - H264Context *hx; - int i; - - if (h->mb_y >= h->mb_height) { - av_log(h->avctx, AV_LOG_ERROR, - "Input contains more MB rows than the frame height.\n"); - return AVERROR_INVALIDDATA; - } - - if (h->avctx->hwaccel) - return 0; - if (context_count == 1) { - return decode_slice(avctx, &h); - } else { - for (i = 1; i < context_count; i++) { - hx = h->thread_context[i]; - hx->er.error_count = 0; - } - - avctx->execute(avctx, decode_slice, h->thread_context, - NULL, context_count, sizeof(void *)); - - /* pull back stuff from slices to master context */ - hx = h->thread_context[context_count - 1]; - h->mb_x = hx->mb_x; - h->mb_y = hx->mb_y; - h->droppable = hx->droppable; - h->picture_structure = hx->picture_structure; - for (i = 1; i < context_count; i++) - h->er.error_count += h->thread_context[i]->er.error_count; - } - - return 0; -} - static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size, int parse_extradata) { @@ -3929,7 +1436,7 @@ again: hx->inter_gb_ptr = &hx->gb; hx->data_partitioning = 0; - if ((err = decode_slice_header(hx, h))) + if ((err = ff_h264_decode_slice_header(hx, h))) break; if (h->sei_recovery_frame_cnt >= 0 && h->recovery_frame < 0) { @@ -3991,7 +1498,7 @@ again: hx->intra_gb_ptr = hx->inter_gb_ptr = NULL; - if ((err = decode_slice_header(hx, h)) < 0) { + if ((err = ff_h264_decode_slice_header(hx, h)) < 0) { /* make sure data_partitioning is cleared if it was set * before, so we don't try decoding a slice without a valid * slice header later */ @@ -4036,7 +1543,7 @@ again: ff_h264_decode_seq_parameter_set(h); } - ret = h264_set_parameter_from_sps(h); + ret = ff_h264_set_parameter_from_sps(h); if (ret < 0) goto end; @@ -4060,7 +1567,7 @@ again: } if (context_count == h->max_contexts) { - execute_decode_slices(h, context_count); + ff_h264_execute_decode_slices(h, context_count); context_count = 0; } @@ -4080,7 +1587,7 @@ again: } } if (context_count) - execute_decode_slices(h, context_count); + ff_h264_execute_decode_slices(h, context_count); end: /* clean up */ @@ -4195,7 +1702,7 @@ out: if (avctx->flags2 & CODEC_FLAG2_CHUNKS) decode_postinit(h, 1); - field_end(h, 0); + ff_h264_field_end(h, 0); *got_frame = 0; if (h->next_output_pic && ((avctx->flags & CODEC_FLAG_OUTPUT_CORRUPT) || @@ -4219,7 +1726,7 @@ av_cold void ff_h264_free_context(H264Context *h) { int i; - free_tables(h, 1); // FIXME cleanup init stuff perhaps + ff_h264_free_tables(h, 1); // FIXME cleanup init stuff perhaps for (i = 0; i < MAX_SPS_COUNT; i++) av_freep(h->sps_buffers + i); @@ -4234,7 +1741,7 @@ static av_cold int h264_decode_end(AVCodecContext *avctx) ff_h264_free_context(h); - unref_picture(h, &h->cur_pic); + ff_h264_unref_picture(h, &h->cur_pic); return 0; } @@ -4270,6 +1777,6 @@ AVCodec ff_h264_decoder = { CODEC_CAP_FRAME_THREADS, .flush = flush_dpb, .init_thread_copy = ONLY_IF_THREADS_ENABLED(decode_init_thread_copy), - .update_thread_context = ONLY_IF_THREADS_ENABLED(decode_update_thread_context), + .update_thread_context = ONLY_IF_THREADS_ENABLED(ff_h264_update_thread_context), .profiles = NULL_IF_CONFIG_SMALL(profiles), }; -- cgit v1.2.3