X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Flibtheoraenc.c;h=6c4eddf301167602a11715a77e6b8133541ce44f;hb=dbb37e77117466edfc146f9f0df4c70bef6239d6;hp=7d62508aa82c1cea20ac3da76494f6601982cd01;hpb=bdc8c48893ba45e30841cc62dfe89cb2b2e85bbb;p=ffmpeg diff --git a/libavcodec/libtheoraenc.c b/libavcodec/libtheoraenc.c index 7d62508aa82..6c4eddf3011 100644 --- a/libavcodec/libtheoraenc.c +++ b/libavcodec/libtheoraenc.c @@ -1,27 +1,27 @@ /* * Copyright (c) 2006 Paul Richards * - * This file is part of FFmpeg. + * This file is part of Libav. * - * FFmpeg is free software; you can redistribute it and/or + * Libav is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * - * FFmpeg is distributed in the hope that it will be useful, + * Libav is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public - * License along with FFmpeg; if not, write to the Free Software + * License along with Libav; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ -/*! - * \file libtheoraenc.c - * \brief Theora encoder using libtheora. - * \author Paul Richards +/** + * @file + * @brief Theora encoder using libtheora. + * @author Paul Richards * * A lot of this is copy / paste from other output codecs in * libavcodec or pure guesswork (or both). @@ -30,21 +30,28 @@ * and o_ prefixes on variables which are libogg types. */ -/* FFmpeg includes */ +/* Libav includes */ +#include "libavutil/common.h" #include "libavutil/intreadwrite.h" #include "libavutil/log.h" +#include "libavutil/base64.h" #include "avcodec.h" +#include "internal.h" /* libtheora includes */ #include typedef struct TheoraContext { th_enc_ctx *t_state; + uint8_t *stats; + int stats_size; + int stats_offset; + int uv_hshift; + int uv_vshift; + int keyframe_mask; } TheoraContext; -/*! - Concatenates an ogg_packet into the extradata. -*/ +/** Concatenate an ogg_packet into the extradata. */ static int concatenate_packet(unsigned int* offset, AVCodecContext* avc_context, const ogg_packet* packet) @@ -78,6 +85,72 @@ static int concatenate_packet(unsigned int* offset, return 0; } +static int get_stats(AVCodecContext *avctx, int eos) +{ +#ifdef TH_ENCCTL_2PASS_OUT + TheoraContext *h = avctx->priv_data; + uint8_t *buf; + int bytes; + + bytes = th_encode_ctl(h->t_state, TH_ENCCTL_2PASS_OUT, &buf, sizeof(buf)); + if (bytes < 0) { + av_log(avctx, AV_LOG_ERROR, "Error getting first pass stats\n"); + return -1; + } + if (!eos) { + h->stats = av_fast_realloc(h->stats, &h->stats_size, + h->stats_offset + bytes); + memcpy(h->stats + h->stats_offset, buf, bytes); + h->stats_offset += bytes; + } else { + int b64_size = AV_BASE64_SIZE(h->stats_offset); + // libtheora generates a summary header at the end + memcpy(h->stats, buf, bytes); + avctx->stats_out = av_malloc(b64_size); + av_base64_encode(avctx->stats_out, b64_size, h->stats, h->stats_offset); + } + return 0; +#else + av_log(avctx, AV_LOG_ERROR, "libtheora too old to support 2pass\n"); + return -1; +#endif +} + +// libtheora won't read the entire buffer we give it at once, so we have to +// repeatedly submit it... +static int submit_stats(AVCodecContext *avctx) +{ +#ifdef TH_ENCCTL_2PASS_IN + TheoraContext *h = avctx->priv_data; + int bytes; + if (!h->stats) { + if (!avctx->stats_in) { + av_log(avctx, AV_LOG_ERROR, "No statsfile for second pass\n"); + return -1; + } + h->stats_size = strlen(avctx->stats_in) * 3/4; + h->stats = av_malloc(h->stats_size); + h->stats_size = av_base64_decode(h->stats, avctx->stats_in, h->stats_size); + } + while (h->stats_size - h->stats_offset > 0) { + bytes = th_encode_ctl(h->t_state, TH_ENCCTL_2PASS_IN, + h->stats + h->stats_offset, + h->stats_size - h->stats_offset); + if (bytes < 0) { + av_log(avctx, AV_LOG_ERROR, "Error submitting stats\n"); + return -1; + } + if (!bytes) + return 0; + h->stats_offset += bytes; + } + return 0; +#else + av_log(avctx, AV_LOG_ERROR, "libtheora too old to support 2pass\n"); + return -1; +#endif +} + static av_cold int encode_init(AVCodecContext* avc_context) { th_info t_info; @@ -106,8 +179,25 @@ static av_cold int encode_init(AVCodecContext* avc_context) t_info.aspect_numerator = 1; t_info.aspect_denominator = 1; } - t_info.colorspace = TH_CS_UNSPECIFIED; - t_info.pixel_fmt = TH_PF_420; + + if (avc_context->color_primaries == AVCOL_PRI_BT470M) + t_info.colorspace = TH_CS_ITU_REC_470M; + else if (avc_context->color_primaries == AVCOL_PRI_BT470BG) + t_info.colorspace = TH_CS_ITU_REC_470BG; + else + t_info.colorspace = TH_CS_UNSPECIFIED; + + if (avc_context->pix_fmt == AV_PIX_FMT_YUV420P) + t_info.pixel_fmt = TH_PF_420; + else if (avc_context->pix_fmt == AV_PIX_FMT_YUV422P) + t_info.pixel_fmt = TH_PF_422; + else if (avc_context->pix_fmt == AV_PIX_FMT_YUV444P) + t_info.pixel_fmt = TH_PF_444; + else { + av_log(avc_context, AV_LOG_ERROR, "Unsupported pix_fmt\n"); + return -1; + } + avcodec_get_chroma_sub_sample(avc_context->pix_fmt, &h->uv_hshift, &h->uv_vshift); if (avc_context->flags & CODEC_FLAG_QSCALE) { /* to be constant with the libvorbis implementation, clip global_quality to 0 - 10 @@ -129,6 +219,7 @@ static av_cold int encode_init(AVCodecContext* avc_context) return -1; } + h->keyframe_mask = (1 << t_info.keyframe_granule_shift) - 1; /* Clear up theora_info struct */ th_info_clear(&t_info); @@ -138,12 +229,21 @@ static av_cold int encode_init(AVCodecContext* avc_context) return -1; } + // need to enable 2 pass (via TH_ENCCTL_2PASS_) before encoding headers + if (avc_context->flags & CODEC_FLAG_PASS1) { + if (get_stats(avc_context, 0)) + return -1; + } else if (avc_context->flags & CODEC_FLAG_PASS2) { + if (submit_stats(avc_context)) + return -1; + } + /* Output first header packet consisting of theora header, comment, and tables. Each one is prefixed with a 16bit size, then they - are concatenated together into ffmpeg's extradata. + are concatenated together into libavcodec's extradata. */ offset = 0; @@ -162,25 +262,35 @@ static av_cold int encode_init(AVCodecContext* avc_context) return 0; } -static int encode_frame(AVCodecContext* avc_context, uint8_t *outbuf, - int buf_size, void *data) +static int encode_frame(AVCodecContext* avc_context, AVPacket *pkt, + const AVFrame *frame, int *got_packet) { th_ycbcr_buffer t_yuv_buffer; TheoraContext *h = avc_context->priv_data; - AVFrame *frame = data; ogg_packet o_packet; - int result, i; - - assert(avc_context->pix_fmt == PIX_FMT_YUV420P); + int result, i, ret; + + // EOS, finish and get 1st pass stats if applicable + if (!frame) { + th_encode_packetout(h->t_state, 1, &o_packet); + if (avc_context->flags & CODEC_FLAG_PASS1) + if (get_stats(avc_context, 1)) + return -1; + return 0; + } /* Copy planes to the theora yuv_buffer */ for (i = 0; i < 3; i++) { - t_yuv_buffer[i].width = FFALIGN(avc_context->width, 16) >> !!i; - t_yuv_buffer[i].height = FFALIGN(avc_context->height, 16) >> !!i; + t_yuv_buffer[i].width = FFALIGN(avc_context->width, 16) >> (i && h->uv_hshift); + t_yuv_buffer[i].height = FFALIGN(avc_context->height, 16) >> (i && h->uv_vshift); t_yuv_buffer[i].stride = frame->linesize[i]; t_yuv_buffer[i].data = frame->data[i]; } + if (avc_context->flags & CODEC_FLAG_PASS2) + if (submit_stats(avc_context)) + return -1; + /* Now call into theora_encode_YUVin */ result = th_encode_ycbcr_in(h->t_state, t_yuv_buffer); if (result) { @@ -200,6 +310,10 @@ static int encode_frame(AVCodecContext* avc_context, uint8_t *outbuf, return -1; } + if (avc_context->flags & CODEC_FLAG_PASS1) + if (get_stats(avc_context, 0)) + return -1; + /* Pick up returned ogg_packet */ result = th_encode_packetout(h->t_state, 0, &o_packet); switch (result) { @@ -215,58 +329,49 @@ static int encode_frame(AVCodecContext* avc_context, uint8_t *outbuf, } /* Copy ogg_packet content out to buffer */ - if (buf_size < o_packet.bytes) { - av_log(avc_context, AV_LOG_ERROR, "encoded frame too large\n"); - return -1; + if ((ret = ff_alloc_packet(pkt, o_packet.bytes)) < 0) { + av_log(avc_context, AV_LOG_ERROR, "Error getting output packet of size %ld.\n", o_packet.bytes); + return ret; } - memcpy(outbuf, o_packet.packet, o_packet.bytes); + memcpy(pkt->data, o_packet.packet, o_packet.bytes); - // HACK: does not take codec delay into account (neither does the decoder though) - avc_context->coded_frame->pts = frame->pts; + // HACK: assumes no encoder delay, this is true until libtheora becomes + // multithreaded (which will be disabled unless explictly requested) + pkt->pts = pkt->dts = frame->pts; + avc_context->coded_frame->key_frame = !(o_packet.granulepos & h->keyframe_mask); + if (avc_context->coded_frame->key_frame) + pkt->flags |= AV_PKT_FLAG_KEY; + *got_packet = 1; - return o_packet.bytes; + return 0; } static av_cold int encode_close(AVCodecContext* avc_context) { - ogg_packet o_packet; TheoraContext *h = avc_context->priv_data; - int result; - const char* message; - result = th_encode_packetout(h->t_state, 1, &o_packet); th_encode_free(h->t_state); + av_freep(&h->stats); av_freep(&avc_context->coded_frame); + av_freep(&avc_context->stats_out); av_freep(&avc_context->extradata); avc_context->extradata_size = 0; - switch (result) { - case 0: /* No packet is ready */ - case -1: /* Encoding finished */ - return 0; - case 1: - /* We have a packet */ - message = "gave us a packet"; - break; - default: - message = "unknown reason"; - break; - } - av_log(avc_context, AV_LOG_ERROR, "theora_encode_packetout failed (%s) [%d]\n", message, result); - return -1; + return 0; } -static const enum PixelFormat supported_pixel_formats[] = { PIX_FMT_YUV420P, PIX_FMT_NONE }; - -/*! AVCodec struct exposed to libavcodec */ -AVCodec libtheora_encoder = { - .name = "libtheora", - .type = CODEC_TYPE_VIDEO, - .id = CODEC_ID_THEORA, +/** AVCodec struct exposed to libavcodec */ +AVCodec ff_libtheora_encoder = { + .name = "libtheora", + .type = AVMEDIA_TYPE_VIDEO, + .id = AV_CODEC_ID_THEORA, .priv_data_size = sizeof(TheoraContext), - .init = encode_init, - .close = encode_close, - .encode = encode_frame, - .pix_fmts = supported_pixel_formats, - .long_name = NULL_IF_CONFIG_SMALL("libtheora Theora"), + .init = encode_init, + .close = encode_close, + .encode2 = encode_frame, + .capabilities = CODEC_CAP_DELAY, // needed to get the statsfile summary + .pix_fmts = (const enum AVPixelFormat[]){ + AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV444P, AV_PIX_FMT_NONE + }, + .long_name = NULL_IF_CONFIG_SMALL("libtheora Theora"), };