git.sesse.net Git - casparcg/blob - modules/ffmpeg/producer/muxer/frame_muxer.cpp

   1 /*
   2 * Copyright (c) 2011 Sveriges Television AB <info@casparcg.com>
   3 *
   4 * This file is part of CasparCG (www.casparcg.com).
   5 *
   6 * CasparCG is free software: you can redistribute it and/or modify
   7 * it under the terms of the GNU General Public License as published by
   8 * the Free Software Foundation, either version 3 of the License, or
   9 * (at your option) any later version.
  10 *
  11 * CasparCG is distributed in the hope that it will be useful,
  12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14 * GNU General Public License for more details.
  15 *
  16 * You should have received a copy of the GNU General Public License
  17 * along with CasparCG. If not, see <http://www.gnu.org/licenses/>.
  18 *
  19 * Author: Robert Nagy, ronag89@gmail.com
  20 */
  21
  22 #include "../../StdAfx.h"
  23
  24 #include "frame_muxer.h"
  25
  26 #include "../filter/filter.h"
  27 #include "../util/util.h"
  28
  29 #include <core/producer/frame_producer.h>
  30 #include <core/frame/draw_frame.h>
  31 #include <core/frame/frame_transform.h>
  32 #include <core/frame/pixel_format.h>
  33 #include <core/frame/frame_factory.h>
  34 #include <core/frame/frame.h>
  35 #include <core/frame/audio_channel_layout.h>
  36
  37 #include <common/env.h>
  38 #include <common/except.h>
  39 #include <common/log.h>
  40
  41 #if defined(_MSC_VER)
  42 #pragma warning (push)
  43 #pragma warning (disable : 4244)
  44 #endif
  45 extern "C"
  46 {
  47         #define __STDC_CONSTANT_MACROS
  48         #define __STDC_LIMIT_MACROS
  49         #include <libavcodec/avcodec.h>
  50         #include <libavformat/avformat.h>
  51 }
  52 #if defined(_MSC_VER)
  53 #pragma warning (pop)
  54 #endif
  55
  56 #include <common/assert.h>
  57 #include <boost/range/algorithm_ext/push_back.hpp>
  58 #include <boost/algorithm/string/predicate.hpp>
  59
  60 #include <deque>
  61 #include <queue>
  62 #include <vector>
  63
  64 using namespace caspar::core;
  65
  66 namespace caspar { namespace ffmpeg {
  67
  68 bool is_frame_format_changed(const AVFrame& lhs, const AVFrame& rhs)
  69 {
  70         if (lhs.format != rhs.format)
  71                 return true;
  72
  73         for (int i = 0; i < AV_NUM_DATA_POINTERS; ++i)
  74         {
  75                 if (lhs.linesize[i] != rhs.linesize[i])
  76                         return true;
  77         }
  78
  79         return false;
  80 }
  81
  82 struct frame_muxer::impl : boost::noncopyable
  83 {
  84         std::queue<core::mutable_frame>                                 video_stream_;
  85         core::mutable_audio_buffer                                              audio_stream_;
  86         std::queue<draw_frame>                                                  frame_buffer_;
  87         display_mode                                                                    display_mode_                   = display_mode::invalid;
  88         const double                                                                    in_fps_;
  89         const video_format_desc                                                 format_desc_;
  90         audio_channel_layout                                                    channel_layout_;
  91
  92         std::vector<int>                                                                audio_cadence_                  = format_desc_.audio_cadence;
  93
  94         spl::shared_ptr<core::frame_factory>                    frame_factory_;
  95         std::shared_ptr<AVFrame>                                                previous_frame_;
  96
  97         std::unique_ptr<filter>                                                 filter_;
  98         const std::wstring                                                              filter_str_;
  99         bool                                                                                    force_deinterlacing_    = env::properties().get(L"configuration.force-deinterlace", true);
 100
 101         impl(
 102                         double in_fps,
 103                         const spl::shared_ptr<core::frame_factory>& frame_factory,
 104                         const core::video_format_desc& format_desc,
 105                         const core::audio_channel_layout& channel_layout,
 106                         const std::wstring& filter_str)
 107                 : in_fps_(in_fps)
 108                 , format_desc_(format_desc)
 109                 , channel_layout_(channel_layout)
 110                 , frame_factory_(frame_factory)
 111                 , filter_str_(filter_str)
 112         {
 113                 // Note: Uses 1 step rotated cadence for 1001 modes (1602, 1602, 1601, 1602, 1601)
 114                 // This cadence fills the audio mixer most optimally.
 115                 boost::range::rotate(audio_cadence_, std::end(audio_cadence_)-1);
 116         }
 117
 118         void push_video(const std::shared_ptr<AVFrame>& video)
 119         {
 120                 if(!video)
 121                         return;
 122
 123                 if (previous_frame_ && video->data[0] && is_frame_format_changed(*previous_frame_, *video))
 124                 {
 125                         // Fixes bug where avfilter crashes server on some DV files (starts in YUV420p but changes to YUV411p after the first frame).
 126                         CASPAR_LOG(info) << L"[frame_muxer] Frame format has changed. Resetting display mode.";
 127                         display_mode_ = display_mode::invalid;
 128                 }
 129
 130                 if(!video->data[0])
 131                 {
 132                         auto empty_frame = frame_factory_->create_frame(this, core::pixel_format_desc(core::pixel_format::invalid), channel_layout_);
 133                         video_stream_.push(std::move(empty_frame));
 134                         display_mode_ = display_mode::simple;
 135                 }
 136                 else
 137                 {
 138                         if(!filter_ || display_mode_ == display_mode::invalid)
 139                                 update_display_mode(video);
 140
 141                         filter_->push(video);
 142                         previous_frame_ = video;
 143                         for (auto& av_frame : filter_->poll_all())
 144                                 video_stream_.push(make_frame(this, av_frame, format_desc_.fps, *frame_factory_, channel_layout_));
 145                 }
 146
 147                 merge();
 148         }
 149
 150         void push_audio(const std::shared_ptr<AVFrame>& audio)
 151         {
 152                 if(!audio)
 153                         return;
 154
 155                 if(!audio->data[0])
 156                 {
 157                         if (channel_layout_ == core::audio_channel_layout::invalid())
 158                                 channel_layout_ = *core::audio_channel_layout_repository::get_default()->get_layout(L"stereo");
 159
 160                         boost::range::push_back(audio_stream_, core::mutable_audio_buffer(audio_cadence_.front() * channel_layout_.num_channels, 0));
 161                 }
 162                 else
 163                 {
 164                         auto ptr = reinterpret_cast<int32_t*>(audio->data[0]);
 165                         audio_stream_.insert(audio_stream_.end(), ptr, ptr + audio->linesize[0]/sizeof(int32_t));
 166                 }
 167
 168                 merge();
 169         }
 170
 171         bool video_ready() const
 172         {
 173                 switch(display_mode_)
 174                 {
 175                 case display_mode::deinterlace_bob_reinterlace:
 176                 case display_mode::interlace:
 177                 case display_mode::half:
 178                         return video_stream_.size() >= 2;
 179                 default:
 180                         return video_stream_.size() >= 1;
 181                 }
 182         }
 183
 184         bool audio_ready() const
 185         {
 186                 switch(display_mode_)
 187                 {
 188                 case display_mode::duplicate:
 189                         return audio_stream_.size() >= static_cast<size_t>(audio_cadence_[0] + audio_cadence_[1 % audio_cadence_.size()]) * channel_layout_.num_channels;
 190                 default:
 191                         return audio_stream_.size() >= static_cast<size_t>(audio_cadence_.front()) * channel_layout_.num_channels;
 192                 }
 193         }
 194
 195         bool empty() const
 196         {
 197                 return frame_buffer_.empty();
 198         }
 199
 200         core::draw_frame front() const
 201         {
 202                 return frame_buffer_.front();
 203         }
 204
 205         void pop()
 206         {
 207                 frame_buffer_.pop();
 208         }
 209
 210         void merge()
 211         {
 212                 while(video_ready() && audio_ready() && display_mode_ != display_mode::invalid)
 213                 {
 214                         auto frame1                     = pop_video();
 215                         frame1.audio_data()     = pop_audio();
 216
 217                         switch(display_mode_)
 218                         {
 219                         case display_mode::simple:
 220                         case display_mode::deinterlace_bob:
 221                         case display_mode::deinterlace:
 222                                 {
 223                                         frame_buffer_.push(core::draw_frame(std::move(frame1)));
 224                                         break;
 225                                 }
 226                         case display_mode::interlace:
 227                         case display_mode::deinterlace_bob_reinterlace:
 228                                 {
 229                                         auto frame2 = pop_video();
 230
 231                                         frame_buffer_.push(core::draw_frame::interlace(
 232                                                 core::draw_frame(std::move(frame1)),
 233                                                 core::draw_frame(std::move(frame2)),
 234                                                 format_desc_.field_mode));
 235                                         break;
 236                                 }
 237                         case display_mode::duplicate:
 238                                 {
 239                                         //boost::range::push_back(frame1.audio_data(), pop_audio());
 240
 241                                         auto second_audio_frame = core::mutable_frame(
 242                                                         std::vector<array<std::uint8_t>>(),
 243                                                         pop_audio(),
 244                                                         frame1.stream_tag(),
 245                                                         core::pixel_format_desc(),
 246                                                         channel_layout_);
 247                                         auto first_frame = core::draw_frame(std::move(frame1));
 248                                         auto muted_first_frame = core::draw_frame(first_frame);
 249                                         muted_first_frame.transform().audio_transform.volume = 0;
 250                                         auto second_frame = core::draw_frame({ core::draw_frame(std::move(second_audio_frame)), muted_first_frame });
 251
 252                                         // Same video but different audio.
 253                                         frame_buffer_.push(first_frame);
 254                                         frame_buffer_.push(second_frame);
 255                                         break;
 256                                 }
 257                         case display_mode::half:
 258                                 {
 259                                         pop_video(); // Throw away
 260
 261                                         frame_buffer_.push(core::draw_frame(std::move(frame1)));
 262                                         break;
 263                                 }
 264                         default:
 265                                 CASPAR_THROW_EXCEPTION(invalid_operation());
 266                         }
 267                 }
 268         }
 269
 270         core::mutable_frame pop_video()
 271         {
 272                 auto frame = std::move(video_stream_.front());
 273                 video_stream_.pop();
 274                 return std::move(frame);
 275         }
 276
 277         core::mutable_audio_buffer pop_audio()
 278         {
 279                 if (audio_stream_.size() < audio_cadence_.front() * channel_layout_.num_channels)
 280                         CASPAR_THROW_EXCEPTION(out_of_range());
 281
 282                 auto begin = audio_stream_.begin();
 283                 auto end   = begin + audio_cadence_.front() * channel_layout_.num_channels;
 284
 285                 core::mutable_audio_buffer samples(begin, end);
 286                 audio_stream_.erase(begin, end);
 287
 288                 boost::range::rotate(audio_cadence_, std::begin(audio_cadence_)+1);
 289
 290                 return samples;
 291         }
 292
 293         void update_display_mode(const std::shared_ptr<AVFrame>& frame)
 294         {
 295                 std::wstring filter_str = filter_str_;
 296
 297                 display_mode_ = display_mode::simple;
 298
 299                 auto mode = get_mode(*frame);
 300                 if(mode == core::field_mode::progressive && frame->height < 720 && in_fps_ < 50.0) // SD frames are interlaced. Probably incorrect meta-data. Fix it.
 301                         mode = core::field_mode::upper;
 302
 303                 auto fps  = in_fps_;
 304
 305                 if(filter::is_deinterlacing(filter_str_))
 306                         mode = core::field_mode::progressive;
 307
 308                 if(filter::is_double_rate(filter_str_))
 309                         fps *= 2;
 310
 311                 display_mode_ = get_display_mode(mode, fps, format_desc_.field_mode, format_desc_.fps);
 312
 313                 if((frame->height != 480 || format_desc_.height != 486) && // don't deinterlace for NTSC DV
 314                                 display_mode_ == display_mode::simple && mode != core::field_mode::progressive && format_desc_.field_mode != core::field_mode::progressive &&
 315                                 frame->height != format_desc_.height)
 316                 {
 317                         display_mode_ = display_mode::deinterlace_bob_reinterlace; // The frame will most likely be scaled, we need to deinterlace->reinterlace
 318                 }
 319
 320                 // ALWAYS de-interlace, until we have GPU de-interlacing.
 321                 if(force_deinterlacing_ && frame->interlaced_frame && display_mode_ != display_mode::deinterlace_bob && display_mode_ != display_mode::deinterlace)
 322                         display_mode_ = display_mode::deinterlace_bob_reinterlace;
 323
 324                 if(display_mode_ == display_mode::deinterlace)
 325                         filter_str = append_filter(filter_str, L"YADIF=0:-1");
 326                 else if(display_mode_ == display_mode::deinterlace_bob || display_mode_ == display_mode::deinterlace_bob_reinterlace)
 327                         filter_str = append_filter(filter_str, L"YADIF=1:-1");
 328
 329                 if(display_mode_ == display_mode::invalid)
 330                 {
 331                         CASPAR_LOG(warning) << L"[frame_muxer] Auto-transcode: Failed to detect display-mode.";
 332                         display_mode_ = display_mode::simple;
 333                 }
 334
 335                 if(frame->height == 480) // NTSC DV
 336                 {
 337                         auto pad_str = L"PAD=" + boost::lexical_cast<std::wstring>(frame->width) + L":486:0:2:black";
 338                         filter_str = append_filter(filter_str, pad_str);
 339                 }
 340
 341                 filter_.reset (new filter(
 342                         frame->width,
 343                         frame->height,
 344                         boost::rational<int>(1000000, static_cast<int>(in_fps_ * 1000000)),
 345                         boost::rational<int>(static_cast<int>(in_fps_ * 1000000), 1000000),
 346                         boost::rational<int>(frame->sample_aspect_ratio.num, frame->sample_aspect_ratio.den),
 347                         static_cast<AVPixelFormat>(frame->format),
 348                         std::vector<AVPixelFormat>(),
 349                         u8(filter_str)));
 350
 351                 CASPAR_LOG(info) << L"[frame_muxer] " << display_mode_ << L" " << print_mode(frame->width, frame->height, in_fps_, frame->interlaced_frame > 0);
 352         }
 353
 354         uint32_t calc_nb_frames(uint32_t nb_frames) const
 355         {
 356                 uint64_t nb_frames2 = nb_frames;
 357
 358                 if(filter_ && filter_->is_double_rate()) // Take into account transformations in filter.
 359                         nb_frames2 *= 2;
 360
 361                 switch(display_mode_) // Take into account transformation in run.
 362                 {
 363                 case display_mode::deinterlace_bob_reinterlace:
 364                 case display_mode::interlace:
 365                 case display_mode::half:
 366                         nb_frames2 /= 2;
 367                         break;
 368                 case display_mode::duplicate:
 369                         nb_frames2 *= 2;
 370                         break;
 371                 }
 372
 373                 return static_cast<uint32_t>(nb_frames2);
 374         }
 375
 376         void clear()
 377         {
 378                 while(!video_stream_.empty())
 379                         video_stream_.pop();
 380
 381                 audio_stream_.clear();
 382
 383                 while(!frame_buffer_.empty())
 384                         frame_buffer_.pop();
 385
 386                 filter_.reset();
 387         }
 388 };
 389
 390 frame_muxer::frame_muxer(
 391                 double in_fps,
 392                 const spl::shared_ptr<core::frame_factory>& frame_factory,
 393                 const core::video_format_desc& format_desc,
 394                 const core::audio_channel_layout& channel_layout,
 395                 const std::wstring& filter)
 396         : impl_(new impl(in_fps, frame_factory, format_desc, channel_layout, filter)){}
 397 void frame_muxer::push_video(const std::shared_ptr<AVFrame>& frame){impl_->push_video(frame);}
 398 void frame_muxer::push_audio(const std::shared_ptr<AVFrame>& frame){impl_->push_audio(frame);}
 399 bool frame_muxer::empty() const{return impl_->empty();}
 400 core::draw_frame frame_muxer::front() const{return impl_->front();}
 401 void frame_muxer::pop(){return impl_->pop();}
 402 void frame_muxer::clear(){impl_->clear();}
 403 uint32_t frame_muxer::calc_nb_frames(uint32_t nb_frames) const {return impl_->calc_nb_frames(nb_frames);}
 404 bool frame_muxer::video_ready() const{return impl_->video_ready();}
 405 bool frame_muxer::audio_ready() const{return impl_->audio_ready();}
 406
 407 }}