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