git.sesse.net Git - casparcg/blob - modules/ffmpeg/producer/util/util.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 "util.h"
  25
  26 #include "flv.h"
  27
  28 #include "../tbb_avcodec.h"
  29 #include "../../ffmpeg_error.h"
  30
  31 #include <tbb/concurrent_unordered_map.h>
  32 #include <tbb/concurrent_queue.h>
  33
  34 #include <core/frame/frame_transform.h>
  35 #include <core/frame/frame_factory.h>
  36 #include <core/frame/frame.h>
  37 #include <core/producer/frame_producer.h>
  38
  39 #include <common/except.h>
  40 #include <common/array.h>
  41
  42 #include <tbb/parallel_for.h>
  43
  44 #include <common/assert.h>
  45 #include <boost/filesystem.hpp>
  46 #include <boost/lexical_cast.hpp>
  47
  48 #include <asmlib.h>
  49
  50 #if defined(_MSC_VER)
  51 #pragma warning (push)
  52 #pragma warning (disable : 4244)
  53 #endif
  54 extern "C"
  55 {
  56         #include <libswscale/swscale.h>
  57         #include <libavcodec/avcodec.h>
  58         #include <libavformat/avformat.h>
  59 }
  60 #if defined(_MSC_VER)
  61 #pragma warning (pop)
  62 #endif
  63
  64 namespace caspar { namespace ffmpeg {
  65
  66 core::field_mode get_mode(const AVFrame& frame)
  67 {
  68         if(!frame.interlaced_frame)
  69                 return core::field_mode::progressive;
  70
  71         return frame.top_field_first ? core::field_mode::upper : core::field_mode::lower;
  72 }
  73
  74 core::pixel_format get_pixel_format(PixelFormat pix_fmt)
  75 {
  76         switch(pix_fmt)
  77         {
  78         case PIX_FMT_GRAY8:                     return core::pixel_format::gray;
  79         case PIX_FMT_RGB24:                     return core::pixel_format::rgb;
  80         case PIX_FMT_BGR24:                     return core::pixel_format::bgr;
  81         case PIX_FMT_BGRA:                      return core::pixel_format::bgra;
  82         case PIX_FMT_ARGB:                      return core::pixel_format::argb;
  83         case PIX_FMT_RGBA:                      return core::pixel_format::rgba;
  84         case PIX_FMT_ABGR:                      return core::pixel_format::abgr;
  85         case PIX_FMT_YUV444P:           return core::pixel_format::ycbcr;
  86         case PIX_FMT_YUV422P:           return core::pixel_format::ycbcr;
  87         case PIX_FMT_YUV420P:           return core::pixel_format::ycbcr;
  88         case PIX_FMT_YUV411P:           return core::pixel_format::ycbcr;
  89         case PIX_FMT_YUV410P:           return core::pixel_format::ycbcr;
  90         case PIX_FMT_YUVA420P:          return core::pixel_format::ycbcra;
  91         default:                                        return core::pixel_format::invalid;
  92         }
  93 }
  94
  95 core::pixel_format_desc pixel_format_desc(PixelFormat pix_fmt, int width, int height)
  96 {
  97         // Get linesizes
  98         AVPicture dummy_pict;
  99         avpicture_fill(&dummy_pict, nullptr, pix_fmt, width, height);
 100
 101         core::pixel_format_desc desc = get_pixel_format(pix_fmt);
 102
 103         switch(desc.format.value())
 104         {
 105         case core::pixel_format::gray:
 106         case core::pixel_format::luma:
 107                 {
 108                         desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0], height, 1));
 109                         return desc;
 110                 }
 111         case core::pixel_format::bgr:
 112         case core::pixel_format::rgb:
 113                 {
 114                         desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0]/3, height, 3));
 115                         return desc;
 116                 }
 117         case core::pixel_format::bgra:
 118         case core::pixel_format::argb:
 119         case core::pixel_format::rgba:
 120         case core::pixel_format::abgr:
 121                 {
 122                         desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0]/4, height, 4));
 123                         return desc;
 124                 }
 125         case core::pixel_format::ycbcr:
 126         case core::pixel_format::ycbcra:
 127                 {
 128                         // Find chroma height
 129                         int size2 = static_cast<int>(dummy_pict.data[2] - dummy_pict.data[1]);
 130                         int h2 = size2/dummy_pict.linesize[1];
 131
 132                         desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0], height, 1));
 133                         desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[1], h2, 1));
 134                         desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[2], h2, 1));
 135
 136                         if(desc.format == core::pixel_format::ycbcra)
 137                                 desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[3], height, 1));
 138                         return desc;
 139                 }
 140         default:
 141                 desc.format = core::pixel_format::invalid;
 142                 return desc;
 143         }
 144 }
 145
 146 core::mutable_frame make_frame(const void* tag, const spl::shared_ptr<AVFrame>& decoded_frame, double fps, core::frame_factory& frame_factory)
 147 {
 148         static tbb::concurrent_unordered_map<int64_t, tbb::concurrent_queue<std::shared_ptr<SwsContext>>> sws_contvalid_exts_;
 149
 150         if(decoded_frame->width < 1 || decoded_frame->height < 1)
 151                 return frame_factory.create_frame(tag, core::pixel_format_desc(core::pixel_format::invalid));
 152
 153         const auto width  = decoded_frame->width;
 154         const auto height = decoded_frame->height;
 155         auto desc                 = pixel_format_desc(static_cast<PixelFormat>(decoded_frame->format), width, height);
 156
 157         if(desc.format == core::pixel_format::invalid)
 158         {
 159                 auto pix_fmt = static_cast<PixelFormat>(decoded_frame->format);
 160                 auto target_pix_fmt = PIX_FMT_BGRA;
 161
 162                 if(pix_fmt == PIX_FMT_UYVY422)
 163                         target_pix_fmt = PIX_FMT_YUV422P;
 164                 else if(pix_fmt == PIX_FMT_YUYV422)
 165                         target_pix_fmt = PIX_FMT_YUV422P;
 166                 else if(pix_fmt == PIX_FMT_UYYVYY411)
 167                         target_pix_fmt = PIX_FMT_YUV411P;
 168                 else if(pix_fmt == PIX_FMT_YUV420P10)
 169                         target_pix_fmt = PIX_FMT_YUV420P;
 170                 else if(pix_fmt == PIX_FMT_YUV422P10)
 171                         target_pix_fmt = PIX_FMT_YUV422P;
 172                 else if(pix_fmt == PIX_FMT_YUV444P10)
 173                         target_pix_fmt = PIX_FMT_YUV444P;
 174
 175                 auto target_desc = pixel_format_desc(target_pix_fmt, width, height);
 176
 177                 auto write = frame_factory.create_frame(tag, target_desc);
 178
 179                 std::shared_ptr<SwsContext> sws_context;
 180
 181                 //CASPAR_LOG(warning) << "Hardware accelerated color transform not supported.";
 182
 183                 int64_t key = ((static_cast<int64_t>(width)                      << 32) & 0xFFFF00000000) |
 184                                           ((static_cast<int64_t>(height)                 << 16) & 0xFFFF0000) |
 185                                           ((static_cast<int64_t>(pix_fmt)                <<  8) & 0xFF00) |
 186                                           ((static_cast<int64_t>(target_pix_fmt) <<  0) & 0xFF);
 187
 188                 auto& pool = sws_contvalid_exts_[key];
 189
 190                 if(!pool.try_pop(sws_context))
 191                 {
 192                         double param;
 193                         sws_context.reset(sws_getContext(width, height, pix_fmt, width, height, target_pix_fmt, SWS_BILINEAR, nullptr, nullptr, &param), sws_freeContext);
 194                 }
 195
 196                 if(!sws_context)
 197                 {
 198                         CASPAR_THROW_EXCEPTION(operation_failed() << msg_info("Could not create software scaling context.") <<
 199                                                                         boost::errinfo_api_function("sws_getContext"));
 200                 }
 201
 202                 spl::shared_ptr<AVFrame> av_frame(avcodec_alloc_frame(), av_free);
 203                 avcodec_get_frame_defaults(av_frame.get());
 204                 if(target_pix_fmt == PIX_FMT_BGRA)
 205                 {
 206                         auto size = avpicture_fill(reinterpret_cast<AVPicture*>(av_frame.get()), write.image_data(0).begin(), PIX_FMT_BGRA, width, height);
 207                         CASPAR_VERIFY(size == write.image_data(0).size());
 208                 }
 209                 else
 210                 {
 211                         av_frame->width  = width;
 212                         av_frame->height = height;
 213                         for(int n = 0; n < target_desc.planes.size(); ++n)
 214                         {
 215                                 av_frame->data[n]               = write.image_data(n).begin();
 216                                 av_frame->linesize[n]   = target_desc.planes[n].linesize;
 217                         }
 218                 }
 219
 220                 sws_scale(sws_context.get(), decoded_frame->data, decoded_frame->linesize, 0, height, av_frame->data, av_frame->linesize);
 221                 pool.push(sws_context);
 222
 223                 return std::move(write);
 224         }
 225         else
 226         {
 227                 auto write = frame_factory.create_frame(tag, desc);
 228
 229                 for(int n = 0; n < static_cast<int>(desc.planes.size()); ++n)
 230                 {
 231                         auto plane            = desc.planes[n];
 232                         auto result           = write.image_data(n).begin();
 233                         auto decoded          = decoded_frame->data[n];
 234                         auto decoded_linesize = decoded_frame->linesize[n];
 235
 236                         CASPAR_ASSERT(decoded);
 237                         CASPAR_ASSERT(write.image_data(n).begin());
 238
 239                         // Copy line by line since ffmpeg sometimes pads each line.
 240                         tbb::affinity_partitioner ap;
 241                         tbb::parallel_for(tbb::blocked_range<int>(0, desc.planes[n].height), [&](const tbb::blocked_range<int>& r)
 242                         {
 243                                 for(int y = r.begin(); y != r.end(); ++y)
 244                                         A_memcpy(result + y*plane.linesize, decoded + y*decoded_linesize, plane.linesize);
 245                         }, ap);
 246                 }
 247
 248                 return std::move(write);
 249         }
 250 }
 251
 252 spl::shared_ptr<AVFrame> make_av_frame(core::mutable_frame& frame)
 253 {
 254         std::array<uint8_t*, 4> data = {};
 255         for(int n = 0; n < frame.pixel_format_desc().planes.size(); ++n)
 256                 data[n] = frame.image_data(n).begin();
 257
 258         return make_av_frame(data, frame.pixel_format_desc());
 259 }
 260
 261 spl::shared_ptr<AVFrame> make_av_frame(std::array<uint8_t*, 4> data, const core::pixel_format_desc& pix_desc)
 262 {
 263         spl::shared_ptr<AVFrame> av_frame(avcodec_alloc_frame(), av_free);
 264         avcodec_get_frame_defaults(av_frame.get());
 265
 266         auto planes              = pix_desc.planes;
 267         auto format              = pix_desc.format.value();
 268
 269         av_frame->width  = planes[0].width;
 270         av_frame->height = planes[0].height;
 271         for(int n = 0; n < planes.size(); ++n)
 272         {
 273                 av_frame->data[n]         = data[n];
 274                 av_frame->linesize[n] = planes[n].linesize;
 275         }
 276
 277         switch(format)
 278         {
 279         case core::pixel_format::rgb:
 280                 av_frame->format = PIX_FMT_RGB24;
 281                 break;
 282         case core::pixel_format::bgr:
 283                 av_frame->format = PIX_FMT_BGR24;
 284                 break;
 285         case core::pixel_format::rgba:
 286                 av_frame->format = PIX_FMT_RGBA;
 287                 break;
 288         case core::pixel_format::argb:
 289                 av_frame->format = PIX_FMT_ARGB;
 290                 break;
 291         case core::pixel_format::bgra:
 292                 av_frame->format = PIX_FMT_BGRA;
 293                 break;
 294         case core::pixel_format::abgr:
 295                 av_frame->format = PIX_FMT_ABGR;
 296                 break;
 297         case core::pixel_format::gray:
 298                 av_frame->format = PIX_FMT_GRAY8;
 299                 break;
 300         case core::pixel_format::ycbcr:
 301         {
 302                 int y_w = planes[0].width;
 303                 int y_h = planes[0].height;
 304                 int c_w = planes[1].width;
 305                 int c_h = planes[1].height;
 306
 307                 if(c_h == y_h && c_w == y_w)
 308                         av_frame->format = PIX_FMT_YUV444P;
 309                 else if(c_h == y_h && c_w*2 == y_w)
 310                         av_frame->format = PIX_FMT_YUV422P;
 311                 else if(c_h == y_h && c_w*4 == y_w)
 312                         av_frame->format = PIX_FMT_YUV411P;
 313                 else if(c_h*2 == y_h && c_w*2 == y_w)
 314                         av_frame->format = PIX_FMT_YUV420P;
 315                 else if(c_h*2 == y_h && c_w*4 == y_w)
 316                         av_frame->format = PIX_FMT_YUV410P;
 317
 318                 break;
 319         }
 320         case core::pixel_format::ycbcra:
 321                 av_frame->format = PIX_FMT_YUVA420P;
 322                 break;
 323         }
 324         return av_frame;
 325 }
 326
 327 bool is_sane_fps(AVRational time_base)
 328 {
 329         double fps = static_cast<double>(time_base.den) / static_cast<double>(time_base.num);
 330         return fps > 20.0 && fps < 65.0;
 331 }
 332
 333 AVRational fix_time_base(AVRational time_base)
 334 {
 335         if(time_base.num == 1)
 336                 time_base.num = static_cast<int>(std::pow(10.0, static_cast<int>(std::log10(static_cast<float>(time_base.den)))-1));
 337
 338         if(!is_sane_fps(time_base))
 339         {
 340                 auto tmp = time_base;
 341                 tmp.den /= 2;
 342                 if(is_sane_fps(tmp))
 343                         time_base = tmp;
 344         }
 345
 346         return time_base;
 347 }
 348
 349 double read_fps(AVFormatContext& context, double fail_value)
 350 {
 351         auto video_index = av_find_best_stream(&context, AVMEDIA_TYPE_VIDEO, -1, -1, 0, 0);
 352         auto audio_index = av_find_best_stream(&context, AVMEDIA_TYPE_AUDIO, -1, -1, 0, 0);
 353
 354         if(video_index > -1)
 355         {
 356                 const auto video_context = context.streams[video_index]->codec;
 357                 const auto video_stream  = context.streams[video_index];
 358
 359                 AVRational time_base = video_context->time_base;
 360
 361                 if(boost::filesystem::path(context.filename).extension().string() == ".flv")
 362                 {
 363                         try
 364                         {
 365                                 auto meta = read_flv_meta_info(context.filename);
 366                                 return boost::lexical_cast<double>(meta["framerate"]);
 367                         }
 368                         catch(...)
 369                         {
 370                                 return 0.0;
 371                         }
 372                 }
 373                 else
 374                 {
 375                         time_base.num *= video_context->ticks_per_frame;
 376
 377                         if(!is_sane_fps(time_base))
 378                         {
 379                                 time_base = fix_time_base(time_base);
 380
 381                                 if(!is_sane_fps(time_base) && audio_index > -1)
 382                                 {
 383                                         auto& audio_context = *context.streams[audio_index]->codec;
 384                                         auto& audio_stream  = *context.streams[audio_index];
 385
 386                                         double duration_sec = audio_stream.duration / static_cast<double>(audio_context.sample_rate);
 387
 388                                         time_base.num = static_cast<int>(duration_sec*100000.0);
 389                                         time_base.den = static_cast<int>(video_stream->nb_frames*100000);
 390                                 }
 391                         }
 392                 }
 393
 394                 double fps = static_cast<double>(time_base.den) / static_cast<double>(time_base.num);
 395
 396                 double closest_fps = 0.0;
 397                 for(int n = 0; n < core::video_format::count; ++n)
 398                 {
 399                         auto format = core::video_format_desc(core::video_format(n));
 400
 401                         double diff1 = std::abs(format.fps - fps);
 402                         double diff2 = std::abs(closest_fps - fps);
 403
 404                         if(diff1 < diff2)
 405                                 closest_fps = format.fps;
 406                 }
 407
 408                 return closest_fps;
 409         }
 410
 411         return fail_value;
 412 }
 413
 414 void fix_meta_data(AVFormatContext& context)
 415 {
 416         auto video_index = av_find_best_stream(&context, AVMEDIA_TYPE_VIDEO, -1, -1, 0, 0);
 417
 418         if(video_index > -1)
 419         {
 420                 auto video_stream   = context.streams[video_index];
 421                 auto video_context  = context.streams[video_index]->codec;
 422
 423                 if(boost::filesystem::path(context.filename).extension().string() == ".flv")
 424                 {
 425                         try
 426                         {
 427                                 auto meta = read_flv_meta_info(context.filename);
 428                                 double fps = boost::lexical_cast<double>(meta["framerate"]);
 429                                 video_stream->nb_frames = static_cast<int64_t>(boost::lexical_cast<double>(meta["duration"])*fps);
 430                         }
 431                         catch(...){}
 432                 }
 433                 else
 434                 {
 435                         auto stream_time = video_stream->time_base;
 436                         auto duration    = video_stream->duration;
 437                         auto codec_time  = video_context->time_base;
 438                         auto ticks               = video_context->ticks_per_frame;
 439
 440                         if(video_stream->nb_frames == 0)
 441                                 video_stream->nb_frames = (duration*stream_time.num*codec_time.den)/(stream_time.den*codec_time.num*ticks);
 442                 }
 443         }
 444 }
 445
 446 spl::shared_ptr<AVPacket> create_packet()
 447 {
 448         spl::shared_ptr<AVPacket> packet(new AVPacket(), [](AVPacket* p)
 449         {
 450                 av_free_packet(p);
 451                 delete p;
 452         });
 453
 454         av_init_packet(packet.get());
 455         return packet;
 456 }
 457
 458 spl::shared_ptr<AVFrame> create_frame()
 459 {
 460         spl::shared_ptr<AVFrame> frame(avcodec_alloc_frame(), av_free);
 461         avcodec_get_frame_defaults(frame.get());
 462         return frame;
 463 }
 464
 465 spl::shared_ptr<AVCodecContext> open_codec(AVFormatContext& context, enum AVMediaType type, int& index)
 466 {
 467         AVCodec* decoder;
 468         index = THROW_ON_ERROR2(av_find_best_stream(&context, type, -1, -1, &decoder, 0), "");
 469         //if(strcmp(decoder->name, "prores") == 0 && decoder->next && strcmp(decoder->next->name, "prores_lgpl") == 0)
 470         //      decoder = decoder->next;
 471
 472         THROW_ON_ERROR2(tbb_avcodec_open(context.streams[index]->codec, decoder), "");
 473         return spl::shared_ptr<AVCodecContext>(context.streams[index]->codec, tbb_avcodec_close);
 474 }
 475
 476 spl::shared_ptr<AVFormatContext> open_input(const std::wstring& filename)
 477 {
 478         AVFormatContext* weak_context = nullptr;
 479         THROW_ON_ERROR2(avformat_open_input(&weak_context, u8(filename).c_str(), nullptr, nullptr), filename);
 480         spl::shared_ptr<AVFormatContext> context(weak_context, av_close_input_file);
 481         THROW_ON_ERROR2(avformat_find_stream_info(weak_context, nullptr), filename);
 482         fix_meta_data(*context);
 483         return context;
 484 }
 485
 486 std::wstring print_mode(int width, int height, double fps, bool interlaced)
 487 {
 488         std::wostringstream fps_ss;
 489         fps_ss << std::fixed << std::setprecision(2) << (!interlaced ? fps : 2.0 * fps);
 490
 491         return boost::lexical_cast<std::wstring>(width) + L"x" + boost::lexical_cast<std::wstring>(height) + (!interlaced ? L"p" : L"i") + fps_ss.str();
 492 }
 493
 494 bool is_valid_file(const std::wstring filename)
 495 {
 496         static const std::vector<std::wstring> invalid_exts = boost::assign::list_of(L".png")(L".tga")(L".bmp")(L".jpg")(L".jpeg")(L".gif")(L".tiff")(L".tif")(L".jp2")(L".jpx")(L".j2k")(L".j2c")(L".swf")(L".ct");
 497         static std::vector<std::wstring>           valid_exts   = boost::assign::list_of(L".m2t")(L".mov")(L".mp4")(L".dv")(L".flv")(L".mpg")(L".wav")(L".mp3")(L".dnxhd")(L".h264")(L".prores");
 498
 499         auto ext = boost::to_lower_copy(boost::filesystem::path(filename).extension().wstring());
 500
 501         if(std::find(valid_exts.begin(), valid_exts.end(), ext) != valid_exts.end())
 502                 return true;
 503
 504         if(std::find(invalid_exts.begin(), invalid_exts.end(), ext) != invalid_exts.end())
 505                 return false;
 506
 507         auto u8filename = u8(filename);
 508
 509         int score = 0;
 510         AVProbeData pb = {};
 511         pb.filename = u8filename.c_str();
 512
 513         if(av_probe_input_format2(&pb, false, &score) != nullptr)
 514                 return true;
 515
 516         std::ifstream file(filename);
 517
 518         std::vector<unsigned char> buf;
 519         for(auto file_it = std::istreambuf_iterator<char>(file); file_it != std::istreambuf_iterator<char>() && buf.size() < 1024; ++file_it)
 520                 buf.push_back(*file_it);
 521
 522         if(buf.empty())
 523                 return nullptr;
 524
 525         pb.buf          = buf.data();
 526         pb.buf_size = static_cast<int>(buf.size());
 527
 528         return av_probe_input_format2(&pb, true, &score) != nullptr;
 529 }
 530
 531 std::wstring probe_stem(const std::wstring stem)
 532 {
 533         auto stem2 = boost::filesystem::path(stem);
 534         auto dir = stem2.parent_path();
 535         for(auto it = boost::filesystem::directory_iterator(dir); it != boost::filesystem::directory_iterator(); ++it)
 536         {
 537                 if(boost::iequals(it->path().stem().wstring(), stem2.filename().wstring()) && is_valid_file(it->path().wstring()))
 538                         return it->path().wstring();
 539         }
 540         return L"";
 541 }
 542 //
 543 //void av_dup_frame(AVFrame* frame)
 544 //{
 545 //      AVFrame* new_frame = avcodec_alloc_frame();
 546 //
 547 //
 548 //      const uint8_t *src_data[4] = {0};
 549 //      memcpy(const_cast<uint8_t**>(&src_data[0]), frame->data, 4);
 550 //      const int src_linesizes[4] = {0};
 551 //      memcpy(const_cast<int*>(&src_linesizes[0]), frame->linesize, 4);
 552 //
 553 //      av_image_alloc(new_frame->data, new_frame->linesize, new_frame->width, new_frame->height, frame->format, 16);
 554 //
 555 //      av_image_copy(new_frame->data, new_frame->linesize, src_data, src_linesizes, frame->format, new_frame->width, new_frame->height);
 556 //
 557 //      frame =
 558 //}
 559
 560 }}