#include "../tbb_avcodec.h"
#include "../../ffmpeg_error.h"
+#include "../../ffmpeg.h"
#include <tbb/concurrent_unordered_map.h>
#include <tbb/concurrent_queue.h>
#pragma warning (push)
#pragma warning (disable : 4244)
#endif
-extern "C"
+extern "C"
{
#include <libswscale/swscale.h>
#include <libavcodec/avcodec.h>
#endif
namespace caspar { namespace ffmpeg {
-
+
core::field_mode get_mode(const AVFrame& frame)
{
if(!frame.interlaced_frame)
core::pixel_format_desc pixel_format_desc(PixelFormat pix_fmt, int width, int height)
{
// Get linesizes
- AVPicture dummy_pict;
+ AVPicture dummy_pict;
avpicture_fill(&dummy_pict, nullptr, pix_fmt, width, height);
core::pixel_format_desc desc = get_pixel_format(pix_fmt);
-
+
switch(desc.format)
{
case core::pixel_format::gray:
case core::pixel_format::luma:
{
- desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0], height, 1));
+ desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0], height, 1));
return desc;
}
case core::pixel_format::bgr:
case core::pixel_format::rgb:
{
- desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0]/3, height, 3));
+ desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0]/3, height, 3));
return desc;
}
case core::pixel_format::bgra:
case core::pixel_format::rgba:
case core::pixel_format::abgr:
{
- desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0]/4, height, 4));
+ desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0]/4, height, 4));
return desc;
}
case core::pixel_format::ycbcr:
case core::pixel_format::ycbcra:
- {
+ {
// Find chroma height
int size2 = static_cast<int>(dummy_pict.data[2] - dummy_pict.data[1]);
- int h2 = size2/dummy_pict.linesize[1];
+ int h2 = size2/dummy_pict.linesize[1];
desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[0], height, 1));
desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[1], h2, 1));
desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[2], h2, 1));
- if(desc.format == core::pixel_format::ycbcra)
- desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[3], height, 1));
+ if(desc.format == core::pixel_format::ycbcra)
+ desc.planes.push_back(core::pixel_format_desc::plane(dummy_pict.linesize[3], height, 1));
return desc;
- }
- default:
+ }
+ default:
desc.format = core::pixel_format::invalid;
return desc;
}
}
-core::mutable_frame make_frame(const void* tag, const spl::shared_ptr<AVFrame>& decoded_frame, double fps, core::frame_factory& frame_factory, const core::audio_channel_layout& channel_layout)
-{
+core::mutable_frame make_frame(const void* tag, const spl::shared_ptr<AVFrame>& decoded_frame, core::frame_factory& frame_factory, const core::audio_channel_layout& channel_layout)
+{
static tbb::concurrent_unordered_map<int64_t, tbb::concurrent_queue<std::shared_ptr<SwsContext>>> sws_contvalid_exts_;
-
+
if(decoded_frame->width < 1 || decoded_frame->height < 1)
return frame_factory.create_frame(tag, core::pixel_format_desc(core::pixel_format::invalid), core::audio_channel_layout::invalid());
const auto width = decoded_frame->width;
const auto height = decoded_frame->height;
auto desc = pixel_format_desc(static_cast<PixelFormat>(decoded_frame->format), width, height);
-
+
if(desc.format == core::pixel_format::invalid)
{
auto pix_fmt = static_cast<PixelFormat>(decoded_frame->format);
target_pix_fmt = PIX_FMT_YUV422P;
else if(pix_fmt == PIX_FMT_YUV444P10)
target_pix_fmt = PIX_FMT_YUV444P;
-
+
auto target_desc = pixel_format_desc(target_pix_fmt, width, height);
auto write = frame_factory.create_frame(tag, target_desc, channel_layout);
std::shared_ptr<SwsContext> sws_context;
//CASPAR_LOG(warning) << "Hardware accelerated color transform not supported.";
-
- int64_t key = ((static_cast<int64_t>(width) << 32) & 0xFFFF00000000) |
- ((static_cast<int64_t>(height) << 16) & 0xFFFF0000) |
- ((static_cast<int64_t>(pix_fmt) << 8) & 0xFF00) |
+
+ int64_t key = ((static_cast<int64_t>(width) << 32) & 0xFFFF00000000) |
+ ((static_cast<int64_t>(height) << 16) & 0xFFFF0000) |
+ ((static_cast<int64_t>(pix_fmt) << 8) & 0xFF00) |
((static_cast<int64_t>(target_pix_fmt) << 0) & 0xFF);
-
+
auto& pool = sws_contvalid_exts_[key];
-
+
if(!pool.try_pop(sws_context))
{
double param;
sws_context.reset(sws_getContext(width, height, pix_fmt, width, height, target_pix_fmt, SWS_BILINEAR, nullptr, nullptr, ¶m), sws_freeContext);
}
-
+
if(!sws_context)
{
- CASPAR_THROW_EXCEPTION(operation_failed() << msg_info("Could not create software scaling context.") <<
+ CASPAR_THROW_EXCEPTION(operation_failed() << msg_info("Could not create software scaling context.") <<
boost::errinfo_api_function("sws_getContext"));
- }
-
- spl::shared_ptr<AVFrame> av_frame(avcodec_alloc_frame(), av_free);
- avcodec_get_frame_defaults(av_frame.get());
+ }
+
+ auto av_frame = create_frame();
if(target_pix_fmt == PIX_FMT_BGRA)
{
auto size = avpicture_fill(reinterpret_cast<AVPicture*>(av_frame.get()), write.image_data(0).begin(), PIX_FMT_BGRA, width, height);
- CASPAR_VERIFY(size == write.image_data(0).size());
+ CASPAR_VERIFY(size == write.image_data(0).size());
}
else
{
}
}
- sws_scale(sws_context.get(), decoded_frame->data, decoded_frame->linesize, 0, height, av_frame->data, av_frame->linesize);
- pool.push(sws_context);
+ sws_scale(sws_context.get(), decoded_frame->data, decoded_frame->linesize, 0, height, av_frame->data, av_frame->linesize);
+ pool.push(sws_context);
return std::move(write);
}
else
{
auto write = frame_factory.create_frame(tag, desc, channel_layout);
-
+
for(int n = 0; n < static_cast<int>(desc.planes.size()); ++n)
{
auto plane = desc.planes[n];
auto result = write.image_data(n).begin();
auto decoded = decoded_frame->data[n];
auto decoded_linesize = decoded_frame->linesize[n];
-
+
CASPAR_ASSERT(decoded);
CASPAR_ASSERT(write.image_data(n).begin());
fast_memcpy(result, decoded, plane.size);
}
}
-
+
return std::move(write);
}
}
spl::shared_ptr<AVFrame> make_av_frame(std::array<uint8_t*, 4> data, const core::pixel_format_desc& pix_desc)
{
- spl::shared_ptr<AVFrame> av_frame(avcodec_alloc_frame(), av_free);
- avcodec_get_frame_defaults(av_frame.get());
-
+ auto av_frame = create_frame();
+
auto planes = pix_desc.planes;
auto format = pix_desc.format;
av_frame->width = planes[0].width;
av_frame->height = planes[0].height;
- for(int n = 0; n < planes.size(); ++n)
+ for(int n = 0; n < planes.size(); ++n)
{
av_frame->data[n] = data[n];
- av_frame->linesize[n] = planes[n].linesize;
+ av_frame->linesize[n] = planes[n].linesize;
}
switch(format)
av_frame->format = PIX_FMT_BGR24;
break;
case core::pixel_format::rgba:
- av_frame->format = PIX_FMT_RGBA;
+ av_frame->format = PIX_FMT_RGBA;
break;
case core::pixel_format::argb:
- av_frame->format = PIX_FMT_ARGB;
+ av_frame->format = PIX_FMT_ARGB;
break;
case core::pixel_format::bgra:
- av_frame->format = PIX_FMT_BGRA;
+ av_frame->format = PIX_FMT_BGRA;
break;
case core::pixel_format::abgr:
- av_frame->format = PIX_FMT_ABGR;
+ av_frame->format = PIX_FMT_ABGR;
break;
case core::pixel_format::gray:
- av_frame->format = PIX_FMT_GRAY8;
+ av_frame->format = PIX_FMT_GRAY8;
break;
case core::pixel_format::ycbcr:
{
AVRational fix_time_base(AVRational time_base)
{
if(time_base.num == 1)
- time_base.num = static_cast<int>(std::pow(10.0, static_cast<int>(std::log10(static_cast<float>(time_base.den)))-1));
-
+ time_base.num = static_cast<int>(std::pow(10.0, static_cast<int>(std::log10(static_cast<float>(time_base.den)))-1));
+
if(!is_sane_fps(time_base))
{
auto tmp = time_base;
}
double read_fps(AVFormatContext& context, double fail_value)
-{
+{
+ auto framerate = read_framerate(context, boost::rational<int>(static_cast<int>(fail_value * 1000000.0), 1000000));
+
+ return static_cast<double>(framerate.numerator()) / static_cast<double>(framerate.denominator());
+}
+
+boost::rational<int> read_framerate(AVFormatContext& context, const boost::rational<int>& fail_value)
+{
auto video_index = av_find_best_stream(&context, AVMEDIA_TYPE_VIDEO, -1, -1, 0, 0);
auto audio_index = av_find_best_stream(&context, AVMEDIA_TYPE_AUDIO, -1, -1, 0, 0);
-
- if(video_index > -1)
+
+ if (video_index > -1)
{
const auto video_context = context.streams[video_index]->codec;
- const auto video_stream = context.streams[video_index];
+ const auto video_stream = context.streams[video_index];
auto frame_rate_time_base = video_stream->avg_frame_rate;
std::swap(frame_rate_time_base.num, frame_rate_time_base.den);
if (is_sane_fps(frame_rate_time_base))
{
- return static_cast<double>(frame_rate_time_base.den) / static_cast<double>(frame_rate_time_base.num);
+ return boost::rational<int>(frame_rate_time_base.den, frame_rate_time_base.num);
}
AVRational time_base = video_context->time_base;
- if(boost::filesystem::path(context.filename).extension().string() == ".flv")
+ if (boost::filesystem::path(context.filename).extension().string() == ".flv")
{
try
{
auto meta = read_flv_meta_info(context.filename);
- return boost::lexical_cast<double>(meta["framerate"]);
+ return boost::rational<int>(static_cast<int>(boost::lexical_cast<double>(meta["framerate"]) * 1000000.0), 1000000);
}
- catch(...)
+ catch (...)
{
- return 0.0;
+ return fail_value;
}
}
else
{
time_base.num *= video_context->ticks_per_frame;
- if(!is_sane_fps(time_base))
- {
+ if (!is_sane_fps(time_base))
+ {
time_base = fix_time_base(time_base);
- if(!is_sane_fps(time_base) && audio_index > -1)
+ if (!is_sane_fps(time_base) && audio_index > -1)
{
auto& audio_context = *context.streams[audio_index]->codec;
- auto& audio_stream = *context.streams[audio_index];
+ auto& audio_stream = *context.streams[audio_index];
double duration_sec = audio_stream.duration / static_cast<double>(audio_context.sample_rate);
-
+
time_base.num = static_cast<int>(duration_sec*100000.0);
- time_base.den = static_cast<int>(video_stream->nb_frames*100000);
+ time_base.den = static_cast<int>(video_stream->nb_frames * 100000);
}
}
}
-
- double fps = static_cast<double>(time_base.den) / static_cast<double>(time_base.num);
- double closest_fps = 0.0;
+ boost::rational<int> fps(time_base.den, time_base.num);
+ boost::rational<int> closest_fps(0);
for (auto video_mode : enum_constants<core::video_format>())
{
auto format = core::video_format_desc(core::video_format(video_mode));
- double diff1 = std::abs(format.fps - fps);
- double diff2 = std::abs(closest_fps - fps);
+ auto diff1 = boost::abs(boost::rational<int>(format.time_scale, format.duration) - fps);
+ auto diff2 = boost::abs(closest_fps - fps);
- if(diff1 < diff2)
- closest_fps = format.fps;
+ if (diff1 < diff2)
+ closest_fps = boost::rational<int>(format.time_scale, format.duration);
}
-
+
return closest_fps;
}
- return fail_value;
+ return fail_value;
}
void fix_meta_data(AVFormatContext& context)
{
auto video_stream = context.streams[video_index];
auto video_context = context.streams[video_index]->codec;
-
+
if(boost::filesystem::path(context.filename).extension().string() == ".flv")
{
try
auto ticks = video_context->ticks_per_frame;
if(video_stream->nb_frames == 0)
- video_stream->nb_frames = (duration*stream_time.num*codec_time.den)/(stream_time.den*codec_time.num*ticks);
+ video_stream->nb_frames = (duration*stream_time.num*codec_time.den)/(stream_time.den*codec_time.num*ticks);
}
}
}
av_free_packet(p);
delete p;
});
-
+
av_init_packet(packet.get());
return packet;
}
spl::shared_ptr<AVFrame> create_frame()
-{
- spl::shared_ptr<AVFrame> frame(avcodec_alloc_frame(), av_free);
- avcodec_get_frame_defaults(frame.get());
+{
+ spl::shared_ptr<AVFrame> frame(av_frame_alloc(), [](AVFrame* p)
+ {
+ av_frame_free(&p);
+ });
return frame;
}
+std::shared_ptr<core::mutable_audio_buffer> flush_audio()
+{
+ static std::shared_ptr<core::mutable_audio_buffer> audio(new core::mutable_audio_buffer());
+ return audio;
+}
+
+std::shared_ptr<core::mutable_audio_buffer> empty_audio()
+{
+ static std::shared_ptr<core::mutable_audio_buffer> audio(new core::mutable_audio_buffer());
+ return audio;
+}
+
+std::shared_ptr<AVFrame> flush_video()
+{
+ static auto video = create_frame();
+ return video;
+}
+
+std::shared_ptr<AVFrame> empty_video()
+{
+ static auto video = create_frame();
+ return video;
+}
+
spl::shared_ptr<AVCodecContext> open_codec(AVFormatContext& context, enum AVMediaType type, int& index, bool single_threaded)
-{
+{
AVCodec* decoder;
- index = THROW_ON_ERROR2(av_find_best_stream(&context, type, -1, -1, &decoder, 0), "");
+ index = THROW_ON_ERROR2(av_find_best_stream(&context, type, index, -1, &decoder, 0), "");
//if(strcmp(decoder->name, "prores") == 0 && decoder->next && strcmp(decoder->next->name, "prores_lgpl") == 0)
// decoder = decoder->next;
}
bool is_valid_file(const std::wstring& filename, bool only_video)
-{
+{
static const auto invalid_exts = {
L".png",
L".tga",
};
auto ext = boost::to_lower_copy(boost::filesystem::path(filename).extension().wstring());
-
+
if(std::find(valid_exts.begin(), valid_exts.end(), ext) != valid_exts.end())
return true;
if (!only_video && std::find(only_audio.begin(), only_audio.end(), ext) != only_audio.end())
return true;
-
+
if(std::find(invalid_exts.begin(), invalid_exts.end(), ext) != invalid_exts.end())
- return false;
+ return false;
if (only_video && std::find(only_audio.begin(), only_audio.end(), ext) != only_audio.end())
return false;
auto u8filename = u8(filename);
-
+
int score = 0;
AVProbeData pb = {};
pb.filename = u8filename.c_str();
buf.push_back(*file_it);
if(buf.empty())
- return nullptr;
+ return false;
pb.buf = buf.data();
pb.buf_size = static_cast<int>(buf.size());
return L"";
}
-core::audio_channel_layout get_audio_channel_layout(const AVCodecContext& codec_context, const std::wstring& channel_layout_spec)
+core::audio_channel_layout get_audio_channel_layout(int num_channels, std::uint64_t layout, const std::wstring& channel_layout_spec)
{
- auto num_channels = codec_context.channels;
-
if (!channel_layout_spec.empty())
{
if (boost::contains(channel_layout_spec, L":")) // Custom on the fly layout specified.
}
else // Preconfigured named channel layout selected.
{
- auto layout = core::audio_channel_layout_repository::get_default()->get_layout(channel_layout_spec);
+ auto channel_layout = core::audio_channel_layout_repository::get_default()->get_layout(channel_layout_spec);
- if (!layout)
+ if (!channel_layout)
CASPAR_THROW_EXCEPTION(user_error() << msg_info(L"No channel layout with name " + channel_layout_spec + L" registered"));
- layout->num_channels = num_channels;
+ channel_layout->num_channels = num_channels;
- return *layout;
+ return *channel_layout;
}
}
- if (!codec_context.channel_layout)
+ if (!layout)
{
if (num_channels == 1)
return core::audio_channel_layout(num_channels, L"mono", L"FC");
return core::audio_channel_layout(num_channels, L"", L""); // Passthru without named channels as is.
}
- // What FFMpeg calls "channel layout" is only the "layout type" of a channel layout in
+ // What FFmpeg calls "channel layout" is only the "layout type" of a channel layout in
// CasparCG where the channel layout supports different orders as well.
// The user needs to provide additional mix-configs in casparcg.config to support more
// than the most common (5.1, mono and stereo) types.
// Based on information in https://ffmpeg.org/ffmpeg-utils.html#Channel-Layout
- switch (codec_context.channel_layout)
+ switch (layout)
{
case AV_CH_LAYOUT_MONO:
return core::audio_channel_layout(num_channels, L"mono", L"FC");
}
// av_get_default_channel_layout does not work for layouts not predefined in ffmpeg. This is needed to support > 8 channels.
-std::int64_t create_channel_layout_bitmask(int num_channels)
+std::uint64_t create_channel_layout_bitmask(int num_channels)
{
if (num_channels > 63)
- CASPAR_THROW_EXCEPTION(invalid_argument() << msg_info(L"FFMpeg cannot handle more than 63 audio channels"));
+ CASPAR_THROW_EXCEPTION(invalid_argument() << msg_info(L"FFmpeg cannot handle more than 63 audio channels"));
const auto ALL_63_CHANNELS = 0x7FFFFFFFFFFFFFFFULL;
auto to_shift = 63 - num_channels;
auto result = ALL_63_CHANNELS >> to_shift;
- return static_cast<std::int64_t>(result);
+ return static_cast<std::uint64_t>(result);
+}
+
+std::string to_string(const boost::rational<int>& framerate)
+{
+ return boost::lexical_cast<std::string>(framerate.numerator())
+ + "/" + boost::lexical_cast<std::string>(framerate.denominator()) + " (" + boost::lexical_cast<std::string>(static_cast<double>(framerate.numerator()) / static_cast<double>(framerate.denominator())) + ") fps";
+}
+
+std::vector<int> find_audio_cadence(const boost::rational<int>& framerate)
+{
+ static std::map<boost::rational<int>, std::vector<int>> CADENCES_BY_FRAMERATE = []
+ {
+ std::map<boost::rational<int>, std::vector<int>> result;
+
+ for (core::video_format format : enum_constants<core::video_format>())
+ {
+ core::video_format_desc desc(format);
+ boost::rational<int> format_rate(desc.time_scale, desc.duration);
+
+ result.insert(std::make_pair(format_rate, desc.audio_cadence));
+ }
+
+ return result;
+ }();
+
+ auto exact_match = CADENCES_BY_FRAMERATE.find(framerate);
+
+ if (exact_match != CADENCES_BY_FRAMERATE.end())
+ return exact_match->second;
+
+ boost::rational<int> closest_framerate_diff = std::numeric_limits<int>::max();
+ boost::rational<int> closest_framerate = 0;
+
+ for (auto format_framerate : CADENCES_BY_FRAMERATE | boost::adaptors::map_keys)
+ {
+ auto diff = boost::abs(framerate - format_framerate);
+
+ if (diff < closest_framerate_diff)
+ {
+ closest_framerate_diff = diff;
+ closest_framerate = format_framerate;
+ }
+ }
+
+ if (is_logging_quiet_for_thread())
+ CASPAR_LOG(debug) << "No exact audio cadence match found for framerate " << to_string(framerate)
+ << "\nClosest match is " << to_string(closest_framerate)
+ << "\nwhich is a " << to_string(closest_framerate_diff) << " difference.";
+ else
+ CASPAR_LOG(warning) << "No exact audio cadence match found for framerate " << to_string(framerate)
+ << "\nClosest match is " << to_string(closest_framerate)
+ << "\nwhich is a " << to_string(closest_framerate_diff) << " difference.";
+
+ return CADENCES_BY_FRAMERATE[closest_framerate];
}
//
projects / casparcg / blobdiff