diff --git a/CMakeLists.txt b/CMakeLists.txt index 47575e0ce5..e1af8b0b06 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -139,7 +139,7 @@ set(OPENMW_LIBS ${OENGINE_ALL}) set(OPENMW_LIBS_HEADER) # Sound setup -set(FFmpeg_FIND_COMPONENTS AVCODEC AVFORMAT AVUTIL SWSCALE) +set(FFmpeg_FIND_COMPONENTS AVCODEC AVFORMAT AVUTIL SWSCALE SWRESAMPLE) find_package(FFmpeg REQUIRED) set(SOUND_INPUT_INCLUDES ${FFMPEG_INCLUDE_DIRS}) set(SOUND_INPUT_LIBRARY ${FFMPEG_LIBRARIES} ${SWSCALE_LIBRARIES}) diff --git a/apps/openmw/mwrender/videoplayer.cpp b/apps/openmw/mwrender/videoplayer.cpp index 2d05f27705..00443fdaa9 100644 --- a/apps/openmw/mwrender/videoplayer.cpp +++ b/apps/openmw/mwrender/videoplayer.cpp @@ -32,10 +32,11 @@ extern "C" #include #include - // From libavformat version 55.0.100 and onward the declaration of av_gettime() is removed from libavformat/avformat.h and moved - // to libavutil/time.h + // From libavformat version 55.0.100 and onward the declaration of av_gettime() is + // removed from libavformat/avformat.h and moved to libavutil/time.h // https://github.com/FFmpeg/FFmpeg/commit/06a83505992d5f49846c18507a6c3eb8a47c650e - #if AV_VERSION_INT(55, 0, 100) <= AV_VERSION_INT(LIBAVFORMAT_VERSION_MAJOR, LIBAVFORMAT_VERSION_MINOR, LIBAVFORMAT_VERSION_MICRO) + #if AV_VERSION_INT(55, 0, 100) <= AV_VERSION_INT(LIBAVFORMAT_VERSION_MAJOR, \ + LIBAVFORMAT_VERSION_MINOR, LIBAVFORMAT_VERSION_MICRO) #include #endif @@ -46,30 +47,27 @@ extern "C" LIBAVUTIL_VERSION_MINOR, LIBAVUTIL_VERSION_MICRO) #include #endif -} -#ifdef _WIN32 - // Decide whether to play binkaudio. - #include - // libavcodec versions 54.10.100 (or maybe earlier) to 54.54.100 potentially crashes Windows 64bit. - // From version 54.56 or higher, there's no sound due to the encoding format changing from S16 to FLTP - // (see https://gitorious.org/ffmpeg/ffmpeg/commit/7bfd1766d1c18f07b0a2dd042418a874d49ea60d and - // http://git.videolan.org/?p=ffmpeg.git;a=commitdiff;h=3049d5b9b32845c86aa5588bb3352bdeb2edfdb2;hp=43c6b45a53a186a187f7266e4d6bd3c2620519f1), - // but does not crash (or at least no known crash). - #if (LIBAVCODEC_VERSION_MAJOR > 54) + // From version 54.56 binkaudio encoding format changed from S16 to FLTP. See: + // https://gitorious.org/ffmpeg/ffmpeg/commit/7bfd1766d1c18f07b0a2dd042418a874d49ea60d + // http://ffmpeg.zeranoe.com/forum/viewtopic.php?f=15&t=872 + #if AV_VERSION_INT(54, 56, 0) <= AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \ + LIBAVCODEC_VERSION_MINOR, LIBAVCODEC_VERSION_MICRO) + #include /* swr_init, swr_alloc, swr_convert, swr_free */ + #include /* av_opt_set_int, av_opt_set_sample_fmt */ #define FFMPEG_PLAY_BINKAUDIO - #else - #ifdef _WIN64 - #if ((LIBAVCODEC_VERSION_MAJOR == 54) && (LIBAVCODEC_VERSION_MINOR >= 55)) - #define FFMPEG_PLAY_BINKAUDIO - #endif - #else - #if ((LIBAVCODEC_VERSION_MAJOR == 54) && (LIBAVCODEC_VERSION_MINOR >= 10)) - #define FFMPEG_PLAY_BINKAUDIO - #endif + #elif defined(_WIN32) && defined(_WIN64) + // Versions up to 54.54.100 potentially crashes on Windows 64bit. + #if ((LIBAVCODEC_VERSION_MAJOR == 54) && (LIBAVCODEC_VERSION_MINOR >= 55)) + #define FFMPEG_PLAY_BINKAUDIO + #endif + #elif defined(_WIN32) + // 54.10.100 is a known working version on 32bit, but earlier ones may also work. + #if ((LIBAVCODEC_VERSION_MAJOR == 54) && (LIBAVCODEC_VERSION_MINOR >= 10)) + #define FFMPEG_PLAY_BINKAUDIO #endif #endif -#endif +} #define MAX_AUDIOQ_SIZE (5 * 16 * 1024) #define MAX_VIDEOQ_SIZE (5 * 256 * 1024) @@ -317,8 +315,12 @@ class MovieAudioDecoder : public MWSound::Sound_Decoder VideoState *mVideoState; AVStream *mAVStream; + SwrContext *mSwr; /* non-zero indicates FLTP format */ + int mSamplesAllChannels; + int mSampleSize; + AutoAVPacket mPacket; - AVFrame *mFrame; + AVFrame *mFrame; /* AVFrame is now defined in libavutil/frame.h (used to be libavcodec/avcodec.h) */ ssize_t mFramePos; ssize_t mFrameSize; @@ -420,7 +422,11 @@ public: MovieAudioDecoder(VideoState *is) : mVideoState(is) , mAVStream(*is->audio_st) +#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(55, 28, 1) , mFrame(avcodec_alloc_frame()) +#else + , mFrame(av_frame_alloc()) +#endif , mFramePos(0) , mFrameSize(0) , mAudioClock(0.0) @@ -429,10 +435,15 @@ public: /* Correct audio only if larger error than this */ , mAudioDiffThreshold(2.0 * 0.050/* 50 ms */) , mAudioDiffAvgCount(0) + , mSwr(0) + , mSamplesAllChannels(0) + , mSampleSize(0) { } virtual ~MovieAudioDecoder() { av_freep(&mFrame); + if(mSwr) + swr_free(&mSwr); } void getInfo(int *samplerate, MWSound::ChannelConfig *chans, MWSound::SampleType * type) @@ -443,6 +454,12 @@ public: *type = MWSound::SampleType_Int16; else if(mAVStream->codec->sample_fmt == AV_SAMPLE_FMT_FLT) *type = MWSound::SampleType_Float32; + else if(mAVStream->codec->sample_fmt == AV_SAMPLE_FMT_FLTP) + { + // resample to a known format for OpenAL_SoundStream + // TODO: allow different size sample format, e.g. Int16 + *type = MWSound::SampleType_Float32; + } else fail(std::string("Unsupported sample format: ")+ av_get_sample_fmt_name(mAVStream->codec->sample_fmt)); @@ -480,17 +497,65 @@ public: } *samplerate = mAVStream->codec->sample_rate; + + // FIXME: error handling + if(mAVStream->codec->sample_fmt == AV_SAMPLE_FMT_FLTP) + { + mSwr = swr_alloc(); + av_opt_set_int(mSwr, "in_channel_layout", mAVStream->codec->channel_layout, 0); + av_opt_set_int(mSwr, "out_channel_layout", mAVStream->codec->channel_layout, 0); + av_opt_set_int(mSwr, "in_sample_rate", mAVStream->codec->sample_rate, 0); + av_opt_set_int(mSwr, "out_sample_rate", mAVStream->codec->sample_rate, 0); + av_opt_set_sample_fmt(mSwr, "in_sample_fmt", AV_SAMPLE_FMT_FLTP, 0); + av_opt_set_sample_fmt(mSwr, "out_sample_fmt", AV_SAMPLE_FMT_FLT, 0); // TODO: Try S16 + swr_init(mSwr); + } + + mSamplesAllChannels = av_get_bytes_per_sample(mAVStream->codec->sample_fmt) * + mAVStream->codec->channels; + mSampleSize = av_get_bytes_per_sample(mAVStream->codec->sample_fmt); } + /* + * stream is a ptr to vector on the stack, see OpenAL_SoundStream::process in + * mwsound/openal_output.cpp (around line 481) + * + * len is the size of the output buffer (i.e. stream) based on the number of + * channels, rate and sample type (as reported by getInfo). + * + * sample_skip is the number of bytes to skip (from all channels) or repeat (i.e. negative) + * + * mFrameSize is the number of bytes decoded audio frame (from all channels), or -1 if finished + * + * + * +---------------------------------------------------------------------------------+ + * | | + * |<------------------------------------------ len -------------------------------->| + * | | + * |<------ mFrameSize ------>| | + * | | + * +---------------------------------------------------------------------------------+ + * ^ + * | + * mFramePos >= 0 + * + * |<----- len1 -------->| + * + */ size_t read(char *stream, size_t len) { int sample_skip = synchronize_audio(); size_t total = 0; + float *outputStream = (float *)&stream[0]; + //uint16_t *intStream = (uint16_t *)&stream[0]; + while(total < len) { if(mFramePos >= mFrameSize) { + // for FLT sample format mFrameSize returned by audio_decode_frame is: + // 1920 samples x 4 bytes/sample x 2 channels = 15360 bytes /* We have already sent all our data; get more */ mFrameSize = audio_decode_frame(mFrame); if(mFrameSize < 0) @@ -508,9 +573,56 @@ public: if(mFramePos >= 0) { len1 = std::min(len1, mFrameSize-mFramePos); - memcpy(stream, mFrame->data[0]+mFramePos, len1); + + if(mSwr) + { + // Convert from AV_SAMPLE_FMT_FLTP to AV_SAMPLE_FMT_FLT + // FIXME: support channel formats other than stereo + // FIXME: support sample formats other than Float32 + float* inputChannel0 = (float*)mFrame->extended_data[0]; + float* inputChannel1 = (float*)mFrame->extended_data[1]; +#if 0 + uint16_t* inputChannel0 = (uint16_t*)mFrame->extended_data[0]; + uint16_t* inputChannel1 = (uint16_t*)mFrame->extended_data[1]; +#endif + inputChannel0 += mFramePos/mSamplesAllChannels; + inputChannel1 += mFramePos/mSamplesAllChannels; + //if(mFramePos > 0) + //std::cout << "mFramePos (bytes): " + std::to_string(mFramePos) + //<< " samples: " + std::to_string(mFramePos/mSamplesAllChannels) << std::endl; + + // samples per channel = len1 bytes / bytes per sample / number of channels + unsigned int len1Samples = len1 / mSamplesAllChannels; + // stream offset = total bytes / bytes per sample + unsigned int totalOffset = total / mSampleSize; + float sample0 = 0; + float sample1 = 0; + for (unsigned int i = 0 ; i < len1Samples ; ++i) + { + sample0 = *inputChannel0++; + sample1 = *inputChannel1++; +#if 0 + if(sample0<-1.0f) + sample0=-1.0f; + else if(sample0>1.0f) + sample0=1.0f; + if(sample1<-1.0f) + sample1=-1.0f; + else if(sample1>1.0f) + sample1=1.0f; + intStream[totalOffset+i*2] = (uint16_t) (sample0 * 32767.0f); + intStream[totalOffset+i*2+1] = (uint16_t) (sample1 * 32767.0f); +#endif + outputStream[totalOffset+i*2] = sample0; + outputStream[totalOffset+i*2+1] = sample1; + } + } + else + { + memcpy(stream, mFrame->data[0]+mFramePos, len1); + } } - else + else // repeat some samples FIXME: support ftlp { len1 = std::min(len1, -mFramePos); @@ -568,7 +680,7 @@ int VideoState::OgreResource_Read(void *user_data, uint8_t *buf, int buf_size) return stream->read(buf, buf_size); } -int VideoState::OgreResource_Write(void *user_data, uint8_t *buf, int buf_size) + int VideoState::OgreResource_Write(void *user_data, uint8_t *buf, int buf_size) { Ogre::DataStreamPtr stream = static_cast(user_data)->stream; return stream->write(buf, buf_size); @@ -768,9 +880,15 @@ void VideoState::video_thread_loop(VideoState *self) AVFrame *pFrame; double pts; +#if LIBAVCODEC_VERSION_INT < AV_VERSION_INT(55, 28, 1) pFrame = avcodec_alloc_frame(); self->rgbaFrame = avcodec_alloc_frame(); +#else + pFrame = av_frame_alloc(); + + self->rgbaFrame = av_frame_alloc(); +#endif avpicture_alloc((AVPicture*)self->rgbaFrame, PIX_FMT_RGBA, (*self->video_st)->codec->width, (*self->video_st)->codec->height); while(self->videoq.get(packet, self) >= 0) diff --git a/cmake/FindFFmpeg.cmake b/cmake/FindFFmpeg.cmake index a3509597b2..d7206c022c 100644 --- a/cmake/FindFFmpeg.cmake +++ b/cmake/FindFFmpeg.cmake @@ -14,6 +14,7 @@ # - AVUTIL # - POSTPROCESS # - SWSCALE +# - SWRESAMPLE # the following variables will be defined # _FOUND - System has # _INCLUDE_DIRS - Include directory necessary for using the headers @@ -32,7 +33,7 @@ include(FindPackageHandleStandardArgs) # The default components were taken from a survey over other FindFFMPEG.cmake files if (NOT FFmpeg_FIND_COMPONENTS) - set(FFmpeg_FIND_COMPONENTS AVCODEC AVFORMAT AVUTIL SWSCALE) + set(FFmpeg_FIND_COMPONENTS AVCODEC AVFORMAT AVUTIL SWSCALE SWRESAMPLE) endif () # @@ -112,6 +113,7 @@ if (NOT FFMPEG_LIBRARIES) find_component(AVUTIL libavutil avutil libavutil/avutil.h) find_component(SWSCALE libswscale swscale libswscale/swscale.h) find_component(POSTPROC libpostproc postproc libpostproc/postprocess.h) + find_component(SWRESAMPLE libswresample swresample libswresample/swresample.h) # Check if the required components were found and add their stuff to the FFMPEG_* vars. foreach (_component ${FFmpeg_FIND_COMPONENTS}) @@ -142,7 +144,7 @@ if (NOT FFMPEG_LIBRARIES) endif () # Now set the noncached _FOUND vars for the components. -foreach (_component AVCODEC AVDEVICE AVFORMAT AVUTIL POSTPROCESS SWSCALE) +foreach (_component AVCODEC AVDEVICE AVFORMAT AVUTIL POSTPROCESS SWSCALE SWRESAMPLE) set_component_found(${_component}) endforeach ()