Release Version 0.6.1

• Improve D3D11 renderer, more pixel formats
• Player.snapshot() supports saving to file
• Support QSV decoder via DXVA2, and 0-copy rendering
• Subtitle WIP.
• Encoder WIP.
• Fix endless wait in loop mode
• Environment var name change: GL_MMAL_FORMAT=>MMAL_GL_FORMAT, GL_VAAPI=>VAAPI_GL, GL_VDPAU=>VDPAU_GL

Release Version 0.6.0

• D3D11 Renderer: now has almost the same features as OpenGL renderer
  • HDR tone mapping support
  • Support UWP. ICoreWindow or ISwapchainPanel can be used as the surface of Player.updateNativeSurface()
  • Zero copy rendering for hardware decoders, including d3d11, dxva2 and cuda based decoders.
  • Add env vars "SWAPCHAIN_BUFFERS", "DXGI_ADAPTER_RENDER" and "GPU_DEBUG" to control swap chain buffers adapter index and debug layer
  • Support feature level 9.1~12.1(default), controlled by env "D3D_FEATURE_LEVEL_RENDER"
  • Save and restore pipeline states before and after rendering
  • Support foreign context provided by user via Player.setRenderAPI(), D3D11RenderAPI.context and D3D11RenderAPI.rtv must be provided
  • Remove restriction of calling Player.setRenderAPI() before any other apis have vo_opaque parameter. updateNativeSurface() still have to be called after setRenderAPI()
  • Support more pixel formats, including nv12, p010 etc. Packed formats with 3 channels are not supported by d3d11 without cpu convertion, leaves them as unsupported.
• MFT: decoder device feature level can be set by property "feature_level"
• Reduce cpu load on pause
• Range loop: fix seek to A failure
• macOS:
  • Fix ui apis called on non-ui thread when creating CGL context(crash on macOS 10.15+)
  • Weak link to libffmpeg.4.dylib because macOS 10.15+ disallow dlopen. also can be replaced by user provided ffmpeg libraries(avutil, avcodec etc.) at runtime via SetGlobalOption("avutil", (void*)handle)
• FFmpeg:
  • Set ffmpeg module handle via SetGlobalOption("name", (void*)handle), where name is avutil, avcodec, avformat etc.
  • Fix crash if custom ffmpeg libraries are not complete. Required modules are "avutil", "avformat", "avcodec"
  • D3D11 decoder improve
  • Fix VAAPI seek crash cause by wrong ref count

release version 0.5.0

• ABI Changes:
  • PrepareCallback return type change in Player.prepare()
  • Player.snapshot() callback return type change. Return a file path to save as file(not implemented yet)
  • Add SeekFlag for prepare(), setNextMedia() and gaplessSwitch()
• APIs Changes:
  • Add RenderAPI.h: provides necessary informations to use a graphics api other than OpenGL
  • Add Player.setRenderAPI(): select platform graphics api for video rendering. Now OpenGL(default) and D3D11 are supported.
  • Add version macros
• New Features:
  • Experimental D3D11 rendering support. Now can render frames from software decoders. Color is not precise as OpenGL. Only the context created internally from a HWND via updateNativeSurface is tested. Foreign contexts may work, not tested.
• Improve seeking performance, fix unchanged display when seeking continuously. Now it's close to mpv, and should be better than most players like potplayer, vlc, mpc-hc etc.
• Fix seek position if media start time > 0
• Accurate seek is the default for the start position of prepare(), setNextMedia() and gaplessSwitch()
• OpenGL: Skip uploading video frame to gpu when redrawing
• setState(State::Stopped) now clears video renderer viewport and releases video renderer resources
• Fix stopping playback is not executed immediately
• setState(State::Stopped) now releases host and gpu resouces and clear background color. Normal playback end does not, and still keep the last frame displayed
• Fix setState() sometimes does not work
• MFT Decoder:
  • Disable low latency by default only for hevc. Previous release turned off low latency mode for all codecs, then h264 decoding may be too slow
  • Add property "ignore_profile" and "ignore_level"
• Add FFmpeg decoder property "drop" to control decoder frame dropping when seeking
• glfwplay example:
  • new seek options: -seek_step, -seek_any
  • support d3d11 rendering using -d3d11 -gfxthread

release version 0.4.1

• macOS: fix glx weak link
• support ubuntu >= 14.04
• load vaapi dynamically
• improve vaapi host map, control via env var VAIMAGE_DERIVE, VAIMAGE_FORMAT
• fix many playback loop bugs when more tracks are enabled
• fix too many packets buffered when decoding the 1st frame
• MFT: disable low latency to fix frames out of order
• subtitle support WIP.
  • decode embedded subtitle tracks via avcodec
  • decode loop and thread
• FFmpeg
  • versioned dso is preferred
  • reload ffmpeg symbols if dso is changed by SetGlobalOption()
  • log level is controlled by SetGlobalOption("ffmpeg.loglevel", val), val can be "quiet", "panic", "fatal", "error", "warning", "info", "verbose", "debug", "trace"

release version 0.4.0

• API:
  • add Player.setRange(a, b)
  • add Player.setLoop(count), onLoop(callback)
• Reduce pixel format storage
• Fix onEvent() can not add multiple callbacks
• Support windows on arm64
• VA-API: support map to host
• GL: support creating ES context from WGL
• avglue plugin is now merged into main library
• FFmpeg code is no longer embedded into mdk by default. External ffmpeg4.0+ is loaded automatically. Distributing mdk sdk is easier especially on linux

release version 0.3.1

This is a bug fix release

• fix user stop block
• fix some bugs at eof, e.g. unable to seek
• fix setLoop(0) can not disable looping
• support clang-cl 9.0
• c++ api: support g++ 4.9
• vaapi: sync surface after exporting drm handle
• improve ci. now nightly packages are uploaded automatically to https://sourceforge.net/projects/mdk-sdk/files/nightly
• now mdk can be installed via nuget in visual studio. https://www.nuget.org/packages/mdk

release version 0.3.0

• support A-B loop via setLoop()
• API:
  • fix player.mediaInfo() crash
  • add chapters description in MediaInfo
  • add setLoop()
• GL:
  • add hdr tone mapping. can be disabled by env var "GL_TONE_MAP=0"
  • can create wgl context with specified version and profile
  • improve glsl
• raspberry pi:
  • fix legacy driver check. gl and mmal works correctly on all drivers
  • load libwayland-egl.so.1 when needed
  • mmal zero copy rendering supports yuv textures. not perfect. not the default. enabled by "GL_MMAL_FORMAT=yuv420p"
• MFT(windows media foundation):
  • fix dolby decoder output format
  • input/output type index can be specified by property, example: decoder name is "MFT:in_type=0:out_type=0"
• add missing metadata in MediaInfo
• improve EOS handling in demux thread
• remove terminal null in log
• videotoolbox: stop decoding if decoded frame is invalid
• stop annexb conversion if error occurs
• fix video output drop rate is not desired value
• fix crash if cuda is not available when using CUDA decoder
• fix xaudio2 crash
• fix iOS 32bit min version

v0.2.3

• muxer: skip unwanted stream packets
• GL:
  • support gbrp formats
  • improve yuv to rgb
  • xyz gamma correction
• generic color space, trc, primaries support. done in renderer, demuxers and decoders(FFmpeg, VT, AMediaCodec, MFT)
• android AudioTrack: do not use deprecated 7.1 channel value, which may not work in new OSes
• FFmpeg:
  • fix sw format after seek
  • check hw pixfmt of decoded frame, fix sw fallback
  • support avcodec options via property, e.g. FFmpeg:thread=2
• fix annexb check for CUDA, MFT, AMediaCodec decoder
• try next decoder set by user if failed to decode a frame
• AMediaCodec decoder:
  • profile, level and capabilities check before using a codec
  • fix nv12 plane line size in copy mode
  • async mode support
  • detect decode error
• VT decoder(VideoToolbox): add property "copy"
• MFT decoder:
  • apply property "copy", "pool" on the fly
  • fix annexb check
• raspberry pi:
  • enable builtin mmal decoder. decoder name is "mmal", while FFmpeg's is "MMAL"