Age | Commit message (Collapse) | Author | |
---|---|---|---|
2017-05-09 | cuvid: replace printf with proper log out | Hendrik Leppkes | |
2017-05-09 | cuvid: output > 8 bit content using P010 | Hendrik Leppkes | |
2017-05-09 | cuvid: support vp9 decoding | Hendrik Leppkes | |
2017-05-09 | cuvid: enable support for hevc main10 decoding | Hendrik Leppkes | |
2017-05-09 | Parse more bitstream properties from the HEVC SPS | Hendrik Leppkes | |
2017-05-08 | cuvid: pass bitdepth to the decoder | Hendrik Leppkes | |
2017-05-08 | cuvid: properly fail decoding if creating a decoder fails | Hendrik Leppkes | |
2017-05-08 | Update NVIDIA NVDEC headers to SDK 8.0 | Hendrik Leppkes | |
2017-05-03 | Update ffmpeg for a flv workaround | Hendrik Leppkes | |
2017-04-28 | Add AVdh mediatype for DNxHR streams | Hendrik Leppkes | |
2017-04-25 | Update build script for recent ffmpeg changes | Hendrik Leppkes | |
2017-04-24 | Support container content light level data | Hendrik Leppkes | |
2017-04-24 | Fix a copy-pasta problem | Hendrik Leppkes | |
2017-04-24 | Export HDR Content Light Level metadata for madVR | Hendrik Leppkes | |
2017-04-23 | avcodec: avoid an extra decoding loop in a double EAGAIN error condition | Hendrik Leppkes | |
2017-04-22 | avcodec: simplification and doc update | Hendrik Leppkes | |
2017-04-22 | avcodec: fully handle EAGAIN returns from avcodec_send_packet | Hendrik Leppkes | |
In theory this code should never be needed, but best to be fully API compliant. Who knows what decoder changes in the future might turn up. | |||
2017-04-22 | Use the native compiler environment, if available. | Hendrik Leppkes | |
This ensures the 64-bit compiler binaries are used when running on a 64-bit system. | |||
2017-04-22 | avcodec: use refcounted packets if we're using a dynamic input allocator | Hendrik Leppkes | |
This avoids copying the input packets into new ref-counted memory. | |||
2017-04-22 | Allow the PacketAllocator to allocate new packets on the fly | Hendrik Leppkes | |
This allows downstream filters to hang on to them for longer without deadlocking the allocator. | |||
2017-04-22 | Set the version in the vpcC header properly | Hendrik Leppkes | |
2017-04-21 | avcodec: always parse mpeg1/2 | Hendrik Leppkes | |
2017-04-21 | avcodec: refactor packet decoding, separate parsing out | Hendrik Leppkes | |
In preparation of future optimizations, properly separate parsing from decoding and switch to the new decoding API. | |||
2017-04-21 | avcodec: don't run the parser if the data is from LAV Splitter | Hendrik Leppkes | |
LAV Splitter already does the same parsing process (and if it doesn't for some file, thats a bug and should be fixed), so avoid doing double the work for no gains. | |||
2017-04-21 | Update changelog for recent changes | Hendrik Leppkes | |
2017-04-21 | Update ffmpeg | Hendrik Leppkes | |
2017-04-21 | Update copyright year to 2017 | Hendrik Leppkes | |
2017-04-21 | Release all DXVA2 resources when DXVA2-Native decoding fails | Hendrik Leppkes | |
2017-04-21 | Simplify the main decoder management | Hendrik Leppkes | |
Remove the dedicated thread the decoder worked on, it only added complexity without tangible performance benefits (could even add overhead). Instead, re-cycle the remnants of the DecodeThread into a much simpler DecodeManager that encapsulates the actual decoder objects. | |||
2017-04-18 | Generate VP8/9 extradata with profile and color information | Hendrik Leppkes | |
This information is modeled after the latest draft of the VP Codec in ISO Media specification (ie. VP8/9 in MP4). | |||
2017-04-15 | cuvid: skip D3D9 init on Windows 10 | Hendrik Leppkes | |
D3D9 CUVID interop is not supported on Windows 10 and will always fail, thus skipping hardware init avoids any overhead or complications. | |||
2017-04-15 | Add helper function to check for Windows 10 | Hendrik Leppkes | |
2017-04-15 | cuvid: set deinterlace mode to weave (disabled) on progressive streams | Hendrik Leppkes | |
In theory the GPU should avoid deinterlacing progressive streams automatically, but it appears at least some device/driver combinations do still attempt to deinterlace progressive streams in some situations, resulting in a degraded image. Based on the report and patch by Lutz Wolter | |||
2017-04-11 | Use unsigned types where appropriate in the generic conversion functions | Hendrik Leppkes | |
2017-04-11 | Fix Y416 conversion to be in the right format. | Hendrik Leppkes | |
Note: this change breaks compatibility with software that expects the old layout (ie. older madVR versions). madVR 0.91.8 or newer use the correct layout provided by this change. | |||
2017-04-04 | Enable reconnect for http streaming | Hendrik Leppkes | |
2017-04-03 | Update ffmpeg | Hendrik Leppkes | |
2017-03-31 | Unblock the network access when closing the demuxer | Hendrik Leppkes | |
This allows protocols like RTSP to send cleanup commands. | |||
2017-03-30 | Update ffmpeg for LATM AAC PCE fix | Hendrik Leppkes | |
2017-03-20 | gpu_memcpy: fix the name of a variable | Hendrik Leppkes | |
2017-03-20 | gpu_memcpy: add a memory barrier to avoid compiler re-ordering | Hendrik Leppkes | |
2017-03-20 | Move DTSinPCM timestamp handling to a better place | Hendrik Leppkes | |
2017-03-20 | Don't use H264 strict decoding with hwaccels | Hendrik Leppkes | |
Use of strict mode can artificially inflate the number of requires surfaces substantially for very little real world benefits. | |||
2017-03-20 | Disable software error concealment with hwaccel decoding | Hendrik Leppkes | |
The hardware performs error concealment, the software cannot access the hardware surfaces. | |||
2017-03-20 | Properly handle timestamps on very short PCM streams | Hendrik Leppkes | |
2017-03-11 | Signal DTS-only HEVC in MP4 when no CTTS atom is present | Hendrik Leppkes | |
2017-03-07 | Support VS2017 | Hendrik Leppkes | |
2017-03-07 | Update ffmpeg | Hendrik Leppkes | |
2017-03-07 | msdk_mvc: Unescape SEI NALs before parsing them | Hendrik Leppkes | |
2017-03-07 | Update libbluray | Hendrik Leppkes | |