Merge pull request #11014 from gnattu/vf-videotoolbox

MediaBrowser.Controller/MediaEncoding/EncodingHelper.cs

@@ -253,6 +253,14 @@ namespace MediaBrowser.Controller.MediaEncoding
                    && _mediaEncoder.SupportsFilterWithOption(FilterOptionType.OverlayVulkanFrameSync);
         }
 
+        private bool IsVideoToolboxFullSupported()
+        {
+            return _mediaEncoder.SupportsHwaccel("videotoolbox")
+                && _mediaEncoder.SupportsFilter("yadif_videotoolbox")
+                && _mediaEncoder.SupportsFilter("overlay_videotoolbox")
+                && _mediaEncoder.SupportsFilter("scale_vt");
+        }
+
         private bool IsHwTonemapAvailable(EncodingJobInfo state, EncodingOptions options)
         {
             if (state.VideoStream is null
@@ -272,7 +280,8 @@ namespace MediaBrowser.Controller.MediaEncoding
                 var isNvdecDecoder = vidDecoder.Contains("cuda", StringComparison.OrdinalIgnoreCase);
                 var isVaapiDecoder = vidDecoder.Contains("vaapi", StringComparison.OrdinalIgnoreCase);
                 var isD3d11vaDecoder = vidDecoder.Contains("d3d11va", StringComparison.OrdinalIgnoreCase);
-                return isSwDecoder || isNvdecDecoder || isVaapiDecoder || isD3d11vaDecoder;
+                var isVideoToolBoxDecoder = vidDecoder.Contains("videotoolbox", StringComparison.OrdinalIgnoreCase);
+                return isSwDecoder || isNvdecDecoder || isVaapiDecoder || isD3d11vaDecoder || isVideoToolBoxDecoder;
             }
 
             return state.VideoStream.VideoRange == VideoRange.HDR
@@ -308,6 +317,21 @@ namespace MediaBrowser.Controller.MediaEncoding
                    && state.VideoStream.VideoRangeType == VideoRangeType.HDR10;
         }
 
+        private bool IsVideoToolboxTonemapAvailable(EncodingJobInfo state, EncodingOptions options)
+        {
+            if (state.VideoStream is null
+                || !options.EnableVideoToolboxTonemapping
+                || GetVideoColorBitDepth(state) != 10)
+            {
+                return false;
+            }
+
+            // Certain DV profile 5 video works in Safari with direct playing, but the VideoToolBox does not produce correct mapping results with transcoding.
+            // All other HDR formats working.
+            return state.VideoStream.VideoRange == VideoRange.HDR
+                   && state.VideoStream.VideoRangeType is VideoRangeType.HDR10 or VideoRangeType.HLG or VideoRangeType.HDR10Plus;
+        }
+
         /// <summary>
         /// Gets the name of the output video codec.
         /// </summary>
@@ -4954,22 +4978,30 @@ namespace MediaBrowser.Controller.MediaEncoding
                 return (null, null, null);
             }
 
-            var swFilterChain = GetSwVidFilterChain(state, options, vidEncoder);
+            var isMacOS = OperatingSystem.IsMacOS();
+            var vidDecoder = GetHardwareVideoDecoder(state, options) ?? string.Empty;
+            var isVtEncoder = vidEncoder.Contains("videotoolbox", StringComparison.OrdinalIgnoreCase);
+            var isVtFullSupported = isMacOS && IsVideoToolboxFullSupported();
+            var isVtOclSupported = isVtFullSupported && IsOpenclFullSupported();
 
-            if (!options.EnableHardwareEncoding)
+            // legacy videotoolbox pipeline (disable hw filters)
+            if (!isVtEncoder
+                || !isVtOclSupported
+                || !_mediaEncoder.SupportsFilter("alphasrc"))
             {
-                return swFilterChain;
+                return GetSwVidFilterChain(state, options, vidEncoder);
             }
 
-            if (_mediaEncoder.EncoderVersion.CompareTo(new Version("5.0.0")) < 0)
-            {
-                // All features used here requires ffmpeg 5.0 or later, fallback to software filters if using an old ffmpeg
-                return swFilterChain;
-            }
+            // preferred videotoolbox + vt/ocl filters pipeline
+            return GetAppleVidFiltersPreferred(state, options, vidDecoder, vidEncoder);
+        }
 
-            var doDeintH264 = state.DeInterlace("h264", true) || state.DeInterlace("avc", true);
-            var doDeintHevc = state.DeInterlace("h265", true) || state.DeInterlace("hevc", true);
-            var doDeintH2645 = doDeintH264 || doDeintHevc;
+        public (List<string> MainFilters, List<string> SubFilters, List<string> OverlayFilters) GetAppleVidFiltersPreferred(
+            EncodingJobInfo state,
+            EncodingOptions options,
+            string vidDecoder,
+            string vidEncoder)
+        {
             var inW = state.VideoStream?.Width;
             var inH = state.VideoStream?.Height;
             var reqW = state.BaseRequest.Width;
@@ -4977,33 +5009,114 @@ namespace MediaBrowser.Controller.MediaEncoding
             var reqMaxW = state.BaseRequest.MaxWidth;
             var reqMaxH = state.BaseRequest.MaxHeight;
             var threeDFormat = state.MediaSource.Video3DFormat;
-            var newfilters = new List<string>();
-            var noOverlay = swFilterChain.OverlayFilters.Count == 0;
-            var supportsHwDeint = _mediaEncoder.SupportsFilter("yadif_videotoolbox");
-            // fallback to software filters if we are using filters not supported by hardware yet.
-            var useHardwareFilters = noOverlay && (!doDeintH2645 || supportsHwDeint);
 
-            if (!useHardwareFilters)
+            var isVtEncoder = vidEncoder.Contains("videotoolbox", StringComparison.OrdinalIgnoreCase);
+
+            var doDeintH264 = state.DeInterlace("h264", true) || state.DeInterlace("avc", true);
+            var doDeintHevc = state.DeInterlace("h265", true) || state.DeInterlace("hevc", true);
+            var doDeintH2645 = doDeintH264 || doDeintHevc;
+            var doVtTonemap = IsVideoToolboxTonemapAvailable(state, options);
+            var doOclTonemap = !doVtTonemap && IsHwTonemapAvailable(state, options);
+
+            var scaleFormat = string.Empty;
+            if (!string.Equals(state.VideoStream.PixelFormat, "yuv420p", StringComparison.OrdinalIgnoreCase))
             {
-                return swFilterChain;
+                // Use P010 for OpenCL tone mapping, otherwise force an 8bit output.
+                scaleFormat = doOclTonemap ? "p010le" : "nv12";
             }
 
-            // ffmpeg cannot use videotoolbox to scale
-            var swScaleFilter = GetSwScaleFilter(state, options, vidEncoder, inW, inH, threeDFormat, reqW, reqH, reqMaxW, reqMaxH);
-            newfilters.Add(swScaleFilter);
+            var hwScaleFilter = GetHwScaleFilter("vt", scaleFormat, inW, inH, reqW, reqH, reqMaxW, reqMaxH);
 
-            // hwupload on videotoolbox encoders can automatically convert AVFrame into its CVPixelBuffer equivalent
-            // videotoolbox will automatically convert the CVPixelBuffer to a pixel format the encoder supports, so we don't have to set a pixel format explicitly here
-            // This will reduce CPU usage significantly on UHD videos with 10 bit colors because we bypassed the ffmpeg pixel format conversion
-            newfilters.Add("hwupload");
+            var hasSubs = state.SubtitleStream is not null && state.SubtitleDeliveryMethod == SubtitleDeliveryMethod.Encode;
+            var hasTextSubs = hasSubs && state.SubtitleStream.IsTextSubtitleStream;
+            var hasGraphicalSubs = hasSubs && !state.SubtitleStream.IsTextSubtitleStream;
+            var hasAssSubs = hasSubs
+                && (string.Equals(state.SubtitleStream.Codec, "ass", StringComparison.OrdinalIgnoreCase)
+                    || string.Equals(state.SubtitleStream.Codec, "ssa", StringComparison.OrdinalIgnoreCase));
+
+            if (!isVtEncoder)
+            {
+                // should not happen.
+                return (null, null, null);
+            }
+
+            /* Make main filters for video stream */
+            var mainFilters = new List<string>();
 
+            // Color override is only required for OpenCL where hardware surface is in use
+            if (doOclTonemap)
+            {
+                mainFilters.Add(GetOverwriteColorPropertiesParam(state, doOclTonemap));
+            }
+
+            // INPUT videotoolbox/memory surface(vram/uma)
+            // this will pass-through automatically if in/out format matches.
+            mainFilters.Add("format=nv12|p010le|videotoolbox_vld");
+            mainFilters.Add("hwupload=derive_device=videotoolbox");
+
+            // hw deint
             if (doDeintH2645)
             {
                 var deintFilter = GetHwDeinterlaceFilter(state, options, "videotoolbox");
-                newfilters.Add(deintFilter);
+                mainFilters.Add(deintFilter);
             }
 
-            return (newfilters, swFilterChain.SubFilters, swFilterChain.OverlayFilters);
+            if (doVtTonemap)
+            {
+                const string VtTonemapArgs = "color_matrix=bt709:color_primaries=bt709:color_transfer=bt709";
+
+                // scale_vt can handle scaling & tonemapping in one shot, just like vpp_qsv.
+                hwScaleFilter = string.IsNullOrEmpty(hwScaleFilter)
+                    ? "scale_vt=" + VtTonemapArgs
+                    : hwScaleFilter + ":" + VtTonemapArgs;
+            }
+
+            // hw scale & vt tonemap
+            mainFilters.Add(hwScaleFilter);
+
+            // ocl tonemap
+            if (doOclTonemap)
+            {
+                // map from videotoolbox to opencl via videotoolbox-opencl interop.
+                mainFilters.Add("hwmap=derive_device=opencl:mode=read");
+
+                var tonemapFilter = GetHwTonemapFilter(options, "opencl", "nv12");
+                mainFilters.Add(tonemapFilter);
+
+                // OUTPUT videotoolbox(nv12) surface(vram/uma)
+                // reverse-mapping via videotoolbox-opencl interop.
+                mainFilters.Add("hwmap=derive_device=videotoolbox:mode=write:reverse=1");
+            }
+
+            /* Make sub and overlay filters for subtitle stream */
+            var subFilters = new List<string>();
+            var overlayFilters = new List<string>();
+
+            if (hasSubs)
+            {
+                if (hasGraphicalSubs)
+                {
+                    var subPreProcFilters = GetGraphicalSubPreProcessFilters(inW, inH, reqW, reqH, reqMaxW, reqMaxH);
+                    subFilters.Add(subPreProcFilters);
+                    subFilters.Add("format=bgra");
+                }
+                else if (hasTextSubs)
+                {
+                    var framerate = state.VideoStream?.RealFrameRate;
+                    var subFramerate = hasAssSubs ? Math.Min(framerate ?? 25, 60) : 10;
+
+                    var alphaSrcFilter = GetAlphaSrcFilter(state, inW, inH, reqW, reqH, reqMaxW, reqMaxH, subFramerate);
+                    var subTextSubtitlesFilter = GetTextSubtitlesFilter(state, true, true);
+                    subFilters.Add(alphaSrcFilter);
+                    subFilters.Add("format=bgra");
+                    subFilters.Add(subTextSubtitlesFilter);
+                }
+
+                subFilters.Add("hwupload=derive_device=videotoolbox");
+                overlayFilters.Add("overlay_videotoolbox=eof_action=pass:repeatlast=0");
+            }
+
+            return (mainFilters, subFilters, overlayFilters);
         }
 
         /// <summary>
@@ -5995,22 +6108,37 @@ namespace MediaBrowser.Controller.MediaEncoding
                                     || string.Equals("yuvj420p", videoStream.PixelFormat, StringComparison.OrdinalIgnoreCase);
             var is8_10bitSwFormatsVt = is8bitSwFormatsVt || string.Equals("yuv420p10le", videoStream.PixelFormat, StringComparison.OrdinalIgnoreCase);
 
+            // Hardware surface only make sense when interop with OpenCL
+            // VideoToolbox's Hardware surface in ffmpeg is not only slower than hwupload, but also breaks HDR in many cases.
+            // For example: https://trac.ffmpeg.org/ticket/10884
+            var useOclToneMapping = !IsVideoToolboxTonemapAvailable(state, options)
+                                    && options.EnableTonemapping
+                                    && state.VideoStream is not null
+                                    && GetVideoColorBitDepth(state) == 10
+                                    && state.VideoStream.VideoRange == VideoRange.HDR
+                                    && (state.VideoStream.VideoRangeType == VideoRangeType.HDR10
+                                        || state.VideoStream.VideoRangeType == VideoRangeType.HLG
+                                        || (state.VideoStream.VideoRangeType == VideoRangeType.DOVI
+                                            && string.Equals(state.VideoStream.Codec, "hevc", StringComparison.OrdinalIgnoreCase)));
+
+            var useHwSurface = useOclToneMapping && IsVideoToolboxFullSupported() && _mediaEncoder.SupportsFilter("alphasrc");
+
             if (is8bitSwFormatsVt)
             {
                 if (string.Equals("avc", videoStream.Codec, StringComparison.OrdinalIgnoreCase)
                     || string.Equals("h264", videoStream.Codec, StringComparison.OrdinalIgnoreCase))
                 {
-                    return GetHwaccelType(state, options, "h264", bitDepth, false);
+                    return GetHwaccelType(state, options, "h264", bitDepth, useHwSurface);
                 }
 
                 if (string.Equals("mpeg2video", videoStream.Codec, StringComparison.OrdinalIgnoreCase))
                 {
-                    return GetHwaccelType(state, options, "mpeg2video", bitDepth, false);
+                    return GetHwaccelType(state, options, "mpeg2video", bitDepth, useHwSurface);
                 }
 
                 if (string.Equals("mpeg4", videoStream.Codec, StringComparison.OrdinalIgnoreCase))
                 {
-                    return GetHwaccelType(state, options, "mpeg4", bitDepth, false);
+                    return GetHwaccelType(state, options, "mpeg4", bitDepth, useHwSurface);
                 }
             }
 
@@ -6019,12 +6147,12 @@ namespace MediaBrowser.Controller.MediaEncoding
                 if (string.Equals("hevc", videoStream.Codec, StringComparison.OrdinalIgnoreCase)
                     || string.Equals("h265", videoStream.Codec, StringComparison.OrdinalIgnoreCase))
                 {
-                    return GetHwaccelType(state, options, "hevc", bitDepth, false);
+                    return GetHwaccelType(state, options, "hevc", bitDepth, useHwSurface);
                 }
 
                 if (string.Equals("vp9", videoStream.Codec, StringComparison.OrdinalIgnoreCase))
                 {
-                    return GetHwaccelType(state, options, "vp9", bitDepth, false);
+                    return GetHwaccelType(state, options, "vp9", bitDepth, useHwSurface);
                 }
             }
 

MediaBrowser.MediaEncoding/Encoder/EncoderValidator.cs

@@ -128,6 +128,8 @@ namespace MediaBrowser.MediaEncoding.Encoder
             "overlay_vulkan",
             // videotoolbox
             "yadif_videotoolbox",
+            "scale_vt",
+            "overlay_videotoolbox",
             // rkrga
             "scale_rkrga",
             "vpp_rkrga",

MediaBrowser.Model/Configuration/EncodingOptions.cs

@@ -28,6 +28,7 @@ public class EncodingOptions
         VaapiDevice = "/dev/dri/renderD128";
         EnableTonemapping = false;
         EnableVppTonemapping = false;
+        EnableVideoToolboxTonemapping = false;
         TonemappingAlgorithm = "bt2390";
         TonemappingMode = "auto";
         TonemappingRange = "auto";
@@ -146,6 +147,11 @@ public class EncodingOptions
     /// </summary>
     public bool EnableVppTonemapping { get; set; }
 
+    /// <summary>
+    /// Gets or sets a value indicating whether videotoolbox tonemapping is enabled.
+    /// </summary>
+    public bool EnableVideoToolboxTonemapping { get; set; }
+
     /// <summary>
     /// Gets or sets the tone-mapping algorithm.
     /// </summary>