Add vaapi hardware acceleration

INSTALLING.md

@@ -60,6 +60,28 @@ stop Kyoo's services. You can then remove the configuration files.
 
 # Hardware Acceleration
 
+## VA-API (intel, amd)
+
+First install necessary drivers on your system, when running `vainfo` you should have something like this:
+```
+libva info: VA-API version 1.20.0
+libva info: Trying to open /run/opengl-driver/lib/dri/iHD_drv_video.so
+libva info: Found init function __vaDriverInit_1_20
+libva info: va_openDriver() returns 0
+vainfo: VA-API version: 1.20 (libva 2.20.1)
+vainfo: Driver version: Intel iHD driver for Intel(R) Gen Graphics - 23.3.5 ()
+vainfo: Supported profile and entrypoints
+      VAProfileH264Main               :	VAEntrypointVLD
+      VAProfileH264Main               :	VAEntrypointEncSlice
+      ...Truncated...
+      VAProfileHEVCSccMain444_10      :	VAEntrypointVLD
+      VAProfileHEVCSccMain444_10      :	VAEntrypointEncSliceLP
+```
+Kyoo will default to use your primary card (located at `/dev/dri/renderD128`). If you need to specify a secondary one, you
+can use the `GOTRANSCODER_VAAPI_RENDERER` env-var to specify `/dev/dri/renderD129` or another one.
+
+Then you can simply run kyoo using `docker compose --profile vaapi up -d` (notice the `--profile vaapi` added)
+
 ## Nvidia
 
 To enable nvidia hardware acceleration, first install necessary drivers on your system.
@@ -70,7 +92,7 @@ follow the instructions on the official webpage or your distribution wiki.
 To test if everything works, you can run `sudo docker run --rm --gpus all ubuntu nvidia-smi`. If your version of docker is older,
 you might need to add `--runtime nvidia` like so: `sudo docker run --rm --runtime=nvidia --gpus all ubuntu nvidia-smi`
 
-After that, you can now use `docker compose --profile nvidia up -d` to start kyoo with nvidia hardware acceleration.
+After that, you can now use `docker compose --profile nvidia up -d` to start kyoo with nvidia hardware acceleration (notice the `--profile nvidia` added).
 
-Note that most nvidia cards have an artifical limit on the number of encodes. You can confirm your card limit [here](https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new).
+Note that most nvidia cards have an artificial limit on the number of encodes. You can confirm your card limit [here](https://developer.nvidia.com/video-encode-and-decode-gpu-support-matrix-new).
-This limit can also be remove by applying an [unofficial patch](https://github.com/keylase/nvidia-patch) to you driver.
+This limit can also be removed by applying an [unofficial patch](https://github.com/keylase/nvidia-patch) to you driver.

docker-compose.dev.yml

@@ -87,6 +87,15 @@ services:
       - GOTRANSCODER_HWACCEL=nvidia
     profiles: ['nvidia']
 
+  transcoder-vaapi:
+    <<: *transcoder-base
+    devices:
+      - /dev/dri:/dev/dri
+    environment:
+      - GOTRANSCODER_HWACCEL=vaapi
+      - GOTRANSCODER_VAAPI_RENDERER=${GOTRANSCODER_VAAPI_RENDERER:-/dev/dri/renderD128}
+    profiles: ['vaapi']
+
   ingress:
     image: nginx
     restart: on-failure

docker-compose.prod.yml

@@ -64,6 +64,15 @@ services:
       - GOTRANSCODER_HWACCEL=nvidia
     profiles: ['nvidia']
 
+  transcoder-vaapi:
+    <<: *transcoder-base
+    devices:
+      - /dev/dri:/dev/dri
+    environment:
+      - GOTRANSCODER_HWACCEL=vaapi
+      - GOTRANSCODER_VAAPI_RENDERER=${GOTRANSCODER_VAAPI_RENDERER:-/dev/dri/renderD128}
+    profiles: ['vaapi']
+
   ingress:
     image: nginx
     restart: unless-stopped

docker-compose.yml

@@ -63,6 +63,15 @@ services:
       - GOTRANSCODER_HWACCEL=nvidia
     profiles: ['nvidia']
 
+  transcoder-vaapi:
+    <<: *transcoder-base
+    devices:
+      - /dev/dri:/dev/dri
+    environment:
+      - GOTRANSCODER_HWACCEL=vaapi
+      - GOTRANSCODER_VAAPI_RENDERER=${GOTRANSCODER_VAAPI_RENDERER:-/dev/dri/renderD128}
+    profiles: ['vaapi']
+
   ingress:
     image: nginx
     restart: on-failure

transcoder/Dockerfile

@@ -33,8 +33,11 @@ RUN go build -o ./transcoder
 # we use trixie (debian's testing because ffmpeg on latest is v5 and we need v6)
 # https://packages.debian.org/bookworm/ffmpeg for version tracking
 FROM debian:trixie-slim
+RUN sed -i -e's/ main/ main contrib non-free/g' /etc/apt/sources.list.d/debian.sources
 RUN apt-get update \
-	&& apt-get install --no-install-recommends --no-install-suggests -y ffmpeg libmediainfo-dev \
+	&& apt-get install --no-install-recommends --no-install-suggests -y \
+	ffmpeg libmediainfo-dev \
+	vainfo mesa-va-drivers intel-media-va-driver-non-free i965-va-driver-shaders \
 	&& apt-get clean autoclean -y \
 	&& apt-get autoremove -y \
 	&& rm -rf /var/lib/apt/lists/*

transcoder/Dockerfile.dev

@@ -23,9 +23,15 @@ ENV SSL_CERT_DIR=/etc/ssl/certs
 RUN update-ca-certificates
 
 RUN go install github.com/bokwoon95/wgo@latest
+RUN sed -i -e's/ main/ main contrib non-free/g' /etc/apt/sources.list.d/debian.sources
 RUN apt-get update \
 	&& apt-get install --no-install-recommends --no-install-suggests -y \
-	ffmpeg libavformat-dev libavutil-dev libswscale-dev libmediainfo-dev \
+	# runtime dependencies
+	ffmpeg libavformat-dev \
+	# build dependencies
+	libavutil-dev libswscale-dev libmediainfo-dev \
+	# hwaccel dependencies
+	vainfo mesa-va-drivers intel-media-va-driver-non-free i965-va-driver-shaders \
 	&& apt-get clean autoclean -y \
 	&& apt-get autoremove -y
 WORKDIR /app

transcoder/src/hwaccel.go

@@ -31,7 +31,6 @@ func DetectHardwareAccel() HwAccelT {
 		return HwAccelT{
 			Name: "nvidia",
 			DecodeFlags: []string{
-				// TODO: check if this can always be enabled (for example with weird video formats)
 				"-hwaccel", "cuda",
 				// this flag prevents data to go from gpu space to cpu space
 				// it forces the whole dec/enc to be on the gpu. We want that.
@@ -47,6 +46,24 @@ func DetectHardwareAccel() HwAccelT {
 			// since we use hwaccel_output_format, decoded data stays in gpu memory so we must not specify it (it errors)
 			ScaleFilter: "scale_cuda=%d:%d",
 		}
+	case "vaapi":
+		return HwAccelT{
+			Name: name,
+			DecodeFlags: []string{
+				"-hwaccel", "vaapi",
+				"-hwaccel_device", GetEnvOr("GOTRANSCODER_VAAPI_RENDERER", "/dev/dri/renderD128"),
+				"-hwaccel_output_format", "vaapi",
+			},
+			EncodeFlags: []string{
+				// h264_vaapi does not have any preset or scenecut flags.
+				"-c:v", "h264_vaapi",
+				// if the hardware decoder could not work and fallbacked to soft decode, we need to instruct ffmpeg to
+				// upload back frames to gpu space (after converting them)
+				// see https://trac.ffmpeg.org/wiki/Hardware/VAAPI#Encoding for more info
+				// "-vf", "format=nv12|vaapi,hwupload",
+			},
+			ScaleFilter: "scale_vaapi=%d:%d",
+		}
 	default:
 		log.Printf("No hardware accelerator named: %s", name)
 		os.Exit(2)