RPJosh
/
memories
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add support for NVIDIA GPU accelerated transcoding 

Add a transcoding profile for ffmpeg suitable for using with NVENC.
monorepo
MB-Finski 2022-11-25 13:59:33 +02:00 committed by GitHub
parent 56767bc56d
commit 9972fc23c6
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 15 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
stream.go
View File

@ -288,12 +288,16 @@ func (s *Stream) transcode(startId int) {
// encoder selection
CV := "libx264"
// no need to transcode h264 streams for max quality
// Check whether hwaccel should be used
if os.Getenv("VAAPI") == "1" {
CV = "h264_vaapi"
extra := "-hwaccel vaapi -hwaccel_device /dev/dri/renderD128 -hwaccel_output_format vaapi"
args = append(args, strings.Split(extra, " ")...)
}
} else if os.Getenv("NVENC") == "1" {
CV = "h264_nvenc"
extra := "-hwaccel cuda -hwaccel_output_format cuda"
args = append(args, strings.Split(extra, " ")...)
}
// Input specs
args = append(args, []string{
@ -309,7 +313,15 @@ func (s *Stream) transcode(startId int) {
// VAAPI
format = "format=nv12|vaapi,hwupload"
scale = fmt.Sprintf("scale_vaapi=w=%d:h=%d:force_original_aspect_ratio=decrease", s.width, s.height)
} else {
} else if CV == "h264_nvenc" {
// NVENC
format = "format=nv12|cuda,hwupload"
if s.width >= s.height {
scale = fmt.Sprintf("scale_cuda=-2:%d", s.height)
} else {
scale = fmt.Sprintf("scale_cuda=%d:-2", s.width)
}
} else {
// x264
format = "format=nv12"
if s.width >= s.height {