feat(server): fully accelerated qsv #9689
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Deploying immich with
|
| Latest commit: |
5279a0e
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://9b8702ea.immich.pages.dev |
| Branch Preview URL: | https://feat-qsv-hw-decoding.immich.pages.dev |
jrasm91
approved these changes
May 23, 2024
| export class VAAPIConfig extends BaseHWConfig { | ||
| getBaseInputOptions() { | ||
| if (this.devices.length === 0) { | ||
| throw new Error('No VAAPI device found'); | ||
| } | ||
|
|
||
| let hwDevice = this.getPreferredHardwareDevice(); | ||
| if (hwDevice === null) { | ||
| if (!hwDevice) { |
mertalev
changed the title
nyanmisaka
reviewed
May 24, 2024
| return [ | ||
| 'hwmap=derive_device=opencl', | ||
| `tonemap_opencl=${tonemapOptions.join(':')}`, | ||
| 'hwmap=derive_device=vaapi:reverse=1', |
There was a problem hiding this comment.
hwmap=derive_device=vaapi:reverse=1
This will be not allowed in the future. It should be hwmap=derive_device=qsv:reverse=1
The qsv encoder's ability to accept vaapi fmt is experimental and was decided by Intel not to be promoted to upstream.
There was a problem hiding this comment.
I used vaapi there because qsv throws an error:
[AVHWDeviceContext @ 0x24e400304c0] 0.0: Intel(R) OpenCL Graphics / Intel(R) Arc(TM) Graphics
[AVHWDeviceContext @ 0x24e400304c0] Intel QSV to OpenCL mapping function found (clCreateFromVA_APIMediaSurfaceINTEL).
[AVHWDeviceContext @ 0x24e400304c0] Intel QSV in OpenCL acquire function found (clEnqueueAcquireVA_APIMediaSurfacesINTEL).
[AVHWDeviceContext @ 0x24e400304c0] Intel QSV in OpenCL release function found (clEnqueueReleaseVA_APIMediaSurfacesINTEL).
[AVHWFramesContext @ 0x24e401d4740] The hardware pixel format 'vaapi' is not supported by the device type 'QSV'
[Parsed_hwmap_2 @ 0x24e40173000] Failed to initialise target frames context: -38.
[Parsed_hwmap_2 @ 0x24e40173000] Failed to configure output pad on Parsed_hwmap_2
Error reinitializing filters!
Failed to inject frame into filter network: Function not implemented
Error while processing the decoded data for stream #0:0
[AVIOContext @ 0x24e401907c0] Statistics: 0 bytes written, 0 seeks, 0 writeouts
Terminating demuxer thread 0
[AVIOContext @ 0x24e401902c0] Statistics: 1982895 bytes read, 2 seeks
Conversion failed!
Is there something else I would need to change to make this work? The full command is this:
ffmpeg -hwaccel qsv -async_depth 4 -threads 1 -i HDR.mp4 \
-c:v h264_qsv -c:a copy -movflags faststart -fps_mode passthrough -map 0:0 -bf 7 -refs 5 -g 256 -v verbose \
-vf hwmap=derive_device=opencl,tonemap_opencl=desat=0:format=nv12:matrix=bt709:primaries=bt709:range=pc:tonemap=hable:transfer=bt709,hwmap=derive_device=qsv:reverse=1 \
-preset 7 -global_quality 23 SDR.mp4
There was a problem hiding this comment.
I think you have to specify the format=qsv to enable the indirect hwmap.
hwmap=derive_device=qsv:reverse=1,format=qsv
There was a problem hiding this comment.
Thanks, that worked! I'll change it to use qsv.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Descripion
This PR adds end-to-end acceleration for Quick Sync transcoding. Similar to NVENC and RKMPP, tone-mapping is done with OpenCL. Based on my testing, OpenCL was even faster than native VPP tone-mapping while having more features and higher quality.
Testing
Tested with hardware decoding enabled and disabled, with tone-mapping enabled and disabled, and scaling set to either 720p or original.
Note: My processor is too new for the
intel-compute-runtimein the server image, so I had to manually install the latest release here. I'll make a PR to apply this in the base image, but in the meantime it should only affect very new processors and only when hardware decoding is enabled.Current transcoding time: 470s
With hardware decoding: 34s