Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Combination of --avsw, --pa [preanalysis] and --vpp-resize when using amf specific scalers causes video to not encode. #103

Closed
DimkaTsv opened this issue Feb 21, 2024 · 3 comments

Comments

@DimkaTsv
Copy link

Full details here:
GPUOpen-LibrariesAndSDKs/AMF#456

Thought it was AMF issue during original testing, but after expanding on it discovered that problem lies mainly in --avsw interaction and --pa, when --vpp-resize [amf_bilinear / amf_bicubic / amf_fsr] scaler is being utilized.

How trace level log looks like:

[AMFPreAnalysisImpl] 2024-02-21 20:06:30.022     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.023     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.025     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.026     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.028     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.029     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.031     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.033     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.034     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.036     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.038     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.039     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.041     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.042     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()
[AMFPreAnalysisImpl] 2024-02-21 20:06:30.044     5500 [AMFPreAnalysisImpl]   Debug: AMFPreAnalysisImpl::QueryOutput()

Also... Side question @rigaya . Is there any particular reason that AV1 codec forces --avsw flag to decoder on RX 7000 GPU's? (and RX 6000 GPU's as well, i assume?). Because HW decoder is capable for decoding AV1 on these GPU's. Or was that done to increase general compatability with older GPU while using autoconfiguration of non-specified parameters?

rigaya added a commit that referenced this issue Jul 18, 2024
HQScalerにHostフレームを投入すると、QueryOutputが返答を返さなくなる。
@rigaya
Copy link
Owner

rigaya commented Jul 18, 2024

Thank you for the details.

This problem can now be avoided by a workaround in VCEEnc 8.23.

It seems like that it gets stuck(freezed) when calling QueryOutput of HQScaler, when PreAnalysis is running and the input frame is coming from avsw via host memory. I found out that by explicitly converting the frame in host memory to device memory, the problem can be avoided ( 342b394 ).

@rigaya
Copy link
Owner

rigaya commented Jul 18, 2024

Is there any particular reason that AV1 codec forces --avsw flag to decoder on RX 7000 GPU's?

Yes. Currently, I have problem that AV1 hw decoder crushes when sending input bitstream (when calling SubmitInput). As I'm not sure of the cause, I am have AV1 hw decoder to be disabled.

@DimkaTsv
Copy link
Author

Can confirm fix, issue closed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants