Skip to content

ASoC: SOF: sof-audio: skip prepared widgets when pipeline is prepared #4240

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.

Sign up for GitHub

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

Closed
wants to merge 1 commit into from
Closed

ASoC: SOF: sof-audio: skip prepared widgets when pipeline is prepared #4240

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions sound/soc/sof/sof-audio.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -611,8 +611,16 @@ sof_walk_widgets_in_order(struct snd_sof_dev *sdev, struct snd_sof_pcm *spcm,
break;
case SOF_WIDGET_PREPARE:
{
struct snd_sof_widget *swidget = widget->dobj.private;
struct snd_pcm_hw_params pipeline_params;

/* In case that feedback buffer is shared between two stream, a prepared
* widget in another stream for producing feedback buffer should be skipped
* since pipeline_params is built based on the current stream, not feedback stream.
*/
if (swidget->prepared)
Copy link
Collaborator

@ranj063 ranj063 Mar 13, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@RanderWang I'm not following the issue here. Even if you didn't have this check here, if the widget were already prepared and so were all it's sink widgets, the function prepare _widgets_in_path() would skip this anyway, right?

Copy link
Author

@RanderWang RanderWang Mar 14, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@ranj063 yes, it is a little tricky. Let me explain it. For DMIC stream only support S32_LE and aec module use S16_LE format, we will use copier module to convert S32_LE to S16_LE. Please check the picture. If the stream is prepared from copier.SSP.10.1, this widget is prepared so it will be skipped and start to prepare google-rtc-aec.18.1. At this time pipeline params use S32_LE but google-rtc-aec.18.1 only support S16_LE, so we will get error "sof_ipc4_init_audio_fmt: Unsupported audio format: 48000Hz, 32bit, 2 channels". But if the stream is prepared from copier.DMIC.14.1, the copier.module.19.1 will change pipeline params to use S16_LE since we define the conversion in copier config, so the next prepared google-rtc-aec.18.1 will match this format. The problem is that we can't start from another pipeline since we may change pipeline params in current stream.

example

Copy link
Member

@plbossart plbossart Mar 14, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@RanderWang I don't understand your explanations. what does the copier.SSP.10.1 provide to the google-rtc-aec.18.1 module? The echo reference? If yes, why can't we demote the output of the copier.SSP.10.1 to match what the AEC needs?

I also don't understand the notion of 'prepared' from different sources that may or may not all be available. what happens if we start/stop the playback, there's no way this sort of topology can adjust.

I also don't think this integration is quite right, as I mentioned in another PR you want the output of the AEC to go to the host AND to the noise suppression.

Copy link
Collaborator

@ranj063 ranj063 Mar 14, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@RanderWang yes, it looks like you've got the wrong audio formats set for output pin 1 for copier.SSP.10.1

Copy link
Author

@RanderWang RanderWang Mar 15, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@RanderWang I don't understand your explanations. what does the copier.SSP.10.1 provide to the google-rtc-aec.18.1 module? The echo reference? If yes, why can't we demote the output of the copier.SSP.10.1 to match what the AEC needs?

SSP.10.1 is used for reference. The Copier.SSP.10.1 is prepared in another stream so it will be skipped by driver, so we use fe_params which uses S32_LE

I also don't understand the notion of 'prepared' from different sources that may or may not all be available. what happens if we start/stop the playback, there's no way this sort of topology can adjust.

Currently we build module params from source (DAI or HOST) and change the params in some modules like SRC or Copier and propagate params to sink module. We can't use incorrect source for prepare. Playback also get such issue if the host copier linked to another stream directly.

I also don't think this integration is quite right, as I mentioned in another PR you want the output of the AEC to go to the host AND to the noise suppression.

please check thesofproject/sof#7237 (comment). And It is also defined in chrome document that AEC is linked to noise suppression. We discussed in my WSR001 2023

PCM99C <----------RTNR ----------< GOOGLE_AEC <---------------DMIC01

Copy link
Member

@plbossart plbossart Mar 15, 2023

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

well, I am not going to move until we agree on the topology. You added 2 instances of AEC in thesofproject/sof#7237 and I don't see how this can fly.

continue;

str = "prepare";
/*
* When walking the list of connected widgets, the pipeline_params for each
Expand Down