thesofproject
[FEATURE] topology2: nocodec: remove redundant copiers
Is your feature request related to a problem? Please describe. The nocodec topology exposes redundant intermediate copiers on the SSP input path
Describe the solution you'd like Removal of those intermediate useless copiers.
@marcinszkudlinski @mwasko do other OSes use the intermediate copiers ? i.e. safe to remove for compatibility reasons. Thanks !
FYI @ujfalusi @singalsu -- in context of the "capture pipeline with branches" discussion.
@marcinszkudlinski @mwasko do other OSes use the intermediate copiers ? i.e. safe to remove for compatibility reasons. Thanks !
It seems that these additional copiers were added by Bard in change : https://github.com/thesofproject/sof/commit/f6f9b62d5a17a416d203ae531ae652ae131837c2 . It seems that one potential problem they can solve is perhaps different formats for gain type modules?
fwiw, in nocodec configurations we have fixed FE PCM format for all streams and while the modules in paths most of the time have two formats supported, one of them is the (static) FE format. This works around the issue that we cannot handle a module copier which does branching correctly. The format cannot be changed at the branch, if there is need for a change, it has to be done on the branch itself after the module copier.
Stable branched for 2.10, nobody assigned to work on this, so moving this to 2.11.
Feature cutoff for v2.12. No owner in 2.12 cycle either, so clearing the milestone for now.
@singalsu is this something you can look at in general (just a quick check)- if its only impacts SSP paths then we can de-prioritize and close. If it impacts other endpoints like SDW or display then we should schedule. Thanks !
@singalsu is this something you can look at in general (just a quick check)- if its only impacts SSP paths then we can de-prioritize and close. If it impacts other endpoints like SDW or display then we should schedule. Thanks !
The SSP nocodec case looks clear, the module-copier looks redundant. But we have one similar extra module-copier in Chromebook topologies with mic input to google-rtc-aec. Need to check if it performs format conversion. The parts of topologies should be not shared (look quite different -- didn't check yet the code) so they can be addressed separately. Since this issue is about nocodec, can be de-priorised and done when we have more time.
