Tools: Topology2: Add build of 1ch DMIC topologies #9763
Conversation
| @@ -16,6 +16,8 @@ EFX_HDA_MIC_TDFB_PARAMS=line2_pass,EFX_HDA_MIC_DRC_PARAMS=passthrough" | |||
| # HDA topology with mixer-based pipelines for HDA and | |||
| # passthrough pipelines for HDMI and | |||
| # 2 or 4 DMIC, no NHLT blob included in topology | |||
| "sof-hda-generic\;sof-hda-generic-1ch\;HDA_CONFIG=mix,NUM_DMICS=1,PDM0_MIC_B_ENABLE=false" | |||
There was a problem hiding this comment.
Should PDM0_MIC_B_ENABLE be 0 instead of false?
And do we need DMIC0_ENHANCED_CAPTURE for the topologies?
There was a problem hiding this comment.
Yes, 0 / 1 seems to be more common. Then maybe better to change.
I dropped DMIC0_ENHANCED_CAPTURE because the beamformer doesn't support mono. The 1ch version would need a topology variant with only DRC. I'd propose to fix this in two steps, first without DRC, and after that add the DRC.
|
Converting to draft, there's a topology build issue. Likely after I added similar DMIC1 stuff and didn't fully rebuild. |
singalsu
force-pushed
the
tplg_dmic_1ch
branch
2 times, most recently
from
854a60c to
cf6c37e
Compare
|
Changed back to proposal, my test with MTL worked. |
|
@singalsu can you rebase/force push again to get CI to rebuild with disk space fix. |
This patch adds to dmic-generic.conf for Intel platforms the mono audio formats for DAI copier, host copier, module copier, IIR equalizer, and gain. The earlier version of this file could not handle the case NUM_DMICS equals 1. Signed-off-by: Seppo Ingalsuo <[email protected]>
singalsu
force-pushed
the
tplg_dmic_1ch
branch
from
cf6c37e to
758426c
Compare
This patch adds the mono formats to dmic1-passthrough.conf for Intel platforms. It also fixes failing capture issue happening due to kernel formats handling changes. The formats are now introduced per NUM_DMICS to avoid kernel DMIC blob format select logic to be unable to select. The rate_min/max is replaced by rates. Signed-off-by: Seppo Ingalsuo <[email protected]>
This patch adds the missing mono formats to dmic1-mfcc.conf for Intel platforms. Signed-off-by: Seppo Ingalsuo <[email protected]>
This patch adds build of topologies - sof-hda-generic-1ch - sof-hda-generic-cavs25-1ch - sof-hda-generic-ace1-1ch - sof-hda-generic-ace3-1ch The three latter topologies contain the NHLT blob for DMIC setup. The first one requires that BIOS NHLT contains the mono DMIC setup. The macro PDM0_MIC_B_ENABLE need to be set to zero for alsa-utils to be able to create a valid mono DMIC blob. Signed-off-by: Seppo Ingalsuo <[email protected]>
singalsu
force-pushed
the
tplg_dmic_1ch
branch
from
758426c to
6615043
Compare
|
The alsa-utils update is needed for MTL and successive platforms to correctly set up mono FIFO, see alsa-project/alsa-utils#286 . |
|
I'm a bit at loss how do we track the alsa-utils dependency after we merge this PR. I can already see myself releasing SOF 2.13 without checking the builder making the tplg binaries has the correct alsa-utils and potentially ship invalid 1ch tplg binaries. Maybe file an issue that once the alsa-utils patch is merged and the github public SOF docker is update, we make a change to tools to make this is a hard dependency (tplg2 build will fail if alsa-utils is too old). Better ideas anyone? |
Use west to locally add alsa-lib and alsa-utils as a module and locally install in $HOME/work/sof |
Local alsa, if git master version, can be at times tricky to compile. But if download some specific release or git hash for build then should be more predictable. We have had with all new platforms problems with not everyone having latest alsa. Including our own CI system. Minimum version check with cmake would make a coarse filter for using system alsa, I'm now at v1.2.13, so once the ACE 1ch fix is merged and new release is made min. would be v1.2.14. Not sure what would be the best way here. |
Yes - that's exactly what west would do. You pick a commit that works and use it, then update when needed. |
No description provided.