Audio: MFCC: Updates to run MFCC in sof-testbench4 by singalsu · Pull Request #10727 · thesofproject/sof

singalsu · 2026-04-27T14:42:25Z

No description provided.

For capture topologies the file write component acts as the host (sink endpoint). Its ipc_config.type was left as SOF_COMP_MODULE_ADAPTER, which caused module_adapter_prepare() to fail with 'no sink present on period size calculation' because the capture host has no downstream sinks in its bsink_list. As the fix set ipc_config.type to SOF_COMP_HOST for the capture host file component (FILE_WRITE with capture direction) so the sink check in module_adapter_prepare() is properly skipped, matching the behavior of SOF_COMP_DAI set for the DAI-side file components. Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>

The files host_io_gateway_pipelines_s16/24/32.conf are obsolete and are not used for current test topologies. Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>

Copilot

Pull request overview

This PR adds MFCC as a selectable component in the topology2 benchmark topologies (HDA + SDW) and updates MFCC tuning/run tooling to work with sof-testbench4.

Changes:

Add MFCC bench include files (widgets, routes, controls) plus 16 kHz host gateway/audio format snippets.
Register MFCC in benchmark CMake targets and benchmark top-level configs (cavs-benchmark-*.conf).
Update testbench file component init logic and refresh MFCC tuning scripts/docs (Octave/Matlab + run script + decoder helper).

Reviewed changes

Copilot reviewed 20 out of 22 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
tools/topology/topology2/include/bench/one_input_output_format_s32_16k.conf	New 32-bit 16 kHz single-format I/O definition for bench components
tools/topology/topology2/include/bench/one_input_output_format_s24_16k.conf	New 24-in-32 16 kHz single-format I/O definition for bench components
tools/topology/topology2/include/bench/one_input_output_format_s16_16k.conf	New 16-bit 16 kHz single-format I/O definition for bench components
tools/topology/topology2/include/bench/mfcc_s32.conf	New MFCC bench component include (s32) wiring widget + host gateway + routes
tools/topology/topology2/include/bench/mfcc_s24.conf	New MFCC bench component include (s24) wiring widget + host gateway + routes
tools/topology/topology2/include/bench/mfcc_s16.conf	New MFCC bench component include (s16) wiring widget + host gateway + routes
tools/topology/topology2/include/bench/mfcc_route.conf	New bench routes connecting MFCC between host-copier and DAI copier
tools/topology/topology2/include/bench/mfcc_controls_playback.conf	New bench playback control block for MFCC bytes control
tools/topology/topology2/include/bench/mfcc_controls_capture.conf	New bench capture control block for MFCC bytes control
tools/topology/topology2/include/bench/host_io_gateway_pipelines_s32.conf	Removed old host+io gateway pipeline definition (appears unused after changes)
tools/topology/topology2/include/bench/host_io_gateway_pipelines_s24.conf	Removed old host+io gateway pipeline definition (appears unused after changes)
tools/topology/topology2/include/bench/host_gateway_pipelines_s32_16k.conf	New 16 kHz host gateway pipelines for s32
tools/topology/topology2/include/bench/host_gateway_pipelines_s24_16k.conf	New 16 kHz host gateway pipelines for s24
tools/topology/topology2/include/bench/host_gateway_pipelines_s16_16k.conf	Switch to BENCH pipeline indices and add explicit 16 kHz rates
tools/topology/topology2/development/tplg-targets-bench.cmake	Add `mfcc` to bench component list and parameters (`BENCH_MFCC_PARAMS`)
tools/topology/topology2/cavs-benchmark-sdw.conf	Include MFCC component + add BENCH_CONFIG entries mfcc16/24/32
tools/topology/topology2/cavs-benchmark-hda.conf	Include MFCC component + add BENCH_CONFIG entries mfcc16/24/32
tools/testbench/file.c	Adjust file write init to set component type HOST for capture direction
src/audio/mfcc/tune/setup_mfcc.m	Update output paths to repo `tools/` dir and adjust addpath/rmpath locations
src/audio/mfcc/tune/run_mfcc.sh	Update to run `sof-testbench4` with a topology2 benchmark MFCC tplg
src/audio/mfcc/tune/decode_ceps.m	New helper to parse MFCC output stream and plot cepstral coefficients
src/audio/mfcc/tune/README.txt	New README documenting MFCC config generation, testbench run, and plotting

Comments suppressed due to low confidence (2)

src/audio/mfcc/tune/run_mfcc.sh:19

OPT already contains “-i $RAW_INPUT -o $RAW_OUTPUT”, but the testbench invocation also appends another “-i … -o …”. This duplicates CLI options and can lead to parsing errors or unintended overrides; keep input/output specified in only one place.
src/audio/mfcc/tune/run_mfcc.sh:16
The script generates a mono raw input (sox uses “-c 1”), but OPT configures the testbench for 2 channels (“-c 2”) and the preceding comment also says “1 channel”. Align channel count across the sox conversion, the testbench options, and the comment.

Copilot

Pull request overview

Copilot reviewed 21 out of 23 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (2)

src/audio/mfcc/tune/run_mfcc.sh:15

The script sets -c 2 in OPT but the sox conversion generates a 1-channel raw file (-c 1). This will misconfigure the testbench run. Align the channel count between sox and the testbench options (and update the nearby comment accordingly).
src/audio/mfcc/tune/run_mfcc.sh:18
OPT already includes -i $RAW_INPUT -o $RAW_OUTPUT, but the testbench invocation appends -i/-o again. This double-specification can lead to confusing behavior depending on argument parsing; keep -i/-o in one place only.

Copilot · 2026-04-27T15:41:16Z

+
+	"mfcc16" {
+		<include/bench/mfcc_s16.conf>
+	}
+
+	"mfcc24" {
+		<include/bench/mfcc_s24.conf>
+	}
+
+	"mfcc32" {
+		<include/bench/mfcc_s32.conf>
+	}
+


The new MFCC bench configs include 16 kHz audio formats (*_16k.conf), but in this SDW benchmark topology the ALH DAI/coprocessor widgets are explicitly constrained to $JACK_RATE (default 48000) for both input/output formats. This will create a hard sample-rate mismatch on routes that connect mfcc.* to $BENCH_*_DAI_COPIER. Consider overriding JACK_RATE to 16000 for the mfcc* BENCH_CONFIGs (or inserting SRC/conversion), or omit MFCC from the SDW benchmark until the rate domains match.

Suggested change

"mfcc16" {

<include/bench/mfcc_s16.conf>

}

"mfcc24" {

<include/bench/mfcc_s24.conf>

}

"mfcc32" {

<include/bench/mfcc_s32.conf>

}

# MFCC benchmark configs are intentionally not enabled in the SDW

# benchmark topology because the SDW ALH DAI/coprocessor side is

# constrained to $JACK_RATE while MFCC bench configs operate in a

# different rate domain. Re-enable only after adding a matching rate

# override or explicit sample-rate conversion.

Right, I'll do as suggested and remove for now MFCC from SDW. It can be done since it's not usable and HDA variant is sufficient for testbench runs.

There was earlier only topology1 for MFCC. This patch adds topologies sof-hda-benchmark-mfcc16/24/32.tplg to be able to test the component in testbench environment. Note: Due to the set 16 kHz DAI rate these can't be used in real devices. It will need a capture pipeline with 48 kHz to 16 kHz SRC. Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>

This patch updates the run script run_mfcc.sh to use topology2 and the current sof-testbech4. Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>

This patch moves the files from tools/tune/mfcc to directory src/audio/mfcc/tune. The patchs are fixed for the new location. Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>

Copilot

Pull request overview

Copilot reviewed 21 out of 23 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (2)

src/audio/mfcc/tune/run_mfcc.sh:12

run_mfcc.sh assumes SOF_WORKSPACE is set when constructing TESTBENCH and TOPOLOGY paths. If it is unset/empty, the script will fail later with confusing file-not-found errors. Consider adding an early check that SOF_WORKSPACE is set (and points to a valid tree) and print a clear error message if not.
src/audio/mfcc/tune/run_mfcc.sh:15
The script converts the input into a 1-channel raw file (sox ... -c 1 ... "$RAW_INPUT") but then invokes sof-testbench4 with -c 2 (input channels). This channel-count mismatch will cause the testbench to interpret the raw stream incorrectly. Align these (either generate 2-channel raw from sox or run the testbench with -c 1 and use a 1-channel topology).

Copilot · 2026-04-27T16:07:06Z

 	"gain"
 	"igo_nr"
 	"level_multiplier"
+	"mfcc"


mfcc is now included in the generic components list, which means this CMake loop will also generate SDW benchmark targets for MFCC. However cavs-benchmark-sdw.conf intentionally does not define any BENCH_CONFIG keys for mfcc16/24/32, so those generated SDW topologies will silently omit MFCC and produce misleading/incorrect artifacts. Consider either excluding MFCC from the SDW target generation here (special-case comp == mfcc) or adding properly rate-matched MFCC entries to the SDW IncludeByKey.BENCH_CONFIG block.

Suggested change

"mfcc"

I don't prefer to do that to not create CMake complexity that would be changed again later. The tests are not running the SDW benchmark variants. The topology built for SDW now is just empty. I'll create in later phase topologies for real devices with SRC in a split pipeline from the module copier.

singalsu added 2 commits April 27, 2026 16:10

Tools: Topology: Delete unused test topology .conf

c6220a8

The files host_io_gateway_pipelines_s16/24/32.conf are obsolete and are not used for current test topologies. Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>

singalsu commented Apr 27, 2026

View reviewed changes

Comment thread tools/topology/topology2/include/bench/host_gateway_pipelines_s16_16k.conf

singalsu force-pushed the mfcc_update branch from 6f676c9 to b30415a Compare April 27, 2026 14:56

singalsu marked this pull request as ready for review April 27, 2026 15:05

Copilot AI review requested due to automatic review settings April 27, 2026 15:05

singalsu requested review from dbaluta, jsarha, kv2019i, lbetlej, lgirdwood, mmaka1, plbossart and ranj063 as code owners April 27, 2026 15:05

Copilot started reviewing on behalf of singalsu April 27, 2026 15:09 View session

Copilot AI reviewed Apr 27, 2026

View reviewed changes

singalsu force-pushed the mfcc_update branch from b30415a to ba71a66 Compare April 27, 2026 15:23

singalsu requested a review from Copilot April 27, 2026 15:24

Copilot started reviewing on behalf of singalsu April 27, 2026 15:33 View session

Copilot AI reviewed Apr 27, 2026

View reviewed changes

singalsu added 3 commits April 27, 2026 18:52

Audio: MFCC: Use sof-testbench4 and topology2 to run the module

e3107ec

This patch updates the run script run_mfcc.sh to use topology2 and the current sof-testbech4. Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>

Audio: MFCC: Move setup and run scripts to module directory

45da55c

This patch moves the files from tools/tune/mfcc to directory src/audio/mfcc/tune. The patchs are fixed for the new location. Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>

singalsu force-pushed the mfcc_update branch from ba71a66 to 45da55c Compare April 27, 2026 15:54

singalsu requested a review from Copilot April 27, 2026 15:56

Copilot AI reviewed Apr 27, 2026

View reviewed changes

lgirdwood approved these changes Apr 27, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Audio: MFCC: Updates to run MFCC in sof-testbench4#10727

Audio: MFCC: Updates to run MFCC in sof-testbench4#10727
singalsu wants to merge 5 commits intothesofproject:mainfrom
singalsu:mfcc_update

singalsu commented Apr 27, 2026

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 27, 2026

Uh oh!

singalsu Apr 27, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 27, 2026

Uh oh!

singalsu Apr 27, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

singalsu commented Apr 27, 2026

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Apr 27, 2026

Choose a reason for hiding this comment

Uh oh!

singalsu Apr 27, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Apr 27, 2026

Choose a reason for hiding this comment

Uh oh!

singalsu Apr 27, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants