Conversation
Fix the build issues related to - libc.a path update to use from xt-clang xtensa - comment ibuffer_allocater.cc entry as file does not exist - Add zephyr compiler options - Introduce assert_func Signed-off-by: Naveen Manohar <naveen.m@intel.com> Signed-off-by: Uday M Bhat <uday.m.bhat@intel.com>
PR thesofproject#11214 dropped ibuffer_allocator.cc from mww_tflm_lib on the grounds that the file does not exist. That is true of the tflite-micro commit pinned by scripts/tensorflow-clone.sh (e86d97b6), where IBufferAllocator is a header-only pure-virtual interface, but not of trees carrying the Xtensa ICF vtable-folding fix, which split IBufferAllocator's 13 virtuals out of line into that .cc. On those trees the removal breaks the link with ~14 undefined references under both toolchains: libmww_tflm_lib.a(micro_allocator.cc.obj):(.literal. _ZN6tflite16IBufferAllocatorC2ERKS0_+0x0): undefined reference to `vtable for tflite::IBufferAllocator' Guard the source on the file's existence so both layouts build. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com> Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
ptl fails to link, on the branch base and with PR thesofproject#11214, under both xt-clang and the Zephyr SDK: libmww_libc_shim.a(libm_math_sf_exp.c.o): sf_exp.c:(.text.expf+0x12): dangerous relocation: call8: call target out of range: __addsf3 The shim is assembled by `ar x`-ing prebuilt members out of the toolchain's libc.a/libm.a. Those objects were compiled without -mlongcalls, so their calls to the libgcc soft-float helpers are plain CALL8 with a +/-512KB reach. Zephyr's Xtensa linker script places plain `.text` (where libgcc lives) via `*(.literal .text)` at the front of the .text output section and per-function `.text.*` via the later `*(.literal.* .text.*)` rule at the back. Measured on wcl, which still links: __addsf3 at 0xa003504c, expf at 0xa0091450 - 378KB apart. ptl's .text is 651KB, so the two ends fall outside CALL8 range. Rename the extracted members' .text.<fn>/.literal.<fn> back to plain .text/.literal so they land in the same early cluster as libgcc. On ptl this brings expf to 0xa00327bc and __addsf3 to 0xa0032d84, 1.5KB apart. Order within a single `*(...)` wildcard follows input order, so each object's .literal still precedes its .text and l32r reach is unaffected. Verified: ptl links and boots on PTL silicon (aphid), topology sof-ptl-dmic-wov-multi-4ch instantiates, DMIC ECNS capture records cleanly and MWW runs inference on all three slots. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com> Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
tlv_value_set_uaol_caps() set link_count to the devicetree device count unconditionally and indexed link_caps[] by device index, so a link whose uaol_get_capabilities() call failed was still advertised to the host as a present link with all-zero capabilities. That matters now that the driver rejects the query when the host has not set UAOLCTL.OFLEN: without this, a firmware that cannot read any link still reports link_count = 1 with zeroed stream counts and FIFO sizes. Count only the links that answered, pack them from index 0, and size the TLV to match. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com> Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
… on IPC4 kpb.c drives both a sel_sink (buf_id 0, the real-time feed to detectors) and a host_sink (buf_id 1, the drain that bursts pre-roll to the host). Declare two sink pins in kpb.toml so the drain path can be bound from topology. Additionally, under IPC4 pipeline state transitions and resets are synchronous; returning -EBUSY on reset breaks the host teardown state machine and causes widget free failures. Guard the deferred -EBUSY drain check under #if !CONFIG_IPC_MAJOR_4. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
…in pipeline ops Add NOTIFIER_ID_D0IX_STATE so components can track host S0iX/D0i3 entry and exit transitions dispatched from ipc4_module_process_d0ix(). In ipc4_pipeline_prepare() and ipc4_pipeline_trigger(), add COMP_STATE_PREPARE as a valid no-op state for RUNNING and RESET transitions. When a pipeline has executed pipeline_prepare(), its state is COMP_STATE_PREPARE; previously omitting it caused spurious IPC4_INVALID_REQUEST (error 7) returns during stream restart and reset. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
…eaming Send SOF_IPC4_NOTIFY_PHRASE_DETECTED to the host upon keyword detection via notify_host_detect(). Update source buffer lookup to dynamically identify the KPB host sink audio buffer, keep the sink silent while listening (WOV_ARB_NO_ACTIVE) so the host PCM stays blocked in read() and allows platform D0i3 entry, and periodically dispatch phrase detected notifications during audio drain to keep the host reader fed. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
…ycle Add CONFIG_COMP_MWW_FAKE_WAKE_MS test aid and app/wov-d0i3-fake-wake.conf to synthesize keyword detections after a specified delay without a physical microphone. In mww_prepare(), arm the fake wake deadline so S0-only tests can trigger directly, and fire the detection inline from mww_process(). Refactor mww_prepare() so model initialization runs once while stream state and fake wake deadlines re-arm cleanly on every prepare even if the module instance remains allocated. Remove mww_notify_score(0) from mww_reset() to prevent host widget lookup race conditions during teardown. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
…oV arbiter Set period 1000 on custom capture pipeline classes so the timer domain scheduler period reaches the firmware. Configure KPB with 2 output pins: - Pin 0 (sel_sink) real-time feed to detector slots via mixin.106.1 - Pin 1 (host_sink) drain route to wov-arbiter.104.1 pin 0 Configure WOV arbiter with 4 inputs (pin 0 KPB audio, pins 1-3 MWW detector features) and 16-bit PCM output. Set capture_compatible_d0i3 on the host copier widget to enable D0i3 stream compatibility under IPC4. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
…ncies Document the multi-slot Wake-on-Voice test case with microWakeWord (MWW) on Panther Lake (PTL): - Update the topology architecture diagram with dual-sink KPB and 4-pin WOV arbiter connections. - Document Linux kernel dependencies (branch wov-ipc4-d0i3 on lgirdwood/linux) and SOF firmware dependencies (branch wcl-uaol-wov-002 on lgirdwood/sof). - Provide build commands for firmware with the fake-wake overlay (CONFIG_COMP_MWW_FAKE_WAKE_MS=5000) and topology compilation. - Include complete source code for the standalone blocking PCM read client (wov_blocking_read.c) and 20-run test suite (run_20_tests.sh). - Document verified hardware test results (20/20 passes, zero IPC/ASoC errors). Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
…elemetry in D0i3 Track D0ix state transitions via NOTIFIER_ID_D0IX_STATE across all MWW detector slots. While in D0i3 (host sleeping in s2idle): - Suppress mww_notify_score() ALSA enum kcontrol notifications so that intermediate probability score updates do not send IPCs that prematurely wake the host out of sleep. - Re-arm the fake wake deadline (CONFIG_COMP_MWW_FAKE_WAKE_MS) from the time D0i3 is entered so that the configured timer duration elapses while the system is suspended. - Disarm the fake wake timer when returning to D0i0. Update topology README with the 5-run D0i3 sleep-and-wake validation runbook and verified Panther Lake test results (5/5 passes, waking via AudioDSP IRQ 193). Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
… TGL, and WCL Add prebuilt topology binary files for the 4-channel native 16 kHz DMIC multi-slot Wake-on-Voice use case: - build/sof-ptl-dmic-wov-multi-4ch.tplg (Panther Lake) - build/sof-tgl-dmic-wov-multi-4ch.tplg (Tiger Lake) - build/sof-wcl-dmic-wov-multi-4ch.tplg (Wildcat Lake) The Panther Lake binary matches the exact md5 (4f76c48bc0234062ae628ec12e5f8dd0) tested and verified on Aphid hardware across S0 and D0i3 wake test suites. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
…ter to fix heap leak
TFLite Micro defines TF_LITE_REMOVE_VIRTUAL_DELETE on MicroMutableOpResolver,
which expands to a no-op 'void operator delete(void* p) {}'. Consequently,
calling 'delete inst->op_resolver' in MWW_Free() never returned the memory
to the heap, leaking ~600 bytes per slot on every stream open/close cycle.
Across multi-slot configurations (3 detector slots), repeated stream cycles
exhausted the DSP system heap, failing on run 10 with -ENOMEM in
MWW_InitOps() and returning -22 (IPC4_INVALID_REQUEST) to snd_pcm_start().
Eliminate heap allocations from MWW model initialization by allocating
both MwwOpResolver and MicroInterpreter in aligned static storage using
placement new and explicit destructor invocation.
Verified with 10 consecutive D0i3 sleep-and-wake test cycles on Panther Lake
hardware (Aphid), achieving 10/10 passes with zero errors.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
…ified PTL topology Align tools/topology/topology2/dmic-wov-multi-wcl-4ch-manifest.conf to match the verified Panther Lake topology manifest exactly, with the target platform set to 'wcl'. This brings the dual-sink KPB architecture (real-time detector feed on sink 0, pre-roll burst drain on sink 1), 4-pin WOV arbiter routing, scheduler period token, and D0i3 capture compatibility flags to Wildcat Lake. Recompiled and updated build/sof-wcl-dmic-wov-multi-4ch.tplg (37636 bytes). Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Disable UAOL on Intel ADSP targets in boards/intel_adsp/Kconfig.defconfig (default n) and explicitly disable CONFIG_UAOL and CONFIG_DAI_INTEL_UAOL in the wov-d0i3-fake-wake overlay. Rebuilt firmware binaries for both Panther Lake (PTL) and Wildcat Lake (WCL), and verified D0i3 sleep-and-wake on Aphid hardware (AudioDSP IRQ 193). Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
…MICs Refactor multi-slot Wake-on-Voice (WoV) pipelines into a reusable feature module that integrates with functional topologies (cavs-sdw, sof-hda-generic) and can also be loaded as a standalone feature topology via snd_sof feature_topologies. - Refactor platform/intel/dmic-wov-multi.conf to use non-colliding pipeline IDs (110-117), parameterized DMIC DAI bindings, and D0i3-compatible host copiers. - Ensure WoV always utilizes hardware PDM DMICs (4 channels, 16kHz) for audio input. - Add multi-slot WoV support to dmic-wov-feature.conf, including PDM DMIC DAI and DAPM virtual widgets required for standalone or overlay usage. - Integrate multi-slot WoV into cavs-sdw.conf and sof-hda-generic.conf with WCL platform support and control class definitions. - Add production CMake targets in tplg-targets-ace3.cmake and tplg-targets-hda-generic.cmake for PTL and WCL. - Hardware verified on Aphid (PTL ACE 3.0): * 10 / 10 S0 WoV detection tests passed. * 10 / 10 D0i3 suspend/resume wake tests passed (IRQ 193). Signed-off-by: Liam Girdwood <liam.r.girdwood@intel.com>
lgirdwood
changed the title
Member
Author
|
@naveen-manohar @harajend @bardliao @ujfalusi this is just for testing on the CI, it has a lot of patches not for upstream, just for debug. Will allow us to roll out WOV test alongside other WOV PR in flow. |
…722 targets
Fix driver error "can't find host comp to bind pcm" when loading SoundWire
topologies with multi-slot WoV enabled.
- cavs-sdw.conf: Align DMIC0_PCM_CAPS to 'Gain Capture 10' so that host-copier.10.capture
matches the stream capability name expected by PCM 10 ("DMIC ECNS Capture").
- dmic-wov-multi.conf: Define DMIC0_PCM_CAPS as "Gain Capture 10" to ensure consistent
widget-to-PCM capability string binding across all parent topology templates.
- tplg-targets-ace3.cmake: Add production CMake targets for sof-wcl-rt721-4ch-wov-multi
and sof-wcl-rt722-4ch-wov-multi.
Signed-off-by: Liam Girdwood <liam.r.girdwood@intel.com>
Contributor
|
Thanks @lgirdwood |
…odules Functional and feature topologies (such as SoundWire RT721/RT722 and DMIC pipelines) require open audio processing components like DRC, TDFB, Multiband DRC, Aria, and MUX. Ensure these modules are enabled in both PTL and WCL board configurations and module overlays so they are built into sof-ptl-openmodules.ri and sof-wcl-openmodules.ri. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
… tables Update embedded DMIC NHLT configuration tables in prebuilt multi-slot WoV topology binaries: - build/sof-ptl-dmic-wov-multi-4ch.tplg - build/sof-tgl-dmic-wov-multi-4ch.tplg - build/sof-wcl-dmic-wov-multi-4ch.tplg The updated NHLT tables expand endpoint format capabilities to support mono (1-channel), stereo (2-channel), and quad (4-channel) audio streams in both 16-bit and 32-bit container depths at 16 kHz and 48 kHz. This prevents ALSA copier bind failures (no matching blob) when capturing mono or 16-bit streams from DMIC endpoints. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Implement an on-demand synthetic wake trigger controlled via a volatile
kcontrol ('wovdebug_$index') per WoV slot instance:
- Setting the switch kcontrol to 1 enables and arms that instance's
fake wake countdown timer.
- Once the stream is prepared/active and running in S0 or upon D0i3
suspend, the countdown timer triggers keyword detection (KPB drain
and arbiter notify).
- Upon firing, the firmware clears the kcontrol back to 0 and sends
an IPC4 module notification event with the ALSA magic value so that
the Linux kernel updates the control and notifies ALSA userspace.
- Define 'wovdebug_$index' mixer control in mww.conf and manifests.
- Recompile prebuilt WoV topologies.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
…gies In alsatplg, mixing named object syntax (Object.PCM.pcm."10") with anonymous array syntax (Object.PCM.pcm [ ... ]) causes alsatplg's internal array index tracker to reset, resulting in subsequent Object.PCM.pcm array definitions (Speaker, HDMI, Jack, Microphone) overwriting indices 0 and 1. This caused the Speaker PCM (PCM ID 2) and others to be omitted from the compiled SoundWire multi-slot WoV topologies. - dmic-wov-multi.conf: Convert Object.PCM.pcm to standard anonymous array definitions for PCM 10 (DMIC ECNS Capture) and PCM 11 (DMIC Multi-WOV). - dmic-generic.conf: Wrap default DMIC0 Object.PCM.pcm in IncludeByKey.INCLUDE_WOV so that multi-slot WoV topologies do not instantiate duplicate PCM 10. - Recompile all SoundWire, HDA generic, and manifest multi-slot WoV topologies, restoring Speaker (PCM 2), Jack Out (PCM 0), Jack In (PCM 1), HDMI 1-3, and Microphone (PCM 4) alongside DMIC ECNS (PCM 10) and Multi-WOV (PCM 11). Signed-off-by: Liam Girdwood <liam.r.girdwood@intel.com>
…CM 12 WoV mono Fan out DMIC DAI capture pipeline (Pipeline 110) via mixin.110.1 (num_output_pins=2) to ECNS mixout.115.1 and new Pipeline 118 (mixout.118.1 -> host-copier.10.capture). Remap ECNS stereo capture to PCM 11 (Pipeline 117) and multi-slot WoV to PCM 12 (Pipeline 114). Resolve duplicate target in tplg-targets-ace3.cmake.
…lsa utilities Add blocking read test utilities for Wake-on-Voice gated capture validation: - wov_blocking_read.c: ALSA-lib non-MMAP blocking read with no-period-wakeup - wov_blocking_read_tinyalsa.c: tinyalsa blocking read using PCM_MMAP | PCM_NOIRQ - Update Makefile with modular build targets
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
This PR introduces multi-slot Wake-on-Voice (WoV) with low-power D0i3 sleep/wake capability, hardware PDM DMIC input, and modular topology feature integration across Panther Lake (PTL), Wildcat Lake (WCL), and Tiger Lake (TGL).
Highlights
Hardware PDM DMIC Wake-on-Voice:
dmic16k, DAI index 1) directly from the SoC PDM decimator, avoiding software resampling.dmic01) and external SoundWire microphones (Capture-SmartMic).Multi-Slot WoV Pipeline Architecture:
DMIC Multi-WOV).D0i3 Low-Power Sleep & Wake:
capture_compatible_d0i3 1to maintain low-power DSP SRAM retention across host S0ix sleep.COMP_STATE_PREPAREand suppress telemetry during D0i3.Topology Integration & Modularity:
dmic-wov-feature.confallowing multi-slot WoV to be included as a feature module into functional topologies (INCLUDE_WOV=multi).sof-ptl-rt721-4ch-wov-multi,sof-ptl-rt722-4ch-wov-multi, and WCL equivalents.Toolchain & Compiler Fixes:
CALL8out-of-range relocation in the libc shim by unifying.textsection placement.ibuffer_allocator.ccto support both upstream TFLM and trees carrying Xtensa ICF vtable fixes.Validation
test_startandwov_blocking_read.alsatplg.