Repository navigation
Conversation
Contributor
There was a problem hiding this comment.
🟡 Changes recommended
Unresolved critical defects in codec configuration, sample-rate handling, and buffering can produce incorrect audio, silently discard frames, or stall playback indefinitely.
11 open issues
What changed in this PR
Extends SOF’s FFmpeg-based compressed playback and capture with SBC/mSBC, G.722, and aptX support, building on the Phase 1 lossless work in #11427.
Changes:
- Adds codec implementations, PCM conversion helpers, and unparsed-stream handling.
- Integrates codec selection into Kconfig, built-in and loadable-module builds, and firmware capability reporting.
- Adds development topology targets for Bluetooth, voice, and lossless codecs.
| File | Description |
|---|---|
tools/topology/topology2/development/tplg-targets.cmake |
Adds compressed playback and capture targets. |
src/kernel/audio/base_fw.c |
Advertises added codec capabilities and BESPOKE aliases. |
src/include/ipc/compress_params.h |
Defines SBC, G.722, and aptX codec IDs. |
src/audio/ffmpeg_dec/llext/CMakeLists.txt |
Includes added codecs in loadable-module builds. |
src/audio/ffmpeg_dec/Kconfig |
Adds decoder and encoder configuration options. |
src/audio/ffmpeg_dec/ffmpeg.cmake |
Enables native codecs and reduces FLAC encoder allocations. |
src/audio/ffmpeg_dec/ffmpeg-encode.c |
Extends encoder dispatch, conversion, and parameter handling. |
src/audio/ffmpeg_dec/ffmpeg-enc-sbc.c |
Implements SBC/mSBC encoding and capture conversion. |
src/audio/ffmpeg_dec/ffmpeg-enc-mp3.c |
Keeps MP3 rate-control validation codec-specific. |
src/audio/ffmpeg_dec/ffmpeg-enc-g722.c |
Implements mono G.722 encoding. |
src/audio/ffmpeg_dec/ffmpeg-enc-flac.c |
Implements FLAC encoding with a stream header. |
src/audio/ffmpeg_dec/ffmpeg-enc-aptx.c |
Implements stereo aptX encoding. |
src/audio/ffmpeg_dec/ffmpeg-enc-alac.c |
Implements ALAC encoding with a setup cookie. |
src/audio/ffmpeg_dec/ffmpeg-enc-aac.c |
Keeps AAC rate-control validation codec-specific. |
src/audio/ffmpeg_dec/ffmpeg_dec.h |
Extends codec interfaces and stream state. |
src/audio/ffmpeg_dec/ffmpeg_dec.c |
Registers codecs and handles selection and reset state. |
src/audio/ffmpeg_dec/ffmpeg_dec-sbc.c |
Defines SBC/mSBC decoder frame sizing. |
src/audio/ffmpeg_dec/ffmpeg_dec-g722.c |
Bounds G.722 decode packet sizes. |
src/audio/ffmpeg_dec/ffmpeg_dec-ffmpeg.c |
Adds codec mappings and unparsed-packet buffering. |
src/audio/ffmpeg_dec/ffmpeg_dec-convert.c |
Supports mono output duplication across sink channels. |
src/audio/ffmpeg_dec/ffmpeg_dec-aptx.c |
Defines aptX packet alignment and size limits. |
src/audio/ffmpeg_dec/ffmpeg_dec-alac.c |
Parses ALAC containers, cookies, and frame consumption. |
src/audio/ffmpeg_dec/CMakeLists.txt |
Includes added codecs in built-in builds. |
Review effort: Balanced
lgirdwood
commented
Nov 10, 2026
Advertise SND_AUDIOCODEC_ALAC decode and SND_AUDIOCODEC_FLAC / SND_AUDIOCODEC_ALAC encode capabilities in base_fw when the corresponding FFmpeg Kconfig options are enabled. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Extend the FFmpeg decode and encode adapter infrastructure to support container-framed and variable-blocksize lossless codecs: - Add frame_bytes_consumed hook and unparsed packet reassembly buffer handling in ffmpeg_dec-ffmpeg.c for codecs without a standalone AVCodecParser (such as ALAC). - Extend ffmpeg_dec_pcm_to_frame() in ffmpeg_dec-convert.c to support AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_S32, and AV_SAMPLE_FMT_S32P sample formats. - Add an optional encoder flush() hook in struct ffmpeg_enc_codec_ops and support native libavcodec encoders in ffmpeg.cmake. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Add Apple Lossless Audio Codec (ALAC, SND_AUDIOCODEC_ALAC) decoder support via libavcodec's integer alac decoder: - Support raw ALAC packets with 36-byte out-of-band magic cookie extradata as well as self-contained Core Audio Format (.caf) streams. - Resize the PCM output staging buffer if the magic cookie's max_samples per frame exceeds the initial allocation. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Add native libavcodec FLAC lossless encoder (SND_AUDIOCODEC_FLAC) support for compress capture streams: - Use fixed integer LPC predictors (FF_LPC_TYPE_FIXED, compression_level 1) with 1152-sample blocks and zero floating-point instructions in the per-frame encode path. - Prepend the 42-byte "fLaC" + STREAMINFO header on stream start and convert input PCM via ffmpeg_dec_pcm_to_frame(). Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Add native libavcodec ALAC lossless encoder (SND_AUDIOCODEC_ALAC)
support for 1..2 channel compress capture streams:
- Configure 1024-sample frame size using pure integer prediction and
adaptive Rice coding.
- Prepend a 64-byte Core Audio Format (.caf) header ("caff" + "desc" +
"kuki" + "data") so captured streams are directly decodable.
Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Add development topology targets for ALAC compress playback and FLAC/ALAC compress capture on CAVS 2.5 and ACE 3.0. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Add SND_AUDIOCODEC_SBC (0x12), SND_AUDIOCODEC_G722 (0x13), and SND_AUDIOCODEC_APTX (0x14) definitions to compress_params.h and advertise them in base_fw.c alongside SND_AUDIOCODEC_BESPOKE when their corresponding Kconfig symbols are enabled. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Extend the FFmpeg decode and encode adapter infrastructure for Bluetooth and wideband voice codecs: - Support SND_AUDIOCODEC_BESPOKE sub-codec dispatch via reserved[0], requested bitrate propagation, and encoder profile selection. - Add limit_packet_size hook in struct ffmpeg_dec_codec_ops and explicit decoder sample rate / G.722 bits_per_codeword initialization in ffmpeg_dec-ffmpeg.c. - Add stateful 13-tap FIR low-pass anti-aliased 3:1 (48->16 kHz) mono decimation helper ffmpeg_dec_pcm_decimate_3to1_s16_mono() in ffmpeg_dec-convert.c. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Add Bluetooth A2DP Subband Codec (SBC) and HFP Wideband Speech (mSBC) decode and encode support (SND_AUDIOCODEC_SBC) via libavcodec's integer sbc codec and parser: - Support standard A2DP SBC (16/32/44.1/48 kHz, 1..2 channels) and HFP mSBC (profile == AV_PROFILE_SBC_MSBC, 16 kHz mono, 120-sample 57-byte frames). - Support 48->16 kHz anti-aliased decimation when capturing 48 kHz pipeline audio into a 16 kHz mSBC stream. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Add ITU-T G.722 7 kHz wideband speech ADPCM decode and encode support (SND_AUDIOCODEC_G722) via libavcodec's integer adpcm_g722 codec: - Support 64/56/48 kbps G.722 playback with 160-byte (320-sample, 20 ms) packet clamping. - Support 16 kHz mono G.722 encoding at 64 kbps, including anti-aliased 3:1 decimation when fed by a 48 kHz capture pipeline. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Add Bluetooth classic aptX stereo audio decode and encode support (SND_AUDIOCODEC_APTX) via libavcodec's integer aptx codec: - Clamp incoming aptX packets to 4-byte codeword boundaries (<= 512 bytes, <= 512 stereo samples per decode call). - Encode 256-sample stereo S32P blocks into 256-byte aptX packets using ffmpeg_dec_pcm_to_frame(). Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
Add development topology targets for SBC, G.722, and aptX compress playback and capture. Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>
lgirdwood
force-pushed
the
topic/ffmpeg-bt-voice
branch
from
November 10, 2026 11:51
0750ea9 to
e789e41
Compare
Contributor
There was a problem hiding this comment.
🟡 Changes recommended
Unresolved buffer-handling issues can stall aptX playback and reject valid ALAC frames, while G.722 setup accepts bitrates the encoder does not honor.
3 open issues
11 resolved since last review
Review effort: Balanced
| return -ENOTSUP; | ||
| } | ||
|
|
||
| return 0; |
Comment on lines
+260
to
+263
| ff->pktbuf_sz = cd->out_frame_bytes ? | ||
| (size_t)cd->out_frame_bytes * 4096 : 65536; | ||
| if (ff->pktbuf_sz < 65536) | ||
| ff->pktbuf_sz = 65536; |
| ff->pktbuf_len = 0; | ||
| return 0; | ||
| } | ||
| if (pkt_len >= cap) { |
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
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.
Summary
Add Phase 2 Bluetooth & Voice Offload codec support to
ffmpeg_dec/ffmpeg_enc(stacked on top of Phase 1 lossless support in #11427):SBC& HFP Wideband SpeechmSBC(SND_AUDIOCODEC_SBC=0x12):CONFIG_FFMPEG_DEC_SBC: nativelibavcodecsbcdecoder + parser (--enable-decoder=sbc --enable-parser=sbc), handling both A2DPSBC(0x9csyncword) and HFPmSBC(0xadsyncword) bitstreams.CONFIG_FFMPEG_ENC_SBC: nativelibavcodecsbcencoder (--enable-encoder=sbc), supporting both standard A2DPSBCand HFP Wideband SpeechmSBC(profile = 1,FF_PROFILE_SBC_MSBC, 16 kHz mono with 3:1 decimation from 48 kHz capture pipelines).G.722Wideband Speech ADPCM (SND_AUDIOCODEC_G722=0x13):CONFIG_FFMPEG_DEC_G722: nativelibavcodecadpcm_g722integer QMF + 2-band ADPCM decoder (--enable-decoder=adpcm_g722), with mono-to-stereo upmixing inffmpeg_dec_convert_frame()when the downstream sink is stereo.CONFIG_FFMPEG_ENC_G722: nativelibavcodecadpcm_g722encoder (--enable-encoder=adpcm_g722), extracting mono S16 samples viaffmpeg_enc_pcm_to_s16_mono().aptX(SND_AUDIOCODEC_APTX=0x14):CONFIG_FFMPEG_DEC_APTX: nativelibavcodecaptxinteger 4-band QMF + ADPCM decoder (--enable-decoder=aptx) with 4-byteblock_alignpacket clamping.CONFIG_FFMPEG_ENC_APTX: nativelibavcodecaptxencoder (--enable-encoder=aptx) with planar S32 conversion viaffmpeg_enc_pcm_to_s32_planar().sof-hda-generic-ffmpeg-{sbc,g722,aptx}-comprandsof-hda-generic-ffmpeg-{sbc,g722,aptx}-enc-comprtargets intools/topology/topology2/development/tplg-targets.cmake.All 6 codecs are pure integer / fixed-point in
libavcodecwith zero floating-point math dependency and ~32–36 KB total image footprint when all 6 are enabled together.Hardware Verification
Verified encode + round-trip decode on both Panther Lake (
ptl/ ACE 3.0, Clang) and Tiger Lake (tgl/ CAVS 2.5, GCC) hardware:SBCencode (comprC0D51) & round-trip decode (comprC0D50)mSBC(profile = 1) encode (comprC0D51) & round-trip decode (comprC0D50)G.722encode (comprC0D51) & round-trip decode (comprC0D50)aptXencode (comprC0D51) & round-trip decode (comprC0D50)ffmpeg -v erroron host.