[lgirdwood]

Summary

This series contains the remaining core, topology, and tooling updates for ffmpeg_dec compressed audio decode support on top of the merged ffmpeg_dec integration (#11284 / #11343):

1. audio: src: align the SRC delay line allocation to SOF_FRAME_BYTE_ALIGN
   • Use mod_alloc_align(..., SOF_FRAME_BYTE_ALIGN) for cd->delay_lines so SIMD circular buffer accesses in the SRC filter do not fault when ffmpeg_dec is paired with an SRC stage.
2. platform: cavs: initialize DP scheduler on primary core
   • Call scheduler_dp_init() unconditionally during platform_init() on CAVS/ACE platforms when CONFIG_SOF_USERSPACE_APPLICATION=y so DP modules (such as ffmpeg_dec) can be scheduled on core 0 as well as secondary cores.
3. topology2: compr-ffmpeg: use COMPR_DECODER_CORE_ID and add Opus target
   • Parameterize the ffmpeg_dec pipeline and component core_id via COMPR_DECODER_CORE_ID, size COMPR_FFMPEG_DEC_HEAP to 256 KiB to cover the 64 KiB PCM staging buffer and decoder working set, and add the sof-hda-generic-ffmpeg-opus-compr build target.
4. topology2: add ffmpeg_dec benchmark test topology
   • Add the playback benchmark topology (ffmpeg_dec32, HDA and SoundWire) with DP scheduling and decoder stack/heap budgets for ffmpeg_dec.
5. tools: test: add ffmpeg_dec FLAC test vector generator script
   • Add tools/test/audio/ffmpeg_dec_prepare.sh referenced by src/audio/ffmpeg_dec/TESTING.md to generate raw FLAC frames, 34-byte STREAMINFO extradata, and reference S32_LE PCM vectors.
6. audio: host: align remaining compress drain bytes to dma_copy_align
   • Round the remaining committed compress drain byte count in host_get_copy_bytes_normal() up to MAX(dma_sample_bytes, hd->dma_copy_align) for the DMA transfer/reload, and clamp the bytes copied into the downstream sink in host_common_update() to the actual payload remainder (consuming any trailing DMA padding from the source DMA buffer).
7. audio: ffmpeg_dec: strip container header before lazy backend configure
   • Embed extradata inline in struct ffmpeg_dec_comp_data (avoiding mod_alloc() on the user-mode DP thread) and strip the container header before lazy backend configuration so stream-embedded setup headers (such as FLAC STREAMINFO) are extracted before avcodec_open2().

[Copilot]

🟡 Changes recommended

Decoder heap undersizing, excess DMA bytes published downstream, and backend initialization on the shared LL thread risk allocation failures, payload corruption, and audio stalls; benchmark routing and SoundWire integration are also incomplete.

5 open issues

• Commit only the actual payload remainder after aligned DMA reads · New
• Increase the Opus decoder heap to cover staging and backend allocations · New
• Set decoder scheduling domains and resource budgets · New
• Add SoundWire decoder integration or exclude SoundWire benchmark targets · New
• Route playback through the decoder for bit-exact comparison · New

What changed in this PR

This PR updates SOF’s ffmpeg_dec runtime integration, topology configuration, and test tooling for compressed audio decoding.

Changes:

• Adjusts SRC allocation alignment, primary-core DP scheduler initialization, and compressed-stream drain alignment.
• Stores codec extradata inline and extracts container headers before lazy backend configuration.
• Parameterizes decoder core selection, adds an Opus target and benchmark fragments, and provides FLAC test-vector preparation.

File	Description
tools/​topology/​topology2/​platform/​intel/​compr-ffmpeg.conf	Uses configurable decoder core selection.
tools/​topology/​topology2/​platform/​intel/​compr-default.conf	Updates comments to cover both decoder implementations.
tools/​topology/​topology2/​include/​bench/​ffmpeg_dec_s32.conf	Adds S32 decoder benchmark instances.
tools/​topology/​topology2/​include/​bench/​ffmpeg_dec_s24.conf	Adds S24 decoder benchmark instances.
tools/​topology/​topology2/​include/​bench/​ffmpeg_dec_s16.conf	Adds S16 decoder benchmark instances.
tools/​topology/​topology2/​include/​bench/​ffmpeg_dec_route.conf	Adds playback and capture benchmark routes.
tools/​topology/​topology2/​include/​bench/​ffmpeg_dec_controls_playback.conf	Adds commented playback control placeholders.
tools/​topology/​topology2/​include/​bench/​ffmpeg_dec_controls_capture.conf	Adds commented capture control placeholders.
tools/​topology/​topology2/​development/​tplg-targets.cmake	Registers the FFmpeg Opus compressed-playback target.
tools/​topology/​topology2/​development/​tplg-targets-bench.cmake	Adds FFmpeg to S32 benchmark target generation.
tools/​topology/​topology2/​cavs-benchmark-hda.conf	Includes the decoder class and S32 benchmark selection.
tools/​test/​audio/​ffmpeg_dec_prepare.sh	Generates FLAC data, STREAMINFO, and reference PCM.
src/​platform/​intel/​cavs/​platform.c	Initializes the enabled DP scheduler during platform startup.
src/​audio/​src/​src_common.c	Requests eight-byte alignment for SRC delay lines.
src/​audio/​host-zephyr.c	Rounds drain remainder to DMA copy alignment.
src/​audio/​ffmpeg_dec/​ffmpeg_dec.h	Replaces the extradata pointer with inline storage.
src/​audio/​ffmpeg_dec/​ffmpeg_dec.c	Updates extradata handling and strips headers before backend configuration.

_{Review effort: Balanced}

[lgirdwood]

Addressed review comments:

1. src/audio/src/src_common.c: Shortened the comment to a single line and used SOF_FRAME_BYTE_ALIGN in mod_alloc_align().
2. src/audio/host-zephyr.c: Kept the DMA transfer/reload aligned to hd->dma_copy_align in host_get_copy_bytes_normal(), and clamped the bytes copied into the downstream sink in host_common_update() to the actual committed payload remainder (consuming any trailing DMA padding from the source DMA buffer).
3. tools/topology/topology2/platform/intel/compr-default.conf & tplg-targets.cmake: Updated COMPR_FFMPEG_DEC_HEAP to 262144 (256 KiB) for the default and Opus compress targets to cover the 64 KiB PCM staging buffer, input bounce buffer, and decoder working set.
4. tools/topology/topology2/include/bench/ffmpeg_dec_s{16,24,32}.conf: Configured scheduler_domain "DP", stack_bytes_requirement 65536, and heap_bytes_requirement 262144 on the benchmark decoder widget.
5. tools/topology/topology2/cavs-benchmark-sdw.conf: Added <ffmpeg_dec.conf> and the ffmpeg_dec32 BENCH_CONFIG entry for the SoundWire benchmark targets.
6. tools/topology/topology2/include/bench/ffmpeg_dec_route.conf: Removed the second decoder instance from the microphone capture path and routed capture directly.

[Copilot]

🟡 Changes recommended

Two critical issues remain: extradata pointer checks break warning-as-error builds, and successful padded DMA copies can leave padding unconsumed and the read pointer misaligned.

2 open issues

• Remove obsolete extradata pointer checks · New
• Consume padding after successful DMA copy · New

5 resolved since last review

• Set decoder scheduling domains and resource budgets
• Increase the Opus decoder heap to cover staging and backend allocations
• Commit only the actual payload remainder after aligned DMA reads
• Route playback through the decoder for bit-exact comparison
• Add SoundWire decoder integration or exclude SoundWire benchmark targets

_{Review effort: Balanced}

[lgirdwood]

Updated to address the review comments and tested on damsel-fly (Panther Lake / ACE 3.0) built with the local LLVM + HiFi compiler:

1. src/audio/ffmpeg_dec/ffmpeg_dec.c & ffmpeg_dec-ffmpeg.c: Removed redundant cd->extradata pointer truthiness checks after switching extradata to an inline array, checking cd->extradata_size only.
2. src/audio/host-zephyr.c: Added a descriptive comment explaining the compress drain EOS clamping in host_common_update(), and changed the return check on dma_buffer_copy_from() to if (ret >= 0 && bytes > copy_bytes) before consuming trailing DMA alignment padding.
3. Additional fixes uncovered during damsel-fly testing:
   • Use source_drop_data() instead of source_release_data() when consuming container bytes in ffmpeg_dec_skip_container() (since the preceding source_release_data(src, 0) resets src->requested_read_frag_size = 0, causing subsequent source_release_data(src, N) calls to return -ENODATA without advancing the read pointer).
   • Free cd->eos_msg with mod_ipc_msg_free(mod, cd->eos_msg) instead of ipc_msg_free() in ffmpeg_dec_free() to match mod_ipc_msg_w_ext_init() and avoid a double-free on module teardown.
4. Hardware verification on damsel-fly (PTL / ACE 3.0):
   • cplay -v -c 0 -d 50 /root/ffmpeg_test/test_short.mp3 -> play_samples: exit success, Finish Playing.... Close Normally
   • cplay -v -I FLAC -c 0 -d 50 /root/ffmpeg_test/test.flac -> play_samples: exit success, Finish Playing.... Close Normally

[lgirdwood]

The last patch needs rebased/squashed with the others. Lets also put our per codec logic in per codec files to keep things simple.

[lgirdwood]

Updated PR to address review comments:

• Rebased onto latest origin/main (80b14b070) and squashed the ffmpeg_dec changes.
• Split per-codec logic (decoder lookup, max frame samples, extradata validation, and container stripping) into dedicated per-codec files behind struct ffmpeg_dec_codec_ops:
  • src/audio/ffmpeg_dec/ffmpeg_dec-flac.c
  • src/audio/ffmpeg_dec/ffmpeg_dec-mp3.c
  • src/audio/ffmpeg_dec/ffmpeg_dec-aac.c
  • src/audio/ffmpeg_dec/ffmpeg_dec-opus.c
• Updated the lazy avcodec_open2() comments in ffmpeg_dec.c and ffmpeg_dec-ffmpeg.c to reflect the current per-core IPC/IDC dispatch flow.
• Retested FLAC, MP3, and AAC compressed playback (cplay -c 0 -d 50) on damsel-fly (PTL / ACE 3.0) built with the local LLVM + HiFi compiler.

[Copilot]

🟡 Changes recommended

Critical issues remain: undersized per-instance heaps cause bind/prepare failures, and explicitly requested codecs disabled in Kconfig silently fall back to an incompatible default decoder.

3 open issues

• Reject explicitly requested codecs without an available decoder · New
• Increase userspace heap budget for decoder buffers and state · New
• Remove obsolete extradata pointer checks

1 resolved since last review

• Consume padding after successful DMA copy

_{Review effort: Balanced}

[lgirdwood]

Updated PR to address the latest review comments and re-trigger CI:

1. src/audio/ffmpeg_dec/ffmpeg_dec.c:
   • Reject explicitly requested IPC snd_codec IDs without an enabled ffmpeg_dec_codec_ops entry with -ENOTSUP in ffmpeg_dec_init() instead of silently falling back to the default decoder.
   • Fix max_frame_bytes calculation in ffmpeg_dec_prepare() to avoid multiplying by out_channels twice (cd->out_frame_bytes from sink_get_frame_bytes(sink) already includes out_channels).
2. tools/topology/topology2/:
   • Increase COMPR_FFMPEG_DEC_HEAP in platform/intel/compr-default.conf and development/tplg-targets.cmake (Opus override) and heap_bytes_requirement in include/bench/ffmpeg_dec_s{16,24,32}.conf from 262144 (256 KiB) to 524288 (512 KiB) so per-instance userspace vregions cover the 192 KiB DP output ring, 64 KiB PCM staging buffer, input/packet scratch buffers, and decoder working set.
3. Rebased onto latest origin/main (414a8a228) to re-run the two CI jobs that hit GitHub Actions runner No space left on device errors.