Skip to content

README.md

Latest commit

 

History

History
88 lines (60 loc) · 4.21 KB

README.md

File metadata and controls

88 lines (60 loc) · 4.21 KB

Module Implementing Interfaces (module_adapter/module)

The module directory contains the concrete module_interface implementations that bridge the core SOF Component API with various audio processing algorithms. These adapters translate the generic initialization, parameter configuration, and process loop calls into the specific operational sequences required by different module frameworks.

The directory primarily houses three distinct module adapters:

  1. Generic Core Adapter (generic.c)
  2. Modules (IADK/Shim) Adapter (modules.c)
  3. Cadence DSP Codec Adapter (cadence.c, cadence_ipc4.c)

1. Generic Core Adapter (generic.c)

The Generic adapter provides the default runtime container for typical native SOF processing modules (e.g., volume, eq, src).

It implements the full struct module_interface contract:

  • Memory Management: It intercepts memory allocation mappings using mod_balloc_align and tracks memory requests in a module-specific resource pool (module_resource). When the module goes out of scope, the framework garbage-collects any leaked allocations automatically via mod_free_all().
  • Configuration Handling: Manages large blob configuration messages across multiple IPC fragments (module_set_configuration). It allocates memory for runtime_params until the blob is fully assembled, then triggers the underlying algorithm with the completed struct.
  • State Machine Enforcement: It wraps process_audio_stream and process_raw_data calls to verify the module is in either MODULE_IDLE or MODULE_PROCESSING states before execution.

2. Modules (IADK Shim) Adapter (modules.c)

The modules.c base is an extension adapter designed specifically to run Intel Audio Development Kit (IADK) 3rd party algorithms.

Unlike the generic modules, the IADK modules are object-oriented C++ architectures linked into a separate library (module_adapter/iadk). This file acts as the primary C entry point wrapper.

  • It utilizes the iadk_wrapper_* C-bridge functions to invoke methods on the C++ intel_adsp::ProcessingModuleInterface classes.
  • It exposes the standard DECLARE_MODULE_ADAPTER(processing_module_adapter_interface) that is bound to the SOF pipeline.

3. Cadence Codec Adapter (cadence.c)

This is a highly specialized adapter used for integrating Xtensa Audio (XA) codecs from Cadence (e.g., MP3, AAC, Vorbis, SBC, DAB).

The cadence.c implementation maps the standard SOF pipeline controls into the Cadence memory buffer management and synchronous execution models.

graph TD
    subgraph SOF Initialization
        INIT["cadence_codec_init"]
        CONFIG["cadence_configure_codec_params"]
        RESOLVE["cadence_codec_resolve_api"]
    end

    subgraph Cadence Library Init XA API
        MEMTABS["cadence_codec_init_memory_tables"]
        GETSIZE["XA_API_CMD_GET_MEM_INFO_SIZE"]
        SETPTR["XA_API_CMD_SET_MEM_PTR"]
    end

    subgraph Data Processing Loop
        PROC["cadence_codec_process_data"]
        FILL["XA_API_CMD_SET_INPUT_BYTES"]
        EXEC["XA_API_CMD_EXECUTE"]
        GETOUT["XA_API_CMD_GET_OUTPUT_BYTES"]
    end

    INIT --> RESOLVE
    RESOLVE --> CONFIG
    CONFIG --> MEMTABS

    MEMTABS -. Iterates Memory Types .-> GETSIZE
    GETSIZE -. Assigns Buffers .-> SETPTR

    SETPTR ==>|Pipeline Trigger| PROC

    PROC --> FILL
    FILL --> EXEC
    EXEC --> GETOUT
Loading

Cadence Memory Tables

Unlike standard modules that directly read from a sof_source API, Cadence codecs require their memory isolated exclusively into exact predefined chunks categorized by "type". cadence_codec_init_memory_tables iterates through the codec's hardware definition to construct these memory areas:

  • XA_MEMTYPE_INPUT
  • XA_MEMTYPE_OUTPUT
  • XA_MEMTYPE_SCRATCH
  • XA_MEMTYPE_PERSIST

The SOF adapter allocates tracking structures via mod_alloc_align for each of these mandatory regions prior to audio playback.

Execution Wrapper

During cadence_codec_process(), the adapter:

  1. Performs source_get_data and mechanically copies audio bytes into the isolated XA_MEMTYPE_INPUT buffer.
  2. Invokes the Xtensa Audio codec API (XA_API_CMD_EXECUTE).
  3. Reads the produced byte count and copies them back out from XA_MEMTYPE_OUTPUT into the sof_sink.