pulp::format::ProcessorNode Class Reference

Runs a real pulp::format::Processor as a node inside the routed graph. More...

#include <processor_node_adapter.hpp>

Public Member Functions
	ProcessorNode (Processor &processor) noexcept
	Wrap a non-owning processor.
bool	prepare (const PrepareContext &context)
	Off-RT preparation: prepares the wrapped processor for the given audio configuration and sizes the block-adapter scratch so process_binding() stays allocation-free.
bool	release () noexcept
void	set_audio_input_provider (AudioInputBusProvider provider) noexcept
	Install a host-owned route provider before prepare().
GraphRuntimeNodeBinding	binding (graph::NodeId node_id, bool required=true) noexcept
	Build the only valid graph binding for this adapter.
Processor &	processor () noexcept
const Processor &	processor () const noexcept

Static Public Member Functions
static bool	process_binding (ProcessBlock &block, const GraphRuntimeNodeProcessContext &ctx, void *user_data) noexcept
	Routed node binding entry point.

Detailed Description

Runs a real pulp::format::Processor as a node inside the routed graph.

ProcessorNode is the in-process adapter between a routed graph node and the legacy Processor::process() ABI. It owns nothing about routing: the graph's GraphRuntimeExecutor still gathers a node's inputs and assigns its output slots; ProcessorNode only translates the per-node routed I/O the executor hands it into a ProcessBlock and forwards to process_processor_block(). There is no second routing path — a ProcessorNode is wired into a snapshot exactly like any other GraphRuntimeNodeBinding.

The binding attaches gathered sparse parameter events, MIDI, and borrowed dense modulation lanes through one EventBlock. Legacy processors still enter through process_processor_block(); processors that declare dense consumption enter through Processor::process_block().

Lifetime: the wrapped Processor is non-owning and must outlive the node and every routed block that references it. prepare() runs off the realtime thread and must precede the first process_binding() call.

Constructor & Destructor Documentation

ProcessorNode()

pulp::format::ProcessorNode::ProcessorNode ( Processor & processor )

inlineexplicitnoexcept

Wrap a non-owning processor.

The processor must already have its StateStore bound (see HeadlessHost / format adapters) before prepare().

Member Function Documentation

binding()

GraphRuntimeNodeBinding pulp::format::ProcessorNode::binding ( graph::NodeId node_id, bool required = true )

noexcept

Build the only valid graph binding for this adapter.

The cached capability selects dense views or the legacy event expansion.

prepare()

bool pulp::format::ProcessorNode::prepare

(

const PrepareContext &

context

)

Off-RT preparation: prepares the wrapped processor for the given audio configuration and sizes the block-adapter scratch so process_binding() stays allocation-free.

Idempotent enough to re-run when the configuration changes; call with the audio thread stopped. Call release() before re-preparing this non-owning adapter.

process_binding()

bool pulp::format::ProcessorNode::process_binding ( ProcessBlock & block, const GraphRuntimeNodeProcessContext & ctx, void * user_data )

staticnoexcept

Routed node binding entry point.

user_data must be the owning ProcessorNode. Reads the executor's gathered mono input view (ctx.node_inputs) and writes this node's mono output view (ctx.node_outputs) by aliasing them into a block-local ProcessBlock over the node's reused bus set, then forwards to process_processor_block(). Allocation-free per block and only valid on the routed path (ctx.routed == true).

processor() [1/2]

const Processor & pulp::format::ProcessorNode::processor ( ) const

inlinenoexcept

processor() [2/2]

Processor & pulp::format::ProcessorNode::processor ( )

inlinenoexcept

release()

bool pulp::format::ProcessorNode::release ( )

noexcept

set_audio_input_provider()

void pulp::format::ProcessorNode::set_audio_input_provider ( AudioInputBusProvider provider )

inlinenoexcept

Install a host-owned route provider before prepare().

The handle is copied as a small immutable POD and read from the audio thread without allocation or locking. Clearing it requires the audio thread to be stopped and the node to be prepared again.

---

The documentation for this class was generated from the following file:

• /home/runner/work/pulp/pulp/core/format/include/pulp/format/processor_node_adapter.hpp