MayaFlux::Kriya::BufferPipeline Class Reference

Coroutine-based execution engine for composable, multi-strategy buffer processing. More...

#include <BufferPipeline.hpp>

Inheritance diagram for MayaFlux::Kriya::BufferPipeline:

Collaboration diagram for MayaFlux::Kriya::BufferPipeline:

Classes
struct	BranchInfo
struct	GraphicsReader

Public Member Functions
BufferPipeline &	branch_if (std::function< bool(uint32_t)> condition, const std::function< void(BufferPipeline &)> &branch_builder, bool synchronous=false, uint64_t samples_per_operation=1)
	Add conditional branch to the pipeline.
	BufferPipeline ()=default
	BufferPipeline (Vruta::TaskScheduler &scheduler, std::shared_ptr< Buffers::BufferManager > buffer_manager=nullptr, std::shared_ptr< IO::IOManager > io_manager=nullptr)
BufferPipeline &	capture_timing (Vruta::DelayContext mode)
	Set timing mode for capture phase.
void	end ()
	End a running pipeline now.
void	execute_buffer_rate (uint64_t max_cycles=0)
	Execute pipeline synchronized to audio hardware cycle boundaries.
void	execute_continuous ()
	Start continuous execution of the pipeline.
void	execute_for_cycles (uint64_t cycles=0)
	Execute the pipeline for a specified number of cycles.
void	execute_frame_rate (uint64_t max_cycles=0, uint64_t frames_per_operation=0)
	Execute the pipeline at graphics frame rate.
void	execute_once ()
	Execute the pipeline for a single cycle.
void	execute_scheduled (uint64_t max_cycles=0, uint64_t samples_per_operation=1)
	Execute pipeline with sample-accurate timing between operations.
void	execute_scheduled_at_rate (uint32_t max_cycles=0, double seconds_per_operation=1)
	Execute pipeline with real-time rate control.
uint32_t	get_current_cycle () const
	Get the current cycle number.
std::shared_ptr< Kakshya::DynamicVideoStream >	get_graphics_stream () const
	The dynamic video stream of the first operation that writes one.
ExecutionStrategy	get_strategy () const
	Get current execution strategy.
bool	has_pending_data () const
	Check if any operations have pending data ready for processing.
bool	is_running () const
	True from an execute call until the pipeline ends or runs out of cycles.
void	mark_data_consumed (uint32_t operation_index)
	Execute pipeline synchronized to audio hardware cycle boundaries.
BufferPipeline &	on_complete (std::function< void()> cb)
	Register a callback fired once when pipeline execution ends.
BufferPipeline &	operator>> (BufferOperation &&operation)
	Chain an operation to the pipeline.
BufferPipeline &	parallel (std::initializer_list< BufferOperation > operations)
	Execute operations in parallel within the current cycle.
BufferPipeline &	process_timing (Vruta::DelayContext mode)
	Set timing mode for process phase.
void	stop_continuous ()
	Stop continuous execution of the pipeline.
BufferPipeline &	with_lifecycle (std::function< void(uint32_t)> on_cycle_start, std::function< void(uint32_t)> on_cycle_end)
	Set lifecycle callbacks for cycle management.
BufferPipeline &	with_strategy (ExecutionStrategy strategy)
	Set the execution strategy for this pipeline.
	~BufferPipeline ()

Static Public Member Functions
static std::shared_ptr< BufferPipeline >	create (Vruta::TaskScheduler &scheduler, std::shared_ptr< Buffers::BufferManager > buffer_manager=nullptr, std::shared_ptr< IO::IOManager > io_manager=nullptr)

Private Types
enum class	DataState : uint8_t { EMPTY , READY , CONSUMED , EXPIRED }

Private Member Functions
void	await_timing (Vruta::DelayContext mode, uint64_t units)
void	capture_operation (BufferOperation &op, uint64_t cycle)
void	cleanup_completed_branches ()
void	cleanup_expired_data ()
std::shared_ptr< Vruta::SoundRoutine >	dispatch_branch_async (BranchInfo &branch, uint64_t cycle)
std::vector< std::shared_ptr< Vruta::Routine > >	dispatch_cycle_branches (bool frame_rate)
std::shared_ptr< Vruta::GraphicsRoutine >	dispatch_frame_branch (BranchInfo &branch)
void	execute_capture_phase (uint64_t cycle_base)
Vruta::GraphicsRoutine	execute_frame_internal (uint64_t max_cycles, uint64_t frames_per_operation)
Vruta::GraphicsRoutine	execute_frame_phased (uint64_t max_cycles, uint64_t frames_per_operation)
Vruta::GraphicsRoutine	execute_frame_streaming (uint64_t max_cycles, uint64_t frames_per_operation)
Vruta::SoundRoutine	execute_internal (uint64_t max_cycles, uint64_t samples_per_operation)
Vruta::SoundRoutine	execute_parallel (uint64_t max_cycles, uint64_t samples_per_operation)
Vruta::SoundRoutine	execute_phased (uint64_t max_cycles, uint64_t samples_per_operation)
void	execute_process_phase (uint64_t cycle)
Vruta::SoundRoutine	execute_reactive (uint64_t max_cycles, uint64_t samples_per_operation)
Vruta::SoundRoutine	execute_streaming (uint64_t max_cycles, uint64_t samples_per_operation)
void	finish_cycle ()
void	flow_through (size_t index, uint64_t cycle)
bool	has_immediate_routing (const BufferOperation &op) const
bool	operation_due (const BufferOperation &op) const
uint32_t	operation_iterations (const BufferOperation &op) const
void	prepare_displays ()
void	process_dispatch (BufferOperation &op, uint64_t cycle)
void	process_fuse (BufferOperation &op, uint64_t cycle)
void	process_load (BufferOperation &op)
void	process_modify (BufferOperation &op, uint64_t cycle)
void	process_operation (BufferOperation &op, uint64_t cycle)
void	process_route (BufferOperation &op)
void	process_transform (BufferOperation &op, uint64_t cycle)
void	queue_graphics_write (BufferOperation &op, const std::shared_ptr< Buffers::VKBuffer > &target, const Kakshya::DataVariant &data)
std::optional< Kakshya::DataVariant >	read_graphics_buffer (BufferOperation &op, const std::shared_ptr< Buffers::VKBuffer > &buffer)
void	reset_accumulated_data ()
void	reset_cycle_state ()
bool	route_image (BufferOperation &op)
void	start_cycle ()

Static Private Member Functions
static bool	any_active (const std::vector< std::shared_ptr< Vruta::Routine > > &tasks)

Private Attributes
std::shared_ptr< BufferPipeline >	m_active_self
std::vector< std::shared_ptr< Vruta::Routine > >	m_branch_tasks
std::vector< BranchInfo >	m_branches
std::shared_ptr< Buffers::BufferManager >	m_buffer_manager
Vruta::DelayContext	m_capture_timing { Vruta::DelayContext::BUFFER_BASED }
bool	m_continuous_execution { false }
uint64_t	m_current_cycle { 0 }
std::function< void(uint32_t)>	m_cycle_end_callback
std::function< void(uint32_t)>	m_cycle_start_callback
std::vector< DataState >	m_data_states
ExecutionStrategy	m_execution_strategy { ExecutionStrategy::PHASED }
std::shared_ptr< IO::IOManager >	m_io_manager
uint64_t	m_max_cycles { 0 }
std::function< void()>	m_on_complete
std::unordered_map< BufferOperation *, Kakshya::DataVariant >	m_operation_data
std::vector< BufferOperation >	m_operations
Vruta::DelayContext	m_process_timing { Vruta::DelayContext::SAMPLE_BASED }
std::map< std::pair< const BufferOperation *, const Buffers::VKBuffer * >, GraphicsReader >	m_readers
std::shared_ptr< Vruta::Routine >	m_routine
Vruta::TaskScheduler *	m_scheduler = nullptr

Detailed Description

Coroutine-based execution engine for composable, multi-strategy buffer processing.

BufferPipeline provides a flexible framework for orchestrating complex data flow patterns through declarative operation chains. It supports multiple execution strategies (phased, streaming, parallel, reactive), sophisticated data accumulation modes, and sample-accurate timing coordination via the coroutine scheduler.

Core Concepts:

• Operations: Composable units (capture, transform, route, modify, fuse) chained via >>
• Execution Strategies: PHASED (capture-then-process), STREAMING (immediate flow-through), PARALLEL (concurrent captures), REACTIVE (data-driven)
• Capture Modes: TRANSIENT (single), ACCUMULATE (concatenate), CIRCULAR (rolling buffer), WINDOWED (overlapping windows), TRIGGERED (conditional)
• Timing Control: Buffer-rate synchronization, sample-accurate delays, or immediate execution

Simple Capture & Route:

auto pipeline = BufferPipeline::create(*scheduler, buffer_manager);
*pipeline
    >> BufferOperation::capture_from(input_buffer)
        .for_cycles(1)
    >> BufferOperation::route_to_buffer(output_buffer);
 
pipeline->execute_buffer_rate(100);  // Run for 100 audio buffer cycles

Accumulation & Batch Processing (PHASED strategy):

auto pipeline = BufferPipeline::create(*scheduler, buffer_manager)
    ->with_strategy(ExecutionStrategy::PHASED)
    ->capture_timing(Vruta::DelayContext::BUFFER_BASED);
 
*pipeline
    >> BufferOperation::capture_from(audio_buffer)
        .for_cycles(20)  // Captures 20 times, concatenates into single buffer
    >> BufferOperation::transform([](const auto& data, uint32_t cycle) {
        const auto& accumulated = std::get<std::vector<double>>(data);
        // Process 20 * buffer_size samples as one batch
        return apply_batch_fft(accumulated);
    })
    >> BufferOperation::route_to_container(output_stream);
 
pipeline->execute_buffer_rate();

Real-Time Streaming Modification (STREAMING strategy):

auto pipeline = BufferPipeline::create(*scheduler, buffer_manager)
    ->with_strategy(ExecutionStrategy::STREAMING);
 
*pipeline
    >> BufferOperation::capture_from(audio_buffer).for_cycles(1)
    >> BufferOperation::modify_buffer(audio_buffer, [noise](auto buf) {
        auto& data = buf->get_data();
        for (auto& sample : data) {
            sample *= noise->process_sample();  // Apply effect in-place
        }
    }).as_streaming();  // Processor stays attached across cycles
 
pipeline->execute_buffer_rate();  // Runs continuously

Circular Buffer for Rolling Analysis:

*pipeline
    >> BufferOperation::capture_from(input_buffer)
        .for_cycles(100)
        .as_circular(2048)  // Maintains last 2048 samples
    >> BufferOperation::transform([](const auto& data, uint32_t cycle) {
        const auto& history = std::get<std::vector<double>>(data);
        return analyze_recent_trends(history);  // Always sees last 2048 samples
    });

Windowed Capture with Overlap:

*pipeline
    >> BufferOperation::capture_from(input_buffer)
        .for_cycles(50)
        .with_window(1024, 0.5f)  // 1024 samples, 50% overlap
    >> BufferOperation::transform([](const auto& data, uint32_t cycle) {
        const auto& window = std::get<std::vector<double>>(data);
        return apply_hann_window_and_fft(window);
    });

Multi-Source Fusion:

*pipeline
    >> BufferOperation::fuse_data(
        {mic_buffer, synth_buffer, file_buffer},
        [](std::vector<Kakshya::DataVariant>& sources, uint32_t cycle) {
            // Combine three audio sources with custom mixing
            return mix_sources(sources, {0.5, 0.3, 0.2});
        },
        output_buffer);

Conditional Branching:

pipeline->branch_if(
    [](uint32_t cycle) { return cycle % 16 == 0; },  // Every 16 cycles
    [](BufferPipeline& branch) {
        branch >> BufferOperation::dispatch_to([](const auto& data, uint32_t cycle) {
            save_analysis_snapshot(data, cycle);
        });
    },
    true  // Synchronous - wait for branch to complete
);

Per-Iteration Routing (Immediate Routing):

// ROUTE directly after CAPTURE → routes each iteration immediately
*pipeline
    >> BufferOperation::capture_from(input_buffer).for_cycles(20)
    >> BufferOperation::route_to_container(stream);  // Writes 20 times (streaming output)
 
// ROUTE after TRANSFORM → routes accumulated result once
*pipeline
    >> BufferOperation::capture_from(input_buffer).for_cycles(20)
    >> BufferOperation::transform(process_fn)
    >> BufferOperation::route_to_container(stream);  // Writes 1 time (batch output)

Lifecycle Callbacks:

pipeline->with_lifecycle(
    [](uint32_t cycle) { std::cout << "Cycle " << cycle << " start\n"; },
    [](uint32_t cycle) { std::cout << "Cycle " << cycle << " end\n"; }
);

Execution Modes:

• execute_buffer_rate(N): Synchronized to audio buffer boundaries for N cycles
• execute_continuous(): Runs indefinitely until stopped
• execute_for_cycles(N): Runs exactly N cycles then stops
• execute_once(): Single cycle execution
• execute_scheduled(N, samples): With sample-accurate delays between operations

Strategy Selection:

• PHASED (default): Capture phase completes, then process phase runs
  • Best for: batch analysis, accumulation, FFT processing
• STREAMING: Operations flow through immediately, minimal latency
  • Best for: real-time effects, low-latency processing, modify_buffer
• PARALLEL: Multiple captures run concurrently (TODO)
  • Best for: multi-source synchronized capture
• REACTIVE: Data-driven execution when inputs available (TODO)
  • Best for: event-driven workflows, complex dependencies

See also

BufferOperation For operation types and configuration

BufferCapture For capture modes and data accumulation strategies

Vruta::TaskScheduler For coroutine scheduling and timing

ExecutionStrategy For execution coordination patterns

Definition at line 171 of file BufferPipeline.hpp.

---

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

• BufferPipeline.hpp
• BufferPipeline.cpp
• BufferPipelineBranches.cpp
• BufferPipelineCapture.cpp
• BufferPipelineExecution.cpp
• BufferPipelineOperations.cpp