Struct ConvNetBlock 

pub struct ConvNetBlock {
    pub conv: Conv1dDyn,
    pub bn: BatchNorm1D,
    pub activation: ActivationType,
    pub state: WaveNetLayerState,
    /* private fields */
}

A single ConvNet feed-forward block.

Processing pipeline:

1. Write input frames into the causal ring buffer
2. Run causal Conv1D on the ring buffer → scratch buffer
3. Apply BatchNorm (affine transform) in-place on scratch
4. Apply activation function in-place on scratch
5. Copy scratch to output

Fields

conv: Conv1dDyn

Causal 1D convolution (runtime dimensions).

bn: BatchNorm1D

Batch normalization (pre-fused affine transform).

activation: ActivationType

Activation function applied after batch norm.

state: WaveNetLayerState

Ring buffer state for causal convolution lookback.

Implementations

impl ConvNetBlock

pub fn new( in_ch: usize, out_ch: usize, kernel: usize, dilation: usize, do_bias: bool, activation: ActivationType, alloc_num: usize, ) -> Result<Self>

Creates a new ConvNetBlock with zero-initialized weights.

Weights must be populated via set_conv_weights / set_conv_bias and set_bn_params by the dispatcher.

pub fn receptive_field(&self) -> usize

Returns the receptive field contribution of this block.

pub fn set_conv_weights(&mut self, weights: &[f32])

Loads convolution weights from a flat f32 slice.

pub fn set_conv_bias(&mut self, bias: &[f32])

Loads convolution bias from a flat f32 slice.

pub fn set_bn_params( &mut self, scale: &[f32], offset: &[f32], ) -> Result<(), NamErrorCode>

Loads pre-fused batch norm parameters.

pub unsafe fn process_block( &mut self, input: &[f32], output: &mut [f32], num_frames: usize, )

Public dispatch wrapper that selects the optimal SIMD path.

Safety

Input and output slices must have sizes compatible with the block dimensions: input.len() == num_frames * in_ch, output.len() == num_frames * out_ch.

pub unsafe fn process_block_internal<M: SimdMath>( &mut self, input: &[f32], output: &mut [f32], num_frames: usize, )

SIMD-dispatched processing kernel.

Safety

input.len() must equal num_frames * in_ch. output.len() must equal num_frames * out_ch.

pub fn prewarm(&mut self)

Public prewarm wrapper with SIMD dispatch.

pub unsafe fn prewarm_internal<M: SimdMath>(&mut self)

Fills the conv state buffer with a single frame of silence replicated backward to cover the entire receptive field.

Safety

Must be called via dispatch_simd! macro. The state buffer must be properly allocated and the ring buffer start pointer must be valid.

Trait Implementations

impl Clone for ConvNetBlock

fn clone(&self) -> ConvNetBlock

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.