Enum ScOp 

pub enum ScOp {
    Input {
        id: ValueId,
        name: String,
        ty: ScType,
    },
    Output {
        id: ValueId,
        name: String,
        source: ValueId,
    },
    Constant {
        id: ValueId,
        value: ScConst,
        ty: ScType,
    },
    Encode {
        id: ValueId,
        prob: ValueId,
        length: usize,
        seed: u16,
    },
    BitwiseAnd {
        id: ValueId,
        lhs: ValueId,
        rhs: ValueId,
    },
    Popcount {
        id: ValueId,
        input: ValueId,
    },
    LifStep {
        id: ValueId,
        current: ValueId,
        leak: ValueId,
        gain: ValueId,
        noise: ValueId,
        params: LifParams,
    },
    DenseForward {
        id: ValueId,
        inputs: ValueId,
        weights: ValueId,
        leak: ValueId,
        gain: ValueId,
        params: DenseParams,
    },
    BitwiseXor {
        id: ValueId,
        lhs: ValueId,
        rhs: ValueId,
    },
    Reduce {
        id: ValueId,
        input: ValueId,
        mode: ReduceMode,
    },
    GraphForward {
        id: ValueId,
        features: ValueId,
        adjacency: ValueId,
        n_nodes: usize,
        n_features: usize,
    },
    SoftmaxAttention {
        id: ValueId,
        q: ValueId,
        k: ValueId,
        v: ValueId,
        dim_k: usize,
    },
    KuramotoStep {
        id: ValueId,
        phases: ValueId,
        omega: ValueId,
        coupling: ValueId,
        dt: f64,
    },
    Scale {
        id: ValueId,
        input: ValueId,
        factor: f64,
    },
    Offset {
        id: ValueId,
        input: ValueId,
        offset: f64,
    },
    DivConst {
        id: ValueId,
        input: ValueId,
        divisor: u64,
    },
}

A single operation in the SC compute graph.

Each variant produces exactly one value identified by id. Input operands reference values produced by earlier operations.

Variants

Input

Module input port. No operands; value comes from external I/O.

Fields

id: ValueId

name: String

ty: ScType

Output

Module output port. Consumes one value; no new value produced. id is a dummy (not referenced by other ops).

Fields

id: ValueId

name: String

source: ValueId

Constant

Compile-time constant.

Fields

id: ValueId

value: ScConst

ty: ScType

Encode

Encode a probability (Rate or FixedPoint) into a Bitstream. Maps to sc_bitstream_encoder in HDL.

Fields

id: ValueId

prob: ValueId

Input probability value.

length: usize

Bitstream length.

seed: u16

LFSR seed parameter name (resolved from graph params).

BitwiseAnd

Bitwise AND of two bitstreams (stochastic multiply). Maps to sc_bitstream_synapse in HDL.

Fields

id: ValueId

lhs: ValueId

rhs: ValueId

Popcount

Population count: count 1-bits in a bitstream. Part of sc_dotproduct_to_current in HDL.

Fields

id: ValueId

input: ValueId

LifStep

Single LIF neuron step. Maps to sc_lif_neuron in HDL.

Fields

id: ValueId

current: ValueId

Input current (FixedPoint).

leak: ValueId

Leak coefficient (FixedPoint).

gain: ValueId

Input gain coefficient (FixedPoint).

noise: ValueId

External noise (FixedPoint, can be zero constant).

params: LifParams

Neuron parameters.

DenseForward

Dense SC layer: N_INPUTS → N_NEURONS with full SC pipeline. Maps to sc_dense_layer_core in HDL.

Fields

id: ValueId

inputs: ValueId

Input values (Vec<Rate> or Vec<FixedPoint>).

weights: ValueId

Weight matrix (Vec<Rate> or Vec<FixedPoint>), row-major [n_neurons × n_inputs].

leak: ValueId

Leak coefficient for all neurons.

gain: ValueId

Gain coefficient for all neurons.

params: DenseParams

Layer parameters.

BitwiseXor

Bitwise XOR of two bitstreams (HDC binding).

Fields

id: ValueId

lhs: ValueId

rhs: ValueId

Reduce

Reduce a vector to a scalar (Sum or Max).

Fields

id: ValueId

input: ValueId

mode: ReduceMode

GraphForward

Graph forward: input features × adjacency → aggregated output.

Fields

id: ValueId

features: ValueId

adjacency: ValueId

n_nodes: usize

n_features: usize

SoftmaxAttention

Softmax attention: Q·K^T/sqrt(d) → softmax → ·V.

Fields

id: ValueId

q: ValueId

k: ValueId

v: ValueId

dim_k: usize

KuramotoStep

Single Kuramoto integration step: dθ/dt = ω + ΣK sin(θ_m - θ_n).

Fields

id: ValueId

phases: ValueId

omega: ValueId

coupling: ValueId

dt: f64

Scale

Scale a value by a constant: output = input * factor.

Fields

id: ValueId

input: ValueId

factor: f64

Offset

Offset a value by a constant: output = input + offset.

Fields

id: ValueId

input: ValueId

offset: f64

DivConst

Integer division by a constant (for rate computation).

Fields

id: ValueId

input: ValueId

divisor: u64

Implementations

impl ScOp

pub fn result_id(&self) -> ValueId

Return the ValueId produced by this operation.

pub fn operands(&self) -> Vec<ValueId>

Return all ValueIds consumed by this operation.

pub fn op_name(&self) -> &'static str

Human-readable operation name for the text format.

Trait Implementations

impl Clone for ScOp

fn clone(&self) -> ScOp

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ScOp

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl PartialEq for ScOp

fn eq(&self, other: &ScOp) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for ScOp