Enum OpKind 

pub enum OpKind {
    Input,
    Param,
    Constant,
    Activation,
    Cast,
    Quantize,
    Dequantize,
    FakeQuantize,
    FakeQuantizeLSQ,
    FakeQuantizeLSQBackwardX,
    FakeQuantizeLSQBackwardScale,
    Binary,
    Compare,
    Where,
    ElementwiseRegion,
    MatMul,
    DotGeneral,
    DenseSolve,
    BatchedDenseSolve,
    LayerNorm,
    LayerNorm2d,
    GroupNorm,
    RmsNorm,
    ResizeNearest2x,
    Attention,
    Rope,
    AxialRope2d,
    Reshape,
    Transpose,
    Narrow,
    Concat,
    Expand,
    Gather,
    Reduce,
    Softmax,
    Cumsum,
    TopK,
    Sample,
    Conv,
    ConvTranspose2d,
    Pool,
    ReluBackward,
    ActivationBackward,
    FakeQuantizeBackward,
    ComplexNormSq,
    ComplexNormSqBackward,
    Conjugate,
    MaxPool2dBackward,
    Conv2dBackwardInput,
    Conv2dBackwardWeight,
    SoftmaxCrossEntropyWithLogits,
    SoftmaxCrossEntropyBackward,
    AttentionBackward,
    LayerNormBackwardInput,
    LayerNormBackwardGamma,
    RmsNormBackwardInput,
    RmsNormBackwardGamma,
    RmsNormBackwardBeta,
    RopeBackward,
    GroupNormBackwardInput,
    GroupNormBackwardGamma,
    GroupNormBackwardBeta,
    CumsumBackward,
    GatherBackward,
    GroupedMatMul,
    DequantGroupedMatMul,
    DequantMoEWeights,
    ScatterAdd,
    LoraMatMul,
    DequantMatMul,
    QMatMul,
    QConv2d,
    SelectiveScan,
    GatedDeltaNet,
    FusedSwiGLU,
    FusedMatMulBiasAct,
    FusedResidualLN,
    FusedResidualRmsNorm,
    FusedAttentionBlock,
    FusedTransformerLayer,
    If,
    While,
    Scan,
    ScanBackward,
    ScanBackwardXs,
    GaussianSplatRender,
    GaussianSplatRenderBackward,
    GaussianSplatPrepare,
    GaussianSplatRasterize,
    Custom,
    CustomFn,
    Fft,
}

PLAN L4: discriminant for each Op variant. Used by Op::kind + the Backend::supported_ops trait method to declare which ops a backend can lower; the LegalizeForBackend pass in rlx-opt checks the graph against this set and fails the compile when an unsupported op is present (instead of silent fallback).

Variants

Input

Param

Constant

Activation

Cast

Quantize

Dequantize

FakeQuantize

FakeQuantizeLSQ

FakeQuantizeLSQBackwardX

FakeQuantizeLSQBackwardScale

Binary

Compare

Where

ElementwiseRegion

MatMul

DotGeneral

DenseSolve

BatchedDenseSolve

LayerNorm

LayerNorm2d

GroupNorm

RmsNorm

ResizeNearest2x

Attention

Rope

AxialRope2d

Reshape

Transpose

Narrow

Concat

Expand

Gather

Reduce

Softmax

Cumsum

TopK

Sample

Conv

ConvTranspose2d

Pool

ReluBackward

ActivationBackward

FakeQuantizeBackward

ComplexNormSq

ComplexNormSqBackward

Conjugate

MaxPool2dBackward

Conv2dBackwardInput

Conv2dBackwardWeight

SoftmaxCrossEntropyWithLogits

SoftmaxCrossEntropyBackward

AttentionBackward

LayerNormBackwardInput

LayerNormBackwardGamma

RmsNormBackwardInput

RmsNormBackwardGamma

RmsNormBackwardBeta

RopeBackward

GroupNormBackwardInput

GroupNormBackwardGamma

GroupNormBackwardBeta

CumsumBackward

GatherBackward

GroupedMatMul

DequantGroupedMatMul

DequantMoEWeights

ScatterAdd

LoraMatMul

DequantMatMul

QMatMul

QConv2d

SelectiveScan

GatedDeltaNet

FusedSwiGLU

FusedMatMulBiasAct

FusedResidualLN

FusedResidualRmsNorm

FusedAttentionBlock

FusedTransformerLayer

If

While

Scan

ScanBackward

ScanBackwardXs

GaussianSplatRender

CPU reference 3D Gaussian splat raster (project → bin → sort → raster). See Op::GaussianSplatRender.

GaussianSplatRenderBackward

Backward of Op::GaussianSplatRender — packed scene parameter gradients.

GaussianSplatPrepare

Project + tile bin + sort + ray grid (strict IR splat stage 1).

GaussianSplatRasterize

Per-pixel raster from prepared buffers (strict IR splat stage 2).

Custom

User-registered op dispatched through op_registry. All custom ops (Sparse-LU, FFT, eigensolve, …) share this kind; the per-op identity lives in Op::Custom::name.

CustomFn

User-defined sub-graph with optional override AD rules. See Op::CustomFn / crate::Graph::custom_fn.

Fft

1D FFT primitive (forward or inverse) — see Op::Fft.

Trait Implementations

impl Clone for OpKind

fn clone(&self) -> OpKind

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for OpKind

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for OpKind

fn deserialize<__D>( __deserializer: __D, ) -> Result<OpKind, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Hash for OpKind

fn hash<__H>(&self, state: &mut __H)

where __H: Hasher,

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for OpKind

fn eq(&self, other: &OpKind) -> 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 Serialize for OpKind

fn serialize<__S>( &self, __serializer: __S, ) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl Copy for OpKind

impl Eq for OpKind

impl StructuralPartialEq for OpKind