Contents

Binary Fusion Patterns

Overview

The Binary category includes operations such as: Add, Subtract, Maximum, Minimum, Multiply, Divide.

oneDNN supports various binary fusion patterns to optimize performance and reduce memory bandwidth requirements. This document describes the supported fusion patterns for Binary operations.

Pattern Structure

oneDNN defines floating-point Binary fusion patterns as follows. The blue nodes are required when defining a Binary fusion pattern while the brown nodes are optional.

Binary pattern

  1. Binary Operation : Performs the corresponding binary operation between the src_0 and src_1 tensors. Refer to the Note in Fusion Patterns.

  2. Epilogue Subgraph : Optional and can include the following operations:

    • Binary Operations.

    • Unary Operations : Refer to the Note in Fusion Patterns.

    Combination Rules:

    epilogue subgraph

    • 0 to 4 Binary or Unary operations are supported in the epilogue subgraph.

Data Types

oneDNN supports the following combinations of data types for src0, src1 and dst:

src_0

src_1

dst

f32,bf16,f16

f32,bf16,f16

f32,bf16,f16

The definition of the data types and support status on different CPU and GPU platforms follow the general description in the Data Types Guide.

Quantized ConvTranspose Fusion Patterns Unary Fusion Patterns