# A JIT Compiler for Neural Network Inference

## TL;DR

This paper introduces a JIT compiler library for neural network inference in C++, achieving high performance on small networks by integrating static properties into generated machine code, with experimental results demonstrating significant improvements on specific hardware.

## Contribution

The paper presents a novel JIT compilation approach for neural network inference that is integrated into a C++ library, enabling optimized runtime performance.

## Key findings

- Outperforms existing implementations on small networks
- Less effective on large networks
- Available as a standalone C++14 library

## Abstract

This paper describes a C++ library that compiles neural network models at runtime into machine code that performs inference. This approach in general promises to achieve the best performance possible since it is able to integrate statically known properties of the network directly into the code. In our experiments on the NAO V6 platform, it outperforms existing implementations significantly on small networks, while being inferior on large networks. The library was already part of the B-Human code release 2018, but has been extended since and is now available as a standalone version that can be integrated into any C++14 code base.