# cuSten -- CUDA Finite Difference and Stencil Library

## TL;DR

cuSten is a CUDA library that simplifies the implementation of finite difference and stencil computations in 2D and batched 1D, accelerating development of GPU-based numerical solvers.

## Contribution

The paper introduces cuSten, a user-friendly CUDA library that streamlines finite difference and stencil computations, with an application to the Cahn-Hilliard equation and performance benchmarking.

## Key findings

- cuSten significantly speeds up numerical code development on GPUs.
- The library demonstrates competitive performance in solving PDEs.
- Benchmark results show advantages over serial implementations.

## Abstract

In this paper we present cuSten, a new library of functions to handle the implementation of 2D and batched 1D finite-difference/stencil programs in CUDA. cuSten wraps data handling, kernel calls and streaming into four easy to use functions that speed up development of numerical codes on GPU platforms. The paper also presents an example of this library applied to solve the Cahn-Hilliard equation utilizing an ADI method with periodic boundary conditions, this solver is also used to benchmark the cuSten library performance against a serial implementation.