# GPU implementation of the Rosenbluth generation method for static Monte Carlo simulations

## TL;DR

This paper introduces a GPU-accelerated parallel implementation of the Rosenbluth self-avoiding walk method for static Monte Carlo simulations, achieving near-linear scaling and enabling efficient large-scale molecular conformation sampling.

## Contribution

It presents the first GPU-based parallel implementation of the Rosenbluth method for both lattice and real space, demonstrating scalability and consistency with serial methods.

## Key findings

- Near-linear scaling with CUDA cores
- Consistent results between serial and parallel implementations
- Effective for large-scale molecular conformation sampling

## Abstract

We present parallel version of Rosenbluth Self-Avoiding Walk generation method implemented on Graphics Processing Units (GPUs) using CUDA libraries. The method scales almost linearly with the number of CUDA cores and the method efficiency has only hardware limitations. The method is introduced in two realizations: on a cubic lattice and in real space. We find a good agreement between serial and parallel implementations and consistent results between lattice and real space realizations of the method for linear chain statistics. The developed GPU implementations of Rosenbluth algorithm can be used in Monte Carlo simulations and other computational methods that require large sampling of molecules conformations.