FlowPM: Distributed TensorFlow Implementation of the FastPM Cosmological N-body Solver

TL;DR

FlowPM is a GPU-accelerated, distributed Particle-Mesh N-body simulation code in Mesh-TensorFlow that offers significant speed improvements and supports advanced cosmological inference tasks.

Contribution

The paper introduces a novel multi-grid scheme for large-scale force computation and demonstrates scalable, efficient cosmological simulations in a distributed environment.

Findings

Achieves 10x speed-up over traditional Python PM code

Validates accuracy of the multi-grid force computation

Enables efficient large-scale cosmological inference

Abstract

We present FlowPM, a Particle-Mesh (PM) cosmological N-body code implemented in Mesh-TensorFlow for GPU-accelerated, distributed, and differentiable simulations. We implement and validate the accuracy of a novel multi-grid scheme based on multiresolution pyramids to compute large scale forces efficiently on distributed platforms. We explore the scaling of the simulation on large-scale supercomputers and compare it with corresponding python based PM code, finding on an average 10x speed-up in terms of wallclock time. We also demonstrate how this novel tool can be used for efficiently solving large scale cosmological inference problems, in particular reconstruction of cosmological fields in a forward model Bayesian framework with hybrid PM and neural network forward model. We provide skeleton code for these examples and the entire code is publicly available at https://github.com/modichirag/flowpm.