Heterogeneous Computing (CPU–GPU) for Pollution Dispersion in an Urban Environment

Abstract

The use of Computational Fluid Dynamics (CFD) to assist in air quality studies in urban environments can provide accurate results for the dispersion of pollutants. However, due to the computational resources needed, simulation domain sizes tend to be limited. This study aims to improve the computational efficiency of an emission and dispersion model implemented in a CPU-based solver by migrating it to a CPU–GPU-based one. The migration of the functions that handle boundary conditions and source terms for the pollutants is explained, as well as the main differences present in the solvers used. Once implemented, the model was used to run simulations with both engines on different platforms, enabling the comparison between them and reaching promising time improvements in favor of the use of GPUs.

Assessment of a heterogeneous computing CFD code in wind farm simulations

“…Integration between the Fortran 90 code running on the CPU and the CUDA code for the GPU kernels is achieved through extensive use of of the Fortran 90 ISO-C-Binding module [4]. The parallelisation of Chamán execution between CPU processes is based on block-structured meshing (domain decomposition), and is performed using the MPI library, in the same manner as caffa3d.…”

“…Recently, scientists have assessed the use of CPU-GPU architectures for solving Computational Fluid Dynamics (CFD) problems, obtaining significant improvement in the computational performance of the codes employed [2][3][4]. These works also show that the redesign of code programming strategies is needed to maximize the benefits of the GPU architecture.…”

“…As a result, the new code combines the use of GPU and the communication protocol Message Passing Interface (MPI) to handle different levels of parallelisms and benefit from massive parallel computing platforms. More details of Chamán are given in [4], where it was used to analyse the pollution dispersion in an urban environment.…”

Turbomachinery GPU Accelerated CFD: An Insight into Performance

“…Computational Fluid Dynamics (CFD) is one of the most time-consuming activities in High-Performance Computing (HPC). Recent research shows that heterogeneous architectures, with Graphics Processing Units (GPUs) as massively parallel co-processors to the Central Processing Unit (CPU), can accelerate computation processes in various CFD applications [1][2][3][4][5][6][7]. Developing a general-purpose CFD software to leverage this powerful hardware is challenging and time-intensive, as evidenced by the cited references.…”

Data-driven control of discrete-phase concentration in a vortex involved two-phase flow with optimized energy consumption

“…Although a computational power is improved in last decade, simulation of multiphase flow still faces the challenge of trade-off between computational speed and accuracy. It was reported that over 8 million cells with 59 milliondegree of freedoms were required for numerical simulation of pollutant dispersion process, and more than 40 minutes required for one main time loop with a four regions GPU (Graphics Processing Unit), and more than 8 hours with a four regions CPU (Central Processing Unit) [46]. It is almost impossible for traditional control algorithm to control a two-phase flow with an efficient computational speed.…”