A three‐stage graphics processing unit‐based finite element analyses matrix generation strategy for unstructured meshes

Abstract

SummaryWith the development of parallel computing architectures, larger and more complex finite element analyses (FEA) are being performed with higher accuracy and smaller execution times. Graphics processing units (GPUs) are one of the major contributors of this computational breakthrough. This work presents a three‐stage GPU‐based FEA matrix generation strategy with the key idea of decoupling the computation of global matrix indices and values by use of a novel data structure referred to as the neighbor matrix. The first stage computes the neighbor matrix on the GPU based on the unstructured mesh. Using this neighbor matrix, the indices and values of the global matrix are computed separately in the second and third stages. The neighbor matrix is computed for three different element types. Two versions for performing numerical integration and assembly in the same or separate kernels are implemented and simulations are run for different mesh sizes having up to three million degrees of freedom on a single GPU. Comparison with GPU‐based parallel implementation from the literature reveals speedup ranging from 4× to 6× for the proposed workload division strategy. Furthermore, the same kernel implementation is found to outperform the separate kernel implementation by 70% to 150% for different element types.

Development of GPU‐based matrix‐free strategies for large‐scale elastoplasticity analysis using conjugate gradient solver

“…There exists a large number of CUDA-based applications that use GPU to accelerate compute-intensive codes in disciplines like scientific computing, graphics rendering, AI and machine learning, image and video processing and so forth. In scientific computing there exist a large number of previous works in literature that demonstrates great speedup on GPU for methods like FEM, [15][16][17][18][19] computational fluid dynamics (CFD) [20][21][22] and molecular dynamics (MD). 23 The GPUs have been found more suitable for tasks that involve high computation and less memory usage.…”

Solving the Discretised Neutron Diffusion Equations using Neural Networks

“…This additional coding task has hindered the take-up of GPUs, although there are examples of this having been done successfully, for example, in computational fluid dynamics [3], for acoustic waves [4] and, in radiation transport, for a Monte Carlo neutron transport code [5] and for eigenvalue problems [6]. With CUDA and OpenCL, GPUs have been used to accelerate generation of finite element matrices for unstructured meshes [7][8][9][10] and for discontinuous Galerkin methods [11]. Recently, new types of processors have been unveiled, which have been designed specifically for tasks associated with AI such as matrix multiplication and vector operations.…”

Solving the discretised neutron diffusion equations using neural networks

“…These additional coding tasks and algorithmic developments have hindered the take-up of GPUs, although there are examples of this having been done successfully, for example, in computational fluid dynamics, 3,7 for acoustic waves 6 and, in radiation transport, for a Monte Carlo neutron transport code 4 and for eigenvalue problems. 8 With CUDA and OpenCL, GPUs have been used to accelerate generation of finite element matrices for unstructured meshes [9][10][11][12] and for discontinuous Galerkin methods. 13 Recently, new types of processors have been unveiled, which have been designed specifically for tasks associated with AI such as matrix multiplication and vector operations.…”