CPU–GPU hybrid parallel strategy for cosmological simulations

Abstract

SUMMARYGadget is a simulation application for N‐body and smoothed particle hydrodynamics problems in cosmology, and it is widely applied in solving series of cosmological problems. N‐body focuses on the motion of the interaction of N particles, and smoothed particle hydrodynamics is a fluid simulation algorithm that studies the movement of fluid through particle simulation. Most scholars focus their attention on accelerating Gadget on multi‐core CPU or graphics processing units (GPUs) platforms. However, these research activities failed to achieve CPU–GPU hybrid computing, which resulted in tremendous waste of CPU computing resources.In this paper, we propose a CPU–GPU hybrid parallel strategy to accelerate Gadget‐2, a massively parallel structure formation code for cosmological simulations. This strategy uses CPU and GPU to process the calculation of short‐range force. To ensure CPU and GPU workload balance, a dynamic task allocation scheme is proposed according to the computational performance difference between the CPU and GPU.Experimental results showed that our CPU–GPU hybrid parallel strategy achieved an overall speedup factor of 18.6 and a partial speedup factor for short‐range force calculation of 28.35 compared with a single‐core CPU implementation for particles in million‐size magnitudes. Moreover, compared with a GPU platform that contained 12 CPU cores and one GPU, our hybrid parallel strategy obtained overall speedup and partial speedup factors of 6% and 20%, respectively. Furthermore, the scalability of the hybrid strategy is very fine – its performance will be enhanced when the problem scale is increasing. However, this strategy also has its limitation that the performance enhancement will be decreasing if the ratio(the number of CPU cores divides that of the GPU cards) reduces. Finally, in our hybrid strategy, the CPU coefficient of utilization improved by 17.14% or better. Copyright © 2013 John Wiley & Sons, Ltd.

GPU based numerical simulation of core shooting process

“…The increase is 40.1 times (92 days was reduced to 2 days 7 hours) and it shows the increase change with the problem scale [20] . Some practical conclusions can be drawn under the simulations of different parameters.…”

An equal‐area triangulated partition method for parallel Xdraw viewshed analysis

“…The further optimization of the performance of communication between threads in shared memory mode such as GPGPU architecture 19 will be our next work. The further optimization of the performance of communication between threads in shared memory mode such as GPGPU architecture 19 will be our next work.…”

“…15 A hybrid MPI/OpenMP approach is applied to parallel geo-computation using multi-core computers for two-level parallelization to implement process-level and thread-level parallelization for the higher speed-up ratio. 19 An efficient modified R3 algorithm is designed, which executed on the GPU by a two-level spatial domain decomposition strategy. 8,17,18 The parallel computing technique based on the GPU is utilized to perform viewshed analysis more efficiently in some cases.…”

“…17 Although these algorithms have gained a notable increase in computation performance, the speedup ratio of them depends on a specific design and a platform-specific optimization to yield significant gains. 19,22 The other is to tackle the non-contiguous data in the real-world applications. 1 Wu et al designed a GPU-accelerated optimization algorithm for viewshed analysis on multipoint viewshed computation.…”

Collaborating CPU and GPU for the electromagnetic simulations with the FDTD algorithm

“…CUBLAS used both CPUs and GPUs to compute different parts of a dense matrix. Wang extended the cosmology simulation package Gadget2 with the CPU–GPU hybrid parallel strategy. The short‐range force calculation was split between CPUs and GPU, while communication and the hydrodynamics computation were performed on CPUs.…”