XFluids: A unified cross-architecture high performance heterogeneous reacting flows simulation solver
Description
We present a cross-architecture high-order heterogeneous Navier-Stokes simulation solver, XFluids, for compressible reacting multicomponent flows on different platforms, where ‘X’ stands for multiple different devices. The multicomponent reacting flows are ubiquitous in many scientific and engineering applications, while their numerical simulations are usually time-consuming to capture the underlying multiscale features. Although heterogeneous accelerated computing is significantly beneficial for large-scale simulations of these flows, the effective utilization of various heterogeneous accelerators with different architectures and programming models in the market remains a challenge. To address this, we develop XFluids by SYCL, to perform acceleration directly targeted to different devices, without translating any source code. This solver has been open-sourced, and tested on multiple graphics processing units (GPUs) from different mainstream vendors, indicating high portability. Through various benchmark cases, including the shock tube, diffusion, autoignition, detonation, and shock-bubble interaction, the accuracy of XFluids is demonstrated, with approximately no efficiency loss compared to popular existing GPU programming models, such as CUDA and HIP. In addition, XFluids show considerable acceleration compared to other open-source multicomponent reacting flow solvers. Then, to extend the solver to multiple GPUs, the Message Passing Interface (MPI) library is employed, with the GPU-aware communication supported. With this, the weak scaling of XFluids for multi-GPU devices is larger than 95% for 1024 GPUs. Last, in order to fully exploit the computational capability of the all devices, the hybrid CPU-GPU heterogeneous simulations are achieved without changing the source code of XFluids.