The current paper reports on the implementation of a numerical solver on the Graphic Processing Units (GPU) to model reactive gas mixture with detailed chemical kinetics. The solver incorporates high-order finite volume methods for solving the fluid dynamical equations coupled with stiff source terms. The chemical kinetics are solved implicitly via an operator-splitting method. We explored different approaches in implementing a fast kinetics solver on the GPU. The detail of the implementation is discussed in the paper. The solver is tested with two high-order shock capturing schemes: MP5 (Suresh, et al. (1997)) and ADERWENO (Titarev, et al. (2005)). Considering only the fluid dynamics calculation, the speed-up factors obtained are 30 for the MP5 scheme and 55 for ADERWENO scheme. For the fully-coupled solver, the performance gain depended on the size of the reaction mechanism. Two different examples of chemistry were explored. The first mechanism consisted of 9 species and 38 reactions, resulting in a speed-up factor up to 35. The second, larger mechanism consisted of 36 species and 308 reactions, resulting in a speed-up factor of up to 40.
Conclusion and Future Works
In the current paper, we described the implementation of a numerical solver for simulating chemically reacting flow on the GPU. The fluid dynamics is modeled using high-order shock-capturing schemes, and the chemical kinetics is solved using an implicit solver. Results of both the fluid dynamics and chemical kinetics are shown. Considering only the fluid dynamics, we obtained a speed-up of 30 and 55 times compared to the CPU version for the MP5 and ADERWENO scheme, respectively. For the chemical kinetics, we presented two different approaches on implementing the Gaussian elimination algorithm on the GPU. The best performance obtained by solving the kinetics problem ranges from 30-40 depending on the size of the reaction mechanism. When the fluid dynamics is coupled with the kinetics, we obtained a speed-up factor of 40 times for a 9-species gas mixture with 38 reactions. The solver is also tested with a larger mechanism (36 species, 308 reactions) and the performance obtained is faster than the small mechanism.
The current work can be extended in different ways. First, since the framework is performing well in shared memory architecture, it is possible to also extend it to distributed memory architecture utilizing Message Passing Interface (MPI). The extension permits using multi-GPU which is attracted for performing large-scale simulations. On the other hand, although the current simulation is done for chemically reacting flow, it is desired to extend it to simulate ionized gas (i.e. plasma) which requires modeling additional physical process (Collisional-Radiative) to characterize different excitation levels of the charged species. In adidition, the governing equations also need to be extended to characterize the thermal non-equilibrium environment of the plasma. Given that the physics has been well established[7, 8, 3], the extension is certainly achievable.
H. P. Le, J.-L. Cambier arXiv:1110.4903v1 [physics.flu-dyn]