Accelerated solution of problems of combustion gas dynamics on GPUs

• Parallel algorithms are built for the purpose of calculations on graphic processors.
• The proposed algorithms allow for 41.66 times faster acceleration on 4 GPU then CPU.
• All computations on the GPU were conducted with double-order accuracy.
• To conduct chemical computation 2.3 Kbytes of GPU memory was needed for each thread.

The paper deals with a solution of gas dynamics problems with chemical kinetics on graphic accelerators. To solve these problems, schemes are analyzed that combine high resolving power in the areas of small perturbations and monotony in areas of strong discontinuities. Problems are studied with consideration of chemical kinetics, which address the processes of transition from combustion to detonation. Parallel algorithms of the analyzed difference models are built for the purpose of calculations on graphic processors using the CUDA technology. The proposed algorithms and programs allow for 41.66 times faster acceleration on four graphic accelerators M 2090 than the CPU.