A hybrid GPU algorithm for modeling axisymmetric reacting flows with stiff chemical kinetics

Title	A hybrid GPU algorithm for modeling axisymmetric reacting flows with stiff chemical kinetics
Authors	V. V. Loginov¹, E. E. Peskova¹ ¹National Research Ogarev Mordovia State University
Annotation	A hybrid computational algorithm is presented for modeling axisymmetric flows of a multicomponent reacting gas with detailed chemical kinetics. The gas-dynamic part is solved on GPU using PyCUDA, while the integration of the stiff system of ordinary differential equations of chemical kinetics is performed with the JAX library (Kvaerno5 solver) on the same GPU. A zero-copy data exchange between PyCUDA and JAX is implemented via DLPack, avoiding unnecessary copying through host memory. The performance of the proposed approach is compared with a C++ implementation using the Radau5 solver, and the impact of zero-copy on data transfer speed is evaluated. The results demonstrate that the hybrid GPU algorithm provides high efficiency for problems with a large number of cells and complex chemistry.
Keywords	axisymmetric flows, reacting gases, chemical kinetics, GPU, PyCUDA, JAX, DLPack
Citation	Loginov V. V., Peskova E. E. ''A hybrid GPU algorithm for modeling axisymmetric reacting flows with stiff chemical kinetics'' [Electronic resource]. Mathematical modeling, numerical methods, and software systems: Collection of materials from the 12th All‑Russian Scientific Youth School‑Seminar named after E. V. Voskresensky (Saransk, July 21–24, 2026). - pp. 130-135. Available at: https://conf.svmo.ru/files/2026/papers/paper89.pdf. - Date of access: 25.09.2026.

PROCEEDINGS OF THE SCIENTIFIC CONFERENCES AND SCHOOL-SEMINARS

A hybrid GPU algorithm for modeling axisymmetric reacting flows with stiff chemical kinetics

Contacts