Dual-resolution GPU asynchronism for direct numerical simulations of turbulent mixing at high Schmidt number

In turbulent mixing at high Schmidt number, scalar fluctuations arise at scales much smaller than the Kolmogorov scale of the velocity field, leading to stringent requirements for resolution in space and time, and hence a strong motivation for dual-resolution approaches. In this talk we describe recent development of a GPU code which computes, with a high degree of asynchronism, the velocity field on a coarser grid and a high Sc scalar on a finer grid. This code is an outgrowth of the ``GESTS'' code used recently to study forced isotropic turbulence at 35-trillion points resolution (Yeung et al., J. Fluid Mech., Vol. 1019, R2 (2025). Discrete point-to-point communication calls are used to send the coarse grid velocity to the finer grid where convective terms for the high-Sc scalar are formed. A second scalar of modest Sc can be included on the coarse grid at virtually no cost. This is relevant in contexts such as oceanic mixing, where differential diffusion between temperature and salinity of disparate Schmidt numbers (7 and 700, respectively) is important. We report on science results obtained to date.