Radiative transfer in decomposed domains

Abstract

Aims. An efficient algorithm for calculating radiative transfer on massively parallel computers using domain decomposition is presented. Methods. The integral formulation of the transfer equation is used to divide the problem into a local but compute-intensive part for calculating the intensity and optical depth integrals, and a nonlocal part for communicating the intensity between adjacent processors. Results. The waiting time of idle processors during the nonlocal communication part does not have a severe impact on the scaling. The wall clock time thus scales nearly linearly with the inverse number of processors. © ESO 2006.