Shende, Nikhil and Balakrishnan, N (2004) New migratory memory algorithm for Implicit Finite Volume Solvers. In: AIAA Journal, 42 (9). pp. 1863-1870.Full text not available from this repository. (Request a copy)
This work deals with the issues regarding the convergence acceleration and the memory management of unstructured data-based cell-center finite volume flow solvers. The convergence acceleration to the steady state is achieved using three implicit relaxation procedures cast in a matrix-free framework, namely, point Jacobi, symmetric Gauss–Seidel, and lower-upper-symmetric Gauss–Seidel. The formulation of the point Jacobi procedure presented is different from the symmetric Gauss–Seidel and the lower-upper-symmetric Gauss–Seidel procedures in the sense that the point Jacobi procedure is cast in terms of a face-based algorithm, whereas the other two are cell-based procedures. It is observed that the performances of all of the three implicit relaxation procedures are comparable. Inspired by the developments in the parallelization of unstructured data-based codes, we have presented a new memory-saving device called the migratory memory algorithm (MMA). It is shown that the MMA drastically reduces the memory requirement of the class of codes just mentioned. The MMA is presented in the context of the least-squares-based linear reconstruction and the point Jacobi procedures. Particularly in the case of computational-fluid-dynamics codes to be used in routine design cycle, the use of MMA can considerably increase the problem size a given machine can handle.
|Item Type:||Journal Article|
|Additional Information:||Copyright of this article belongs to AIAA.|
|Department/Centre:||Division of Mechanical Sciences > Aerospace Engineering (Formerly, Aeronautical Engineering)|
|Date Deposited:||28 Jun 2007|
|Last Modified:||27 Aug 2008 12:33|
Actions (login required)