Parallelization of an Unsteady ALE Solver with Deforming Mesh Using OpenACC

Algorithm 2

Three-dimensional parallel coefficient matrix assembly (3DPCMA) algorithm.

Input: () Unstructured grid metrics (UGM), including the coordinates of the mesh nodes

     and the indices of the nodes that form the elements; adjacent data structure of grid

   () Row array, , which stores the starting positions of rows.

   () Inner node identifier and the GPU thread identifier .

Output:

   () Column array, , which stores the positions of the no-zero values in each row.

   () Value array, , which stores the no-zero values in each row.

   () Right hand vector of the linear system of equations, .

Procedure: Thread is responsible for the following tasks

    () initialize vector that is to calculate: ;

    () fetch the global and local indices of the neighbors of node :

    () fill : ;

    () For each tetrahedron Do

    ()   consider each face that contains node , face for example:

    ()   

    ()   ;

    ()   

    ()   IF ( is a boundary node) then

   ()   ;

   ()  ELSE

   ()   ;

   ()  End IF

   ()  

   ()  similar steps are performed like previous ()–() lines for node

   ()  

   ()  similar steps are performed like previous ()–() lines for node

   () End For