Element-by-element techniques and Parallel Programming - The Boundary Element Method with Programming - page 214

before. At the end of the loop, the element matrices are stored. The section adding the
azimuthal integral for infinite regions is unchanged. In the next loop F is augmented
from Diag as before, but of course there is no Lhs .
We now proceed to a new section for solving the equation system element-by-
element. This contains two new subroutines form_lhs and get_km . The first of these
combines element dUe and dTe matrices appropriately and the second gets the
appropriate part of Diag for its addition into the matrix*vector multiplication. In the first
case the “gather” vector is Ldeste and in the second it is g .
So in the BiCGStab process there are three matrix*vector products: Elements_2 , to
start the process and Elements_3 and Elements_4 in the “ell” loop. The usual value of
ell is taken to be 4 but this can be changed by the user.
The results are collected from the global result vector u1 in exactly the same manner
as in the previous program.

8.3.1 Sample input file

The input file is precisely the same as for the previous program.

8.3.2 Sample output file

Project:
Square excavation 3D
Cartesian_dimension: 3
Elasticity Problem
Infinite Region
No symmetry
Linear Elements
Modulus: 1000.00000000000
Poissons ratio: 0.000000000000000E+000
Number of Nodes of System: 602
Number of Elements of System: 600
Node 1 Coor 10.00 0.00 0.00
Node 602 Coor 10.00 9.00 9.00
Incidences:
EL 1 Inci 1 2 3 4
EL 600 Inci 602 141 143 331
Elements with Dirichlet BC´s:
Elements with Neuman BC´s:
Element 600 Prescribed values:
-10.000 0.000E+000 0.000E+000
-10.000 0.000E+000 0.000E+000