FEAP User Forum
FEAP => Parallel FEAP => Topic started by: jonwong on August 01, 2012, 12:50:35 AM
-
Before, I had only used simple initial boundary value displacements with parallel feap. It seems that the jacobi preconditioner and cg solver do not converge for the finite element I have. 'Iteration' seems to work find with serial feap, but the problem doesn't seem to play nicely with the ksp solvers in FEAP.
Are they any suggestions on what to do to get this to solve in parallel? Perhaps a different solver?
Below are some of the values from one of the output files.
Computing solution at time 2.0000E+00: Total proportional load 1.0000E+00
Residual norm = 1.7112276E+03 1.0000000E+00 t= 2.51 0.03
Residual norm = 2.8047157E+03 1.6390080E+00 t= 2.86 0.03
Residual norm = 1.6679214E+04 9.7469292E+00 t= 3.22 0.04
Residual norm = 1.0023124E+06 5.8572709E+02 t= 3.58 0.04
Residual norm = 7.6881301E+11 4.4927572E+08 t= 3.97 0.04
-
It seems the convergence issue happened because I had an extraneous "opti" command that was not necessary in my solve.XXX file. Now it converges properly.
-
OPTI and parFEAP are incompatible with each other. We should probably put in a check to disable opti in parallel.