1. Principle of the test

This test validates the (computer) parallelism of elementary calculations and assemblies. The advantage of this test is to run in a parallel MPI version ** (on 4 processors for example). This test validates programming MPI but does not prove anything about performance.

It tests all keyword combinations:

• AFFE_MODELE/DISTRIBUTION/METHODE =

*/”CENTRALISE”

*/”GROUP_ELEM”

*/”MAIL_CONTIGU”

*/”MAIL_DISPERSE”

• */”SOUS_DOMAINE”

• MATR_DISTRIBUE = “OUI”/”NON”

• ELIM_LAGR2 = “OUI”/”NON”

• Kinematic conditions:

*/dualized (AFFE_CHAR_MECA)

*/eliminated (AFFE_CHAR_CINE)

*/ »straddling » multiple processors

It also tests the possibilities of « modifying » the partitioning before executing post-processing (command MODI_MODELE).