A parallel algorithm for stiffness matrix assembling in a shared memory environment

This paper describes an algorithm for parallel assembling of the stiffness matrix in structural analysis in shared memory environments. In this algorithm the stiffness matrix is assembled by groups of lines related to each node of the finite element mesh. Therefore as each processor will only assemble the lines related to a specific group of nodes, no synchronization is required to avoid that two or more processor try to update the same part of the memory concurrently and the task can be equally distributed among processors. Some examples are presented and they show that the proposed algorithm has a very good speed up rate and efficiency.