The parallelization of implicit code for springback analysis

The parallelization of methods of implicit code for springback analysis have been carried out by using NX library. The program based on Fortran uses two large one-dimensional arrays for efficient memory usage, but this scheme makes the parallelization more difficult. Two main parts are parallelized: one is the formation of the global stiffness and force matrix from element matrices; and the other is the parallelization of a skyline solver based on Crout factorization. By distributing the skylined matrix in a slabbed and interleaving way, we succeeded in making an improvement to the load balance between processors and increasing parallelization efficiency. The benchmarking results show that the parallelized program running on 32 processors gives a speedup of about 20 times faster than the serial program running on an Intel Paragon