Vectorizing the right-hand side assembly in an explicit finite element program
暂无分享,去创建一个
Abstract The major architectural difference between supercomputers and large mainframes is the ability of supercomputers to process large blocks of independent data with only a single instruction. The blocks of independent data are called “vectors,” and supercomputers are sometimes referred to as “vector processors”. Most operations within a finite element program may be vectorized almost trivially [1] with the exception of the assembly of the right-hand side force vector. This technical note demonstrates how this operation may be vectorized and gives benchmark results on a Cray-XMP-48.
[1] Thomas J. R. Hughes,et al. LARGE-SCALE VECTORIZED IMPLICIT CALCULATIONS IN SOLID MECHANICS ON A CRAY X-MP/48 UTILIZING EBE PRECONDITIONED CONJUGATE GRADIENTS. , 1986 .
[2] J. O. Hallquist,et al. DYNA3D user's manual: (Nonlinear dynamic analysis of structures in three dimensions): Revision 5 , 1987 .
[3] J. O. Hallquist,et al. Recent developments in large-scale finite-element Lagrangian hydrocode technology. [Dyna 20/dyna 30 computer code] , 1981 .