W. Gropp. Parallel Computing and Domain Decomposition. Argonne National Laborabory Technical Report MCS-P257-0891 (1991).  Home/Search   Document Not in Database   Summary   Related Articles

.... Theta p N array of elements is divided into P subdomains of q N P Theta q N P elements each. Note that each subdomain contains exactly N P elements. In this decomposition, a subdomain may share nodes with as many as eight other subdomains as illustrated in Figure 2. As discussed in [8], this decomposition can be implemented by sending only four messages by piggybacking the small amount of information shared between diagonally adjacent subdomains on the messages sent between horizontally and vertically adjacent subdomains. This, however, requires substantial coordination of ....

....(P ) for different machines. of messages sent to four under some conditions; ffl The use of the brick decomposition, a variation of the block decomposition that reduces the number of messages sent at the cost of a small increase in the total message size. 6. 1 Coordinating Messages In [8], Gropp described a method for implementing the 2D block decomposition with each processor sending at most four messages rather than the expected eight. This is done by piggybacking the information for the single nodes shared by diagonally adjacent subdomains on the messages being sent between ....

W. Gropp. Parallel Computing and Domain Decomposition. Argonne National Laborabory Technical Report MCS-P257-0891 (1991).

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC