Several possibilities for the reduction of indirect addressing operations within edge-based solvers are discussed. It is shown that indirect addressing operations may be reduced by a factor in excess of 1 : 4 if edges are grouped and linked together. Tests of benchmark loops on two common engineering computer platforms reveal that CPU saving factors ranging from 1.3 to 3.5 may be realized when switching from edge-based to superedge or chain-based loops.
Diff selection: Mark the radio boxes of the revisions to compare and hit enter or the button at the bottom.
Legend: (cur) = difference with latest revision, (prev) = difference with preceding revision, m = minor edit.
Published on 01/01/1994
DOI: 10.1016/0045-7825(94)90133-3Licence: CC BY-NC-SA license
Views 1Recommendations 0
Are you one of the authors of this document?