Matrix computations (3rd ed.)
Hi-index | 0.00 |
Implementation of standard algorithms coupled with immature optimization techniques often cause VAX/Ada's run-time performance to be less than optimal. An analysis of a VAX/Ada LINPACK benchmark test revealed that the storage scheme of two dimensional arrays and the column addressing of matrix elements leads to a disproportionate number of index computation multiplications. Further analysis shows how the LINPACK decomposition algorithm can be rewritten so that VAX/Ada run-time performance is optimized.