Measuring the Component of a Divide and Conquer Skeleton
Alicia Marcela Printista, F. Saez · Americanae (AECID Library) · 2009
Abstract — Current performance prediction analytical models try to characterize the performance behavior of actual machines through a small set of parameters. Due to different factors, the predicted times suffer sub-stantial deviations. A natural approach is to associate a different proportionality constant with each basic block of computation. In particular, the paper deals with a skeleton designed for parallel divide and conquer algo-rithms that provide hypercubical communications among processes. Our proposal is to introduce different kinds of components to the analytical model by associating a performance constant for each conceptual block of a skeleton. The trace files obtained from the execution of the resulting code using the programming skeleton are used by lineal regression techniques giving us, among other information, the values of the parameters of those blocks. The accuracy of the proposed model is analyzed by means of two instances of skeleton.