Cost Optimality and Predictability of Parallel Programming With Skeletons

Abstract

Skeletons are reusable, parameterized program components with well-defined semantics and pre-packaged efficient parallel implementation. This paper develops a new, provably cost-optimal implementation of the DS (double-scan) skeleton for programming divide-and-conquer algorithms. Our implementation is based on a novel data structure called plist (pointed list); implementation's performance is estimated using an analytical model. We demonstrate the use of the DS skeleton for parallelizing a tridiagonal system solver and report experimental results for its MPI implementation on a Cray T3E and a Linux cluster: they confirm the performance improvement achieved by the cost-optimal implementation and demonstrate its good predictability by our performance model.

Resource analyses for parallel and distributed coordination

“…For our array summing problem, the analysis would return a concrete prediction, composed similarly to the discussion in 6.5 but embedding concrete BSP cost values for the chosen architecture. Meanwhile, in a more informal setting reflecting the approach of [40], Bischof et al [43] report on a BSP-based, extended BMF derivation of a program for the solution of tridiagonal systems of linear equations. Once again, good correlation between (hand-generated) predictions and real implementation is reported, with no more than 12% error across a range of problem sizes.…”

Characterising Effective Resource Analyses for Parallel and Distributed Coordination

“…Meanwhile, in a more informal setting reflecting the approach of [31], [3] reports upon a BSP based, extended BMF derivation of a program for the solution of tridiagonal systems of linear equations. Once again good correlation between (hand generated) predictions and real implementation is reported, with no more than 12% error across a range of problem sizes.…”