[Beowulf] interesting paper on experimental computer science

m at pavis.biodec.com m at pavis.biodec.com
Fri Nov 23 05:37:35 PST 2007


Hi all
	the November issue of the Communications of the ACM has an
nice paper of Basili and Zelkowitz where they report the following 
data, from the paper 

Hochstein, L., Carver, J., Shull, F., Asgari, A., Basili, V.,
Hollingsworth, J., and Zelkowitz, M. Parallel programmer productivity: A
case study of novice HPC programmers. In Proceedings of SC05 (Seattle,
Nov. 12-18). ACM Press, New York, 2005, 1-9.

the sentence of interest is the following:

``Although validity may be threatened by using students in classroom
experiments, there is support for the following hypotheses among this
(interviewed) population:

    * OpenMP offers greater speedup for novices in a shorter amount of
      time when the problem is more computationally based than
      communication based;

    * OpenMP saves 35%-75% of effort over MPI on most problems,
      UPC/CAF saves approximately 40% of effort over MPI, and XMT-C
      saves approximately 50% of effort over MPI;

    * The programming model has a greater influence on
      performance than on experience with the problem to be solved; and

    * When performance is the goal, experts and students
      spend the same amount of time, but experts produce significantly
      better performance.''

-- 
 .*.                            finelli
 /V\
(/ \) --------------------------------------------------------------
(   )       Linux: Friends dont let friends use Piccolosoffice
^^-^^ --------------------------------------------------------------



More information about the Beowulf mailing list