TACO (Topologies and Collections) is a template library that introduces the flavour of distributed data parallel processing by means of reusable topology classes and C++ templates. This paper introduces TACO's basic abstractions and provides a performance analysis for basic collective operations on various cluster architectures with several different networks.
Citation:
Jörg Nolte, Mitsuhisa Sato, Yutaka Ishikawa, "TACO-Exploiting Cluster Networks for High-Level Collective Operations," ccgrid, pp.253, First IEEE International Symposium on Cluster Computing and the Grid (CCGrid'01), 2001