Parallelizing dense linear algebra operations with task queues in 11c

Abstract

11c is a language based on C where parallelism is expressed using compiler directives. The 11c compiler produces MPI code which can be ported to both shared and distributed memory systems. In this work we focus our attention in the 11c implementation of the Workqueuing Model. This model is an extension of the OpenMP standard that allows an elegant implementation of irregular parallelism. We evaluate our approach by comparing the OpenMP and 11c parallelizations of the symmetric rank-k update operation on shared and distributed memory parallel platforms. © Springer-Verlag Berlin Heidelberg 2007.