Unveiling the performance-energy trade-off in iterative linear system solvers for multithreaded processors

摘要

In this paper, we analyze the interactions occurring in the triangle performance-power-energy for the executionrnof a pivotal numerical algorithm, the iterative conjugate gradient (CG) method, on a diverse collection ofrnparallel multithreaded architectures. This analysis is especially timely in a decade where the power wall hasrnarisen as a major obstacle to build faster processors. Moreover, the CG method has recently been proposedrnas a complement to the LINPACK benchmark, as this iterative method is argued to be more archetypicalrnof the performance of today’s scientific and engineering applications. To gain insights about the benefits ofrnhands-on optimizations we include runtime and energy efficiency results for both out-of-the-box usage relyingrnexclusively on compiler optimizations, and implementations manually optimized for target architectures,rnthat range from general-purpose and digital signal multicore processors to manycore graphics processingrnunits, all representative of current multithreaded systems.