Modeling power and energy consumption of dense matrix factorizations on multicore processors

摘要

In this paper, we propose a model for the energy consumption of the concurrent execution of three keyrndense matrix factorizations, with task parallelism leveraged via the Symmetric Multi-Processing Superscalarrn(SMPSs) runtime, on a multicore processor. Our model decomposes the power dissipation into the system,rnstatic and dynamic components, with the former two being estimated from basic, off-line experiments. Therndynamic power, on the other hand, requires significantly more care, and we introduce a contention-awarernmodel that accommodates for the variability of power consumption due to memory contention. Experimentalrnresults on an Intel Xeon E5504 processor with four cores, using an internal powermeter that samples thernpower drawn by the mainboard with a frequency of 1 KHz, show the reliability of the energy model for thernCholesky, LU, and QR factorizations on this platform.