English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Energy proportionality is the key design goal followed by architects ofmodern multicore CPUs. One of its implications is that optimization of anapplication for performance will also optimize it for energy. In this work, weshow that energy proportionality does not hold true for multicore CPUs. Thisfinding creates the opportunity for bi-objective optimization of applicationsfor performance and energy. We propose and study the first application-levelmethod for bi-objective optimization of multithreaded data-parallelapplications for performance and energy. The method uses two decisionvariables, the number of identical multithreaded kernels (threadgroups)executing the application and the number of threads in each threadgroup, withthe workload always partitioned equally between the threadgroups. Weexperimentally demonstrate the efficiency of the method using four highlyoptimized multithreaded data-parallel applications, 2D fast Fourier transformbased on FFTW and Intel MKL, and dense matrix-matrix multiplication usingOpenBLAS and Intel MKL. Four modern multicore CPUs are used in the experiments.The experiments show that optimization for performance alone results in theincrease in dynamic energy consumption by up to 89% and optimization fordynamic energy alone degrades the performance by up to 49%. By solving thebi-objective optimization problem, the method determines up to 11 globallyPareto-optimal solutions. Finally, we propose a qualitative dynamic energymodel employing performance monitoring counters as parameters, which we use toexplain the discovered energy nonproportionality and the Pareto-optimalsolutions determined by our method. The model shows that the energynonproportionality in our case is due to the activity of the data translationlookaside buffer (dTLB), which is disproportionately energy expensive.

Modern Multicore CPUs are not Energy Proportional: Opportunity for Bi-objective Optimization for Performance and Energy