From bf41694afb1b02585ece2f87942338199a6d959b Mon Sep 17 00:00:00 2001 From: Rizwana Begum Date: Fri, 14 Aug 2015 16:24:15 -0400 Subject: [PATCH] edits --- optimal_performance.tex | 2 +- performance_clusters.tex | 8 ++++---- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/optimal_performance.tex b/optimal_performance.tex index ebb5e59..8365f8f 100644 --- a/optimal_performance.tex +++ b/optimal_performance.tex @@ -97,7 +97,7 @@ algorithm presented by CoScale~\cite{deng2012coscale} takes 5us to find optimal frequency settings, time taken by PLLs to change voltage and frequency in commercial processors is in the order of 10s of microseconds. Reducing the frequency at which tuning algorithms need to re-tune is critical to -reduce the cost of tuning overhead on application performance. +reduce the impact of tuning overhead on application performance. %\item \noindent \textit{Limited energy performance trade-off options.} Choosing the diff --git a/performance_clusters.tex b/performance_clusters.tex index 0c605a5..fa0ca95 100644 --- a/performance_clusters.tex +++ b/performance_clusters.tex @@ -26,7 +26,7 @@ the system. We search for the performance clusters using an algorithm that is similar to the approach we used to find the optimal settings. We first filter the settings that fall within a given inefficiency budget and then search for the optimal settings in the first pass. In the second pass, we find all of the -settings that have a speedup within the specified \textit{cluster threshold} of the optimal performance. +settings that have a speedup within the specified cluster threshold of the optimal performance. \begin{figure}[t] \centering \includegraphics[width=\columnwidth]{./figures/plots/496/stable_line_plots/lbm_stable_lineplot_annotated_5.pdf} @@ -96,13 +96,13 @@ Figures~\ref{clusters-gobmk}(c),~\ref{clusters-gobmk}(d) plot the performance clusters for \textit{gobmk} for inefficiency budget of 1.3 and cluster thresholds of 1\% and 5\% respectively. As we observed in Figure~\ref{gobmk-optimal}, the optimal settings for \textit{gobmk} change -every sample (of length 10 million instructions) and follows +every sample (of length 10 million instructions) at inefficiency of 1.3 and follow application phases (CPI). Figure~\ref{clusters-gobmk}(c) shows that by allowing just 1\% performance degradation, the number of settings available to choose from increase. For example, for sample 11, the -optimal settings were at 1000MHz CPU and 500MHz memory. With 1\% +optimal settings were at 920MHz CPU and 580MHz memory. With 1\% cluster threshold, the range of available frequencies increases to -970MHz-1000MHz for CPU and 420MHz-580MHz for memory. With a 5\% +900MHz-9200MHz for CPU and 420MHz-580MHz for memory. With a 5\% cluster threshold, the range of available frequencies increases further as shown in Figure~\ref{clusters-gobmk}(d). With an increase in number of available settings, the probability of finding common settings in two consecutive samples -- libgit2 0.22.2