Commit 44d949a6a8976e121bc6ec0e024ba1bdac7017e9

Authored by Rizwana Begum
1 parent 31e336b9

update

inefficiency.tex
... ... @@ -198,7 +198,7 @@ to use inefficiency as a constraint to their
198 198 system~\cite{david2011memory,deng2012multiscale,deng2011memscale,diniz2007limiting,lebeck2000power,malladi2012towards,sudan2010micro,zheng2009decoupled}.
199 199 %
200 200 While most of the existing multi-component energy management approaches work
201   -under performance constraints, some have potential to be modified to work
  201 +under performance constraints, some have the potential to be modified to work
202 202 under energy constraints and thus could operate under
203 203 inefficiency budget~\cite{bitirgen2008coordinated,deng2012coscale,chen2011coordinating,fan2005synergy,felter2005performance,li2007cross,raghavendra2008no}.
204 204 %
... ...
performance_clusters.tex
... ... @@ -230,9 +230,9 @@ number of transitions made by \textit{lbm} decreases with an increase in cluster
230 230 threshold, however, the absolute number of transitions compared to other
231 231 benchmarks does not decrease significantly as it doesn't have too many
232 232 transition to start with at 3\%. Like our previous observation, the number of
233   -transitions also decreases with increase in inefficiency for these two
234   -benchmarks showing that there is a high number of consecutive samples that have
235   -similar performance but different inefficiency at same CPU and memory frequency
  233 +transitions also decreases with increasing inefficiency for these two
  234 +benchmarks. This shows that there is a high number of consecutive samples that have
  235 +similar performance but different inefficiency at the same CPU and memory frequency
236 236 settings. A decrease in the number of transitions is a result of an increase in
237 237 the length of stable regions.
238 238  
... ...