#34 advances with RAPl

paper/cplusplus-energy.bib

@@ -5,16 +5,10 @@ @book{alfred2007compilers
   publisher={pearson Education}
 }
 
-@inproceedings{david2010rapl,
-  title={RAPL: Memory power estimation and capping},
-  author={David, Howard and Gorbatov, Eugene and Hanebutte, Ulf R and Khanna, Rahul and Le, Christian},
-  booktitle={Proceedings of the 16th ACM/IEEE international symposium on Low power electronics and design},
-  pages={189--194},
-  year={2010}
-}
+
 
 @inproceedings{de2010new,
-  title={The new linux’perf’tools},
+  title={The new {Linux} ’perf’ tools},
   author={De Melo, Arnaldo Carvalho},
   booktitle={Slides from Linux Kongress},
   volume={18},

paper/energy-sac-2025.Rnw

@@ -137,6 +137,8 @@ At any rate, these papers, especially the last one, establish the state of the a
 
 First and foremost, we need to decide the instrumentation that we are going to use to measure energy. The instruments needed to measure them are roughly divided in hardware and software tools \cite{abdurachmanov2015techniques}, with the former being physical instruments that draw information from the power source or else sit between the main electricity source and the machine, thus accurately measuring the energy consumption of the machine. These are expensive scientific instruments, that besides have a synchronization issue to accurately reflect the start and end of the process we are measuring. The latter, however, are simply applications that run alongside the process under measurement, or run the process writing measurements when it ends, which has the disadvantage of introducing a small overhead, but on the other hand is perfectly synchronized. There are also many free, even free software, options, so cost and precision have made us choose this option for our research.
 
+These applications need to have some way of {\em reading} the sensors that the computer includes and that measure energy consumption in different devices. The sensors, and the API to read them, is processor and architecture specific; however, in Intel and AMD based processors a standard proposed by Intel called RAPL (Running Average Power Limit) \cite{rapl,david2010rapl} is available. This API provides a series of virtual registers that receive information on power consumed by different parts of the computer, usually called {\em domains}. We will specially pay attention to what is called the {\em package} (or PKG) domain, which includes the energy consumed by the whole processor package, all the cores, but also other so called {\em uncore} components: the on-chip caches, for instance, but also in many cases the memory controller. This implies that although the energy the memory consumes is not really considered here, memory operations will actually contribute to the measurements found here. This PKG is important for our purposes, not only because it might include the bulk of the energy consumption, but also because it is the minimum granularity level that is common to Intel and AMD processors. In general, PKG will include other sub-sensors in the case or Intel processors, but it will not do so in the case of AMD processors, so it is one of the measures that we can use to compare energy consumption in processors from different manufacturers \cite{khan2018rapl}.
+
 \section{Results}\label{sec:results}
 
 \section{Conclusions}\label{sec:conclusions}

paper/energy.bib

@@ -472,6 +472,15 @@ @CONFERENCE{Abdelhafez2019121
 	note = {Cited by: 2}
 }
 
+@inproceedings{david2010rapl,
+  title={RAPL: Memory power estimation and capping},
+  author={David, Howard and Gorbatov, Eugene and Hanebutte, Ulf R and Khanna, Rahul and Le, Christian},
+  booktitle={Proceedings of the 16th ACM/IEEE international symposium on Low power electronics and design},
+  pages={189--194},
+  year={2010}
+}
+
+
 @Misc{cruz21,
   author =    {Luís Cruz},
   title =     {Tools to Measure Software Energy Consumption from your Computer},