source: docs/HPCA2011/06-scalability.tex @ 1320

Last change on this file since 1320 was 1302, checked in by lindanl, 8 years ago

Create a directory for HPCA

File size: 3.0 KB
Line 
1\section{Scalability}
2\subsection{Performance}
3Figure \ref{Scalability} (a) demonstrates the average XML well-formedness checking performance of Parabix2 for each of the workloads and as executed on each of the processor cores --- \CO\, \CITHREE\ and \SB{}.
4Processing time is shown in terms of bit stream based operations executed in `bit-space' and postprocessing operations executed in `byte-space'.
5In the Parabix2 parser, bit-space parallel bit stream parser operations consist primarily of SIMD instructions; byte-space operations
6consist of byte comparisons across arrays of values. Executing Parabix2 on \CITHREE{} over \CO\ results in an average performance improvement
7of 17\% in bit stream processing whereas migrating Parabix2 from \CITHREE{} to \SB{} results in a 22\% average performance gain. Bit space measurements are stable and consistent across each of the source inputs and cores. Postprocessing operations demonstrate data dependent variance. Performance gains from 18\% to 31\% performance are observered in migrating Parabix2 from \CO\ to \CITHREE{}; 0\% to 17\% performance
8from \CITHREE\ to \SB{}. For the purpose of comparison, Figure \ref{Scalability} (b) shows the performance of the Expat parser on each of the processor cores.
9A performance improvement of less than 5\% is observed when executing Expat on \CITHREE\ over \CO\
10and less than 10\% on \SB\ over \CITHREE{}.
11
12Overall, Parabix2 scales better than Expat. Simply executing identical Parabix2 object code on \SB\ results in an overall performance improvement
13up to 26\%. Additional performance aspects of Parabix2 on \SB\ with AVX instructions are discussed in the following sections.
14
15\begin{figure}
16\centering
17\subfigure[Parabix2]{
18\includegraphics[width=0.40\textwidth]{plots/P2_scalability.pdf}
19}
20\subfigure[Expat]{
21\includegraphics[width=0.40\textwidth]{plots/Expat_scalability.pdf}
22}
23\caption{Average Performance Parabix vs. Expat (y-axis: CPU Cycles per kB)}
24\label{Scalability}
25\end{figure}
26
27
28\subsection{Power and Energy}
29
30Figure \ref{power_Parabix2} shows the average power consumption of Parabix2 over each workload and as executed on each of the processor cores --- \CO{}, \CITHREE\ and \SB{}.
31Average power consumption on \CO{} is 32 watts. Execution on \CITHREE\ results in 30\% power saving over \CO{}.
32\SB\ saves 25\% of the power compared with \CITHREE\ and consumes only 15 watts.
33
34In XML parsing we observe energy consumption is dependent on processing time. That is, a reduction in processing time results in a directly proportional reduction in energy consumption.
35With newer processor cores comes improvements in application performance. As a result, Parabix2 executed on \SB\ consumes 72\% to 75\% less energy than Parabix2 on \CO{}.
36
37\begin{figure}
38\begin{center}
39\includegraphics[width=85mm]{plots/power_Parabix2.pdf}
40\end{center}
41\caption{Average Power of Parabix2 (watts)}
42\label{power_Parabix2}
43\end{figure}
44
45\begin{figure}
46\begin{center}
47\includegraphics[width=85mm]{plots/energy_Parabix2.pdf}
48\end{center}
49\caption{Energy consumption of Parabix2 (nJ/B)}
50\label{energy_Parabix2}
51\end{figure}
Note: See TracBrowser for help on using the repository browser.