\section{CPU Requirements \label{sec:speed}}

We tested the speed of the extractors by running them on an Intel\ Pentium\ IV, 
2.4\,GHz~CPU machine at IFAE and measured 
the number of executed events per seconds. This was done using the CPU-time measure features which incorporates each task in MARS. 
\par
The results could easily differ by about 20\% from one try to another (using the 
same extractor), but in general, 
the differences between the extractors are much bigger than the intrinsic fluctuation. 
Table~\ref{tab:cpu} shows the obtained results. The numbers in this list have to be compared to the 
I/O speed of about 400~evts/sec. of the {\textit{\bf MRead}}-task which performs the reading 
(and de-compression) of the merpped root-files in MARS. Thus, for time being, every extractor being faster 
than this reference number does not limit the total event reconstruction speed. Only the integrating spline 
extractors lie below this limit and would need to be optimized further.

\begin{table}[htp]
\centering
\begin{tabular}{|c|c|c|c|}
\hline
\hline
\multicolumn{4}{|c|}{\large Measured Extraction Speed} \rule{0mm}{6mm} \rule[-2mm]{0mm}{6mm} \hspace{-3mm}\\
\hline
\hline
Nr. & Name         &  Events/sec. & comments\\
    &              &  (CPU)       & \\
\hline                                                     
\hline                                                     
1   & Fixed Win. (4,4)   & 3500--4500 & no time \\   
2   & Fixed Win. (4,6)   & 3500--4500 & no time \\   
3   & Fixed Win. (6.6)   & 3200--4000 & no time \\   
4   & Fixed Win. (8,8)   & 3200--4000 & no time \\   
5   & Fixed W. (14,10)   & 2700--3300 & no time \\   
\hline                          
6   & Fix. Win. Spline (4,4) & 1700--2200 &  no time \\   
7   & Fix. Win. Spline (4,6) & 1700--2200 &  no time \\   
8   & Fix. Win. Spline (6.6) & 1700--2100 &  no time \\   
9   & Fix. Win. Spline (8,8) & 1300--1800 &  no time \\   
10  & Fix. Win. Spl. (14,10) &  600--1200 &  no time \\   
\hline                          
11  & Fix. Win. Peak S. (2,2) & 2300--2900 &    no time \\   
12  & Fix. Win. Peak S. (4,4) & 2300--2900 &    no time \\   
13  & Fix. Win. Peak S. (4.6) & 2300--2900 &    no time \\   
14  & Fix. Win. Peak S. (6,6) & 2300--2900 &    no time \\   
15  & Fix. Win. Peak S. (8,8) & 2200--2800 &    no time \\   
16  & Fix. Win. Pk S. (14,10) & 2000--2500 &    no time \\   
\hline                          
17  & Slid. Win. (2,2) & 400--700 & \\   
18  & Slid. Win. (4,4) & 500--800 & \\   
19  & Slid. Win. (4.6) & 500--800 & \\   
20  & Slid. Win. (6,6) & 1000--1300 & \\   
21  & Slid. Win. (8,8) & 1100--1400 & \\   
22  & Slid. W. (14,10) & 1000--1800 & \\   
\hline                          
23  & Spline Amplitude  & 700--1000 & \\
24  & Spline Int. (1,1.5)& 300--500 & \\
25  & Spline Int. (2,3) & 200--400 & \\
26  & Spline Int. (4,6) & 150--200 & to be optimized \\
27  & Spline Int. (6,9) & 80--120 & to be optimized \\
\hline                               
28  & Dig. Filt. (6,6)  & 700--900 &   \\
29  & Dig. Filt. (4,4)  & 700--900 &   \\
34  & Dig. Filt. Pk S.  & 500--600 &   \\
\hline
\hline
\end{tabular}
\caption{The extraction speed measured for every extractor.}
\label{tab:cpu}
\end{table}


%%% Local Variables: 
%%% mode: latex
%%% TeX-master: "MAGIC_signal_reco"
%%% End: