[R-sig-Finance] Backtesting Speed

Adrian Trapletti a.trapletti at swissonline.ch
Fri Jul 7 10:22:38 CEST 2006


Wojciechowski, William wrote:

>Hi,
>
>What type of algorithms/computations would you
>program in C/C++?
>  
>
The only general comment I can make about your question is: You have to 
find out where speed really matters for your application. As already 
mentioned here on the list, typically a very small portion of the code 
uses almost all cpu and/or I/O time. And improving this code or porting 
it to a faster environment helps. However, sometimes it is better to 
implement the whole application in a faster and maybe more primitive 
environment as switching between high-level and low-level environments 
is too expensive (like calling C/C++ code from R). And sometimes the 
"cheapest" way is just to improve your hardware, e.g. if I/O behaviour 
is really important, then a faster hardware solution is maybe far 
cheaper than having a software engineer redesign your whole application.

Best regards
Adrian

>Thanks,
>
>William C. Wojciechowski
>Koch Quantitative Trading
>20 East Greenway Plaza
>Houston, TX 77046
>713-544-5093
>william.wojciechowski at kochind.com
>
>-----
>
>The companies I (used to) work for use intraday data, sometimes
>tick-by-tick data, to backtest trading strategies. In that case speed is
>a very important issue and interpreted language environments like R are
>slow in comparison to compiled language environments like C/C++. 
>
>Generic environment for quant trading might look like the following: 
>
>time critical computations like backtesting and realtime computing in
>C/C++, client/server, GUI etc. in Java, and statistical analysis and
>diagnostics, prototyping etc. in Matlab or R or similar.
>
>Regards
>
>Adrian
>
>
>
>  
>



More information about the R-SIG-Finance mailing list