An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-finance/attachments/20080924/eac89861/attachment.pl>
Use of sorting in Financial Domain
4 messages · subramanian R, Brian G. Peterson, Joshua Reich +1 more
On Wed, 2008-09-24 at 11:37 +0530, subramanian R wrote:
Hi All, The *sort() *command in R has two methods (method specification is an optional parameter) 1.shell sort(default) 2.quick sort and order() has 1 extra method 3.radix sort. In our multicore architecture we have some optimized implementations of several sorting methods like Bitonic sort that is not supported by R.It is faster than quick sort in most cases in our architecture. I have two questions in this regard with respect to financial domain. How much sorting is used in this domain? In those cases what type of sorting is used? Will it help if a seperate R command for bitonic sort is implemented and override the default one? Are there any specific packages used by Financial sector where optimized sorting techniques have been implemented?
Sorting is often used in financial time series: e.g. to examine drawdowns, or to extract a quantile of an observed series or a monte carlo simulation. Implementation of additional parallel sort algorithms in R such as Batcher's bitonic sort seems as though it could be valuable in financial time series. I'm not aware of additional sorting implementations in R specific to finance, most finance specific functionality uses R general functions such as sort() and order() It seems that your implementation would be best proposed as a patch to one of these functions on the r-devel list, or alternately contained in an optimized sorting package that could overload the standard functions with a new implementation. Regards, - Brian
Extending the question further, I assume that bitonic sorting makes sense when you have access to a SIMD platform, such as in a GPU. I have been playing with porting some of my code to NVIDA's CUDA platform, and I'd love to hear if other people have had much success. If I recall correctly, I had discussed this on the list in the past and we discussed using CUDA BLAS over default installations. Has anyone moved CUDA into a production environment? If so, how has it fared? If not, why not? Josh Reich -----Original Message----- From: r-sig-finance-bounces at stat.math.ethz.ch [mailto:r-sig-finance-bounces at stat.math.ethz.ch] On Behalf Of Brian G. Peterson Sent: Wednesday, September 24, 2008 6:08 AM To: subramanian R Cc: r-sig-finance at stat.math.ethz.ch Subject: Re: [R-SIG-Finance] Use of sorting in Financial Domain
On Wed, 2008-09-24 at 11:37 +0530, subramanian R wrote:
Hi All, The *sort() *command in R has two methods (method specification is an optional parameter) 1.shell sort(default) 2.quick sort and order() has 1 extra method 3.radix sort. In our multicore architecture we have some optimized implementations
of
several sorting methods like Bitonic sort that is not supported by
R.It is
faster than quick sort in most cases in our architecture. I have two questions in this regard with respect to financial domain. How much sorting is used in this domain? In those cases what type of sorting is used? Will it help if a seperate R command for bitonic sort is implemented
and
override the default one? Are there any specific packages used by Financial sector where
optimized
sorting techniques have been implemented?
Sorting is often used in financial time series: e.g. to examine drawdowns, or to extract a quantile of an observed series or a monte carlo simulation. Implementation of additional parallel sort algorithms in R such as Batcher's bitonic sort seems as though it could be valuable in financial time series. I'm not aware of additional sorting implementations in R specific to finance, most finance specific functionality uses R general functions such as sort() and order() It seems that your implementation would be best proposed as a patch to one of these functions on the r-devel list, or alternately contained in an optimized sorting package that could overload the standard functions with a new implementation. Regards, - Brian _______________________________________________ R-SIG-Finance at stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-finance -- Subscriber-posting only. -- If you want to post, subscribe first.
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-finance/attachments/20080925/e38a1c98/attachment.pl>