NAG on Fortran for finance quants
beliavsky_at_aol.com
Date: 01/21/04
- Next message: PetitTrot: "oops - standard end of 2004 (?)"
- Previous message: Steve Lionel: "Re: MS Fortran"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: 21 Jan 2004 13:06:20 -0800
http://fenews.com/fen35/where_num_matters/where_num_matters.html
Malcolm Cohen of NAG has written an article on the advantages of a
matrix-oriented language like Fortran 2003 for quantitative finance
(QF), an area where C++ is now dominant.
I am glad he wrote the article and hope that NAG is successful in
promoting Fortran in QF. However, NAG would be more persuasive in
promoting Fortran 95 and later versions if it fully adopted the
standard itself in its math libraries. The NAG Fortran 77 library
continues to be provide substantially more functionality than the
Fortran 90 library. It would not be difficult to include all the
functionality of the Fortran 77 library in the Fortran 90 library,
simply by writing wrappers for the F77 code that eliminate arguments
specifying array sizes and providing workspace arrays.
One NAG procedure I have considered using, G13DCF, which estimates
vector ARMA time series models, has 27 (!) arguments, at least 11 of
which are worksspace or array dimension arguments that could be easily
removed. I am sure many other arguments could be made OPTIONAL. There
is no exact equivalent in their F90 library.
NAG has not only failed to make the entire functionality of their
traditional F77 library available in their F90 library, they are
adding NEW procedures ONLY to the F77 and C libraries. An example is
their GARCH subroutines, use to model time series with changing
variance, a hot topic in finance. (The Nobel prize in economics this
year was awarded to an inventor of GARCH). Subroutines such as G05HKF
in F77 have no equivalent in their F90 library.
Numerical Recipes got it right: they made available a Fortran 90
version of their full library soon after F90 was standardized. What I
am talking about is not hard to do. Buyers can still use their F77
library if they prefer, of course.
NAG might be unsuccessful in promoting modern Fortran in quantitative
finance even if it did things properly, due to the bad connotations of
the "F-word" in that field. But its current approach almost ensures
that its market share in quantitative finance will be miniscule,
compared to more user-friendly products like Matlab and S-Plus.
- Next message: PetitTrot: "oops - standard end of 2004 (?)"
- Previous message: Steve Lionel: "Re: MS Fortran"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
