Ralph said:
No, it doesn't sound like you would 'gain' much with PCode. <g>
The VB runtime has a few highly optimized functions, but in general uses the
standard math library in the crt and does it thru an extra layer of
indirection.
Who's fortran are you using? Have you ever investigated some of the 3rd
party math libraries out there? The only package I have ever used was
Statisica. The performance gained using a dedicated math library has always
amazed me.
W/ VB since it is by definition "Winders only" I almost exclusively will
develop the Fortran using CVF (Compaq (nee Digital) "Visual" Fortran)
which, unfortunately, is no longer being supported as Compaq sold the
compiler group to Intel. At some point, I may migrate to another
compiler but so far what known bugs are in CVF are not show-stoppers.
When F2003-compatible compilers become available then there may be
incentive enough w/ newer features to upgrade at that point.
W/ the Professional version of CVF came included the Visual Numerics
(nee IMSL) math and stat libraries which I use extensively. The math
libraries in particular are tuned version of BLAS/LINPACK code w/ a
standardized error-handling interface that is convenient.
There is, of course, a "veritable plethora" of Fortran code catalogued
at Netlib accessible there or through the NIST/ORNL GAMS interface that
can help in many problem areas. For VB'ers, I have often suggested that
as a source of computational algorithms/software as (particularly the
F77 dialect) Fortran syntax is so similar to "pure" BASIC and w/ the
column major array structure of both languages, interlanguage calls are
quite easy.
I know of Statistica but have never used it so can't really comment
knowledgeably about it. I became aware of and began to use IMSL back in
the mainframe days in the mid/late 60s when I first went to work out of
school and have used it almost continuously since on CDC/Cray/DEC
(10/20, VAX, PDP)/IBM/Amdahl until the first PC/AT 16-bit installation
from which gradually more and more PC compute power led to moving more
and more apps to the desktop. Consequently, at the present time,
familiarity trumps alternate routes.
Matlab is another of my key tools especially for fast prototyping and
data display/visualization although I have never used it for a
production system as many do.