I would like to work with developer to improve solve in Calc (Jean-Baptiste Faure)
probabilisticdatainterp at gmail.com
Mon Apr 20 21:42:47 UTC 2020
I'm open to reworking the Fortran and using the Fortran-C binding.
Regarding the Libre Office interface, I would argue that this can be
simplified to setting up Target cells (any number) returning specified
values to be obtained by varying Source-cells (any number). In the
optimization, GN would minimize the sum of squares (Target minus
specified values, called "residuals"). Note that GN makes use of the
residuals themselves, not just the sum of squares. Limiting conditions
of the form of less-than or greater-than constraints can be placed on
Source cells. The variables in GN are real numbers, so the integer and
binary constraints don't apply; however, it would seem in this case that
all values of the source cells with these constraints could be checked
to find the lowest sum of squares.
You do not need to translate the Fortran in another language, you could
embed it in C/C++ program using Fortran-C binding (ISO_C_BINDING
That said the Fortran code of GN could be improved to make it more
portable. I can help if you decide to keep the Fortran code.
On 4/19/2020 6:00 AM, libreoffice-request at lists.freedesktop.org wrote:
> Re: I would like to work with developer to improve solve in
> Calc (Jean-Baptiste Faure)
More information about the LibreOffice