Re: Math Package to Solve Linear Equations?
On Fri, Oct 05, 2007 at 09:45:11PM -0400, Kamaraju S Kusumanchi wrote:
> Thomas H. George wrote:
>
> > I'm feeling stupid. I used to have a math package which inverted
> > matrices to solve systems of linear equations - i.e. enter the matrix
> > and the y values and the program inverts the matrix and reports the x
> > values. I know how to do it manually but it is laborious for large
> > matrices. Perhaps Openoffice.calc/solver does this but it is not clear
> > to me how to enter data for solver to do this. apt-cache search matrix
> > |grep inversion doesn't turn up anything.
> >
> > Would someone please beat me over the head and point me to a simple
> > package to do this job?
>
> Libraries (to call from C, C++, Fortran 90 codes):
>
> Petsc - pretty much everyone in the numerical analysis community rely on
> this for solving large scale linear problems. Has good parallelization
> (MPI) support, lot of sparse, direct solvers etc.,
>
> lapack95 - Good if you are using fortran as your programming language. Has
> been around for sometime. Code is well tested but does not have as many
> routines as petsc. No parallel solvers or sparse solvers AFAIK.
>
> IMSL - proprietary.
>
> Software suites:
>
> Matlab - proprietary, comes with a good GUI, lots of routines to plot
> graphs, contours etc., contains lot of sparse solvers.
>
> Octave - Use versions > 2.9.12 for better experience. Older versions are
> incompatible with newer versions. Does not have a GUI provided by the
> upstream, plotting capabilities are kind of limited. Does not have that
> many sparse solvers. But if it works, it does the job well and the code is
> actively developed. It integrates well with Linux (for example it uses
> gnuplot for plotting, lapack for solving system of matrices etc.,).
Add scipy to that list. The scientific extension to the python language.
Oli
Reply to: