LAPACK is written in Fortran 90 and provides routines for solving systems of simultaneous linear equations, least-squares solutions of linear systems of equations, eigenvalue problems, and singular value problems. The associated matrix factorizations (LU, Cholesky, QR, SVD, Schur, generalized Schur) are also provided, as are related computations such as reordering of the Schur factorizations and estimating condition numbers. Dense and banded matrices are handled, but not general sparse matrices. In all areas, similar functionality is provided for real and complex matrices, in both single and double precision.
I have been writing code professionally in C# since around 2001 and I love the language. I also love math and the thought of recreating algorithms in C# that somebody had already coded seemed like a big waste of time.
I stumbled upon LAPACK since I was working on a mapping app that needed to solve thousands of simultaneous equations. I figured why not use code that had been battle tested for years.
One problem it was written in Fortran. So I boned up on Fortran using the gfortran compiler on Linux and used the C/PInvoke feature of C#. (See my first foray at the git below:) to call Fortran routines which use the iso_c_binding so that the subroutines can be called by name from c (Fortran name mangles subroutines by default)
From Fortran Wiki:
iso_c_binding is a standard intrinsic module which defines named constants, types, and procedures for C interoperability.
The code in this git exploits this interoperability to call Fortran routines from C# using P/Invoke (Platform Invocation Services). Below is a Fortran subroutine we wish to call:
https://gist.github.com/5eb3cd43a409391bdced3ea1b909646c
This example can solve a series of simultaneous equations.
Below is the C# code which will call the Fortran subroutine.
undefined
The assumption is that you have gfortran, make and the .NET 5.0 SDK installed either on Linux or on Windows in the WSL Linux subsystem. If not, directions for Ubuntu are given below.
https://gist.github.com/08e34dcfd2307925ccafc4277ae036cf
https://gist.github.com/3e71a2c24b9eaeddd7cb2e4fc27095fd
Let's walk through the steps to create a C# .NET 5.0 project to call the above Fortran subroutine.
Open a terminal window and enter the following (I am doing it from my HOME directory):
https://gist.github.com/f70c0480bac259cd4fc2d0f6bd4f5ac6
Change directories to CSharpCallingLaPack
https://gist.github.com/380007b99c59b63afdba5d11fc689eb8
With your trusty editor (I am using VSCode) edit the Program.cs and cut and paste the code from below into it:
https://gist.github.com/5dfbcdd3a6d6655391baed579fa34bda
I have provided the libraries on my github site so you can just wget them into your /usr/local/lib directory.
Enter the following into a terminal window.
undefined
Change back to the CSharpCallingLaPack directory and with your favorite editor cut and paste the Fortran code below and save it.
https://gist.github.com/53f067696811d69263ad8038ebe8427d
Enter the following in your terminal window to create the Fortran shared library.
https://gist.github.com/d5085cf4de3144e07d1df1fb9ad4fecd
Enter the following in your terminal window to build the C# project and copy the Fortran module
https://gist.github.com/90d9a40c9291a008572751b92f448641
Run the following from the terminal window to run the app.
https://gist.github.com/ad951ab2b0a526027b1806ebbbe336fc
You should see the following output:
https://gist.github.com/1df0aaca0b6eb5a219b91df6dd201fe4
I was surprised by how much I enjoyed working with Fortran and the huge amount of mathematical library code that is written in it. The cool thing about Fortran is that optimizing compilers (Intel) can really speed up operations that involve huge arrays and matrices.
The source code for the above project and other projects I am working on can be found at:
Thanks for reading!