Gosl is a set of tools for developing scientific simulations using the Go language. We mainly consider the development of numerical methods and solvers for differential equations. We also present some functions for fast Fourier transforms, the generation of random numbers, probability distributions, and computational geometry.
This library contains essential functions for linear algebra computations (operations between all combinations of vectors and matrices, eigenvalues and eigenvectors, linear solvers) and the development of numerical methods (e.g., numerical quadrature).
We link Gosl with existing libraries written in C and Fortran, such as OpenBLAS, LAPACK, UMFPACK, MUMPS, QUADPACK, and FFTW3. These libraries have been fundamental for the development of high-performant simulations over many years. We believe that it is nearly impossible to rewrite these libraries in native Go and achieve the same speed delivered by them. For reference, a naive implementation of matrix-matrix multiplication in Go is more than 100 times slower than OpenBLAS.
Because of the other libraries, Docker is the easiest way to work with Gosl. Having Docker and VS Code installed, you can start developing powerful numerical simulations using Gosl in a matter of minutes. Furthermore, it works out of the box on Windows, Linux, and MacOS.
Done. And your system will "remain clean."
First, install Go as explained in https://golang.org/doc/install
Second, install some libraries:
sudo apt-get install \
gcc \
gfortran \
libfftw3-dev \
liblapacke-dev \
libmetis-dev \
libmumps-seq-dev \
libopenblas-dev \
libsuitesparse-dev
Finally, download and compile Gosl:
git clone https://github.com/cpmech/gosl.git
cd gosl
./all.bash
Done. Installation completed.
Gosl includes the following essential packages:
Gosl includes the following main packages:
(see each subdirectory for more information)
We have removed the previous mpi
sub-package for maintenance (see next section). However, if you plan to use MPI, we recommend the external library gompi.
The previous version, including more packages, is available here and can be used with the Docker image 1.1.3 as in this hello gosl example.
These other packages, such as machine learning, plotting, etc., have been removed because they do not depend on CGO and may be developed independently. We can now more efficiently maintain the core of Gosl, which focuses on the foundation for other scientific codes.