Skip to content

Latest commit

 

History

History
78 lines (64 loc) · 2.41 KB

INSTALLATION.md

File metadata and controls

78 lines (64 loc) · 2.41 KB

Installation Instructions

Enabling WSL on Windows 10

(Skip this step if you're on Ubuntu).

Run on Powershell with administrator rights:

Enable-WindowsOptionalFeature -Online -FeatureName Microsoft-Windows-Subsystem-Linux

Then restart when prompted.

Required Packages

Install the following packages:

sudo apt-get install gcc g++ gfortran subversion patch wget cmake build-essential

(OPTIONAL) Install Intel Math Kernal Library for Linux

Using Intel MKL results in better performance. Instructions on installing it is here.

Once you're done with the Setting up the Repository steps, install the intel-mkl package.

Then, once done installing, in ~/.profile, add the line (adjusting to to whatever version of the toolset you're using):

source /opt/intel/parallel_studio_xe_2018.3.051/psxevars.sh intel64 -platform linux

Finally either restart or run source ~/.profile to continue.

Install Ipopt

  • We'll download Ipopt: 
    svn co https://projects.coin-or.org/svn/Ipopt/stable/3.12 CoinIpopt
  • Next, we'll download third party licenses. From within the downloaded folder, run: 
    cd ThirdParty/ASL
./get.ASL
cd ../Blas
./get.Blas
cd ../Lapack
./get.Lapack
cd ../Metis
./get.Metis
cd ../Mumps
./get.Mumps
cd ../../
  • (OPTIONAL) Next, aquire HSL.
    • Download the source code for the Linux version.
    • Extract the folder
    • Rename the folder to coinhsl/ and place such that it looks like ThirdParty/HSL/coinhsl/
    • HSL is used in a particular linear solver that we're found to work well. If you cannot get HSL, you'll have to change the solver in our examples.
  • Next, in the root folder of the Ipopt run: 
    mkdir build
cd build
  • If you're using the Intel compiler, run: 
    ../configure --prefix=/usr/local/ ADD_CFLAGS=-fopenmp ADD_FFLAGS=-fopenmp ADD_CXXFLAGS=-fopenmp --with-blas="-L$MKLROOT/lib/intel64 -lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lm"
  • Otherwise, run: 
    ../configure
  • Then run make or make -j4 for 4 threads (or however many you want).
  • Then, run sudo make install
  • Then, run sudo ldconfig

Install JSONlab

Go here to isntall JSONlab on for Matlab.