|
Ipopt Documentation
|
|
|
Ipopt Documentation
|
|
The following sections describe the installation procedures on UNIX/Linux, macOS, and Windows systems.
The configuration script and Makefiles in the Ipopt distribution have been created using GNU's autoconf and automake. In general, you can see the list of options and variables that can be set for the configure script by typing configure --help.
Many Linux distributions will come with all necessary tools. All you should need to do is check the compiler versions. On a Debian-based distribution, you can obtain all necessary tools with the following command:
sudo apt-get install gcc g++ gfortran git patch wget pkg-config liblapack-dev libmetis-dev
Replace apt-get with your relevant package manager, e.g. dnf for RedHat-based distributions, pacman for Arch, etc. The g++ and gfortran compilers may need to be specified respectively as gcc-c++ and gcc-gfortran with some package managers.
If Building for 64-bit integers, a variant of Lapack that uses 64-bit integers is required. On some Linux distributions (e.g., Ubuntu), this is available by installing package liblapack64-dev.
You need either the Xcode Command Line Tools or a community alternative such as Homebrew to install the GNU compilers:
brew update brew install bash gcc brew link --overwrite gcc
Further, pkg-config is required, which can be installed via Homebrew by
brew install pkg-config
If you have Xcode installed, the Command Line Tools are available under Preferences, Downloads. These items unfortunately do not come with a Fortran compiler, but you can get gfortran from http://gcc.gnu.org/wiki/GFortranBinaries#MacOS. In the past, we have been able to compile Ipopt using default Xcode versions of gcc and g++ and a newer version of gfortran from this link, but consistent version numbers may be an issue in future cases.
If you intend to use MUMPS or one of the HSL linear solvers, then they may profit from an installation of METIS. This can be accomplished via
brew install metis
For Windows, one can compile Ipopt under MSYS2/MinGW, which is a light-weight UNIX-like environment for Windows.
MSYS2/MinGW can be obtained from https://www.msys2.org/. After installation, a number of additional packages need to be installed:
pacman -S binutils diffutils git grep make patch pkg-config
(I might have forgotten something.) If you also want to use GNU compilers, then install them, too:
pacman -S mingw-w64-x86_64-gcc mingw-w64-x86_64-gcc-fortran
When using MinGW, it is convenient to install MinGW's Lapack and METIS as well:
pacman -S mingw-w64-x86_64-lapack mingw-w64-x86_64-metis
To use MSYS2/MinGW to compile Ipopt with native MSVC/Intel compilers, the basic version MSYS2 version without MinGW is sufficient. To use the GNU compilers, MinGW is required.
If you want to use the native MSVC or Intel compiler, you need to make sure that environment variables are setup for the compilers. One way is to create a shell script that executes the batch files that are provided by Microsoft and Intel for this purpose before starting MSYS2. For using MSVC, this script may have a form similar to
set MSYS2_PATH_TYPE=inherit call "C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\vcvarsall.bat" C:\msys64\msys2.exe exit 0
For using Intel C/C++ and Fortran compilers, the 2nd line may be replaced by something like
call "C:\Program Files (x86)\IntelSWTools\compilers_and_libraries\windows\bin\ipsxe-comp-vars.bat" intel64 vs2017
You might have to search around a bit. The important thing is that, after your change, you can type cl and it finds the Microsoft C++ compiler and, if installed, you can type ifort and it finds the Intel Fortran compiler.
--enable-msvc as argument to configure below.Ipopt uses a few external packages that are not included in the Ipopt source code distribution, for example ASL (the AMPL Solver Library if you want to compile the Ipopt AMPL solver executable), Blas, Lapack.
Ipopt also requires at least one linear solver for sparse symmetric indefinite matrices. There are different possibilities, see below. It is important to keep in mind that usually the largest fraction of computation time in the optimizer is spent for solving the linear system, and that your choice of the linear solver impacts Ipopt's speed and robustness. It might be worthwhile to try different linear solver to experiment with what is best for your application.
Since this third party software is released under different licenses than Ipopt, we cannot distribute their code together with the Ipopt packages and have to ask you to go through the hassle of obtaining it yourself. Keep in mind that it is still your responsibility to ensure that your downloading and usage of the third party components conforms with their licenses.
For ASL, HSL, and MUMPS, COIN-OR provides specialized build systems that produce libraries which are easy to use with the build system of Ipopt. Some of the specialized build systems also come with a script that makes it easy to download the code using wget or curl.
Note that you only need to obtain the ASL if you intend to use Ipopt from AMPL. It is not required if you want to specify your optimization problem in a programming language (e.g., C++, C, or Fortran).
The COIN-OR Tools project ThirdParty-ASL can be used to download the ASL code and build an ASL library that is recognized by Ipopt. To do so, execute
git clone https://github.com/coin-or-tools/ThirdParty-ASL.git cd ThirdParty-ASL ./get.ASL ./configure make sudo make install
Ipopt's build system tries to find an installation of BLAS and LAPACK in your system. On Ubuntu, such libraries can be made available by installing liblapack-dev. Other Linux distributions typically also offer a way to install Blas and Lapack via its package manager. On macOS, Ipopt's build system should pick up the Blas and Lapack from the Accelerate framework.
--with-lapack-lflags flag when running configure (see Compiling and Installing Ipopt): ./configure --with-lapack-lflags="-L$HOME/lib -lf77blas -lcblas -latlas"The configure script already tries to find Intel MKL libraries on some systems. If that fails, e.g., because the libraries are not available in a standard search path, then also use the
--with-lapack-lflags flag to explictly state what flags are necessary to use MKL, e.g., ./configure --with-lapack-lflags="-L${MKLROOT}/lib/intel64 -Wl,--no-as-needed -lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lm"
There are two versions of HSL available:
To obtain the HSL code, you can follow the following steps:
You may either:
To compile the HSL code via the COIN-OR Tools project ThirdParty-HSL, run
git clone https://github.com/coin-or-tools/ThirdParty-HSL.git cd ThirdParty-HSL
Now unpack the HSL sources archive, move and rename the resulting directory so that it becomes ThirdParty-HSL/coinhsl. Then, in ThirdParty-HSL, configure, build, and install the HSL sources:
./configure make sudo make install
CoinHslConfig.h header file and the linker flags for this library with the flags --with-hsl-cflags and --with-hsl-lflags flags, respectively, when running configure of Ipopt (see Compiling and Installing Ipopt).Alternatively to linking against HSL routines when Ipopt is build, Ipopt can also load a shared library with MA27, MA57, HSL_MA77, HSL_MA86, HSL_MA97, or MC19 at runtime. This means, if you obtain one of those solvers after you compiled Ipopt, you don't need to recompile Ipopt to use it. Instead, you can just place a shared library onto your system and ensure that Ipopt can find it at runtime. Option hsllib can be set to the name of the library from which to load HSL routines if a HSL solver is selected. The name can contain a path; otherwise, the shared library search path (e.g., LD_LIBRARY_PATH) will be used.
You can also use the (public domain) sparse linear solver MUMPS. Please visit the MUMPS home page for more information about the solver. MUMPS is provided as Fortran 90 and C source code. You need to have a Fortran 90 compiler (e.g., the GNU compiler gfortran) to be able to use it.
The COIN-OR Tools project ThirdParty-Mumps can be used to download the MUMPS code and build a MUMPS library that is recognized by Ipopt. To do so, execute
git clone https://github.com/coin-or-tools/ThirdParty-Mumps.git cd ThirdParty-Mumps ./get.Mumps ./configure make sudo make install
MPI_Init() and MPI_Finalize() when loaded and unloaded, respectively. To prevent this, use the configure flag --disable-mpiinit. Note that the Ipopt executable, tests, and examples will not work in this case (if using MUMPS as linear solver).stable/2.1 of project ThirdParty-Mumps can be used to build a library of MUMPS 4 that is usable with Ipopt. Initial experiments on the CUTEst testset have shown that performance with MUMPS 5.2.x is worse than with MUMPS 4.10.0 on average, but note that MUMPS 4 is not longer maintained.To compile Ipopt with a precompiled MUMPS library, you need to specify the -I flag to have the compiler find the directory containing the MUMPS header files with the --with-mumps-cflags flag of Ipopt's configure, e.g.,
--with-mumps-cflags="-I$HOME/MUMPS/include"
and you also need to provide the linker flags for MUMPS with the --with-mumps-lflags flag.
If you would like to use Ipopt with Pardiso from Pardiso Project, you need to obtain the Pardiso library from http://www.pardiso-project.org for your operating system.
From http://www.pardiso-project.org, you can obtain a limited time license of Pardiso for academic or evaluation purposes or buy a non-profit or commercial license. Make sure you read the license agreement before filling out the download form.
To avoid conflicts with Pardiso from Intel MKL, Ipopt only offers the option to load Pardiso from Pardiso Project at runtime. Therefore, the Ipopt interface to Pardiso from Pardiso Project is always build and the value of option pardisolib is used as name (possibly including a path) for the library to be loaded. The configure option --with-pardiso can be used to specify the default value for option pardisolib. Further, if --with-pardiso is specified and HSL codes are not available at compile time, then the default of option linear_solver is set to pardiso. For example, if you want that Ipopt loads the GCC version of Pardiso on a GNU/Linux system and this library is located in $HOME/lib, you should add the flag
--with-pardiso="$HOME/lib/libpardiso720-GNU831-X86-64.so"
--with-pardiso used to specify the corresponding linker flags. This has been changed with Ipopt 3.14!If you are using the parallel version of Pardiso, you need to specify the number of processors it should run on with the environment variable OMP_NUM_THREADS, as described in the Pardiso manual.
If you want to compile Ipopt with the Pardiso library that is included in Intel MKL, it is sufficient to ensure that MKL is used for the linear algebra routines (Blas/Lapack), see BLAS and LAPACK.
If configure finds that Pardiso is available with Lapack, option linear_solver can be set to pardisomkl to use Pardiso from Intel MKL. Specifying --disable-pardisomkl as option for configure disables the check for Pardiso from Intel MKL.
pardiso. With Ipopt 3.14, value pardiso is reserved for Pardiso from Pardiso Project and value pardisomkl has to be used to select Pardiso from Intel MKL.To compile Ipopt with the linear solver SPRAL, SPRAL should first be downloaded and compiled according to the associated SPRAL compilation guide. Alternatively, consider using the LANL ANSI fork of SPRAL. The maintainer of this fork test whether their current version works with Ipopt.
After compilation of SPRAL, while configuring Ipopt, you need to specify the link and include directories for the library with the --with-spral-lflags and --with-spral-cflags flags, including required additional libraries and flags. For example, if you want to compile Ipopt with a purely CPU-parallel version of SPRAL (libs in /path/to/spral/lib and headers in /path/to/spral/include) on a GNU/Linux system, you should add the following flags to the configure command:
--with-spral-lflags="-L/path/to/spral/lib -lspral -lgfortran -lhwloc -lm -lcoinmetis -lopenblas -lstdc++ -fopenmp" --with-spral-cflags="-I/path/to/spral/include"
For a CPU and graphics processing unit- (GPU-) parallel version of SPRAL, the following flags should be used, instead:
--with-spral-lflags="-L/path/to/spral/lib -lspral -lgfortran -lhwloc -lm -lcoinmetis -lopenblas -lstdc++ -fopenmp -lcudadevrt -lcudart -lcuda -lcublas" --with-spral-cflags="-I/path/to/spral/include"
The above assume a SPRAL compilation using OpenBLAS libraries. For best performance, use SPRAL together with linear algebra routines from the Intel MKL (see BLAS and LAPACK).
export OMP_CANCELLATION=TRUE export OMP_NESTED=TRUE export OMP_PROC_BIND=TRUE
A guide for building Ipopt with Spral on Ubuntu Linux is also avaiable here (some GitHub URLs need to be adjusted to use coin-or/Ipopt instead of lanl-ansi/Ipopt). More information on the Spral interface of Ipopt can be found in [9].
If you would like to compile Ipopt with WSMP, you need to obtain the WSMP library for your operating system. Information about WSMP can be found at http://www.research.ibm.com/projects/wsmp.
At this website you can download the library for several operating systems including a trial license key for 90 days that allows you to use WSMP for "educational, research, and benchmarking purposes by non-profit academic institutions" or evaluation purposes by commercial organizations; make sure you read the license agreement before using the library. Once you obtained the library and license, please check if the version number of the library matches the one on the WSMP website.
If a newer version is announced on that website, you can (and probably should) request the current version by sending a message to wsmp@us.ibm.com. Please include the operating system and other details to describe which particular version of WSMP you need.
To compile Ipopt with the linear solver WSMP, you need to specify the linker flags for the library with the --with-wsmp flag, including required additional libraries and flags. For example, if you want to compile Ipopt with WSMP (located in $HOME/lib) on an Intel IA32 Linux system, you may use the configure flag
--with-wsmp="$HOME/lib/wsmp/wsmp-Linux/lib/IA32/libwsmp.a -lpthread -lblas -llapack -lpthread -lm -lgfortran"
But the actual flags depend on WSMP library and your preference for the Blas/Lapack libraries.
Ipopt is available from the COIN-OR group at GitHub. You can either download the code using git or simply retrieve a tarball (compressed archive file). While the tarball is an easy method to retrieve the code, using the git system allows users the benefits of the version control system, including easy updates and revision control.
Of course, the git client must be installed on your system if you want to obtain the code this way (the executable is called git). Information about git and how to download it can be found at http://git-scm.com.
To obtain the Ipopt source code via git, change into the directory in which you want to create a subdirectory Ipopt with the Ipopt source code. Then follow the steps below:
git clone https://github.com/coin-or/Ipopt.git
cd Ipopt
In the following, $IPOPTDIR will refer to the directory in which you are right now (output of pwd).
To use the tarball, follow the steps below:
Ipopt-x.y.z.tgz, where x.y.z is the version number, such as 3.12.0.gunzip Ipopt-x.y.z.tgz tar xvf Ipopt-x.y.z.tar
mv Ipopt-x.y.z Ipopt
cd Ipopt
In the following, $IPOPTDIR will refer to the directory in which you are right now (output of pwd).
Ipopt can be easily compiled and installed with the usual configure, make, make install commands. We follow the procedure that is used for most of the COIN-OR projects, based on the GNU autotools.
Below are the basic steps for the Ipopt compilation that should work on most systems. For special compilations and for some troubleshooting see Flags to configure before creating an issue or sending a message to the mailing list.
Create a directory where you want to compile Ipopt, for example
mkdir $IPOPTDIR/build
and change into this directory
cd $IPOPTDIR/build
Run the configure script:
$IPOPTDIR/configure
One might have to give options to the configure script, e.g., in order to choose a non-default compiler, or to tell it where to install Ipopt eventually, see Flags to configure.
If the last output line reads configure: Main configuration of Ipopt successful then everything worked fine. Otherwise, look at the screen output, have a look at the config.log output file and/or consult Flags to configure.
The default configure (without any options) is sufficient for most users that downloaded and installed the source code for the linear solver and other dependencies. If you want to see the configure options, consult Flags to configure.
Build the code:
make
Note: If you are using GNU make, you can also try to speed up the compilation by using the -jN flag (e.g., make -j3), where N is the number of parallel compilation jobs. A good number for N is the number of available processors plus one. Under some circumstances, this fails, and you might have to re-issue the command, or omit the -j flag.
If you want, you can run a short test to verify that the compilation was successful. For this, you just enter
make test
This will test if the AMPL solver executable works (if you got the ASL code) and if the included C++, C, Java, and Fortran examples work.
configure again, and this time specify all C++ runtime libraries with the CXXLIBS variable (see also Flags to configure).Install Ipopt:
sudo make install
sudo has been added here as make install installs files into /usr/local, where only system administrators have write permissions. This installs
sIpopt AMPL solver executables (if ASL source was downloaded) in /usr/local/bin,sIpopt libraries (libipopt.so, libipopt.a or similar) in /usr/local/lib,sIpopt libraries (ipopt.pc, sipopt.pc) in /usr/local/lib/pkgconfig.sIpopt header files /usr/local/include/coin-or,/usr/local/share/java.You can change the default installation directory (here /usr/local) by using the --prefix switch for configure. If this is a directory that you have write permissions for, sudo can be omitted for make install.
In general, you can see the list of options and variables that can be set for the configure script by typing configure --help.
Below a few particular options are discussed:
If you want to specify that you want to use particular compilers, you can do so by adding the variables definitions for CXX, CC, and F77 to the ./configure command line, to specify the C++, C, and Fortran 77 compiler, respectively. For example,
./configure CXX=g++-8.0.0 CC=gcc-8.0.0 F77=gfortran-8.0.0
In order to set the compiler flags, you should use the variables CXXFLAGS, CFLAGS, FFLAGS. Note, that the Ipopt code uses dynamic_cast. Therefore it is necessary that the C++ code is compiled including RTTI (Run-Time Type Information). Some compilers need to be given special flags to do that (e.g., -qrtti=dyna for the AIX xlC compiler).
--disable-shared flag. If you want to compile both shared and static libraries, you should specify the --enable-static flag. This is no possible on Windows.It is possible to compile the Ipopt library in a debug configuration, by specifying --enable-debug. This switches the default compiler flags (CFLAGS, CXXFLAGS, FFLAGS) to instruct the compiler to include debug information into object files and disables compiler optimization.
Also, you can tell Ipopt to do some additional runtime sanity checks, by specifying the flag --with-ipopt-checklevel.
This usually leads to a significant slowdown of the code, but might be helpful when debugging something.
The checklevel value has the following effect:
If you want to link the Ipopt library with a main program written in C or Fortran, the C and Fortran compiler doing the linking of the executable needs to be told about the C++ runtime libraries. Unfortunately, the build system does not provide a functionality for automatic detection of those libraries. We have hard-coded some default values for some systems and compilers, but this might not work all the time.
If you have problems linking your Fortran or C code with the Ipopt library libipopt.a and the linker complains about missing symbols from C++ (e.g., the standard template library), you should specify the C++ libraries with the CXXLIBS variable. To find out what those libraries are, it is probably helpful to link a simple C++ program with verbose compiler output.
For example, for the Intel compilers on a Linux system, you might need to specify something like
./configure CC=icc F77=ifort CXX=icpc CXXLIBS='-L/usr/lib/gcc-lib/i386-redhat-linux/3.2.3 -lstdc++'
./configure AR='ar -X64' NM='nm -X64' CC='xlc -q64' F77='xlf -q64' CXX='xlC -q64' CFLAGS='-O3 -bmaxdata:0x3f0000000' FFLAGS='-O3 -bmaxdata:0x3f0000000' CXXFLAGS='-qrtti=dyna -O3 -bmaxdata:0x3f0000000'(Alternatively, a simpler solution for AIX is to set the environment variable
OBJECT_MODE to 64.)Ipopt's Java interface JIpopt uses the Java Native Interface (JNI), which is a programming framework that allows Java code running in the Java Virtual Machine (JVM) to call and be called by native applications and libraries written in languages such as C and C++.
configure tries to find a Java compiler and the necessary compiler flags to locate the JNI header files automatically. You can influence this automated check by setting the environment variable JAVA_HOME to the directory that contains your JDK.
C:\Program Files\Java, try setting JAVA_HOME to the DOS equivalent C:\Progra~1\Java (or similar).You can disable the checks for Java and build of the Java interface by specifying the configure flag --disable-java. This is also the default if JAVA_HOME contains a space or the build of shared libraries has been disabled as JIpopt needs to be able to load the Ipopt library dynamically at runtime.
--disable-linear-solver-loader.--disable-sipopt.Configure flags specific to the use of Ipopt's dependencies were discussed in the corresponding sections of Download, build, and install dependencies.
A few flags that influence which Ipopt code is compiled are not controlled by parameters for configure. We list some of these flags in the following.
IPOPT_THREAD_LOCAL: Keyword to be used to declare a thread-local variable.IPOPT_DEPRECATED: Keyword to be used to mark a function as deprecated.FUNNY_MA57_FINT: If defined, then assume that ptrdiff_t corresponds to the integer type used by HSL MA57, e.g., 64-bit integers on 64-bit systems. This flag was useful when using MA57 from libraries that come with Matlab.MA97_DUMP_MATRIX: If defined, then option ma97_dump_matrix is available. This requires linking against a HSL library (not loading at runtime) that includes function dump_mat_csc.MUMPS_DUMP_MATRIX: If defined, then matrices are written to stdout in the MUMPS interface.IPOPT_MUMPS_NOMUTEX: If defined, concurrent calls of Ipopt into MUMPS from different threads are not prevented by a mutex.On sufficiently prepared systems, an alternative way to download, build, and install Ipopt with dependencies ASL, HSL, and MUMPS is to use the coinbrew script from https://coin-or.github.io/coinbrew/
After obtaining the script, run
/path/to/coinbrew fetch Ipopt --no-prompt /path/to/coinbrew build Ipopt --prefix=/dir/to/install --test --no-prompt --verbosity=3 /path/to/coinbrew install Ipopt --no-prompt
The coinbrew script will take care of building and installing the dependencies ASL and MUMPS before building Ipopt. To use HSL sources, change to the ThirdParty/HSL directory created by coinbrew fetch and follow the download instructions (i.e., skip the build step) from HSL (Harwell Subroutines Library). Then proceed with the coinbrew build step.
More details on using coinbrew can be found at the instructions on Getting Started with the COIN-OR Optimization Suite.
The ipoptr interface can be build after Ipopt has been build and installed. In the best case, it is sufficient to execute the following command in R:
install.packages('$IPOPTDIR/build/contrib/RInterface', repos=NULL, type='source')
After installation of the ipoptr package, it should be possible to load the package in R and to view the help page:
library('ipoptr')
?ipoptr
Ipopt by default uses double-precision floating point arithmetic. Using the configure flag --with-precision=float, it is possible to build a variant of Ipopt that uses single-precision floating point arithmetic. It is not possible to build for both single- and double-precision simultaneously. In the single-precision configuration, the types Ipopt::Number and ipnumber are defined to be float and the C preprocessor variable IPOPT_SINGLE is defined in IpoptConfig.h.
If the Ipopt code is compiled for single-precision, then also third-party codes that interface with or are interfaced by Ipopt need to support this floating point format. Regarding included code that interfaces to Ipopt, the situation is as follows:
float.float.real instead of double precision. A single-precision variant of the HS071 example is available in examples/hs071_f/hs071_fs.f.float instead of double for real numbers. A single-precision variant of the HS071 example is available in examples/hs071_java/HS071s.java. Note, that for the single-precision version of class org.coinor.Ipopt, all mention of double are replaced by float before compilation.Regarding dependencies of Ipopt, the situation is as follows:
--with-lapack-lflags="-L/usr/local/opt/lapack/lib -llapack -lblas") or try out this vecLib wrapper.When building Ipopt for single-precision arithmetic, the default for option tol is changed to 1e-5.
Ipopt by default defines its types Ipopt::Index (C++) and ipindex (C) to be int, which on many systems has a size of 32-bit. This limits the size of problems to be solved to 2^31 variables, constraints, nonzero elements in matrices, etc. Using the configure flag --with-intsize=64, it is possible to build a variant of Ipopt that defines Ipopt::Index/ipindex to be int64_t, i.e., an integer of size 64 bit. Further, the C preprocessor variable IPOPT_INT64 is defined in IpoptConfig.h.
Note that using 64-bit integers increases the memory usage of Ipopt significantly since type Ipopt::Index is used for almost all integers in Ipopt. It is not possible to build for both 32- and 64-bit integers simultaneously.
If the Ipopt code is compiled for 64-bit integers, then also third-party codes that interface with or are interfaced by Ipopt may need to support this integer size. Regarding included code that interfaces to Ipopt, the situation is as follows:
int64_t.int64_t.INTEGER.ASL_big_goff defined (this define also need to be present in the ASL header files that the Ipopt build includes). If using ThirdParty-ASL, then this achieved by selecting configure option --with-intsize=64, see also the ThirdParty-ASL docu. Note that problems instances are still limited to at most 2^31 variables and constraints - this is a limitation of AMPL or ASL.Regarding dependencies of Ipopt, the situation is as follows:
--with-intsize=64 is seletected, then the check in configure by default only looks for Intel MKL and tries -llapack64 -lblas64, see also the BuildTools docu.