Skip to content

Latest commit

 

History

History
452 lines (349 loc) · 20.4 KB

README.md

File metadata and controls

452 lines (349 loc) · 20.4 KB

SuiteSparse: A Suite of Sparse matrix packages at http://suitesparse.com

May 17, 2021. SuiteSparse VERSION 5.10.1

Now includes GraphBLAS, SLIP_LU, and a new interface to the SuiteSparse
Matrix Collection (ssget), via MATLAB and a Java GUI, to
http://sparse.tamu.edu.

Primary author of SuiteSparse (codes and algorithms, excl. METIS): Tim Davis

Code co-authors, in alphabetical order (not including METIS):

Patrick Amestoy, David Bateman, Jinhao Chen.  Yanqing Chen, Iain Duff,
Les Foster, William Hager, Scott Kolodziej, Chris Lourenco, Stefan
Larimore, Erick Moreno-Centeno, Ekanathan Palamadai, Sivasankaran
Rajamanickam, Sanjay Ranka, Wissam Sid-Lakhdar, Nuri Yeralan.

Additional algorithm designers: Esmond Ng and John Gilbert.

Refer to each package for license, copyright, and author information. All codes are authored or co-authored by Timothy A. Davis.


About the BLAS and LAPACK libraries

NOTE: Use of the Intel MKL BLAS is strongly recommended. A recent OpenBLAS can result in severe performance degradation. The reason for this is being investigated, and this may be resolved in the near future. Ignore the comments about OpenBLAS in the various user guides; those are out of date.


SuiteSparse/README

Packages in SuiteSparse, and files in this directory:

GraphBLAS   graph algorithms in the language of linear algebra.
            https://graphblas.org
            A stand-alone package that uses cmake to compile; see
            GraphBLAS/README.txt.  The rest of SuiteSparse still uses
            'make'.  A cmake setup for all of SuiteSparse is in progress.
            author: Tim Davis

SLIP_LU     solves sparse linear systems in exact arithmetic.
            Requires the GNU GMP and MPRF libraries.

AMD         approximate minimum degree ordering.  This is the built-in AMD
            function in MATLAB.
            authors: Tim Davis, Patrick Amestoy, Iain Duff

bin         where the metis-5.1.0 programs are placed when METIS is compiled

BTF         permutation to block triangular form
            authors: Tim Davis, Ekanathan Palamadai

CAMD        constrained approximate minimum degree ordering
            authors: Tim Davis, Patrick Amestoy, Iain Duff, Yanqing Chen

CCOLAMD     constrained column approximate minimum degree ordering
            authors: Tim Davis, Sivasankaran Rajamanickam, Stefan Larimore.
                Algorithm design collaborators: Esmond Ng, John Gilbert
                (for COLAMD)

ChangeLog   a summary of changes to SuiteSparse.  See */Doc/ChangeLog
            for details for each package.

CHOLMOD     sparse Cholesky factorization.  Requires AMD, COLAMD, CCOLAMD,
            the BLAS, and LAPACK.  Optionally uses METIS.  This is chol and
            x=A\b in MATLAB.
            author for all modules: Tim Davis 
            CHOLMOD/Modify module authors: Tim Davis and William W. Hager

COLAMD      column approximate minimum degree ordering.  This is the
            built-in COLAMD function in MATLAB.
            authors (of the code): Tim Davis and Stefan Larimore
            Algorithm design collaborators: Esmond Ng, John Gilbert

Contents.m  a list of contents for 'help SuiteSparse' in MATLAB.

CSparse     a concise sparse matrix package, developed for my
            book, "Direct Methods for Sparse Linear Systems",
            published by SIAM.  Intended primarily for teaching.
            It does have a 'make install' but I recommend using
            CXSparse instead.  In particular, both CSparse and CXSparse
            have the same include filename: cs.h.

            This package is used for the built-in DMPERM in MATLAB.
            author: Tim Davis

CSparse_to_CXSparse
            a Perl script to create CXSparse from CSparse and
            CXSparse_newfiles
            author: David Bateman, Motorola

CXSparse    CSparse Extended.  Includes support for complex matrices
            and both int or long integers.  Use this instead of CSparse
            for production use; it creates a libcsparse.so (or *dylib on
            the Mac) with the same name as CSparse.  It is a superset
            of CSparse.  Any code that links against CSparse should
            also be able to link against CXSparse instead.
            author: Tim Davis, David Bateman

CXSparse_newfiles
            Files unique to CXSparse
            author: Tim Davis, David Bateman

share       'make' places documentation for each package here

include     'make' places user-visible include fomes for each package here

KLU         sparse LU factorization, primarily for circuit simulation.
            Requires AMD, COLAMD, and BTF.  Optionally uses CHOLMOD,
            CAMD, CCOLAMD, and METIS.
            authors: Tim Davis, Ekanathan Palamadai

LDL         a very concise LDL' factorization package
            author: Tim Davis

lib         'make' places shared libraries for each package here

Makefile    to compile all of SuiteSparse
            make            compiles SuiteSparse libraries and runs demos
            make install    compiles SuiteSparse and installs in the
                            current directory (./lib, ./include).
                            Use "sudo make INSTALL=/usr/local" to install
                            in /usr/local/lib and /usr/local/include.
            make uninstall  undoes 'make install'
            make library    compiles SuiteSparse libraries (not demos)
            make distclean  removes all files not in distribution, including
                            ./bin, ./share, ./lib, and ./include.
            make purge      same as 'make distclean'
            make clean      removes all files not in distribution, but
                            keeps compiled libraries and demoes, ./lib,
                            ./share, and ./include.
            make config     displays parameter settings; does not compile

            Each individual package also has each of the above 'make'
            targets.  Doing 'make config' in each package */Lib directory
            displays the exact shared and static library names.

            Things you don't need to do:
            make cx         creates CXSparse from CSparse
            make docs       creates user guides from LaTeX files
            make cov        runs statement coverage tests (Linux only)
            make metis      compiles METIS (also done by 'make')

MATLAB_Tools    various m-files for use in MATLAB
            author: Tim Davis (all parts)
            for spqr_rank: author Les Foster and Tim Davis

            Contents.m      list of contents
            dimacs10        loads matrices for DIMACS10 collection
            Factorize       object-oriented x=A\b for MATLAB
            find_components finds connected components in an image
            GEE             simple Gaussian elimination
            getversion.m    determine MATLAB version
            gipper.m        create MATLAB archive
            hprintf.m       print hyperlinks in command window
            LINFACTOR       predecessor to Factorize package
            MESHND          nested dissection ordering of regular meshes
            pagerankdemo.m  illustrates how PageRank works
            SFMULT          C=S*F where S is sparse and F is full
            shellgui        display a seashell
            sparseinv       sparse inverse subset
            spok            check if a sparse matrix is valid
            spqr_rank       SPQR_RANK package.  MATLAB toolbox for rank
                            deficient sparse matrices: null spaces,
                            reliable factorizations, etc.  With Leslie
                            Foster, San Jose State Univ.
            SSMULT          C=A*B where A and B are both sparse
            SuiteSparseCollection    for the SuiteSparse Matrix Collection
            waitmex         waitbar for use inside a mexFunction

            The SSMULT and SFMULT functions are the basis for the
            built-in C=A*B functions in MATLAB.

Mongoose    graph partitioning.
            authors: Nuri Yeralan, Scott Kolodziej, William Hager, Tim Davis

metis-5.1.0 a modified version of METIS.  See the README.txt files for
            details.
            author: George Karypis; not an integral component of
            SuiteSparse, however.  This is just a copy included with
            SuiteSparse via the open-source license provided by
            George Karypis

RBio        read/write sparse matrices in Rutherford/Boeing format
            author: Tim Davis

README.txt  this file

SPQR        sparse QR factorization.  This the built-in qr and x=A\b in
            MATLAB.
            author of the CPU code: Tim Davis
            author of GPU modules: Tim Davis, Nuri Yeralan,
                Wissam Sid-Lakhdar, Sanjay Ranka

            SPQR/GPUQREngine: GPU support package for SPQR
            (not built into MATLAB, however)
            authors: Tim Davis, Nuri Yeralan, Sanjay Ranka,
                Wissam Sid-Lakhdar

SuiteSparse_config    configuration file for all the above packages.  The
            SuiteSparse_config/SuiteSparse_config.mk is included in the
            Makefile's of all packages.  CSparse and MATLAB_Tools do not
            use SuiteSparse_config.
            author: Tim Davis

SuiteSparse_GPURuntime      GPU support package for SPQR and CHOLMOD
            (not builtin to MATLAB, however).

SuiteSparse_install.m       install SuiteSparse for MATLAB

SuiteSparse_test.m          exhaustive test for SuiteSparse in MATLAB

ssget       MATLAB interface to the SuiteSparse Matrix Collection
            (formerly called the UF Sparse Matrix Collection).
            Includes a UFget function for backward compatibility.
            author: Tim Davis

UMFPACK     sparse LU factorization.  Requires AMD and the BLAS.
            This is the built-in lu and x=A\b in MATLAB.
            author: Tim Davis
            algorithm design collaboration: Iain Duff

Some codes optionally use METIS 5.1.0. This package is located in SuiteSparse in the metis-5.1.0 directory. Its use is optional, so you can remove it before compiling SuiteSparse, if you desire. The use of METIS will improve the ordering quality. METIS has been slightly modified for use in SuiteSparse; see the metis-5.1.0/README.txt file for details. SuiteSparse can use the unmodified METIS 5.1.0, however. To use your own copy of METIS, or a pre-installed copy of METIS use 'make MY_METIS_LIB=-lmymetis' or 'make MY_METIS_LIB=/my/stuff/metis-5.1.0/whereeveritis/libmetis.so MY_METIS_INC=/my/stuff/metis-5.1.0/include'. If you want to use METIS in MATLAB, however, you MUST use the version provided here, in SuiteSparse/metis-5.1.0. The MATLAB interface to METIS required some small changes in METIS itself to get it to work. The original METIS 5.1.0 will segfault MATLAB.

Refer to each package for license, copyright, and author information. All codes are authored or co-authored by Timothy A. Davis. email: [email protected]

Licenses for each package are located in the following files, all in PACKAGENAME/Doc/License.txt:

AMD/Doc/License.txt
BTF/Doc/License.txt
CAMD/Doc/License.txt
CCOLAMD/Doc/License.txt
CHOLMOD/Doc/License.txt
COLAMD/Doc/License.txt
CSparse/Doc/License.txt
CXSparse/Doc/License.txt
GPUQREngine/Doc/License.txt
KLU/Doc/License.txt
LDL/Doc/License.txt
MATLAB_Tools/Doc/License.txt
Mongoose/Doc/License.txt
RBio/Doc/License.txt
SPQR/Doc/License.txt
SuiteSparse_GPURuntime/Doc/License.txt
ssget/Doc/License.txt
UMFPACK/Doc/License.txt
GraphBLAS/Doc/License.txt

These files are also present, but they are simply copies of the above license files for CXSparse and ssget:

CXSparse_newfiles/Doc/License.txt
CSparse/MATLAB/ssget/Doc/License.txt
CXSparse/MATLAB/ssget/Doc/License.txt

METIS 5.0.1 is distributed with SuiteSparse, and is Copyright (c) by George Karypis. Please refer to that package for its License.


QUICK START FOR MATLAB USERS (Linux, Mac, or Windows):

Uncompress the SuiteSparse.zip or SuiteSparse.tar.gz archive file (they contain the same thing). Next, compile the GraphBLAS library (see instructions in GraphBLAS/Doc). Then in the MATLAB Command Window, cd to the SuiteSparse directory and type SuiteSparse_install. All packages will be compiled, and several demos will be run. To run a (long!) exhaustive test, do SuiteSparse_test.


QUICK START FOR THE C/C++ LIBRARIES:

For just GraphBLAS, do this:

cd GraphBLAS/build ; cmake .. ; make ; cd ../Demo ; ./demo 
cd ../build ; sudo make install

For all other packages, type 'make' in this directory. All libraries will be created and copied into SuiteSparse/lib. All include files need by the applications that use SuiteSparse are copied into SuiteSparse/include. All user documenation is copied into SuiteSparse/share/doc.

Be sure to first install all required libraries: BLAS and LAPACK for UMFPACK, CHOLMOD, and SPQR, and GMP and MPFR for SLIP_LU. Be sure to use the latest libraries; SLIP_LU requires MPFR 4.0 for example.

When compiling the libraries, do NOT use the INSTALL=... options for installing. Just do:

make

or to compile just the libraries without running the demos, do:

make library

Any program that uses SuiteSparse can thus use a simpler rule as compared to earlier versions of SuiteSparse. If you add /home/myself/SuiteSparse/lib to your library search patch, you can do the following (for example):

S = /home/myself/SuiteSparse
cc myprogram.c -I$(S)/include -lumfpack -lamd -lcholmod -lsuitesparseconfig -lm

To change the C and C++ compilers, and to compile in parallel use:

AUTOCC=no CC=gcc CX=g++ JOBS=32 make

for example, which changes the compiler to gcc and g++, and runs make with 'make -j32', in parallel with 32 jobs.

Now you can install the libraries, if you wish, in a location other than SuiteSparse/lib, SuiteSparse/include, and SuiteSparse/share/doc, using 'make install INSTALL=...'

Do 'make install' if you want to install the libraries and include files in SuiteSparse/lib and SuiteSparse/include, and the documentation in SuiteSparse/doc/suitesparse-VERSION. This will work on Linux/Unix and the Mac. It should automatically detect if you have the Intel compilers or not, and whether or not you have CUDA. If this fails, see the SuiteSparse_config/SuiteSparse_config.mk file. There are many options that you can either list on the 'make' command line, or you can just edit that file. For example, to compile with your own BLAS:

make BLAS=-lmyblaslibraryhere

NOTE: Use of the Intel MKL BLAS is strongly recommended. The OpenBLAS can result in severe performance degradation, in CHOLMOD in particular.

To list all configuration options (but not compile anything), do:

make config

Any parameter you see in the output of 'make config' with an equal sign can be modified at the 'make' command line.

If you do "make install" by itself, then the packages are all installed in SuiteSparse/lib (libraries), SuiteSparse/include (include .h files), and SuiteSparse/doc/suitesparse-VERSION (documentation). To install in /usr/local, the default location for Linux, do:

make library
sudo make install INSTALL=/usr/local

If you want to install elsewhere, say in /my/path, first ensure that /my/path is in your LD_LIBRARY_PATH. How to do that depends on your system, but in the bash shell, add this to your ~/.bashrc file:

LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/my/path
export LD_LIBRARY_PATH

You may also need to add SuiteSparse/lib to your path. If your copy of SuiteSparse is in /home/me/SuiteSparse, for example, then add this to your ~/.bashrc file:

LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/me/SuiteSparse/lib:/my/path
export LD_LIBRARY_PATH

Then do the following (use "sudo make ..." if needed):

make library
make install INSTALL=/my/path

which puts the files in /my/path/lib, /my/path/include, and /my/path/doc. If you want to selectively put the libraries, include files, and doc files in different locations, do:

make install INSTALL_LIB=/my/libs INSTALL_INCLUDE=/myotherstuff/include INSTALL_DOC=/mydocs

for example. Any term not defined will be set to its default, so if you don't want to install the documentation, but wish to install the libraries and includes in /usr/local/lib and /usr/local/include, do:

make install INSTALL_DOC=/tmp/doc

which copies the documentation to /tmp/doc where you can then remove it later.

Both the static (.a) and shared (.so) libraries are compiled. The lib.a libraries are left in the package Lib folder (AMD/Lib/libamd.a for example). The main exception to this rule is the SuiteSparse_config library, which is in SuiteSparse/libsuiteSparseconfig.a. SuiteSparse_config is required by all packages. The (extremely) optional xerbla library is also an exception, but it is highly unlikely that you need that library.

The 'make uninstall' takes the same command-line arguments.


Step-by-step details:

(1) Use the right BLAS and LAPACK libraries. Determine where your BLAS and LAPACK libraries are. If the default 'make' does not find them, use 'make BLAS=-lmyblaslibraryhere LAPACK=-lmylapackgoeshere'

(2) Install Intel's Threading Building Blocks (TBB). This is optionally used by SuiteSparseQR. Refer to the User Guide in SuiteSparse/SPQR/Doc/spqr_user_guide.pdf for details.

(3) Determine what other command line options you need for 'make'. All options can be set at the 'make' command line without the need to edit this file. Browse that file to see what options you can control. If you choose different options and wish to recompile, be sure to do 'make distclean' in this directory first, to remove all files not in the original distribution.

(4) Type "make" in this directory. All packages will be be compiled. METIS 5.1.0 will be compiled if you have it (note that METIS require CMake to build it). Several demos will be run. To compile just the libraries, without running any demos, use "make library". The libraries will appear in /Lib/.so.* (*.dylib for the Mac). Include files, as needed by user programs that use CHOLMOD, AMD, CAMD, COLAMD, CCOLAMD, BTF, KLU, UMFPACK, LDL, etc. are in /Include/.h. The include files required by user programs are then copied into SuiteSparse/include, and the compiled libraries are copied into SuiteSparse/lib. Documentation is copied into SuiteSparse/doc. The GraphBLAS libraries are created by cmake and placed in GraphBLAS/build. NOTE: on Linux, you may see some errors when you compile METIS ('make: *** No rule to make target 'w'.). You can safely ignore those errors.

(6) To install, type "make install". This will place copies of all libraries in SuiteSparse/lib, and all include files in SuiteSparse/include, and all documentation in SuiteSparse/doc/suitesparse-VERSION. You can change the install location by "make install INSTALL=/my/path" which puts the libraries in /my/path/lib, the include files in /my/path/include, and documentation in /my/path/doc. These directories are created if they do not already exist.

(7) To uninstall, type "make uninstall", which reverses "make install" by removing the SuiteSparse libraries, include files, and documentation from the place they were installed. If you pass INSTALL_***= options to 'make install', you must pass the same to 'make uninstall'.