Install GCC: Difference between revisions

From Micro and Nano Mechanics Group
Jump to navigation Jump to search
 
(44 intermediate revisions by the same user not shown)
Line 1: Line 1:
(written by Alfredo Correa)

GCC stands for GNU Compiler Collection. The latest version is GCC 4.3.
GCC stands for GNU Compiler Collection. The latest version is GCC 4.3.
The use of this tutorial is to install GCC 4, which is not available in some linux distributions. In those systems GCC 3 is available instead but I suppose that for some reason you need GCC 4.
The use of this tutorial is to install GCC 4, which is not available in some linux distributions. In those systems GCC 3 is available instead but I suppose that for some reason you need GCC 4.
Line 13: Line 15:
This tutorial is mainly oriented to install in a campus cluster. If you want a bleeding edge compiler in your Linux computer it is better to just use an up to date distribution instead of going through these instructions. For example Ubuntu 9.10 has as default compiler gcc 4.4, just by doing:
This tutorial is mainly oriented to install in a campus cluster. If you want a bleeding edge compiler in your Linux computer it is better to just use an up to date distribution instead of going through these instructions. For example Ubuntu 9.10 has as default compiler gcc 4.4, just by doing:


sudo apt-get install gcc
sudo aptitude install gcc g++

''If you are lucky, and not interested in the details you can donwload this [[Media:gcc_install.sh.txt|script]] and run it in your computer to install GCC from scratch. It downloads, compiles and installs to your ~/usr directory''. (It takes ~30 minutes in a dual core plus the download time.)


= Preparation and Downloads =
= Preparation and Downloads =
''
If you are not interested about the details and feel lucky you can donwload this [[Media:gcc_install.sh.txt|script]] and run it in your computer.''


The tutorial assumes that you want to install GCC/g++ in your userspace directories (i.e. in your home directory). To do that create the following directories:
The tutorial assumes that you want to install GCC/g++ in your userspace directories (i.e. in your home directory). To do that create the following directories:
Line 23: Line 25:
mkdir ~/soft
mkdir ~/soft
mkdir ~/usr
mkdir ~/usr

== MP libraries ==


GCC depends on several libraries and tools most of which are usually in the system already, except for [http://gmplib.org/ GMP library], [http://www.mpfr.org/ MPFR] and [http://www.multiprecision.org/ MPC].
GCC depends on several libraries and tools most of which are usually in the system already, except for [http://gmplib.org/ GMP library], [http://www.mpfr.org/ MPFR] and [http://www.multiprecision.org/ MPC].
Line 30: Line 34:


cd ~/soft
cd ~/soft
export GCC_VER=4.4.2
export GCC_VER=4.5.0

wget wget http://gcc-uk.internet.bs/releases/gcc-$GCC_VER/gcc-$GCC_VER.tar.gz
* from internet:
tar -jxvf gcc-$GCC_VER.tar.bz2

mv gcc-$GCC_VER gcc-source
wget http://gcc-uk.internet.bs/releases/gcc-$GCC_VER/gcc-$GCC_VER.tar.gz
tar -zxvf gcc-$GCC_VER.tar.gz
mv gcc-$GCC_VER gcc-$GCC_VER-source
Check for last release at http://gcc-uk.internet.bs/releases/

* or from SVN:
svn checkout svn://gcc.gnu.org/svn/gcc/trunk


And then download and uncompress GMP, MPFR and MPC
and then download and uncompress GMP, MPFR and MPC


cd ~/soft
cd ~/soft
export GMP_VER=4.3.1
export GMP_VER=5.0.2
wget ftp://ftp.gmplib.org/pub/gmp-4.3.1/gmp-$GMP_VER.tar.gz
wget ftp://ftp.gmplib.org/pub/gmp-$GMP_VER/gmp-$GMP_VER.tar.gz
tar -zxvf gmp-$GMP_VER.tar.gz
tar -zxvf gmp-$GMP_VER.tar.gz
mv gmp-$GMP_VER gcc-source/gmp
mv gmp-$GMP_VER gcc-$GCC_VER-source/gmp
export MPFR_VER=2.4.2
export MPFR_VER=3.0.1
wget http://www.mpfr.org/mpfr-current/mpfr-$MPFR_VER.tar.gz
wget http://www.mpfr.org/mpfr-current/mpfr-$MPFR_VER.tar.gz
tar -zxvf mpfr-$MPFR_VER.tar.gz
tar -zxvf mpfr-$MPFR_VER.tar.gz
mv mpfr-$MPFR_VER gcc-source/mpfr
mv mpfr-$MPFR_VER gcc-$GCC_VER-source/mpfr
export MPC_VER=0.8.1
export MPC_VER=0.8.1
wget http://www.multiprecision.org/mpc/download/mpc-$MPC_VER.tar.gz
wget http://www.multiprecision.org/mpc/download/mpc-$MPC_VER.tar.gz
tar -zxvf mpc-$MPC_VER.tar.gz
tar -zxvf mpc-$MPC_VER.tar.gz
mv mpc-$MPC_VER gcc-source/mpc
mv mpc-$MPC_VER gcc-$GCC_VER-source/mpc


The source directory is ready to for compilation.
The source directory is ready to for compilation.

== GNU Binutils ==

In some systems binutils is so old (for example, v2.17) that it won't work with the new compiler. In this case you need to install Binutils in order to succeed in the compilation of a program

cd ~/soft
wget http://ftp.gnu.org/gnu/binutils/binutils-2.20.tar.gz
tar -zxvf binutils-2.20.tar.gz
cd binutils-*
./configure --prefix=$HOME/usr
make
make install

This will install the linker (ld), the library archive (ar) and other programs in ~/usr/bin. Later they will be used by the compiler, for example, by ~/usr/bin/c++.


= Compilation =
= Compilation =
Line 61: Line 86:


cd ~/soft
cd ~/soft
mkdir gcc-build
mkdir gcc-$GCC_VER-build
cd gcc-build
cd gcc-$GCC_VER-build


Then we will 'configure' from that directory. Everything will be installed in ~/usr, including the executable compilers in ~/usr/bin and the library files in ~/usr/lib. To specify that you do
Then we will 'configure' from that directory. Everything will be installed in ~/usr, including the executable compilers in ~/usr/bin and the library files in ~/usr/lib. To specify that you do


cd ~/soft/gcc-build
cd ~/soft/gcc-$GCC_VER-build
../gcc-source/configure --prefix=$HOME/usr --enable-languages=c++,fortran --enable-shared
../gcc-$GCC_VER-source/configure --prefix=$HOME/usr --bindir=$HOME/usr/bin$CLUSTER --prefix=$HOME/usr/lib$CLUSTER --enable-languages=c++,fortran --enable-shared


If you don't need the Fortran compiler specify --enable-languages=c++. (C compiler is enabled by default.) Even if you don't use Fortran it can be useful to compile Lapack for example.
If you don't need the Fortran compiler specify --enable-languages=c++. (C compiler is enabled by default.) Even if you don't use Fortran it can be useful to compile Lapack for example.
Line 73: Line 98:
Check that the configure works (if it doesn't, report it in this wiki) before doing:
Check that the configure works (if it doesn't, report it in this wiki) before doing:


make -j 2 #takes ~60 minutes
make -j 8 #takes ~60 minutes


and finally:
and finally:
Line 118: Line 143:


Now you are powerful and can tell your friends that you compiled a compiler. Now seriously, you can build programs that you could not compile or compiled with bugs with the system compiler.
Now you are powerful and can tell your friends that you compiled a compiler. Now seriously, you can build programs that you could not compile or compiled with bugs with the system compiler.

If you get an error that looks like

/tmp/ccw54L0X.s: Assembler messages:
/tmp/ccw54L0X.s:9: Error: unknown pseudo-op: `.cfi_personality'

the install [GNU Binutils], as described above.

= Check GCC version during compilation =

When we install our own version of GCC is because we need a certain feature of the new version that is not available in the old/system version. In general that means that the code using new features will not compile with the old compiler, and sometimes it will give some undecipherable syntax error. Because of that you may want to check that portions of your code actually do not compile with a certain old version of GCC, the macro language can verify conditions on the GCC version. For example you can add the following code at the beginning of a code file or inside a function that uses certain feature

void very_modern_function(){
#if not((__GNUC__>=4) and (__GNUC_MINOR__ >1))
#error "expected gcc version > 4.1, got " ## __GNUC__ ## "." ## __GNUC_MINOR__
#endif
... code with new feautures ...
}

If you use a compiler with version 4.1 or earlier, the output will show

./example.cpp:23: error: #error "expected gcc version > 4.1"

before any other error message related to features used within the function. (You can make the behaviour less severe by replacing "#error" for "#warning" so that compilation can eventually succeed.)

Alternatively (if you use Boost) you can have it in one line

BOOST_STATIC_ASSERT((__GNUC__>=4) and (__GNUC_MINOR__ >1));

with the error message

./example.cpp:22: error: invalid application of ‘sizeof’ to incomplete type ‘boost::STATIC_ASSERTION_FAILURE<false>’

This trick avoid the use the wrong compiler version.


= Libraries =
= Libraries =
Line 127: Line 186:
This is the procedure to compile and install the [http://www.mcs.anl.gov/research/projects/mpich2/ MPICH2 library] and environment. The instructions are taken from [http://www.mcs.anl.gov/research/projects/mpich2/documentation/files/mpich2-1.0.8-installguide.pdf the official documentation] and from [http://source.ggy.bris.ac.uk/wiki/Install_and_configure_MPI#Download_MPI this wiki-page].
This is the procedure to compile and install the [http://www.mcs.anl.gov/research/projects/mpich2/ MPICH2 library] and environment. The instructions are taken from [http://www.mcs.anl.gov/research/projects/mpich2/documentation/files/mpich2-1.0.8-installguide.pdf the official documentation] and from [http://source.ggy.bris.ac.uk/wiki/Install_and_configure_MPI#Download_MPI this wiki-page].


mkdir ~/usr
mkdir ~/usr
mkdir ~/soft
mkdir ~/soft
cd ~/soft
cd ~/soft
export MPICH2_VER=1.2.1
wget http://www.mcs.anl.gov/research/projects/mpich2/downloads/tarballs/1.0.8/mpich2-1.0.8.tar.gz
wget http://www.mcs.anl.gov/research/projects/mpich2/downloads/tarballs/$MPICH2_VER/mpich2-$MPICH2_VER.tar.gz
tar -zxvf mpich2-1.0.8.tar.gz
tar -zxvf mpich2-$MPICH2_VER.tar.gz


Like GCC, MPICH2 should be compiled in a different directory than the location of the sources (this is not the case for version 1.1.:
Like GCC, MPICH2 should be compiled in a different directory than the location of the sources:


mkdir mpich2-1.0.8-build
mkdir mpich2-$MPICH2_VER-build
cd mpich2-1.0.8-build
cd mpich2-$MPICH2_VER-build
../mpich2-1.0.8/configure --prefix=$HOME/usr |& tee configure.log
../mpich2-$MPICH2_VER/configure --prefix=$HOME/usr |& tee configure.log
make |& tee make.log
time make |& tee make.log
make install
make install


The compilation takes ~20 minutes. (Do not try <code>make -j 2</code> or <code>make install</code> only). The compiler wrappers (i.e. scripts to compile with MPI settings) are installed in ~/usr/bin, for example ~/usr/bin/mpicxx for the C++ MPI wrapper.
The compilation takes ~20 minutes. (Do not try <code>make -j 2</code> or <code>make install</code> only). The compiler wrappers (i.e. scripts to compile with MPI settings) are installed in ~/usr/bin, for example ~/usr/bin/mpicxx for the C++ MPI wrapper.
Line 145: Line 205:
Now set up the environment, in your home directory create a configuration file. This need to be done only once.
Now set up the environment, in your home directory create a configuration file. This need to be done only once.


cd ~/
cd ~/
touch .mpd.conf
touch .mpd.conf
chmod 600 .mpd.conf
chmod 600 .mpd.conf
echo 'MPD_SECRETWORD=nano' > .mpd.conf
echo 'MPD_SECRETWORD=nano' > .mpd.conf


=== Test example ===
=== Test example ===
Line 167: Line 227:


~/usr/bin/mpirun -np 2 ./mpich2_example
~/usr/bin/mpirun -np 2 ./mpich2_example

== Install GSL ==

For quick installation [http://www.gnu.org/software/gsl/ the GSL library] in Ubuntu try <code>sudo aptitude install libgsl0-dev</code> which install 1.13 in U10.04. To install from scratch:

mkdir ~/soft
cd ~/soft
wget http://mirrors.kernel.org/gnu/gsl/gsl-1.14.tar.gz
tar -zxvf gsl-1.14.tar.gz
cd gsl-1.14
./configure --prefix=$HOME/usr
make --jobs=2
make check > log 2>&1
make install

It takes 2 minutes to compile and one minute to run a few tests.

The header files will be located in <code>~/usr/include/gsl/*.h</code> and the libraries (static and dynamic) will be located at <code>~/usr/libgsl*.{a,la,so}</code>

A program using GSL must be linked to libgsl and other libraries (gslblas and lm):
gcc program.c -L$HOME/usr/lib -lgsl -lgslcblas -lm
This line can be generated automatically by pkgconfig
gcc program.c `pkgconfig --libs gsl`
if we have set (e.g. in bashrc) the variable:
export PKG_CONFIG_PATH=$HOME/usr/lib/pkgconfig

== Install GMP, MPFR and GMPFRXX ==

GMP is necessary for the compilation of the C++ compiler, in the instructions above it is installed automatically. It can also be installed for general use:

# see above for download
cd ~/sotf/gmp-$GMP_VER
./configure --prefix=$HOME/usr --bindir=$HOME/usr/bin$CLUSTER --libdir=$HOME/usr/lib$CLUSTER
time make
make install

(1 minute)

To uninstall
rm -f ~/usr/include/gmp.h ~/usr/include/gmpxx.h ~/usr/lib/libgmp*.* ~/usr/info/gmp*

MPFR has a similar installation

# see above for download
cd ~/soft/mpfr-$MPFR_VER
./configure --prefix=$HOME/usr --libdir=$HOME/usr/lib$CLUSTER
make install

In addition, the a GMP/MPFR C++ wrapper with automatic rounding (used by Boost.Math bindings, http://www.boost.org/doc/libs/1_43_0/libs/math/doc/sf_and_dist/html/math_toolkit/using_udt/use_mpfr.html) can be installed (more manual)

cd ~/soft
wget http://math.berkeley.edu/~wilken/code/gmpfrxx/gmpfrxx.zip
unzip gmpfrxx.zip
cd gmpfrxx/
# add lines #include<cstdio> \\ #include<cstdlib>
mv gmpfrxx.h tmp.h
echo '#include<cstdio>' > gmpfrxx.h
echo '#include<cstdlib>' >> gmpfrxx.h
cat tmp.h >> gmpfrxx.h
make libgmpfrxx.a
cp gmpfrxx.h ~/usr/include
cp libgmpfrxx.a ~/usr/lib$CLUSTER

Latest revision as of 20:12, 14 October 2011

(written by Alfredo Correa)

GCC stands for GNU Compiler Collection. The latest version is GCC 4.3. The use of this tutorial is to install GCC 4, which is not available in some linux distributions. In those systems GCC 3 is available instead but I suppose that for some reason you need GCC 4. (Of course we will use GCC 3, or any other available C compiler to build GCC 4.) The tutorial is focused in obtaining 'g++' (GNU C++) among all the available compilers in the collection. This tutorial is based on this other guide, although this version is much more straightforward.

You can check which version of GCC is currently installed by running

 $gcc -v
 ...
 gcc version 3.4.6 20060404 (Red Hat 3.4.6-3)

This tutorial is mainly oriented to install in a campus cluster. If you want a bleeding edge compiler in your Linux computer it is better to just use an up to date distribution instead of going through these instructions. For example Ubuntu 9.10 has as default compiler gcc 4.4, just by doing:

sudo aptitude install gcc g++

If you are lucky, and not interested in the details you can donwload this script and run it in your computer to install GCC from scratch. It downloads, compiles and installs to your ~/usr directory. (It takes ~30 minutes in a dual core plus the download time.)

Preparation and Downloads

The tutorial assumes that you want to install GCC/g++ in your userspace directories (i.e. in your home directory). To do that create the following directories:

mkdir ~/soft
mkdir ~/usr

MP libraries

GCC depends on several libraries and tools most of which are usually in the system already, except for GMP library, MPFR and MPC. Fortunately we don't need to install these three libraries explicitly. The GCC installation takes care of their compilation as long as the sources directories are put in the GCC source directory. The three directories should be named gmp, mpfr and mpc.

So, first download the GCC sources.

cd ~/soft
export GCC_VER=4.5.0
  • from internet:
wget http://gcc-uk.internet.bs/releases/gcc-$GCC_VER/gcc-$GCC_VER.tar.gz
tar -zxvf gcc-$GCC_VER.tar.gz
mv gcc-$GCC_VER gcc-$GCC_VER-source

Check for last release at http://gcc-uk.internet.bs/releases/

  • or from SVN:
svn checkout svn://gcc.gnu.org/svn/gcc/trunk

and then download and uncompress GMP, MPFR and MPC

cd ~/soft

export GMP_VER=5.0.2
wget ftp://ftp.gmplib.org/pub/gmp-$GMP_VER/gmp-$GMP_VER.tar.gz
tar -zxvf gmp-$GMP_VER.tar.gz
mv gmp-$GMP_VER gcc-$GCC_VER-source/gmp

export MPFR_VER=3.0.1
wget http://www.mpfr.org/mpfr-current/mpfr-$MPFR_VER.tar.gz
tar -zxvf mpfr-$MPFR_VER.tar.gz
mv mpfr-$MPFR_VER gcc-$GCC_VER-source/mpfr

export MPC_VER=0.8.1
wget http://www.multiprecision.org/mpc/download/mpc-$MPC_VER.tar.gz
tar -zxvf mpc-$MPC_VER.tar.gz
mv mpc-$MPC_VER gcc-$GCC_VER-source/mpc

The source directory is ready to for compilation.

GNU Binutils

In some systems binutils is so old (for example, v2.17) that it won't work with the new compiler. In this case you need to install Binutils in order to succeed in the compilation of a program

cd ~/soft
wget http://ftp.gnu.org/gnu/binutils/binutils-2.20.tar.gz
tar -zxvf binutils-2.20.tar.gz
cd binutils-*
./configure --prefix=$HOME/usr
make
make install

This will install the linker (ld), the library archive (ar) and other programs in ~/usr/bin. Later they will be used by the compiler, for example, by ~/usr/bin/c++.

Compilation

This is important and different from other usual compilation procedures: GCC should be compiled in a directory different from the source directory, in this case we will create a "build" directory.

cd ~/soft
mkdir gcc-$GCC_VER-build
cd gcc-$GCC_VER-build

Then we will 'configure' from that directory. Everything will be installed in ~/usr, including the executable compilers in ~/usr/bin and the library files in ~/usr/lib. To specify that you do

cd ~/soft/gcc-$GCC_VER-build
../gcc-$GCC_VER-source/configure --prefix=$HOME/usr --bindir=$HOME/usr/bin$CLUSTER --prefix=$HOME/usr/lib$CLUSTER --enable-languages=c++,fortran --enable-shared

If you don't need the Fortran compiler specify --enable-languages=c++. (C compiler is enabled by default.) Even if you don't use Fortran it can be useful to compile Lapack for example.

Check that the configure works (if it doesn't, report it in this wiki) before doing:

make -j 8                   #takes ~60 minutes

and finally:

make install

To test the version installed:

$ ~/usr/bin/g++ -v
Using built-in specs.
Target: x86_64-unknown-linux-gnu
Configured with: ../gcc-4.3.3/configure --prefix=$HOME/usr --with-local-prefix=$HOME/usr/local --with-gmp=/home/correaa/usr --with-mpfr=/home/correaa/usr --enable-languages=c++
Thread model: posix
gcc version 4.3.3 (GCC)

and do

~/usr/bin/g++ --print-search-dirs

to see which libraries and directories will be used by default.

The installation also includes the C++ Standard Template Library.

Basic Usage

This is the "Hello, world!" program

 #include <iostream>
 int main(){
   std::cout << "Hello, world!\n";
 }

which can be compiled with our brand new compiler:

 cd /tmp
 wget http://micro.stanford.edu/mediawiki-1.11.0/images/Hello_world.cpp.tar -o hello_world.cpp.tar
 tar -xvf hello_word.cpp.tar
 ~/usr/bin/g++ -Wl,-rpath=$HOME/usr/lib:$HOME/usr/lib64 hello_world.cpp -o hello_world
 ./hello_world

Note that you have to specify the tilde (~) in order to run that specific compiler and not the default system one. The rpath option tells the compiler to use your local (home) version of the runtime and standard libraries instead of the default ones (in /usr/lib and /usr/lib64).

In any case you can check which libraries are being used by doing:

 ldd ./hello_world

Now you are powerful and can tell your friends that you compiled a compiler. Now seriously, you can build programs that you could not compile or compiled with bugs with the system compiler.

If you get an error that looks like

/tmp/ccw54L0X.s: Assembler messages:
/tmp/ccw54L0X.s:9: Error: unknown pseudo-op: `.cfi_personality'

the install [GNU Binutils], as described above.

Check GCC version during compilation

When we install our own version of GCC is because we need a certain feature of the new version that is not available in the old/system version. In general that means that the code using new features will not compile with the old compiler, and sometimes it will give some undecipherable syntax error. Because of that you may want to check that portions of your code actually do not compile with a certain old version of GCC, the macro language can verify conditions on the GCC version. For example you can add the following code at the beginning of a code file or inside a function that uses certain feature

void very_modern_function(){
  #if not((__GNUC__>=4) and (__GNUC_MINOR__ >1)) 
  #error "expected gcc version > 4.1, got " ## __GNUC__ ## "." ## __GNUC_MINOR__
  #endif
  ... code with new feautures ...
}

If you use a compiler with version 4.1 or earlier, the output will show

./example.cpp:23: error: #error "expected gcc version > 4.1"

before any other error message related to features used within the function. (You can make the behaviour less severe by replacing "#error" for "#warning" so that compilation can eventually succeed.)

Alternatively (if you use Boost) you can have it in one line

BOOST_STATIC_ASSERT((__GNUC__>=4) and (__GNUC_MINOR__ >1));

with the error message

./example.cpp:22: error: invalid application of ‘sizeof’ to incomplete type ‘boost::STATIC_ASSERTION_FAILURE<false>’ 

This trick avoid the use the wrong compiler version.

Libraries

By following this tutorial you should have the C and C++ compilers of the GNU Compiler Collection, with them you can use any decent C or C++ library available, for example the Standard Template Library (included with GCC), FFTW, BOOST and HDF5.

Install MPICH2

This is the procedure to compile and install the MPICH2 library and environment. The instructions are taken from the official documentation and from this wiki-page.

mkdir ~/usr
mkdir ~/soft
cd ~/soft
export MPICH2_VER=1.2.1
wget http://www.mcs.anl.gov/research/projects/mpich2/downloads/tarballs/$MPICH2_VER/mpich2-$MPICH2_VER.tar.gz
tar -zxvf mpich2-$MPICH2_VER.tar.gz

Like GCC, MPICH2 should be compiled in a different directory than the location of the sources:

mkdir mpich2-$MPICH2_VER-build
cd mpich2-$MPICH2_VER-build
../mpich2-$MPICH2_VER/configure --prefix=$HOME/usr |& tee configure.log
time make |& tee make.log
make install

The compilation takes ~20 minutes. (Do not try make -j 2 or make install only). The compiler wrappers (i.e. scripts to compile with MPI settings) are installed in ~/usr/bin, for example ~/usr/bin/mpicxx for the C++ MPI wrapper.

Now set up the environment, in your home directory create a configuration file. This need to be done only once.

cd ~/
touch .mpd.conf
chmod 600 .mpd.conf
echo 'MPD_SECRETWORD=nano' > .mpd.conf

Test example

Now you are ready to compile the example (based on this original example)

 cd ~/tmp
 wget http://micro.stanford.edu/mediawiki-1.11.0/images/Mpich2_example.tar -o mpich2_example.tar
 tar -xvf mpich2_example.tar 
 cd mpich2_example
 ~/usr/bin/mpicxx mpich2_example.cpp -o mpich2_example

Before executing a MPI program you will need to start the MPI daemon

 ~/usr/bin/mpd &

And then the program can be run

 ~/usr/bin/mpirun -np 2 ./mpich2_example

Install GSL

For quick installation the GSL library in Ubuntu try sudo aptitude install libgsl0-dev which install 1.13 in U10.04. To install from scratch:

mkdir ~/soft
cd ~/soft
wget http://mirrors.kernel.org/gnu/gsl/gsl-1.14.tar.gz
tar -zxvf gsl-1.14.tar.gz
cd gsl-1.14
./configure --prefix=$HOME/usr
make --jobs=2
make check > log 2>&1
make install

It takes 2 minutes to compile and one minute to run a few tests.

The header files will be located in ~/usr/include/gsl/*.h and the libraries (static and dynamic) will be located at ~/usr/libgsl*.{a,la,so}

A program using GSL must be linked to libgsl and other libraries (gslblas and lm):

gcc program.c -L$HOME/usr/lib -lgsl -lgslcblas -lm  

This line can be generated automatically by pkgconfig

gcc program.c `pkgconfig --libs gsl`

if we have set (e.g. in bashrc) the variable:

export PKG_CONFIG_PATH=$HOME/usr/lib/pkgconfig

Install GMP, MPFR and GMPFRXX

GMP is necessary for the compilation of the C++ compiler, in the instructions above it is installed automatically. It can also be installed for general use:

# see above for download
cd ~/sotf/gmp-$GMP_VER
./configure --prefix=$HOME/usr --bindir=$HOME/usr/bin$CLUSTER --libdir=$HOME/usr/lib$CLUSTER
time make 
make install

(1 minute)

To uninstall

rm -f ~/usr/include/gmp.h ~/usr/include/gmpxx.h ~/usr/lib/libgmp*.* ~/usr/info/gmp*

MPFR has a similar installation

# see above for download
cd ~/soft/mpfr-$MPFR_VER
./configure --prefix=$HOME/usr --libdir=$HOME/usr/lib$CLUSTER
make install

In addition, the a GMP/MPFR C++ wrapper with automatic rounding (used by Boost.Math bindings, http://www.boost.org/doc/libs/1_43_0/libs/math/doc/sf_and_dist/html/math_toolkit/using_udt/use_mpfr.html) can be installed (more manual)

cd ~/soft
wget http://math.berkeley.edu/~wilken/code/gmpfrxx/gmpfrxx.zip
unzip gmpfrxx.zip
cd gmpfrxx/
# add lines #include<cstdio> \\ #include<cstdlib>
mv gmpfrxx.h tmp.h
echo '#include<cstdio>' > gmpfrxx.h
echo '#include<cstdlib>' >> gmpfrxx.h
cat tmp.h >> gmpfrxx.h
make libgmpfrxx.a
cp  gmpfrxx.h ~/usr/include
cp libgmpfrxx.a ~/usr/lib$CLUSTER