| 1 | Installation of the Mumps linear solver within COIN. |
|---|
| 2 | |
|---|
| 3 | Mumps is a MUltifrontal Massively Parallel sparse direct Solver. |
|---|
| 4 | |
|---|
| 5 | Please visit the home page: |
|---|
| 6 | |
|---|
| 7 | Home page: http://graal.ens-lyon.fr/MUMPS/ |
|---|
| 8 | |
|---|
| 9 | and make sure you read the usage conditions at |
|---|
| 10 | |
|---|
| 11 | http://graal.ens-lyon.fr/MUMPS/avail.html |
|---|
| 12 | |
|---|
| 13 | ********************************************************************** |
|---|
| 14 | Note: It is YOUR RESPONSIBILITY to ensure that you are entitled to |
|---|
| 15 | download and use this third party package. |
|---|
| 16 | ********************************************************************** |
|---|
| 17 | |
|---|
| 18 | If you are working under Unix/Linux-like system and you have `wget' |
|---|
| 19 | installed, you can downloaded the required source code for MUMPS |
|---|
| 20 | by executing the included script `get.Mumps' in this directory. |
|---|
| 21 | |
|---|
| 22 | Note, however, that the script might not be looking for the most recent |
|---|
| 23 | version of Mumps, and it might fail if it tries to download an outdated |
|---|
| 24 | one. In that case, please check the Mumps home page for the most recent |
|---|
| 25 | version number and change the definition of the mumps_ver variable at the |
|---|
| 26 | beginning of the get.Mumps script. |
|---|
| 27 | |
|---|
| 28 | If you don't have wget available, follow those instructions: |
|---|
| 29 | |
|---|
| 30 | Check the Mumps home page for the most recent version of the 4.9 branch. |
|---|
| 31 | You may then download |
|---|
| 32 | |
|---|
| 33 | http://mumps.enseeiht.fr/MUMPS_4.9.tar.gz (or try |
|---|
| 34 | http://graal.ens-lyon.fr/MUMPS/MUMPS_4.9.tar.gz if first URL doesn't work) |
|---|
| 35 | |
|---|
| 36 | and unpack it in this directory: |
|---|
| 37 | |
|---|
| 38 | gunzip MUMPS_4.9.tar.gz |
|---|
| 39 | tar xf MUMPS_4.9.tar |
|---|
| 40 | |
|---|
| 41 | Then, rename the MUMPS_4.9 directory just to be MUMPS: |
|---|
| 42 | |
|---|
| 43 | mv MUMPS_4.9 MUMPS |
|---|
| 44 | |
|---|
| 45 | Further, you may apply a patch that fixes problems with matrices of size 1: |
|---|
| 46 | |
|---|
| 47 | patch -p0 < mumps.patch |
|---|
| 48 | |
|---|
| 49 | ********************************************************************** |
|---|
| 50 | Additional instructions for building with Msys and MinGW |
|---|
| 51 | ********************************************************************** |
|---|
| 52 | |
|---|
| 53 | If you are trying to build Mumps in Microsoft Windows using the MSys shell |
|---|
| 54 | environment and gnu fortran compilers, you will first need to install the |
|---|
| 55 | MinGW gcc 4.2 technology preview in order obtain a Fortran 95 compliant |
|---|
| 56 | Fortran compiler. GCC 4.2 is available here: |
|---|
| 57 | |
|---|
| 58 | http://sourceforge.net/project/showfiles.php?group_id=2435&package_id=241304 |
|---|
| 59 | |
|---|
| 60 | You need to install the gcc-core, gcc-g++, and gcc-gfortran packages. |
|---|
| 61 | Note the executable names in the gcc packages have the suffix "sjlj" and will |
|---|
| 62 | not be found by the configuration script of whatever project you are trying to |
|---|
| 63 | build, so you must either (1) change their names by deleting the suffix |
|---|
| 64 | (beware that if you do this, you may create a conflict with other version of |
|---|
| 65 | gcc already installed) or (2) configure with |
|---|
| 66 | |
|---|
| 67 | ./configure CC=gcc-sjlj CXX=g++-sjlj F77=gfortran-sjlj |
|---|
| 68 | |
|---|
