-------------------------------------
FILES included in the distribution :
-------------------------------------
DOC/ : User's manual (pspases-manual.ps).
Makefile
Makefile.in
PSPASES/ : PSPASES source code.
README.INSTALL : This file!
README.USAGE : Usage notes and matrix mormat descriptions.
TEST/ : Test code illustrating example usage of various functions.
matrices/ : Test matrices and some format conversion codes.
metis/ : METIS source code.
parMetis/ : ParMETIS source code.
version : Version Info for this distribution.
-------------------------
How to Make the library ?
-------------------------
1. Get the compressed tar file pspases.v1.0.tar.gz from URL :
http://www.cs.umn.edu/~mahesh/pspases
2. Uncompress and untar the file.
("gunzip -c pspases.v1.0.tar.gz | tar -xvf -")
3. do "cd PSPASES.v1.0"
4. Edit Makefile.in to make appropriate changes as indicated in the file.
5. do "make >& make.log".
6. This will create :
libpspases.a in PSPASES.v1.0/PSPASES directory.
libparmetis.a in PSPASES.v1.0/parMetis directory.
libmetis.a in PSPASES.v1.0/metis directory.
--------------------------------------------------
How to test if library has been built properly ?
--------------------------------------------------
1. cd PSPASES.v1.0/TEST
2. Make appropriate changes to Makefile to reflect appropriate compiler
and linker, flags, and the location of BLAS library. Sometimes (e.g.
on SGI), you will need to indicate the MPI library that will be used.
3. do "make".
4. This will create three drivers. for different formats of input files;
cpspases_test : For reading .bin format files,
fcpspases_test : For reading .fcc format files,
frpspases_test : For reading .rsa format files.
NOTE:
There are four different formats accepted :
.fcc : full compressed column format - also the PSPASES internal
format for ordering and symbolic phases (ASCII format).
.bin : binary full compressed column format.
.rsa : (HB) Harwell-Boeing Real Symmetric Assembled compressed
column format.
(RB) Rutherford-Boeing Real Symmetric Assembled compressed
column format.
For details of each format, read the README.USAGE file. There are
two conversion subroutines provided in matrices directory, to convert
from .fcc to .bin, and .rsa to .fcc formats. Issuing "make" in matrices
directory will create fcc2bin and rsa2fcc files (You may need to edit
the Makefile in that directory first).
4. Edit and run the "testrun" shellscript, which will test all these drivers
for the sample matrices provided in PSPASES.v1.0/matrices directory.
5. If the results obtained for each of the runs in the "testrun" script,
look similar to the following, with a very small max |B-AX| error,
then the library has been installed properly.
------- TEST OUTPUT for the fcpspases_test run in the "testrun" script.
*-*-*-*-*-*-*-*-*-* Test Using fcpspases_test *-*-*-*-*-*-*-*-*-*
* Testing ../matrices/fort50.fcc
Dimension = 21901 #processors = 4
Parameters: blk = 64 nrhs = 1 br = 1
symm = 0 sort = 0 serialorder = 0
-------> Testing PSPACEO+PSPACEY
calling PSPACEO (Compute fill-reducing Ordering)..
calling PSPACEY (sYmbolic Factorization)..
Max Min Sum
Memory Consumed by PSPCOMM = 3.066 2.966 12.127 MB
More Factor Memory Needed = 4.048 3.681 15.431 MB
Total Factor Memory Needed = 7.078 6.648 27.558 MB
NNZ_Lower_Triangle_of_A = 0.870E+05 NNZ_L = 0.4994E+06
Tree_Opcount_Imbalance = 1.10 Factor_Opcount = 0.2502E+08
Order Time = 0.898 sec
sYmbolic Time = 0.100 sec
calling PSPACEC (Cleanup option = 0 )..
-------> Testing PSPACEO+PSPACEY+DPSPACEN+DPSPACET
calling PSPACEO (Compute fill-reducing Ordering)..
calling PSPACEY (sYmbolic Factorization)..
Max Min Sum
Memory Consumed by PSPCOMM = 3.066 2.966 12.127 MB
More Factor Memory Needed = 4.048 3.681 15.431 MB
Total Factor Memory Needed = 7.078 6.648 27.558 MB
NNZ_Lower_Triangle_of_A = 0.870E+05 NNZ_L = 0.4994E+06
Tree_Opcount_Imbalance = 1.10 Factor_Opcount = 0.2502E+08
Order Time = 0.680 sec
sYmbolic Time = 0.099 sec
calling DPSPACEN (Numerical Factorization)..
DPSPACEN Time = 0.210 sec
Numerical Factor Perf = 119.137 MFLOPS
calling PSPACEC (Cleanup option = 1 )..
calling DPSPACET (Triangular Systems Solution)..
DPSPACET Time = 0.083 sec
Triangular Solution Perf = 24.163 MFLOPS
Total Solver Time = 1.072 sec
calling PSPACEC (Cleanup option = 0 )..
all stages of PSPASES are done! checking B-AX ..
max |B - AX| = 8.88178419700125232E-16
-------> Testing DPSPACEF+DPSPACET
calling DPSPACEF (Ordering+sYmbolic+Numerical)..
Max Min Sum
Memory Consumed by PSPCOMM = 3.127 2.960 12.300 MB
NNZ_Lower_Triangle_of_A = 0.870E+05 NNZ_L = 0.4994E+06
Tree_Opcount_Imbalance = 1.10 Factor_Opcount = 0.2502E+08
DPSPACEF Time (Ftime) = 1.049 sec
Factor_Opcount / Ftime = 23.848 MFLOPS
calling PSPACEC (Cleanup option = 1 )..
calling DPSPACET (Triangular Systems Solution)..
DPSPACET Time = 0.104 sec
Triangular Solution Perf = 19.163 MFLOPS
Total Solver Time = 1.153 sec
calling PSPACEC (Cleanup option = 0 )..
all stages of PSPASES are done! checking B-AX ..
max |B - AX| = 8.88178419700125232E-16
C/*****************************************************************************/
C/* */
C/* (C) Copyright IBM Corporation, 1997 */
C/* (C) Copyright Regents of the University of Minnesota, 1997 */
C/* */
C/* PSPASES : Scalable Parallel Direct Solver Library for Sparse */
C/* Symmetric Positive Definite Linear Systems. */
C/* (version 1.0) */
C/* */
C/* --------- */
C/* */
C/* ParMETIS and METIS, which are supplied with the PSPASES package as */
C/* the default ordering libraries, are solely owned by the University of */
C/* Minnesota and this notice does not apply to them. */
C/* */
C/* --------- */
C/* */
C/* Written by Mahesh Joshi, U of MN, Anshul Gupta, IBM Corp, and George */
C/* Karypis, U of MN. */
C/* */
C/*****************************************************************************/
C/* */
C/* This code is meant to be used solely for educational, research, and */
C/* benchmarking purposes by non-profit institutions and US government */
C/* agencies only. Use by any other organization requires prior written */
C/* permission from both IBM Corporation and the University of Minnesota. */
C/* The software may not be sold or redistributed. One may make copies */
C/* of the software or modify it for their use provided that the copies, */
C/* modified or otherwise, are not sold or distributed, are used under the */
C/* same terms and conditions, and this notice and any part of the source */
C/* code that follows this notice are not separated. */
C/* */
C/* As unestablished research software, this code is provided on an */
C/* ``as is'' basis without warranty of any kind, either expressed or */
C/* implied, including but not limited to implied warranties of */
C/* merchantability and fitness for a particular purpose. IBM does not */
C/* warrant that the functions contained in this software will meet the */
C/* user's requirements or that the operation of its routines will be */
C/* uninterrupted or error-free. Acceptance and use of this program */
C/* constitutes the user's understanding that he/she will have no recourse */
C/* to IBM for any actual or consequential damages, including, but not */
C/* limited to, lost profits or savings, arising out of the use or inability */
C/* to use these libraries. Even if the user informs IBM of the possibility */
C/* of such damages, IBM expects the user to accept the risk of any such */
C/* harm, or the user shall not attempt to use these libraries for any */
C/* purpose. */
C/* */
C/* The downloading, compiling, or executing any part of this software */
C/* constitutes an implicit agreement to these terms. These terms and */
C/* conditions are subject to change at any time without prior notice. */
C/* */
C/*****************************************************************************/
C/* $Id: README.INSTALL,v 1.2 1998/05/24 02:15:32 mjoshi Exp $ */
C/*****************************************************************************/
The views and opinions expressed in this page are strictly those of the page author.
The contents of this page have not been reviewed or approved by the University of Minnesota.