zhufenggnss / psblas3 Goto Github PK

This directory contains  the PSBLAS library, version 3.5.


New features: additional Krylov methods  FCG and GCR.


Existing codes will work untouched with version 3.5, but they will
have to be recompiled; new methods have been added for the sake of GPU
computations, plus the usual bunch of bug fixes.


This version requires a working Fortran 2003 compiler; we do not use
all of the language features (specifically, so far we did not employ
FINAL subroutines), but the features we use were sufficient in
identifying bugs in ALL compilers we tried (all the bugs have been
reported and mostly fixed by the respective vendors). 


The new internals have been completely overhauled, and in many cases
rewritten; they now enable a much better interfacing with user-defined
storage formats. If the user is only interested in the predefined
formats, then the user's guide should be sufficient; what is somewhat
lacking is documentation on how to add to the library, i.e. a
developers' guide; stay tuned. 


 
The architecture of the Fortran 2003 sparse BLAS is described in 
S. Filippone, A. Buttari:
Object-Oriented Techniques for Sparse Matrix Computations in Fortran
2003, 
ACM Trans. on Math. Software, vol. 38, No. 4, 2012.

The ideas are explored further with the paper:

V. Cardellini, S. Filippone and D. Rouson
Design Patterns for  sparse-matrix computations on
hybrid CPU/GPU platforms, 
Scientific Programming, 22(2014), pp.1-19.


Version 1.0 of the library was described in:
S. Filippone, M. Colajanni
PSBLAS: A library for parallel linear algebra computation on sparse matrices
ACM Trans. on Math. Software, 26(4), Dec. 2000, pp. 527-550.


UTILITIES
The test/util directory contains some utilities to convert to/from
Harwell-Boeing and MatrixMarket file formats. 


DOCUMENTATION

See docs/psblas-3.4.pdf; an HTML version of the same document is
available in docs/html. 
Please consult the sample programs, especially 
test/pargen/ppde[23]d.f90



OTHER SOFTWARE CREDITS 

We originally included a modified implementation of some of the
Sparker (serial sparse BLAS)  material; this has been completely
rewritten, way beyond the intention(s) and responsibilities of the
original developers. 
The main reference for the serial sparse BLAS is: 
Duff, I., Marrone, M., Radicati, G., and Vittoli, C.
Level 3 basic linear algebra subprograms for sparse matrices: a user 
level interface
ACM Trans. Math. Softw., 23(3), 379-401, 1997.


INSTALLING

To compile and run our software you will need the following
prerequisites (see also SERIAL below):

1. A working version of MPI

2. A version of the BLAS; if you don't have a specific version for
   your platform you may try ATLAS available from
   http://math-atlas.sourceforge.net/ 

3. We have had good results with  the METIS library, from 
   http://www-users.cs.umn.edu/~karypis/metis/metis/main.html 
   This is optional; it is  used in the util and test/fileread
   directories but only if you specify --with-metis.

4. If you have the AMD package of Davis, Duff and Amestoy, you can
   specify  --with-amd (see ./configure --help for more details). 

The configure script will generate a Make.inc file suitable for
building the library.  
The script is capable of recognizing the needed libraries with their
default names; if they are in unusual places consider adding the paths
with --with-libs, or explicitly specifying the names in --with-blas, etc. 
Please note that a common way for the configure script to fail is to
specify inconsistent MPI vs. plain compilers, either directly or
indirectly via environment variables; e.g. specifying the Intel
compiler with FC=ifort while at the same time having an  MPIFC=mpif90
which points to GNU Fortran. 
The best way to avoid this situation is (in our opinion) to use the
environment modules package (see http://modules.sourceforge.net/), and
load the relevant variables with (e.g.)  
module load gnu46 openmpi
This will delegate to the modules setup to make sure that the version
of openmpi in use is the one compiled with the gnu46 compilers. 
After the configure script has completed you can always tweak the
Make.inc  file yourself. 

After you have Make.inc fixed,  run 
 make 
to  compile the library; go to the test directory and its
subdirectories to get test programs done.   
If you specify --prefix=/path you can do make install and the
libraries will be installed under /path/lib, while the module files
will be installed under /path/include.

SERIAL

Configuring with --enable-serial will provide a fake MPI stub library
that enables running in pure serial mode; no MPI installation is
needed in this case (but note that the fake MPI stubs are only
guaranteed to cover what we use internally, it's not a complete
replacement). 

EXPERIMENTAL

We have an experimental flag --enable-long-integers that will enable having
8-byte integer data, allowing an index space larger than 2G; some
small cases have been tested but we do not offer full guarantee (yet). 


COMPILER NOTES.

This code is confirmed to work with the following compilers (or
later versions thereof): 
       NAGware 5.2;
       GNU 4.7.3;  
       Cray CCE 8.0.1;

They are all recognized by the configure script. 

To make the script work with the Cray CCE environment, it is
recommended to use the following:
./configure FC=ftn F77=ftn CC=cc MPF90=ftn MPF77=ftn MPCC=cc
with both CCE and GNU lower-level compilers. 

XLF 13.1 configures correctly, but then fails with ICEs (Internal
Compiler Error) at build time. We do not yet know whether XLF 14
compiles correctly. 

For the GNU compilers 4.6.x we are aware of a number of memory management
issues that might surface in your applications; most of them (that
we're aware of) are solved in version 4.7.2 or later.  

The Intel compiler up to version 12.1 fails to compile, as of the last
version we got access to. 



KNOWN ISSUES.

An annoying problem exists with some versions of MPI: if the configure
script sets -DMPI_MOD the MPI call interfaces will be resolved by
using the MPI Fortran module. 
However usage of the  module may cause compilation to fail when
coupled with the debugging option -g, because the compiler complains
that it cannot find a matching interface for some of the
communication routines.
The solution: either take out the -g option, or, if you really need to
debug, edit Make.inc to force -DMPI_H in place of -DMPI_MOD. 


TODO:
Fix all reamining bugs. Bugs? We dont' have any ! ;-)


The PSBLAS team. 

RELATED SOFTWARE

If you are looking for more sophisticated preconditioners, you may be
interested in the package  MLD2P4 from http://www.mld2p4.it 


Contact: Salvatore Filippone     [email protected]

Credits for version 3:
Salvatore  Filippone
Alfredo    Buttari
Ambra	   Abdullahi Hassan
Pasqua	   D'Ambra

Credits for version 2.2:
Michele Martone contributed the initial version of the configure
machinery.  

Credits for version 2.0:
Salvatore  Filippone
Alfredo    Buttari


Credits for version 1.0:

Salvatore  Filippone
Michele    Colajanni
Fabio      Cerioni
Stefano    Maiolatesi
Dario      Pascucci