Soeren Gebbert
cfdf708cb0
Added OpenMP pragma to cholesky sband decomposition
|
13 gadi atpakaļ | |
|---|---|---|
| .. | ||
| test | 15 gadi atpakaļ | |
| ATLAS_wrapper_blas_level_1.c | 14 gadi atpakaļ | |
| DEPEND | 17 gadi atpakaļ | |
| Makefile | 16 gadi atpakaļ | |
| README | 17 gadi atpakaļ | |
| TODO | 15 gadi atpakaļ | |
| blas_level_1.c | 13 gadi atpakaļ | |
| blas_level_2.c | 13 gadi atpakaļ | |
| blas_level_3.c | 14 gadi atpakaļ | |
| brent.c | 14 gadi atpakaļ | |
| ccmath_grass_wrapper.c | 15 gadi atpakaļ | |
| dalloc.c | 14 gadi atpakaļ | |
| del2g.c | 16 gadi atpakaļ | |
| eigen_tools.c | 16 gadi atpakaļ | |
| fft.c | 14 gadi atpakaļ | |
| findzc.c | 14 gadi atpakaļ | |
| gauss.c | 15 gadi atpakaļ | |
| getg.c | 17 gadi atpakaļ | |
| gmathlib.dox | 15 gadi atpakaļ | |
| ialloc.c | 14 gadi atpakaļ | |
| la.c | 16 gadi atpakaļ | |
| lu.c | 16 gadi atpakaļ | |
| max_pow2.c | 17 gadi atpakaļ | |
| mult.c | 16 gadi atpakaļ | |
| rand1.c | 17 gadi atpakaļ | |
| solvers_classic_iter.c | 15 gadi atpakaļ | |
| solvers_direct.c | 13 gadi atpakaļ | |
| solvers_direct_cholesky_band.c | 13 gadi atpakaļ | |
| solvers_krylov.c | 15 gadi atpakaļ | |
| sparse_matrix.c | 15 gadi atpakaļ | |
| symmetric_band_matrix.c | 15 gadi atpakaļ | |
README
Date: Wed, 27 Sep 2000 15:41:24 +0000
From: David D Gray
I have checked in the gmath stuff to CVS. This directory might later
contain other kinds of generic routine (fft is one that comes to mind in
view of the investigations into NR). But for now there is only the
BLAS/LAPACK wrappers and supporting structure in la.c.
I thought to follow the libc in separating maths stuff to a separate
directory, even though they are G_*() functions.
I believe it is fairly intuitive to use - developers should use
G_matrix_init() to create a matrix. A vector can be built with setting
cols (or rows as appropriate) to 1. A G_vector_init() might be useful,
but it would have to distinguish between a row and column vector, (is it
needed?). Then use G_[set,get]_matrix_element() to set or retrieve
elements. No-one should need to ever worry about Fortran peculiarities
(like declaring doublereal variables, or calling a foo_() ).
Currently there are: add matrices A+B (doesn't use BLAS), multiply A.B,
invert A^-1, solve simple linear system A.X=B for X given B. Others are
easy to add as required.
David
-----------------------------------
Getting BLAS/LAPACK (one package):
http://netlib.bell-labs.com/netlib/master/readme.html
http://www.netlib.org
http://www.netlib.org/lapack/
Compiling of LAPACK (be sure to have "." in PATH):
Linux:
cd LAPACK
cp INSTALL/make.inc.LINUX make.inc
#now edit Makefile and
#comment line 11 and uncomment line 12 (to include "blaslib" compilation)
#compile with
make
Solaris:
cd LAPACK
cp INSTALL/make.inc.SUN4SOL2 make.inc
#now edit Makefile and
#comment line 11 and uncomment line 12 (to include "blaslib" compilation)
#compile with
make
etc.
See INSTALL/ for further platforms.
--------------------------------------------------------------------
Usage:
Sample Makefile:
MODULE_TOPDIR = /path/to/grass
PGM = test
LIBES = $(GMATHLIB) $(GISLIB)
DEPENDENCIES = $(GMATHLIB) $(GISLIB)
include $(MODULE_TOPDIR)/include/Make/Module.make
Sample module source:
#include
#include
#include
int main (int argc, char *argv[])
{
mat_struct *matrix;
matrix = G_matrix_init (3, 3, 1);
G_matrix_free (matrix);
return 0;
}
From: David D Gray
I have checked in the gmath stuff to CVS. This directory might later
contain other kinds of generic routine (fft is one that comes to mind in
view of the investigations into NR). But for now there is only the
BLAS/LAPACK wrappers and supporting structure in la.c.
I thought to follow the libc in separating maths stuff to a separate
directory, even though they are G_*() functions.
I believe it is fairly intuitive to use - developers should use
G_matrix_init() to create a matrix. A vector can be built with setting
cols (or rows as appropriate) to 1. A G_vector_init() might be useful,
but it would have to distinguish between a row and column vector, (is it
needed?). Then use G_[set,get]_matrix_element() to set or retrieve
elements. No-one should need to ever worry about Fortran peculiarities
(like declaring doublereal variables, or calling a foo_() ).
Currently there are: add matrices A+B (doesn't use BLAS), multiply A.B,
invert A^-1, solve simple linear system A.X=B for X given B. Others are
easy to add as required.
David
-----------------------------------
Getting BLAS/LAPACK (one package):
http://netlib.bell-labs.com/netlib/master/readme.html
http://www.netlib.org
http://www.netlib.org/lapack/
Compiling of LAPACK (be sure to have "." in PATH):
Linux:
cd LAPACK
cp INSTALL/make.inc.LINUX make.inc
#now edit Makefile and
#comment line 11 and uncomment line 12 (to include "blaslib" compilation)
#compile with
make
Solaris:
cd LAPACK
cp INSTALL/make.inc.SUN4SOL2 make.inc
#now edit Makefile and
#comment line 11 and uncomment line 12 (to include "blaslib" compilation)
#compile with
make
etc.
See INSTALL/ for further platforms.
--------------------------------------------------------------------
Usage:
Sample Makefile:
MODULE_TOPDIR = /path/to/grass
PGM = test
LIBES = $(GMATHLIB) $(GISLIB)
DEPENDENCIES = $(GMATHLIB) $(GISLIB)
include $(MODULE_TOPDIR)/include/Make/Module.make
Sample module source:
#include
#include
#include
int main (int argc, char *argv[])
{
mat_struct *matrix;
matrix = G_matrix_init (3, 3, 1);
G_matrix_free (matrix);
return 0;
}