pastix.h

Summary

pastix.h
PaStiX interface	Parallel Sparse matriX solver interface
Main PaStiX functions
pastix	Computes steps of the resolution of Ax=b linear system, using direct methods.
Example	from file simple.c :
dpastix	Computes steps of the resolution of Ax=b linear system, using direct methods.
Example	from file simple_dist.c :
Thread functions
pastix_bindThreads	Set bindtab in pastix_data, it gives for each thread the CPU to bind in to.
Checking the matrix.
pastix_checkMatrix	Check the matrix :
Getting solver distribution.
pastix_getLocalNodeNbr	Return the node number in the new distribution computed by analyze step.
pastix_getLocalNodeLst	Fill in nodelst with the list of local nodes/columns.
pastix_getLocalUnknownNbr	Return the unknown number in the new distribution computed by analyze step.
pastix_getLocalUnknownLst	Fill in unknownlst with the list of local unknowns/clumns.
About the Schur complement.
pastix_setSchurUnknownList	Set the list of unknowns to isolate at the end of the matrix via permutations.
pastix_getSchurLocalNodeNbr	Compute the number of nodes in the local part of the Schur.
pastix_getSchurLocalUnkownNbr	Compute the number of unknowns in the local part of the Schur.
pastix_getSchurLocalNodeList	Compute the list of nodes in the local part of the Schur.
pastix_getSchurLocalUnkownList	Compute the list of unknowns in the local part of the Schur.
pastix_setSchurArray	Give user memory area to store Schur in PaStiX.
pastix_getSchur	Get the Schur complement from PaStiX.
About parameters.
pastix_initParam	Sets default parameters for iparm and dparm
About Scaling.	Working on it...

PaStiX interface

ComplexFloat_

Parallel Sparse matriX solver interface

Authors

Astrid CASADEI	astrid..nosp@m.casadei@inri.nosp@m.a.fr
Mathieu FAVERGE	fav.nosp@m.erge@labr.nosp@m.i.fr
Xavier LACOSTE	xavier..nosp@m.lacoste@inri.nosp@m.a.fr
Pierre RAMET	ra.nosp@m.met@labr.nosp@m.i.fr

Blabla

Main PaStiX functions

pastix

void pastix( pastix_data_t ** pastix_data,
MPI_Comm pastix_comm,
INT n,
INT * colptr,
INT * row,
FLOAT * avals,
INT * perm,
INT * invp,
FLOAT * b,
INT rhs,
INT * iparm,
double * dparm )

Computes steps of the resolution of Ax=b linear system, using direct methods.

The matrix is given in CSC format.

Parameters

pastix_data	Data used for a step by step execution.
pastix_comm	MPI communicator which compute the resolution.
n	Size of the system.
colptr	Tabular containing the start of each column in row and avals tabulars.
row	Tabular containing the row number for each element sorted by column.
avals	Tabular containing the values of each element sorted by column.
perm	Permutation tabular for the renumerotation of the unknowns.
invp	Reverse permutation tabular for the renumerotation of the unknowns.
b	Right hand side vector(s).
rhs	Number of right hand side vector(s).
iparm	Integer parameters given to pastix.
dparm	Double parameters given to pâstix.

Summary

Example

from file simple.c :

Example

from file simple.c :

/\*******************************************\/
/\*    Check Matrix format                  *\/
/\*******************************************\/
/\*
 * Matrix needs :
 *    - to be in fortran numbering
 *    - to have only the lower triangular part in symmetric case
 *    - to have a graph with a symmetric structure in unsymmetric case
 *\/
pastix_checkMatrix( MPI_COMM_WORLD, verbosemode,
                    (MTX_ISSYM(type) ? API_SYM_YES : API_SYM_NO),
                    API_YES,
                    ncol, &colptr, &rows, &values, NULL);

/\*******************************************\/
/\* Initialize parameters to default values *\/
/\*******************************************\/
iparm[IPARM_MODIFY_PARAMETER] = API_NO;
pastix(&pastix_data, MPI_COMM_WORLD,
       ncol, colptr, rows, values,
       perm, invp, rhs, 1, iparm, dparm);

/\*******************************************\/
/\*       Customize some parameters         *\/
/\*******************************************\/
iparm[IPARM_THREAD_NBR] = nbthread;
if (MTX_ISSYM(type)) {
  :wiparm[IPARM_SYM]           = API_SYM_YES;
  iparm[IPARM_FACTORIZATION] = API_FACT_LDLT;
}
else{
  iparm[IPARM_SYM]           = API_SYM_NO;
  iparm[IPARM_FACTORIZATION] = API_FACT_LU;
}
iparm[IPARM_START_TASK]          = API_TASK_ORDERING;
iparm[IPARM_END_TASK]            = API_TASK_CLEAN;

/\*******************************************\/
/\*           Save the rhs                  *\/
/\*    (it will be replaced by solution)    *\/
/\*******************************************\/
rhssaved = malloc(ncol*sizeof(pastix_float_t));
memcpy(rhssaved, rhs, ncol*sizeof(pastix_float_t));

/\*******************************************\/
/\*           Call pastix                   *\/
/\*******************************************\/
perm = malloc(ncol*sizeof(pastix_int_t));
invp = malloc(ncol*sizeof(pastix_int_t));

pastix(&pastix_data, MPI_COMM_WORLD,
 ncol, colptr, rows, values,
 perm, invp, rhs, 1, iparm, dparm);

dpastix

Computes steps of the resolution of Ax=b linear system, using direct methods. Here the matrix is given distributed.

The matrix is given in CSCD format.

Parameters

pastix_data	Data used for a step by step execution.
pastix_comm	MPI communicator which compute the resolution.
n	Size of the system.
colptr	Tabular containing the start of each column in row and avals tabulars.
row	Tabular containing the row number for each element sorted by column.
avals	Tabular containing the values of each element sorted by column.
loc2glob	Global column number of the local columns.
perm	Permutation tabular for the renumerotation of the unknowns.
invp	Reverse permutation tabular for the renumerotation of the unknowns.
b	Right hand side vector(s).
rhs	Number of right hand side vector(s).
iparm	Integer parameters given to pastix.
dparm	Double parameters given to pâstix.

Summary

Example	from file simple_dist.c :
Thread functions

Example

from file simple_dist.c :

/\*******************************************\/
/\*    Check Matrix format                  *\/
/\*******************************************\/
/\*
 * Matrix needs :
 *    - to be in fortran numbering
 *    - to have only the lower triangular part in symmetric case
 *    - to have a graph with a symmetric structure in unsymmetric case
 *\/
pastix_checkMatrix( MPI_COMM_WORLD, verbosemode,
                    (MTX_ISSYM(type) ? API_SYM_YES : API_SYM_NO),
                    API_YES,
                    ncol, &colptr, &rows, &values, &loc2glob, 1);

/\*******************************************\/
/\* Initialize parameters to default values *\/
/\*******************************************\/
iparm[IPARM_MODIFY_PARAMETER] = API_NO;
dpastix(&pastix_data, MPI_COMM_WORLD,
        ncol, colptr, rows, values, loc2glob,
        perm, invp, rhs, 1, iparm, dparm);

/\*******************************************\/
/\*       Customize some parameters         *\/
/\*******************************************\/
iparm[IPARM_THREAD_NBR] = nbthread;
if (MTX_ISSYM(type))
  {
    iparm[IPARM_SYM]           = API_SYM_YES;
    iparm[IPARM_FACTORIZATION] = API_FACT_LDLT;
  }
else
  {
    iparm[IPARM_SYM]           = API_SYM_NO;
    iparm[IPARM_FACTORIZATION] = API_FACT_LU;
  }
iparm[IPARM_MATRIX_VERIFICATION] = API_NO;

iparm[IPARM_START_TASK]          = API_TASK_ORDERING;
iparm[IPARM_END_TASK]            = API_TASK_BLEND;

/\*******************************************\/
/\*           Call pastix                   *\/
/\*******************************************\/
perm = malloc(ncol*sizeof(pastix_int_t));
/\* No need to allocate invp in dpastix *\/

dpastix(&pastix_data, MPI_COMM_WORLD,
        ncol, colptr, rows, NULL, loc2glob,
        perm, NULL, NULL, 1, iparm, dparm);

/\* Redistributing the matrix *\/

ncol2 = pastix_getLocalNodeNbr(&pastix_data);

if (NULL == (loc2glob2 = malloc(ncol2 * sizeof(pastix_int_t))))
  {
    fprintf(stderr, "Malloc error\n");
    return EXIT_FAILURE;
  }

pastix_getLocalNodeLst(&pastix_data, loc2glob2);

if (EXIT_SUCCESS != cscd_redispatch(ncol,   colptr,   rows,
                                    values,    rhs,  loc2glob,
                                    ncol2, &colptr2, &rows2,
                                    &values2, &rhs2, loc2glob2,
                                    MPI_COMM_WORLD))
  return EXIT_FAILURE;

free(colptr);
free(rows);
free(values);
free(rhs);
free(loc2glob);
free(perm);

iparm[IPARM_START_TASK]          = API_TASK_NUMFACT;
iparm[IPARM_END_TASK]            = API_TASK_CLEAN;


dpastix(&pastix_data, MPI_COMM_WORLD,
        ncol2, colptr2, rows2, values2, loc2glob2,
        perm, invp, rhs2, 1, iparm, dparm);

Thread functions

pastix_bindThreads

Set bindtab in pastix_data, it gives for each thread the CPU to bind in to. bindtab follows this organisation :

bindtab[threadnum] = cpu to set thread threadnum.

Parameters

pastix_data	Structure de donnée pour l’utilisation step by step
thrdnbr	Nombre de threads / Taille du tableau
bindtab	Tableau de correspondance entre chaque thread et coeur de la machine

Summary

Checking the matrix.

pastix_checkMatrix

Check the matrix :

Renumbers in Fortran numerotation (base 1) if needed (base 0)
Can scale the matrix if compiled with -DMC64 -DSCALING (untested)
Checks the symetry of the graph in non symmetric mode. With non distributed matrices, with flagcor == API_YES, tries to correct the matrix.
sort the CSC.

Parameters

pastix_comm	PaStiX MPI communicator
verb	Level of prints (API_VERBOSE_[NOT\|NO\|YES])
flagsym	Indicate if the given matrix is symetric (API_SYM_YES or API_SYM_NO)
flagcor	Indicate if we permit the function to reallocate the matrix.
n	Size of the matrix.
colptr	First element of each row in row and avals.
row	Row of each element of the matrix.
avals	Value of each element of the matrix.
loc2glob	Global column number of local columns.
dof	Number of degrees of freedom.

Summary

Getting solver distribution.

pastix_getLocalNodeNbr

Return the node number in the new distribution computed by analyze step. Needs analyze step to be runned with pastix_data before.

Parameters

pastix_data

Data used for a step by step execution.

Returns

Number of local nodes/columns in new distribution.

pastix_getLocalNodeLst

Fill in nodelst with the list of local nodes/columns. Needs nodelst to be allocated with nodenbr*sizeof(pastix_int_t), where nodenbr has been computed by pastix_getLocalNodeNbr.

Parameters

pastix_data	Data used for a step by step execution.
nodelst	An array where to write the list of local nodes/columns.

pastix_getLocalUnknownNbr

Return the unknown number in the new distribution computed by analyze step. Needs analyze step to be runned with pastix_data before.

Parameters

pastix_data

Data used for a step by step execution.

Returns

Number of local unknowns/columns in new distribution.

pastix_getLocalUnknownLst

Fill in unknownlst with the list of local unknowns/clumns. Needs unknownlst to be allocated with unknownnbr*sizeof(pastix_int_t), where unknownnbr has been computed by pastix_getLocalUnknownNbr.

Parameters

pastix_data	Data used for a step by step execution.
unknownlst	An array where to write the list of local unknowns/columns.

Summary

About the Schur complement.

pastix_setSchurUnknownList

Set the list of unknowns to isolate at the end of the matrix via permutations.

Has to be called if using IPARM_SCHUR = API_YES.

Parameters

pastix_data	Data used for a step by step execution.
n	Number of unknowns.
list	List of unknowns.

pastix_getSchurLocalNodeNbr

Compute the number of nodes in the local part of the Schur.

Parameters

pastix_data	Common data structure for PaStiX calls.
nodeNbr	(out) Number of nodes in Schur (local).

Returns

NO_ERR

For the moment

TODO: Error management.

pastix_getSchurLocalUnkownNbr

Compute the number of unknowns in the local part of the Schur.

Parameters

pastix_data	Common data structure for PaStiX calls.
unknownNbr	(out) Number of unknowns in Schur (local).

Returns

NO_ERR

For the moment

TODO: Error management.

pastix_getSchurLocalNodeList

Compute the list of nodes in the local part of the Schur.

Parameters

pastix_data	Common data structure for PaStiX calls.
nodes	(out) Nodes in Schur (local).

Returns

NO_ERR

For the moment

TODO: Error management.

pastix_getSchurLocalUnkownList

Compute the list of unknowns in the local part of the Schur.

Parameters

pastix_data	Common data structure for PaStiX calls.
unknowns	(out) Unknowns in Schur (local).

Returns

NO_ERR

For the moment

TODO: Error management.

pastix_setSchurArray

Give user memory area to store Schur in PaStiX.

Parameters

pastix_data	Common data structure for PaStiX calls.
array	Memory area to store the Schur.

Returns

NO_ERR

For the moment

TODO: Error management.

pastix_getSchur

Get the Schur complement from PaStiX.

Schur complement is a dense block in a column scheme.

Parameters

pastix_data	Data used for a step by step execution.
schur	Array to fill-in with Schur complement.

Summary

About parameters.

pastix_initParam

Sets default parameters for iparm and dparm

Parameters

iparm	tabular of IPARM_SIZE integer parameters.
dparm	tabular of DPARM_SIZE double parameters.

About Scaling.

Working on it...