xref: /petsc/src/mat/graphops/partition/impls/chaco/chaco.c (revision 8be712e46db5d855f641c6bd97b4543e0efe65bd)

#include <../src/mat/impls/adj/mpi/mpiadj.h> /*I "petscmat.h" I*/

#if defined(PETSC_HAVE_UNISTD_H)
  #include <unistd.h>
#endif

#if defined(PETSC_HAVE_CHACO_INT_ASSIGNMENT)
  #include <chaco.h>
#else
/* Older versions of Chaco do not have an include file */
PETSC_EXTERN int interface(int nvtxs, int *start, int *adjacency, int *vwgts, float *ewgts, float *x, float *y, float *z, char *outassignname, char *outfilename, short *assignment, int architecture, int ndims_tot, int mesh_dims[3], double *goal, int global_method, int local_method, int rqi_flag, int vmax, int ndims, double eigtol, long seed);
#endif

extern int FREE_GRAPH;

/*
int       nvtxs;                number of vertices in full graph
int      *start;                start of edge list for each vertex
int      *adjacency;            edge list data
int      *vwgts;                weights for all vertices
float    *ewgts;                weights for all edges
float    *x, *y, *z;            coordinates for inertial method
char     *outassignname;        name of assignment output file
char     *outfilename;          output file name
short    *assignment;           set number of each vtx (length n)
int       architecture;         0 => hypercube, d => d-dimensional mesh
int       ndims_tot;            total number of cube dimensions to divide
int       mesh_dims[3];         dimensions of mesh of processors
double   *goal;                 desired set sizes for each set
int       global_method;        global partitioning algorithm
int       local_method;         local partitioning algorithm
int       rqi_flag;             should I use RQI/Symmlq eigensolver?
int       vmax;                 how many vertices to coarsen down to?
int       ndims;                number of eigenvectors (2^d sets)
double    eigtol;               tolerance on eigenvectors
long      seed;                 for random graph mutations
*/

typedef struct {
  PetscBool         verbose;
  PetscInt          eignum;
  PetscReal         eigtol;
  MPChacoGlobalType global_method; /* global method */
  MPChacoLocalType  local_method;  /* local method */
  MPChacoEigenType  eigen_method;  /* eigensolver */
  PetscInt          nbvtxcoarsed;  /* number of vertices for the coarse graph */
} MatPartitioning_Chaco;

#define SIZE_LOG 10000 /* size of buffer for mesg_log */

static PetscErrorCode MatPartitioningApply_Chaco(MatPartitioning part, IS *partitioning)
{
  int                    cerr;
  PetscInt              *parttab, *locals, i, nb_locals, M, N;
  PetscMPIInt            size, rank;
  Mat                    mat = part->adj, matAdj, matSeq, *A;
  Mat_MPIAdj            *adj;
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;
  PetscBool              flg;
  IS                     isrow, iscol;
  int                    nvtxs, *start, *adjacency, *vwgts, architecture, ndims_tot;
  int                    mesh_dims[3], global_method, local_method, rqi_flag, vmax, ndims;
#if defined(PETSC_HAVE_CHACO_INT_ASSIGNMENT)
  int *assignment;
#else
  short *assignment;
#endif
  double eigtol;
  long   seed;
  char  *mesg_log;
#if defined(PETSC_HAVE_UNISTD_H)
  int fd_stdout, fd_pipe[2], count;
#endif

  PetscFunctionBegin;
  PetscCheck(!part->use_edge_weights, PetscObjectComm((PetscObject)part), PETSC_ERR_SUP, "Chaco does not support edge weights");
  FREE_GRAPH = 0; /* otherwise Chaco will attempt to free memory for adjacency graph */
  PetscCallMPI(MPI_Comm_size(PetscObjectComm((PetscObject)mat), &size));
  PetscCallMPI(MPI_Comm_rank(PetscObjectComm((PetscObject)mat), &rank));
  PetscCall(PetscObjectTypeCompare((PetscObject)mat, MATMPIADJ, &flg));
  if (size > 1) {
    if (flg) {
      PetscCall(MatMPIAdjToSeq(mat, &matSeq));
    } else {
      PetscCall(PetscInfo(part, "Converting distributed matrix to sequential: this could be a performance loss\n"));
      PetscCall(MatGetSize(mat, &M, &N));
      PetscCall(ISCreateStride(PETSC_COMM_SELF, M, 0, 1, &isrow));
      PetscCall(ISCreateStride(PETSC_COMM_SELF, N, 0, 1, &iscol));
      PetscCall(MatCreateSubMatrices(mat, 1, &isrow, &iscol, MAT_INITIAL_MATRIX, &A));
      PetscCall(ISDestroy(&isrow));
      PetscCall(ISDestroy(&iscol));
      matSeq = *A;
      PetscCall(PetscFree(A));
    }
  } else {
    PetscCall(PetscObjectReference((PetscObject)mat));
    matSeq = mat;
  }

  if (!flg) { /* convert regular matrix to MPIADJ */
    PetscCall(MatConvert(matSeq, MATMPIADJ, MAT_INITIAL_MATRIX, &matAdj));
  } else {
    PetscCall(PetscObjectReference((PetscObject)matSeq));
    matAdj = matSeq;
  }

  adj = (Mat_MPIAdj *)matAdj->data; /* finally adj contains adjacency graph */

  /* arguments for Chaco library */
  nvtxs         = mat->rmap->N;         /* number of vertices in full graph */
  start         = adj->i;               /* start of edge list for each vertex */
  vwgts         = part->vertex_weights; /* weights for all vertices */
  architecture  = 1;                    /* 0 => hypercube, d => d-dimensional mesh */
  ndims_tot     = 0;                    /* total number of cube dimensions to divide */
  mesh_dims[0]  = part->n;              /* dimensions of mesh of processors */
  global_method = chaco->global_method; /* global partitioning algorithm */
  local_method  = chaco->local_method;  /* local partitioning algorithm */
  rqi_flag      = chaco->eigen_method;  /* should I use RQI/Symmlq eigensolver? */
  vmax          = chaco->nbvtxcoarsed;  /* how many vertices to coarsen down to? */
  ndims         = chaco->eignum;        /* number of eigenvectors (2^d sets) */
  eigtol        = chaco->eigtol;        /* tolerance on eigenvectors */
  seed          = 123636512;            /* for random graph mutations */

  PetscCall(PetscMalloc1(mat->rmap->N, &assignment));
  PetscCall(PetscMalloc1(start[nvtxs], &adjacency));
  for (i = 0; i < start[nvtxs]; i++) adjacency[i] = (adj->j)[i] + 1; /* 1-based indexing */

    /* redirect output to buffer */
#if defined(PETSC_HAVE_UNISTD_H)
  fd_stdout = dup(1);
  PetscCheck(!pipe(fd_pipe), PETSC_COMM_SELF, PETSC_ERR_SYS, "Could not open pipe");
  close(1);
  dup2(fd_pipe[1], 1);
  PetscCall(PetscMalloc1(SIZE_LOG, &mesg_log));
#endif

  /* library call */
  cerr = interface(nvtxs, start, adjacency, vwgts, NULL, NULL, NULL, NULL, NULL, NULL, assignment, architecture, ndims_tot, mesh_dims, NULL, global_method, local_method, rqi_flag, vmax, ndims, eigtol, seed);

#if defined(PETSC_HAVE_UNISTD_H)
  PetscCall(PetscFFlush(stdout));
  count = read(fd_pipe[0], mesg_log, (SIZE_LOG - 1) * sizeof(char));
  if (count < 0) count = 0;
  mesg_log[count] = 0;
  close(1);
  dup2(fd_stdout, 1);
  close(fd_stdout);
  close(fd_pipe[0]);
  close(fd_pipe[1]);
  if (chaco->verbose) PetscCall(PetscPrintf(PetscObjectComm((PetscObject)mat), "%s", mesg_log));
  PetscCall(PetscFree(mesg_log));
#endif
  PetscCheck(!cerr, PETSC_COMM_SELF, PETSC_ERR_LIB, "Chaco failed");

  PetscCall(PetscMalloc1(mat->rmap->N, &parttab));
  for (i = 0; i < nvtxs; i++) parttab[i] = assignment[i];

  /* creation of the index set */
  nb_locals = mat->rmap->n;
  locals    = parttab + mat->rmap->rstart;
  PetscCall(ISCreateGeneral(PetscObjectComm((PetscObject)part), nb_locals, locals, PETSC_COPY_VALUES, partitioning));

  /* clean up */
  PetscCall(PetscFree(parttab));
  PetscCall(PetscFree(adjacency));
  PetscCall(PetscFree(assignment));
  PetscCall(MatDestroy(&matSeq));
  PetscCall(MatDestroy(&matAdj));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningView_Chaco(MatPartitioning part, PetscViewer viewer)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;
  PetscBool              isascii;

  PetscFunctionBegin;
  PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &isascii));
  if (isascii) {
    PetscCall(PetscViewerASCIIPrintf(viewer, "  Global method: %s\n", MPChacoGlobalTypes[chaco->global_method]));
    PetscCall(PetscViewerASCIIPrintf(viewer, "  Local method: %s\n", MPChacoLocalTypes[chaco->local_method]));
    PetscCall(PetscViewerASCIIPrintf(viewer, "  Number of vertices for the coarse graph: %" PetscInt_FMT "\n", chaco->nbvtxcoarsed));
    PetscCall(PetscViewerASCIIPrintf(viewer, "  Eigensolver: %s\n", MPChacoEigenTypes[chaco->eigen_method]));
    PetscCall(PetscViewerASCIIPrintf(viewer, "  Tolerance for eigensolver: %g\n", chaco->eigtol));
    PetscCall(PetscViewerASCIIPrintf(viewer, "  Number of eigenvectors: %" PetscInt_FMT "\n", chaco->eignum));
  }
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*@
  MatPartitioningChacoSetGlobal - Set the global method for Chaco partitioner.

  Collective

  Input Parameters:
+ part   - the partitioning context
- method - one of `MP_CHACO_MULTILEVEL`, `MP_CHACO_SPECTRAL`, `MP_CHACO_LINEAR`,
            `MP_CHACO_RANDOM` or `MP_CHACO_SCATTERED`

  Options Database Key:
. -mat_partitioning_chaco_global <method> - the global method

  Level: advanced

  Note:
  The default is the multi-level method. See Chaco documentation for
  additional details.

.seealso: `MatPartitioning`, `MatPartioningSetType()`, `MatPartitioningType`, `MATPARTITIONINGCHACO`, `MatPartitioningChacoSetLocal()`, `MatPartitioningChacoGetGlobal()`
@*/
PetscErrorCode MatPartitioningChacoSetGlobal(MatPartitioning part, MPChacoGlobalType method)
{
  PetscFunctionBegin;
  PetscValidHeaderSpecific(part, MAT_PARTITIONING_CLASSID, 1);
  PetscValidLogicalCollectiveEnum(part, method, 2);
  PetscTryMethod(part, "MatPartitioningChacoSetGlobal_C", (MatPartitioning, MPChacoGlobalType), (part, method));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningChacoSetGlobal_Chaco(MatPartitioning part, MPChacoGlobalType method)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  switch (method) {
  case MP_CHACO_MULTILEVEL:
  case MP_CHACO_SPECTRAL:
  case MP_CHACO_LINEAR:
  case MP_CHACO_RANDOM:
  case MP_CHACO_SCATTERED:
    chaco->global_method = method;
    break;
  default:
    SETERRQ(PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Chaco: Unknown or unsupported option");
  }
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*@
  MatPartitioningChacoGetGlobal - Get the global method used by the Chaco partitioner.

  Not Collective

  Input Parameter:
. part - the partitioning context

  Output Parameter:
. method - the method

  Level: advanced

.seealso: `MatPartitioningType`, `MATPARTITIONINGCHACO`, `MatPartitioningChacoSetGlobal()`
@*/
PetscErrorCode MatPartitioningChacoGetGlobal(MatPartitioning part, MPChacoGlobalType *method)
{
  PetscFunctionBegin;
  PetscValidHeaderSpecific(part, MAT_PARTITIONING_CLASSID, 1);
  PetscAssertPointer(method, 2);
  PetscTryMethod(part, "MatPartitioningChacoGetGlobal_C", (MatPartitioning, MPChacoGlobalType *), (part, method));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningChacoGetGlobal_Chaco(MatPartitioning part, MPChacoGlobalType *method)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  *method = chaco->global_method;
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*@
  MatPartitioningChacoSetLocal - Set the local method for the Chaco partitioner.

  Collective

  Input Parameters:
+ part   - the partitioning context
- method - one of `MP_CHACO_KERNIGHAN` or `MP_CHACO_NONE`

  Options Database Key:
. -mat_partitioning_chaco_local <method> - the local method

  Level: advanced

  Note:
  The default is to apply the Kernighan-Lin heuristic. See Chaco documentation
  for additional details.

.seealso: `MatPartitioningType`, `MATPARTITIONINGCHACO`, `MatPartitioningChacoSetGlobal()`, `MatPartitioningChacoGetLocal()`
@*/
PetscErrorCode MatPartitioningChacoSetLocal(MatPartitioning part, MPChacoLocalType method)
{
  PetscFunctionBegin;
  PetscValidHeaderSpecific(part, MAT_PARTITIONING_CLASSID, 1);
  PetscValidLogicalCollectiveEnum(part, method, 2);
  PetscTryMethod(part, "MatPartitioningChacoSetLocal_C", (MatPartitioning, MPChacoLocalType), (part, method));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningChacoSetLocal_Chaco(MatPartitioning part, MPChacoLocalType method)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  switch (method) {
  case MP_CHACO_KERNIGHAN:
  case MP_CHACO_NONE:
    chaco->local_method = method;
    break;
  default:
    SETERRQ(PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Chaco: Unknown or unsupported option");
  }
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*@
  MatPartitioningChacoGetLocal - Get local method used by the Chaco partitioner.

  Not Collective

  Input Parameter:
. part - the partitioning context

  Output Parameter:
. method - the method

  Level: advanced

.seealso: `MatPartitioningType`, `MATPARTITIONINGCHACO`, `MatPartitioningChacoSetLocal()`
@*/
PetscErrorCode MatPartitioningChacoGetLocal(MatPartitioning part, MPChacoLocalType *method)
{
  PetscFunctionBegin;
  PetscValidHeaderSpecific(part, MAT_PARTITIONING_CLASSID, 1);
  PetscAssertPointer(method, 2);
  PetscUseMethod(part, "MatPartitioningChacoGetLocal_C", (MatPartitioning, MPChacoLocalType *), (part, method));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningChacoGetLocal_Chaco(MatPartitioning part, MPChacoLocalType *method)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  *method = chaco->local_method;
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*@
  MatPartitioningChacoSetCoarseLevel - Set the coarse level parameter for the
  Chaco partitioner.

  Collective

  Input Parameters:
+ part  - the partitioning context
- level - the coarse level in range [0.0,1.0]

  Options Database Key:
. -mat_partitioning_chaco_coarse <l> - Coarse level

  Level: advanced

.seealso: `MatPartitioningType`, `MatPartitioning`, `MATPARTITIONINGCHACO`
@*/
PetscErrorCode MatPartitioningChacoSetCoarseLevel(MatPartitioning part, PetscReal level)
{
  PetscFunctionBegin;
  PetscValidHeaderSpecific(part, MAT_PARTITIONING_CLASSID, 1);
  PetscValidLogicalCollectiveReal(part, level, 2);
  PetscTryMethod(part, "MatPartitioningChacoSetCoarseLevel_C", (MatPartitioning, PetscReal), (part, level));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningChacoSetCoarseLevel_Chaco(MatPartitioning part, PetscReal level)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  PetscCheck(level >= 0.0 && level < 1.0, PETSC_COMM_SELF, PETSC_ERR_ARG_OUTOFRANGE, "Chaco: level of coarsening out of range [0.0-1.0]");
  chaco->nbvtxcoarsed = (PetscInt)(part->adj->cmap->N * level);
  if (chaco->nbvtxcoarsed < 20) chaco->nbvtxcoarsed = 20;
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*@
  MatPartitioningChacoSetEigenSolver - Set the eigensolver method for Chaco partitioner.

  Collective

  Input Parameters:
+ part   - the partitioning context
- method - one of `MP_CHACO_LANCZOS` or `MP_CHACO_RQI`

  Options Database Key:
. -mat_partitioning_chaco_eigen_solver <method> - the eigensolver

  Level: advanced

  Note:
  The default is to use a Lanczos method. See Chaco documentation for details.

.seealso: `MatPartitioningType`, `MatPartitioning`, `MATPARTITIONINGCHACO`, `MatPartitioningChacoSetEigenTol()`, `MatPartitioningChacoSetEigenNumber()`,
          `MatPartitioningChacoGetEigenSolver()`
@*/
PetscErrorCode MatPartitioningChacoSetEigenSolver(MatPartitioning part, MPChacoEigenType method)
{
  PetscFunctionBegin;
  PetscValidHeaderSpecific(part, MAT_PARTITIONING_CLASSID, 1);
  PetscValidLogicalCollectiveEnum(part, method, 2);
  PetscTryMethod(part, "MatPartitioningChacoSetEigenSolver_C", (MatPartitioning, MPChacoEigenType), (part, method));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningChacoSetEigenSolver_Chaco(MatPartitioning part, MPChacoEigenType method)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  switch (method) {
  case MP_CHACO_LANCZOS:
  case MP_CHACO_RQI:
    chaco->eigen_method = method;
    break;
  default:
    SETERRQ(PETSC_COMM_SELF, PETSC_ERR_ARG_WRONG, "Chaco: Unknown or unsupported option");
  }
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*@
  MatPartitioningChacoGetEigenSolver - Get the eigensolver used by the Chaco partitioner.

  Not Collective

  Input Parameter:
. part - the partitioning context

  Output Parameter:
. method - the method

  Level: advanced

.seealso: `MatPartitioningType`, `MatPartitioning`, `MATPARTITIONINGCHACO`, `MatPartitioningChacoSetEigenSolver()`
@*/
PetscErrorCode MatPartitioningChacoGetEigenSolver(MatPartitioning part, MPChacoEigenType *method)
{
  PetscFunctionBegin;
  PetscValidHeaderSpecific(part, MAT_PARTITIONING_CLASSID, 1);
  PetscAssertPointer(method, 2);
  PetscUseMethod(part, "MatPartitioningChacoGetEigenSolver_C", (MatPartitioning, MPChacoEigenType *), (part, method));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningChacoGetEigenSolver_Chaco(MatPartitioning part, MPChacoEigenType *method)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  *method = chaco->eigen_method;
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*@
  MatPartitioningChacoSetEigenTol - Sets the tolerance for the eigensolver used by Chaco

  Collective

  Input Parameters:
+ part - the partitioning context
- tol  - the tolerance

  Options Database Key:
. -mat_partitioning_chaco_eigen_tol <tol> - Tolerance for eigensolver

  Note:
  Must be positive. The default value is 0.001.

  Level: advanced

.seealso: `MatPartitioningType`, `MatPartitioning`, `MATPARTITIONINGCHACO`, `MatPartitioningChacoSetEigenSolver()`, `MatPartitioningChacoGetEigenTol()`
@*/
PetscErrorCode MatPartitioningChacoSetEigenTol(MatPartitioning part, PetscReal tol)
{
  PetscFunctionBegin;
  PetscValidHeaderSpecific(part, MAT_PARTITIONING_CLASSID, 1);
  PetscValidLogicalCollectiveReal(part, tol, 2);
  PetscTryMethod(part, "MatPartitioningChacoSetEigenTol_C", (MatPartitioning, PetscReal), (part, tol));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningChacoSetEigenTol_Chaco(MatPartitioning part, PetscReal tol)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  if (tol == PETSC_DEFAULT) chaco->eigtol = 0.001;
  else {
    PetscCheck(tol > 0.0, PetscObjectComm((PetscObject)part), PETSC_ERR_ARG_OUTOFRANGE, "Tolerance must be positive");
    chaco->eigtol = tol;
  }
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*@
  MatPartitioningChacoGetEigenTol - Gets the eigensolver tolerance used by Chaco

  Not Collective

  Input Parameter:
. part - the partitioning context

  Output Parameter:
. tol - the tolerance

  Level: advanced

.seealso: `MatPartitioningType`, `MatPartitioning`, `MATPARTITIONINGCHACO`, `MatPartitioningChacoSetEigenTol()`
@*/
PetscErrorCode MatPartitioningChacoGetEigenTol(MatPartitioning part, PetscReal *tol)
{
  PetscFunctionBegin;
  PetscValidHeaderSpecific(part, MAT_PARTITIONING_CLASSID, 1);
  PetscAssertPointer(tol, 2);
  PetscUseMethod(part, "MatPartitioningChacoGetEigenTol_C", (MatPartitioning, PetscReal *), (part, tol));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningChacoGetEigenTol_Chaco(MatPartitioning part, PetscReal *tol)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  *tol = chaco->eigtol;
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*@
  MatPartitioningChacoSetEigenNumber - Sets the number of eigenvectors to compute by Chaco during partitioning
  during partitioning.

  Collective

  Input Parameters:
+ part - the partitioning context
- num  - the number of eigenvectors

  Options Database Key:
. -mat_partitioning_chaco_eigen_number <n> - Number of eigenvectors

  Note:
  Accepted values are 1, 2 or 3, indicating partitioning by bisection,
  quadrisection, or octosection.

  Level: advanced

.seealso: `MatPartitioningType`, `MatPartitioning`, `MATPARTITIONINGCHACO`, `MatPartitioningChacoSetEigenSolver()`, `MatPartitioningChacoGetEigenTol()`
@*/
PetscErrorCode MatPartitioningChacoSetEigenNumber(MatPartitioning part, PetscInt num)
{
  PetscFunctionBegin;
  PetscValidHeaderSpecific(part, MAT_PARTITIONING_CLASSID, 1);
  PetscValidLogicalCollectiveInt(part, num, 2);
  PetscTryMethod(part, "MatPartitioningChacoSetEigenNumber_C", (MatPartitioning, PetscInt), (part, num));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningChacoSetEigenNumber_Chaco(MatPartitioning part, PetscInt num)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  if (num == PETSC_DEFAULT) chaco->eignum = 1;
  else {
    PetscCheck(num >= 1 && num <= 3, PetscObjectComm((PetscObject)part), PETSC_ERR_ARG_OUTOFRANGE, "Can only specify 1, 2 or 3 eigenvectors");
    chaco->eignum = num;
  }
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*@
  MatPartitioningChacoGetEigenNumber - Gets the number of eigenvectors used by Chaco.

  Not Collective

  Input Parameter:
. part - the partitioning context

  Output Parameter:
. num - number of eigenvectors

  Level: advanced

.seealso: `MatPartitioningType`, `MatPartitioning`, `MATPARTITIONINGCHACO`, `MatPartitioningChacoSetEigenNumber()`
@*/
PetscErrorCode MatPartitioningChacoGetEigenNumber(MatPartitioning part, PetscInt *num)
{
  PetscFunctionBegin;
  PetscValidHeaderSpecific(part, MAT_PARTITIONING_CLASSID, 1);
  PetscAssertPointer(num, 2);
  PetscUseMethod(part, "MatPartitioningChacoGetEigenNumber_C", (MatPartitioning, PetscInt *), (part, num));
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningChacoGetEigenNumber_Chaco(MatPartitioning part, PetscInt *num)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  *num = chaco->eignum;
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningSetFromOptions_Chaco(MatPartitioning part, PetscOptionItems *PetscOptionsObject)
{
  PetscInt               i;
  PetscReal              r;
  PetscBool              flag;
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;
  MPChacoGlobalType      global;
  MPChacoLocalType       local;
  MPChacoEigenType       eigen;

  PetscFunctionBegin;
  PetscOptionsHeadBegin(PetscOptionsObject, "Chaco partitioning options");
  PetscCall(PetscOptionsEnum("-mat_partitioning_chaco_global", "Global method", "MatPartitioningChacoSetGlobal", MPChacoGlobalTypes, (PetscEnum)chaco->global_method, (PetscEnum *)&global, &flag));
  if (flag) PetscCall(MatPartitioningChacoSetGlobal(part, global));
  PetscCall(PetscOptionsEnum("-mat_partitioning_chaco_local", "Local method", "MatPartitioningChacoSetLocal", MPChacoLocalTypes, (PetscEnum)chaco->local_method, (PetscEnum *)&local, &flag));
  if (flag) PetscCall(MatPartitioningChacoSetLocal(part, local));
  PetscCall(PetscOptionsReal("-mat_partitioning_chaco_coarse", "Coarse level", "MatPartitioningChacoSetCoarseLevel", 0.0, &r, &flag));
  if (flag) PetscCall(MatPartitioningChacoSetCoarseLevel(part, r));
  PetscCall(PetscOptionsEnum("-mat_partitioning_chaco_eigen_solver", "Eigensolver method", "MatPartitioningChacoSetEigenSolver", MPChacoEigenTypes, (PetscEnum)chaco->eigen_method, (PetscEnum *)&eigen, &flag));
  if (flag) PetscCall(MatPartitioningChacoSetEigenSolver(part, eigen));
  PetscCall(PetscOptionsReal("-mat_partitioning_chaco_eigen_tol", "Eigensolver tolerance", "MatPartitioningChacoSetEigenTol", chaco->eigtol, &r, &flag));
  if (flag) PetscCall(MatPartitioningChacoSetEigenTol(part, r));
  PetscCall(PetscOptionsInt("-mat_partitioning_chaco_eigen_number", "Number of eigenvectors: 1, 2, or 3 (bi-, quadri-, or octosection)", "MatPartitioningChacoSetEigenNumber", chaco->eignum, &i, &flag));
  if (flag) PetscCall(MatPartitioningChacoSetEigenNumber(part, i));
  PetscCall(PetscOptionsBool("-mat_partitioning_chaco_verbose", "Show library output", "", chaco->verbose, &chaco->verbose, NULL));
  PetscOptionsHeadEnd();
  PetscFunctionReturn(PETSC_SUCCESS);
}

static PetscErrorCode MatPartitioningDestroy_Chaco(MatPartitioning part)
{
  MatPartitioning_Chaco *chaco = (MatPartitioning_Chaco *)part->data;

  PetscFunctionBegin;
  PetscCall(PetscFree(chaco));
  /* clear composed functions */
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetGlobal_C", NULL));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoGetGlobal_C", NULL));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetLocal_C", NULL));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoGetLocal_C", NULL));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetCoarseLevel_C", NULL));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetEigenSolver_C", NULL));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoGetEigenSolver_C", NULL));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetEigenTol_C", NULL));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoGetEigenTol_C", NULL));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetEigenNumber_C", NULL));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoGetEigenNumber_C", NULL));
  PetscFunctionReturn(PETSC_SUCCESS);
}

/*MC
   MATPARTITIONINGCHACO - Creates a partitioning context that uses the external package Chaco.

   Level: beginner

   Note:
   Does not use the `MatPartitioningSetUseEdgeWeights()` option

   References:
.   * -  http://www.cs.sandia.gov/CRF/chac.html

.seealso: `MatPartitioningSetType()`, `MatPartitioningType`
M*/

PETSC_EXTERN PetscErrorCode MatPartitioningCreate_Chaco(MatPartitioning part)
{
  MatPartitioning_Chaco *chaco;

  PetscFunctionBegin;
  PetscCall(PetscNew(&chaco));
  part->data = (void *)chaco;

  chaco->global_method = MP_CHACO_MULTILEVEL;
  chaco->local_method  = MP_CHACO_KERNIGHAN;
  chaco->eigen_method  = MP_CHACO_LANCZOS;
  chaco->nbvtxcoarsed  = 200;
  chaco->eignum        = 1;
  chaco->eigtol        = 0.001;
  chaco->verbose       = PETSC_FALSE;

  part->ops->apply          = MatPartitioningApply_Chaco;
  part->ops->view           = MatPartitioningView_Chaco;
  part->ops->destroy        = MatPartitioningDestroy_Chaco;
  part->ops->setfromoptions = MatPartitioningSetFromOptions_Chaco;

  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetGlobal_C", MatPartitioningChacoSetGlobal_Chaco));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoGetGlobal_C", MatPartitioningChacoGetGlobal_Chaco));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetLocal_C", MatPartitioningChacoSetLocal_Chaco));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoGetLocal_C", MatPartitioningChacoGetLocal_Chaco));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetCoarseLevel_C", MatPartitioningChacoSetCoarseLevel_Chaco));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetEigenSolver_C", MatPartitioningChacoSetEigenSolver_Chaco));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoGetEigenSolver_C", MatPartitioningChacoGetEigenSolver_Chaco));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetEigenTol_C", MatPartitioningChacoSetEigenTol_Chaco));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoGetEigenTol_C", MatPartitioningChacoGetEigenTol_Chaco));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoSetEigenNumber_C", MatPartitioningChacoSetEigenNumber_Chaco));
  PetscCall(PetscObjectComposeFunction((PetscObject)part, "MatPartitioningChacoGetEigenNumber_C", MatPartitioningChacoGetEigenNumber_Chaco));
  PetscFunctionReturn(PETSC_SUCCESS);
}