ML Version of the Day
Public Attributes
ML_Aggregate_Struct Struct Reference

#include <ml_aggregate.h>

Collaboration diagram for ML_Aggregate_Struct:
Collaboration graph
[legend]

List of all members.

Public Attributes

int ML_id
double print_flag
int max_coarse_size
int ordering
int min_nodes_per_aggregate
int max_neigh_already_selected
double threshold
double curr_threshold
double drop_tol_for_smoothing
int attach_scheme
double smoothP_damping_factor
int * smoothP_damping_sweeps
int smoothP_type
int coarsen_scheme
int * coarsen_scheme_level
int num_PDE_eqns
int nullspace_dim
double * nullspace_vect
int nullspace_corrupted
int * aggr_count
int keep_agg_information
int ** aggr_info
int max_levels
int begin_level
int cur_level
double operator_complexity
double fine_complexity
int nvblocks
int * vblock_info
int keep_P_tentative
struct ML_Operator_Struct ** P_tentative
int smooth_existing_P_tentative
int use_transpose
int Restriction_smoothagg_transpose
void * aggr_options
void * aggr_viz_and_stats
void * field_of_values
int block_scaled_SA
double phase3_agg_creation
double ** nodal_coord
int N_dimensions
void * vblock_data
int minimizing_energy
int cheap_minimizing_energy
double minimizing_energy_droptol
double * old_RowOmegas
int coarsen_rate
int semicoarsen_levels

Detailed Description

definition of the aggregate structure


Member Data Documentation

no. aggregates at each level

node to aggregate map

option about METIS and ParMETIS

information about the aggregates only if the user explicitely < requires them

aggregate shape control

finest grid level

= 1 indicates that the prolongator smoother should use block diagonal scaling (blocksize = num_PDE_eqns)

Uncoupled, Coupled, MIS

temporary variable

adjusted for levels

self-explanatory

nnz of the finest A

1: keep aggr_info otherwise: free it

keeping tentative prolongator

maximum size of coarsest grid

maximum number of levels

complexity control

aggregate size control

Coordinates of fine-grid nodes and aggregates

indicates whether fine grid nullspace has been overwritten

self-explanatory

for null space vectors

block size

for variable blocks (finest)

sum of nnz for all A's

natural, random, graph

so it can be reused later.

Steers how the MIS and Uncoupled handle phase 3 of aggregation. Values near 0 create few additional aggregates.Large values create many additional aggregates. Convergence can be improve convergence by new aggregates but nonzero fill-in increases on coarse meshes. Default: .5

smoothed aggregation on A^T

already have P tent, don't create it

for prolongator smoother

#prolongator smoother sweeps

point, block

for pruning matrix

Used to build restriction by doing

holds data structure aggr_vblock

for variable blocks (finest)


The documentation for this struct was generated from the following file:
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Friends