METIS is a library for partitioning and ordering matrices and graphs. It is used by Solver to reorder the original matrix to reduce fill-ins in the factored matrix.
ParMETIS is an MPI-based parallel library that implements a variety of algorithms for partitioning unstructured graphs and for computing fill-reducing orderings of sparse matrices.
The ParMETIS in MathKeisan is based on the original version 3.1 developed at University of Minnesota and Army HPC research center by George Karypis and Vipin Kumar.
User interface information is available from several sources:
| Name | Description |
|---|---|
METIS_PartGraphRecursive
| Partition mesh using multilevel recursive bisection. |
METIS_PartGraphKway
| Partition mesh using multilevel k-way partitioning to minimize edge cut. |
METIS_PartGraphVKway
| Partition mesh using multilevel k-way partitioning to minimize communication volume. |
METIS_mCPartGraphRecursive
| Partition mesh using multi-constraint multilevel recursive bisection to minimize edge cut. |
METIS_mCPartGraphKway
| Partition mesh using multi-constrained multilevel k-way partitioning algorithm to minimize edge-cut. |
METIS_WPartGraphRecursive
| PartGraphRecursive with prescribed partition weights.
|
METIS_WPartGraphKway
| PartGraphKway with prescribed partition weights.
|
METIS_WPartGraphVKway
| PartGraphVKway with prescribed partition weights.
|
METIS_PartMeshNodal
| Converts mesh to nodal graph, then uses METIS_PartGraphKway.
|
METIS_PartMeshDual
| Converts mesh to dual graph, then uses METIS_PartGraphKway.
|
METIS_EdgeND
| Computes fill reducing ordering of sparse matrices using multilevel nested dissection algorithm. Use only for large matrices from 3-d finite elements. |
METIS_NodeND
| Computes fill reducing ordering of sparse matrices using multilevel nested dissection algorithm. Generally preferred over METIS_EdgeND.
|
METIS_NodeWND
| Similar to METIS_NodeWND, but assumes compression performed prior to calling this routine.
|
METIS_MeshToNodal
| Converts mesh to nodal graph. |
METIS_MeshToDual
| Converts mesh to dual graph. |
METIS_EstimateMemory
| Estimates amount of memory that will be used by Metis. |
| Name | Description |
|---|---|
ParMETIS_PartKway
| Compute p-way partition of graph on p processors using multilevel k-way partitioning algorithm |
ParMETIS_PartGeomKway
| Compute p-way partition of graph on p processors combining coordinate-based partitioning and multilevel k-way partitioning algorithm. |
ParMETIS_PartGeom
| Compute p-way partition of graph on p processors using coordinate-based space-filling curves. |
ParMETIS_RepartLDiffusion
| Balance work load of graph corresponding to mesh that has been adaptively refined. Utilizes a diffusive scheme based on local load-imbalance information. |
ParMETIS_RepartGDiffusion
| Balance work load of graph corresponding to mesh that has been adaptively refined. Utilizes a diffusive scheme based on global load-imbalance information. |
ParMETIS_RepartRemap
| Balance work load of graph corresponding to mesh that has been adaptively refined. Based on new p-way partitioning and intelligently mapping it to the original partitioning so the migration cost is minimized. |
ParMETIS_RepartMLRemap
| Balance work load of graph corresponding to mesh that has been adaptively refined. Based on new p-way partitioning of coarsest graph, intelligently mapping it to original partitioning so migration cost is minimized, then refining using multilevel paradigm. |
ParMETIS_RefineKway
| Improve quality of existing p-way partition on p processors using multilevel p-way refinement algorithm. |
ParMETIS_NodeND
| Compute fill-reducing ordering of matrix using multilevel nested dissection. |