Mapping and Ordering

The Mapping class

class Mapping(*args)

The Mapping constructor calls libscotch’s mapAlloc function. Any argument is conveyed to the init() method. If there are no arguments, the init() method is not called.

first being the argument given to the init() method, the following func methods all proxy to first.map_func(self, *args), except for re_compute() and re_fixed_compute() which proxy the remap_compute and remap_fixed_compute methods, respectively. For now first is always a :class: Graph instance, can be used to factorize Mesh routines later on. We will define the methods in the way in which they are called, using a graph :class: Graph instance as the calling object :


init(first, mapp, arch, parttab)

Calls first.map_init(first, mapp, arch, parttab), self being the Mapping instance. first.map_init(first, mapp, arch, parttab) can also be called directly.

Parameters:

  • first (Graph) – The first argument is the object to be mapped.

  • mapp (Mapping) – The mapping to be initialized.

  • arch (Arch) – The architecture to be used for mapping.

  • parttab (A numpy array) – The partition table.


graph.map_exit(mapp)

This routine cleans-up all fields of the instance. Is the opposite of the init() routine. Called automatically when the object is destroyed. Calling explicitly this method is only required when an init() call follows.

Parameters:

mapp (A Mapping) – The mapping to be freed.


graph.map_compute(mapp, strat)

This routine computes a mapping of the API mapping structure mapping with respect to the given strategy strat .

Parameters:

  • mapping (Mapping) – The mapping to be computed.

  • strat (Strat) – Strategy to be used for the mapping .


graph.map_fixed_compute(mapp, start)

This routine computes a mapping with fixed vertices of the API mapping structure with respect to the given strategy.

Parameters:

  • mapping (Mapping) – The mapping to be computed.

  • strat (Strat) – Strategy to be used for the mapping.


graph.remap_compute(mapp, mapo, emrval, vmlotab, start)

This routine computes a remapping of the API mapping structure mapping with respect to the given strategy strat.

Parameters:

  • mapping (Mapping) – The mapping to be computed.

  • mapo (Any kind of iterable or a numpy array) – The already computed mapping array used to account for migration costs.

  • emraval (Floating-point number) – Individual migration cost.

  • vmlotab (Any kind of iterable or a numpy array) – Vertex migration cost array.

  • strat (Strat) – Strategy to be used for the mapping.


graph.remap_fixed_compute(mapp, mapo, emrval, vmlotab, start)

This routine computes a remapping with fixed vertices of the API mapping structure mapping with respect to the given strategy strat.

Parameters:

  • mapping (Mapping) – The mapping to be computed.

  • mapo (Any kind of iterable or a numpy array) – The already computed mapping array used to account for migration costs.

  • emraval (Floating-point number) – Individual migration cost.

  • vmlotab (Any kind of iterable or a numpy array) – Vertex migration cost array.

  • strat (Strat) – Strategy to be used for the mapping.


graph.map_save(stream)

This routine saves the contents of the given user mapping mapping to the given stream stream.

Parameters:

  • mapping (Mapping) – The mapping to be saved.

  • stream (Either a file object (result of an open(), don’t forget to close it) or a filename, as a string or as a bytes object (such as b"file/name")) – Input file or stream to write to.


graph.map_load(stream)

This routine loads the contents of the given user mapping mapping from the given stream stream.

Parameters:

  • mapping (Mapping) – The mapping to be loaded.

  • stream (Either a file object (result of an open(), don’t forget to close it) or a filename, as a string or as a bytes object (such as b"file/name")) – Input file or stream to read from.


graph.map_view(*args)

This routine writes mapping statistics (load of target processors, number of neigh- boring domains, average dilation and expansion, edge cut size, distribution of edge dilations) to the given stream.

Parameters:

  • mapping (Mapping) – The mapping to be viewed.

  • stream (Either a file object (result of an open(), don’t forget to close it) or a filename, as a string or as a bytes object (such as b"file/name")) – Input file or stream to write to.


map_alloc()

Recreates the mapAlloc function, returning a non-initialized Mapping instance. It will need the init() method to be called before the instance be built or loaded upon.

The Ordering class

class Ordering(*args)

The Ordering constructor calls libscotch’s orderAlloc function. Any argument is conveyed to the init() method. If there are no arguments, the init() method is not called.

first being the argument given to the init() method, the following func methods all proxy to first.map_func(self, *args). For now first is always a :class: Graph instance, can be used to factorize Mesh routines later on. We will define the methods in the way in which they are called, using a graph :class: Graph instance as the calling object :

init(first, ordering, permtab=None, peritab=None, cblknbr=None, rangtab=None, treetab=None)

Calls first.map_init(ordering, permtab=None, peritab=None, cblknbr=None, rangtab=None, treetab=None), self being the Mapping instance. first.map_init(ordering, permtab=None, peritab=None, cblknbr=None, rangtab=None, treetab=None) can also be called directly.

Parameters:

  • first (Graph) – The first argument is the object to be ordered.

  • ordering (Ordering) – Ordering to be initialized.

  • permtab (Any kind of iterable or a numpy array) – Array holding the permutation of the reordered matrix. If given, the values of the returned array are copied in this one. (Optional)

  • peritab (Any kind of iterable or a numpy array) – Inverse permutation of the reordered matrix. If given, the values of the returned array are copied in this one. (Optional)

  • cblknbr (Integer) – Number of column blocks in the block ordering. If given, the value is copied in this one. (Optional)

  • rangtab (Any kind of iterable or a numpy array) – Array of ranges for the column blocks. If given, the values of the returned array are copied in this one. (Optional)

  • treetab (Any kind of iterable or a numpy array) – Array of ascendants of permuted column blocks in the separators tree. If given, the values of the returned array are copied in this one. (Optional)


graph.order_exit(ordering)

This routine cleans-up all fields of the instance. Is the opposite of the init() routine. Called automatically when the object is destroyed. Calling explicitly this method is only required when an init() call follows.

Parameters:

ordering (Ordering) – The ordering to be freed.


graph.order_check(ordering)

This routine checks the consistency of the ordering instance.

Parameters:

ordering (Ordering) – Ordering to be checked.


graph.order_compute(ordering, strat)

This routine computes a block ordering of the graph itself with respect to the given strategy strat. Stores the result in the given ordering instance ordering.

Parameters:

  • ordering (Ordering) – The ordering to be computed.

  • strat (Strat) – Strategy to be used for the ordering.


graph.order_compute_list(ordering, listtab, strat)

This routine computes a block ordering of the graph itself with respect to the given strategy strat. Stores the result in the given ordering instance ordering. The induced subgraph is described by means of a vertex list: listnbr holds the number of vertices to keep in the induced subgraph, the indices of which are given, in any order, in the listtab array.

Parameters:

  • ordering (Ordering) – Ordering to be computed.

  • listtab (Any kind of iterable or a numpy array) – The list of vertices to keep in the induced subgraph.

  • strat (Strat) – Strategy to be used for the ordering.


graph.order_load(ordering, stream)

This routine loads the contents of the given ordering instance ordering from the given stream stream.

Parameters:

  • ordering (Ordering) – Ordering to be loaded.

  • stream (Either a file object (result of an open(), don’t forget to close it) or a filename, as a string or as a bytes object (such as b"file/name")) – Input file or stream to read from.


graph.order_save(ordering, stream)

This routine saves the contents of the given ordering instance ordering to the given stream stream.

Parameters:

  • ordering (Ordering) – Ordering to be saved.

  • stream (Either a file object (result of an open(), don’t forget to close it) or a filename, as a string or as a bytes object (such as b"file/name")) – Input file or stream to write to.


graph.order_save_map(ordering, stream)

This routine saves the contents of the given ordering instance ordering to the given stream stream in the Scotch mapping format.

Parameters:

  • ordering (Ordering) – Ordering to be saved.

  • stream (Either a file object (result of an open(), don’t forget to close it) or a filename, as a string or as a bytes object (such as b"file/name")) – Input file or stream to write to.


graph.order_save_tree(ordering, stream)

This routine saves the contents of the given ordering instance ordering to the given stream stream in the tree output format format.

Parameters:

  • ordering (Ordering) – Ordering to be saved.

  • stream (Either a file object (result of an open(), don’t forget to close it) or a filename, as a string or as a bytes object (such as b"file/name")) – Input file or stream to write to.


order_alloc()

Recreates the orderAlloc function, returning a non-initialized Ordering instance. It will need the init() method to be called before the instance be built or loaded upon.