pddistribute-aux3d.c File Reference

#include "superlu_ddefs.h"

Include dependency graph for pddistribute-aux3d.c:

Functions
void	dpropagate_A_to_LU3d (dLUstruct_t *LUstruct, int_t *xa, int_t *asub, double *a, superlu_dist_options_t *options, gridinfo3d_t *grid3d, int_t nsupers, float *mem_use)
int_t	computeLDAspa_Ilsum (int_t nsupers, int_t *ilsum, dLUstruct_t *LUstruct, gridinfo3d_t *grid3d)
void	propagateDataThroughMatrixBlocks (int_t nsupers, Glu_freeable_t *Glu_freeable, dLUstruct_t *LUstruct, gridinfo3d_t *grid3d, int_t *Urb_length, int_t *rb_marker, int_t *Urb_fstnz, int_t *Ucbs, int **ToSendR, int *ToSendD, int *ToRecv)
int_t	checkDist3DLUStruct (dLUstruct_t *LUstruct, gridinfo3d_t *grid3d)

Function Documentation

checkDist3DLUStruct()

int_t checkDist3DLUStruct	(	dLUstruct_t *	LUstruct,
		gridinfo3d_t *	grid3d
	)

Here is the call graph for this function:

computeLDAspa_Ilsum()

int_t computeLDAspa_Ilsum	(	int_t	nsupers,
		int_t *	ilsum,
		dLUstruct_t *	LUstruct,
		gridinfo3d_t *	grid3d
	)

Function: computeLDAspa_Ilsum

This function computes the Local Distributed Storages (LDS) for the L factor in the LU decomposition. It also updates the input array 'ilsum' to hold the prefix sum of the size of each supernode. Compute ldaspa and ilsum[].

dpropagate_A_to_LU3d()

void dpropagate_A_to_LU3d	(	dLUstruct_t *	LUstruct,
		int_t *	xa,
		int_t *	asub,
		double *	a,
		superlu_dist_options_t *	options,
		gridinfo3d_t *	grid3d,
		int_t	nsupers,
		float *	mem_use
	)

Propagates the new values of A into the existing L and U data structures.

Parameters

Llu	The local L and U data structures.
xa	The index array of A.
asub	The row subscripts of A.
a	The numerical values of A.
options	The options array.
grid	The process grid.
mem_use	The memory usage.

Returns

void

Here is the call graph for this function:

propagateDataThroughMatrixBlocks()

void propagateDataThroughMatrixBlocks	(	int_t	nsupers,
		Glu_freeable_t *	Glu_freeable,
		dLUstruct_t *	LUstruct,
		gridinfo3d_t *	grid3d,
		int_t *	Urb_length,
		int_t *	rb_marker,
		int_t *	Urb_fstnz,
		int_t *	Ucbs,
		int **	ToSendR,
		int *	ToSendD,
		int *	ToRecv
	)

The function propagate_blocks performs data propagation in the form of values through blocks of a supernodal sparse matrix. It sends values to particular blocks and marks blocks to receive new values depending on certain conditions. It also counts the number of non-zeros in each block row and keeps track of the number of column blocks in each block row.

Input: nsupers: Total number of supernodes, i.e., the number of block columns in the original matrix. grid: The process grid. This grid contains the mapping of the matrix blocks to the process grid. xusub: An array of indices, each pointing to the start of each column in the usub array. usub: An array containing the row indices of non-zero elements of the supernodal matrix. ToSendR: A 2D flag array indicating which blocks need to be sent to other processes, organized by column. ToSendD: A flag array indicating which blocks are to be sent to other processes. Urb_length: An array containing the number of non-zero elements in each block row of the matrix. rb_marker: A marker array to track the last block column visited in each block row. Urb_fstnz: An array containing the number of first non-zero elements in each block row of the matrix. Ucbs: An array counting the number of column blocks in each block row. ToRecv: A flag array indicating which blocks are to be received from other processes.

Output: ToSendR, ToSendD, Urb_length, rb_marker, Urb_fstnz, Ucbs, ToRecv are modified in place. Their new values depend on the conditions checked during the loops.

Returns

void