sparsevectormodule Module Reference

A module for handling compressed vectors. Compressed vectors are stored in two lists. The first is a list of indices, the second a list of values. This module can add two of those vectors together. More...

Functions/Subroutines
pure subroutine, public	addsparsevectors (inner_index_a, values_a, inner_index_b, values_b, inner_index_c, values_c, total_values_c, alpha_in, threshold_in)
	Add together two sparse vectors. C = A + alpha*B. More...
pure real(ntreal) function, public	dotsparsevectors (inner_index_a, values_a, inner_index_b, values_b)
	product = dot(A,B) More...
pure subroutine, public	pairwisemultiplyvectors (inner_index_a, values_a, inner_index_b, values_b, inner_index_c, values_c, total_values_c)
	Pairwise Multiply Vectors C = A Mult B. More...

Detailed Description

A module for handling compressed vectors. Compressed vectors are stored in two lists. The first is a list of indices, the second a list of values. This module can add two of those vectors together.

Function/Subroutine Documentation

addsparsevectors()

pure subroutine, public sparsevectormodule::addsparsevectors	(	integer, dimension(:), intent(in)	inner_index_a,
		real(ntreal), dimension(:), intent(in)	values_a,
		integer, dimension(:), intent(in)	inner_index_b,
		real(ntreal), dimension(:), intent(in)	values_b,
		integer, dimension(:), intent(out)	inner_index_c,
		real(ntreal), dimension(:), intent(out)	values_c,
		integer, intent(out)	total_values_c,
		real(ntreal), intent(in), optional	alpha_in,
		real(ntreal), intent(in), optional	threshold_in
	)

Add together two sparse vectors. C = A + alpha*B.

Parameters

[in]	inner_index_a	list of indices for A.
[in]	values_a	list of values for A.
[in]	inner_index_b	list of indices for B.
[in]	values_b	list of values for B.
[out]	inner_index_c	list of indices computed for C.
[out]	values_c	list of values computed for C.
[out]	total_values_c	this is the total number of values in C.
[in]	alpha_in	value to scale VecB by. Optional, default is 1.0.
[in]	threshold_in	for flushing values to zero. Default value is 0.0. The values that are returned for C are only valid in the range (1:total_values_c). We do not do an automatic shrinking of the array to keep this routine low in overhead.

Todo:

in principal this can be done without any branching.

dotsparsevectors()

pure real(ntreal) function, public sparsevectormodule::dotsparsevectors	(	integer, dimension(:), intent(in)	inner_index_a,
		real(ntreal), dimension(:), intent(in)	values_a,
		integer, dimension(:), intent(in)	inner_index_b,
		real(ntreal), dimension(:), intent(in)	values_b
	)

product = dot(A,B)

Parameters

[in]	inner_index_a	list of indices for A.
[in]	values_a	list of values for A.
[in]	inner_index_b	list of indices for B.
[in]	values_b	list of values for B.

pairwisemultiplyvectors()

pure subroutine, public sparsevectormodule::pairwisemultiplyvectors	(	integer, dimension(:), intent(in)	inner_index_a,
		real(ntreal), dimension(:), intent(in)	values_a,
		integer, dimension(:), intent(in)	inner_index_b,
		real(ntreal), dimension(:), intent(in)	values_b,
		integer, dimension(:), intent(out)	inner_index_c,
		real(ntreal), dimension(:), intent(out)	values_c,
		integer, intent(out)	total_values_c
	)

Pairwise Multiply Vectors C = A Mult B.

Parameters

[in]	inner_index_a	list of indices for A.
[in]	values_a	list of values for A.
[in]	inner_index_b	list of indices for B.
[in]	values_b	list of values for B.
[out]	inner_index_c	list of indices computed for C.
[out]	values_c	list of values computed for C.
[out]	total_values_c	this is the total number of values in C.