Single- and multivector view creation.
[Public Interface]

Detailed Description

Vector views are wrappers around user-declared dense array matrix representations.

These wrappers encapsulate information about the stride (leading dimension) and data layout of the matrix in its corresponding array. The following routines create, destroy, and duplicate vector views.


Defines

#define STRIDE_UNIT   ((oski_index_t)1)
 Symbolic constant for "unit stride".

Typedefs

typedef oski_vecstruct_t * oski_vecview_t
 Multivector view handle.

Functions

oski_vecview_t oski_CreateMultiVecView (oski_value_t *X, oski_index_t length, oski_index_t num_vecs, oski_storage_t orient, oski_index_t stride)
 Creates a multivector view on $k$ dense column vectors $X = \left(\begin{array}{c} x_1 \cdots x_k \end{array}\right)$, stored as a submatrix in the user's data.

Function Documentation

oski_vecview_t oski_CreateMultiVecView (  oski_value_t *  X,
oski_index_t  length,
oski_index_t  num_vecs,
oski_storage_t  orient,
oski_index_t  stride
) 
 

Creates a multivector view on $k$ dense column vectors $X = \left(\begin{array}{c} x_1 \cdots x_k \end{array}\right)$, stored as a submatrix in the user's data.

Parameters:
[in] X Multivector (matrix) data.
[in] length Logical length of each (column) vector, i.e., number of logical rows in the matrix.
[in] num_vecs Number of vectors, i.e., number of logical columns in the matrix.
[in] orient Row vs. column major layout.
[in] stride Leading dimension, i.e., distance between vectors. stride must be >= length for column-major storage, or >= num_vecs for row-major storage.
Returns:
A multivector view object (handle), or INVALID_VEC on error.

Generated on Wed Sep 19 16:41:24 2007 for BeBOP Optimized Sparse Kernel Interface Library by  doxygen 1.4.6