shaq: shaq

In kazaam: Tools for Tall Distributed Matrices

Description

Constructor for shaq objects.

Usage

shaq(Data, nrows, ncols, checks = TRUE)

## S3 method for class 'matrix'
shaq(Data, nrows, ncols, checks = TRUE)

## S3 method for class 'numeric'
shaq(Data, nrows, ncols, checks = TRUE)

Arguments

Data	The local submatrix.
nrows, ncols	The GLOBAL number of rows and columns.
checks	Logical. Should some basic dimension checks be performed? Note that these require communication, and with many MPI ranks, could be expensive.

Details

If nrows and/or ncols is missing, then it will be imputed. This means one must be especially careful to manually provide ncols if some of ranks have "placeholder data" (a 0x0 matrix), which is typical when reading from a subset of processors and then broadcasting out to the remainder.

Communication

If checks=TRUE, a check on the global number of rows is performed. This amounts to an allgather operation on a logical value (the local dimension check).

See Also

shaq-class

kazaam documentation built on May 2, 2019, 8:55 a.m.