codistributor

Create codistributor object for codistributed arrays

Syntax

codist = codistributor()
codist = codistributor('1d')
codist = codistributor('1d',dim)
codist = codistributor('1d',dim,part)
codist = codistributor('2dbc')
codist = codistributor('2dbc',lbgrid)
codist = codistributor('2dbc',lbgrid,blksize)

Description

There are two schemes for distributing arrays. The scheme denoted by the character vector '1d' distributes an array along a single specified subscript, the distribution dimension, in a noncyclic, partitioned manner. The scheme denoted by '2dbc', employed by the parallel matrix computation software ScaLAPACK, applies only to two-dimensional arrays, and varies both subscripts over a rectangular computational grid of labs (workers) in a blocked, cyclic manner.

codist = codistributor(), with no arguments, returns a default codistributor object with zero-valued or empty parameters, which can then be used as an argument to other functions to indicate that the function is to create a codistributed array if possible with default distribution. For example, 

Z = zeros(..., codistributor())
R = randn(..., codistributor())

codist = codistributor('1d') is the same as codist = codistributor().

codist = codistributor('1d',dim) also forms a codistributor object with codist.Dimension = dim and default partition.

codist = codistributor('1d',dim,part) also forms a codistributor object with codist.Dimension = dim and codist.Partition = part.

codist = codistributor('2dbc') forms a 2-D block-cyclic codistributor object. For more information about '2dbc' distribution, see 2-Dimensional Distribution.

codist = codistributor('2dbc',lbgrid) forms a 2-D block-cyclic codistributor object with the lab grid defined by lbgrid and with default block size.

codist = codistributor('2dbc',lbgrid,blksize) forms a 2-D block-cyclic codistributor object with the lab grid defined by lbgrid and with a block size defined by blksize.

codist = getCodistributor(D) returns the codistributor object of codistributed array D.

Examples

On four workers, create a 3-dimensional, 2-by-6-by-4 array with distribution along the second dimension, and partition scheme [1 2 1 2]. In other words, worker 1 contains a 2-by-1-by-4 segment, worker 2 a 2-by-2-by-4 segment, etc.

spmd
    dim = 2; % distribution dimension
    codist = codistributor('1d',dim,[1 2 1 2],[2 6 4]);
    if mod(labindex,2)
        L = rand(2,1,4);
    else
        L = rand(2,2,4);
    end
    A = codistributed.build(L,codist)
end
A

On four workers, create a 20-by-5 codistributed array A, distributed by rows (over its first dimension) with a uniform partition scheme. 

spmd
    dim = 1; % distribution dimension
    partn = codistributor1d.defaultPartition(20);
    codist = codistributor('1d',dim,partn,[20 5]);
    L = magic(5) + labindex;
    A = codistributed.build(L,codist)
end
A

See Also

| | | | |

Introduced in R2008b

Was this topic helpful?
Parallel Computing Toolbox Documentation
Other Documentation
Support