Examples

Reduction

We may compute the elementwise sum of arrays on each processor as

using MPIMapReduce
using MPI
MPI.Init()

y1 = pmapreduce(x -> ones(2) * x, +, 1:5)

if MPI.Comm_rank(MPI.COMM_WORLD) == 0
    show(stdout, MIME"text/plain"(), y1)
    println()
end

This produces the output

2-element Array{Float64,1}:
 15.0
 15.0

We may compute the elementwise product of arrays on each processor as

using MPIMapReduce
using MPI
MPI.Init()

y1 = pmapreduce(x -> ones(2, 2) * x, *, 1:4)

if MPI.Comm_rank(MPI.COMM_WORLD) == 0
    show(stdout, MIME"text/plain"(), y1)
    println()
end

This produces the output

2×2 Array{Float64,2}:
 24.0  24.0
 24.0  24.0

Concatenation

The function pmapgatherv performs concatenates the values returned by the map. The operators vcat and hcat may be provided as the reduction operator to carry out concatenations along the first or the second axis respectively.

For example, we may concatenate Vectors along the second dimension as

using MPIMapReduce
using MPI
MPI.Init()

y = pmapgatherv(x -> ones(2) * x^2, hcat, 1:5)

if MPI.Comm_rank(MPI.COMM_WORLD) == 0
    show(stdout, MIME"text/plain"(), y)
    println()
end

This leads to the output

2×5 Array{Float64,2}:
 1.0  4.0  9.0  16.0  25.0
 1.0  4.0  9.0  16.0  25.0

For more general concatenations, use the reduction operator Cat(dims), where dims refers to the dimensions along which the concatenation is to be carried out. For example, we may concatenate numbers along the 1st and 2nd dimensions to generate a diagonal matrix as

using MPIMapReduce
using MPI
MPI.Init()

y = pmapgatherv(x -> x^2, Cat([1,2]), 1:3)

comm = MPI.COMM_WORLD

if MPI.Comm_rank(comm) == 0
    show(stdout, MIME"text/plain"(), y)
    println()
end

which leads to the output

3×3 Array{Int64,2}:
 1  0  0
 0  4  0
 0  0  9