CCES Unicamp

ChunkSplitters.jl: easy split the elements or indices of a collection into chunks.

https://juliafolds2.github.io/ChunkSplitters.jl/stable/#ChunkSplitters.jl

ChunkSplitters.jl makes it easy to split the elements or indices of a collection into chunks:

julia> using ChunkSplitters

julia> x = [1.2, 3.4, 5.6, 7.8, 9.1, 10.11, 11.12];

julia> for inds in index_chunks(x; n=3)
           @show inds
       end
inds = 1:3
inds = 4:5
inds = 6:7

julia> for c in chunks(x; n=3)
           @show c
       end
c = [1.2, 3.4, 5.6]
c = [7.8, 9.1]
c = [10.11, 11.12]

This can be useful in many areas, one of which is multithreading, where we can use chunking to control the number of spawned tasks:

function parallel_sum(x; ntasks=nthreads())
    tasks = map(chunks(x; n=ntasks)) do chunk_of_x
        @spawn sum(chunk_of_x)
    end
    return sum(fetch, tasks)
end

Working with chunks and their respective indices also improves thread-safety compared to a naive parallelisation approach based on threadid()

 

Related posts

Converting scripts into reproducible Workflow Research Objects

escience

Workshop “Actor Databases and Data-driven Scientific Applications”

escience

On the Accuracy of the Direct Method to Calculate pKa from Electronic Structure Calculations

cces cces