Skip to content

make default worker pool an AbstractWorkerPool #49101

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Apr 6, 2023
Merged

make default worker pool an AbstractWorkerPool #49101

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
aef75fd
make default worker pool an AbstractWorkerPool
tanmaykm Mar 22, 2023
111f991
Merge branch 'master' into tan/par
tanmaykm Apr 6, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions stdlib/Distributed/src/pmap.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@ struct BatchProcessingError <: Exception
end

"""
pgenerate([::WorkerPool], f, c...) -> iterator
pgenerate([::AbstractWorkerPool], f, c...) -> iterator

Apply `f` to each element of `c` in parallel using available workers and tasks.

Expand All @@ -18,14 +18,14 @@ Note that `f` must be made available to all worker processes; see
[Code Availability and Loading Packages](@ref code-availability)
for details.
"""
function pgenerate(p::WorkerPool, f, c)
function pgenerate(p::AbstractWorkerPool, f, c)
if length(p) == 0
return AsyncGenerator(f, c; ntasks=()->nworkers(p))
end
batches = batchsplit(c, min_batch_count = length(p) * 3)
return Iterators.flatten(AsyncGenerator(remote(p, b -> asyncmap(f, b)), batches))
end
pgenerate(p::WorkerPool, f, c1, c...) = pgenerate(p, a->f(a...), zip(c1, c...))
pgenerate(p::AbstractWorkerPool, f, c1, c...) = pgenerate(p, a->f(a...), zip(c1, c...))
pgenerate(f, c) = pgenerate(default_worker_pool(), f, c)
pgenerate(f, c1, c...) = pgenerate(a->f(a...), zip(c1, c...))

Expand Down
15 changes: 13 additions & 2 deletions stdlib/Distributed/src/workerpool.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -239,12 +239,14 @@ perform a `remote_do` on it.
"""
remote_do(f, pool::AbstractWorkerPool, args...; kwargs...) = remotecall_pool(remote_do, f, pool, args...; kwargs...)

const _default_worker_pool = Ref{Union{WorkerPool, Nothing}}(nothing)
const _default_worker_pool = Ref{Union{AbstractWorkerPool, Nothing}}(nothing)

"""
default_worker_pool()

[`WorkerPool`](@ref) containing idle [`workers`](@ref) - used by `remote(f)` and [`pmap`](@ref) (by default).
[`AbstractWorkerPool`](@ref) containing idle [`workers`](@ref) - used by `remote(f)` and [`pmap`](@ref)
(by default). Unless one is explicitly set via `default_worker_pool!(pool)`, the default worker pool is
initialized to a [`WorkerPool`](@ref).

# Examples
```julia-repl
Expand All @@ -267,6 +269,15 @@ function default_worker_pool()
return _default_worker_pool[]
end

"""
default_worker_pool!(pool::AbstractWorkerPool)

Set a [`AbstractWorkerPool`](@ref) to be used by `remote(f)` and [`pmap`](@ref) (by default).
"""
function default_worker_pool!(pool::AbstractWorkerPool)
_default_worker_pool[] = pool
end

"""
remote([p::AbstractWorkerPool], f) -> Function

Expand Down
13 changes: 13 additions & 0 deletions stdlib/Distributed/test/distributed_exec.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -701,6 +701,19 @@ wp = CachingPool(workers())
clear!(wp)
@test length(wp.map_obj2ref) == 0

# default_worker_pool! tests
wp_default = Distributed.default_worker_pool()
try
wp = CachingPool(workers())
Distributed.default_worker_pool!(wp)
@test [1:100...] == pmap(x->x, wp, 1:100)
@test !isempty(wp.map_obj2ref)
clear!(wp)
@test isempty(wp.map_obj2ref)
finally
Distributed.default_worker_pool!(wp_default)
end

# The below block of tests are usually run only on local development systems, since:
# - tests which print errors
# - addprocs tests are memory intensive
Expand Down