Skip to content

broadcast: split ComposedFunction on AbstractGPUArrayStyle #719

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
maleadt merged 2 commits into from
May 14, 2026
Merged

broadcast: split ComposedFunction on AbstractGPUArrayStyle #719

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
2ac41d2
broadcast: split ComposedFunction on AbstractGPUArrayStyle
ChrisRackauckas May 6, 2026
e7be961
minimize comment.
maleadt May 14, 2026
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
5 changes: 5 additions & 0 deletions src/host/broadcast.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,11 @@ using Base.Broadcast

using Base.Broadcast: BroadcastStyle, Broadcasted, AbstractArrayStyle, instantiate

# split `ComposedFunction` so the kernel closure doesn't hit its kwarg-accepting
# call, whose kwsorter dispatch GPUCompiler can't resolve statically.
@inline Broadcast.broadcasted(S::AbstractGPUArrayStyle, c::ComposedFunction, args...) =
Broadcast.broadcasted(S, c.outer, Broadcast.broadcasted(S, c.inner, args...))

# but make sure we don't dispatch to the optimized copy method that directly indexes
function Broadcast.copy(bc::Broadcasted{<:AbstractGPUArrayStyle{0}})
ElType = Broadcast.combine_eltypes(bc.f, bc.args)
Expand Down
19 changes: 19 additions & 0 deletions test/testsuite/broadcasting.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
broadcasting(AT, eltypes)
vec3(AT, eltypes)
unknown_wrapper(AT, eltypes)
composed_function(AT, eltypes)
end

test_idx(idx, A::AbstractArray{T}) where T = A[idx] * T(2)
Expand Down Expand Up @@ -228,3 +229,21 @@ function unknown_wrapper(AT, eltypes)
end
end
end

function composed_function(AT, eltypes)
sq(x) = x*x
for ET in eltypes
@testset "ComposedFunction $ET" begin
a = AT(rand(ET, 8))
b = AT(rand(ET, 8))
ca, cb = Array(a), Array(b)

@test Array(broadcast(sq ∘ (+), a, b)) ≈ (ca .+ cb).^2
@test Array((sq ∘ (+)).(a, b)) ≈ (ca .+ cb).^2
@test Array((sq ∘ sq ∘ (+)).(a, b)) ≈ ((ca .+ cb).^2).^2
@test Array((sq ∘ identity).(a)) ≈ ca.^2
@test Array((sq ∘ (+)).(a, Ref(ET(2)))) ≈ (ca .+ ET(2)).^2
@test Array((identity ∘ (-)).(a, b)) ≈ ca .- cb
end
end
end