Skip to content

Synchronize before indexing host-accessible arrays from the CPU #649

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 1 commit into from
Sep 30, 2026
+14 −1
Merged

Synchronize before indexing host-accessible arrays from the CPU #649

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
7 changes: 6 additions & 1 deletion src/array.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -370,14 +370,19 @@ end

## indexing

# Host-accessible arrays can be indexed from CPU, bypassing GPUArrays restrictions
# Host-accessible arrays can be indexed from CPU, bypassing GPUArrays restrictions.
# Wait for queued work first, e.g., the kernel computing the result of a reduction. This
# only synchronizes the current task's stream; work submitted by other tasks, or to an
# explicitly created queue, needs to be synchronized explicitly.
function Base.getindex(x::oneArray{<:Any, <:Any, <:Union{oneL0.HostBuffer, oneL0.SharedBuffer}}, I::Int)
@boundscheck checkbounds(x, I)
synchronize(global_stream(context(x), device()))
return unsafe_load(pointer(x, I; type = oneL0.HostBuffer))
end

function Base.setindex!(x::oneArray{<:Any, <:Any, <:Union{oneL0.HostBuffer, oneL0.SharedBuffer}}, v, I::Int)
@boundscheck checkbounds(x, I)
synchronize(global_stream(context(x), device()))
return unsafe_store!(pointer(x, I; type = oneL0.HostBuffer), v)
end

Expand Down
8 changes: 8 additions & 0 deletions test/array.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -103,6 +103,14 @@ end
@test b == [100, 200]
end

@testset "reductions of host-accessible arrays" begin
for B in (oneL0.SharedBuffer, oneL0.HostBuffer)
a = oneArray{Float32, 1, B}(fill(1.0f0, 1024))
@test sum(a) == 1024
@test maximum(a) == 1
end
end

# https://github.com/JuliaGPU/CUDA.jl/issues/2191
@testset "preserving buffer types" begin
a = oneVector{Int,oneL0.SharedBuffer}([1])
Expand Down
Loading