Skip to content

DArray/stencil: Allocate HaloArray outside spawn_datadeps #703

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

Open
jpsamaroo wants to merge 7 commits into
base: master
Choose a base branch
from
Open

DArray/stencil: Allocate HaloArray outside spawn_datadeps #703

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
fd5f4d2
DArray/stencil: Allocate HaloArray outside spawn_datadeps
jpsamaroo May 22, 2026
f595669
DArray/stencil: Reuse HaloArray allocations
jpsamaroo May 22, 2026
0af8688
test/stencil: Skip 4D stencil on GPUs
jpsamaroo May 23, 2026
7a1416b
DArray/stencil: Reduce memory allocations
jpsamaroo May 26, 2026
617cc01
DArray/stencil: Fixes to HaloArray allocate/fill ordering
jpsamaroo May 29, 2026
4ca299c
test/stencil: Use div, not /
jpsamaroo May 29, 2026
9af65c5
system_uuid: Make robust to multi-worker races
jpsamaroo May 30, 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
6 changes: 4 additions & 2 deletions ext/CUDAExt.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -370,11 +370,13 @@ CuArray(H::Dagger.HaloArray) = convert(CuArray, H)
Base.convert(::Type{C}, H::Dagger.HaloArray) where {C<:CuArray} =
Dagger.HaloArray(C(H.center),
C.(H.halos),
H.halo_width)
H.halo_width;
own_center=H.own_center)
Adapt.adapt_structure(to::CUDA.KernelAdaptor, H::Dagger.HaloArray) =
Dagger.HaloArray(adapt(to, H.center),
adapt.(Ref(to), H.halos),
H.halo_width)
H.halo_width;
own_center=H.own_center)
function Dagger.inner_stencil_proc!(::CuArrayDeviceProc, f, output, read_vars)
Dagger.Kernel(_inner_stencil!)(f, output, read_vars; ndrange=size(output))
return
Expand Down
6 changes: 4 additions & 2 deletions ext/IntelExt.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -322,11 +322,13 @@ oneArray(H::Dagger.HaloArray) = convert(oneArray, H)
Base.convert(::Type{C}, H::Dagger.HaloArray) where {C<:oneArray} =
Dagger.HaloArray(C(H.center),
C.(H.halos),
H.halo_width)
H.halo_width;
own_center=H.own_center)
Adapt.adapt_structure(to::oneAPI.KernelAdaptor, H::Dagger.HaloArray) =
Dagger.HaloArray(adapt(to, H.center),
adapt.(Ref(to), H.halos),
H.halo_width)
H.halo_width;
own_center=H.own_center)
function Dagger.inner_stencil_proc!(::oneArrayDeviceProc, f, output, read_vars)
Dagger.Kernel(_inner_stencil!)(f, output, read_vars; ndrange=size(output))
return
Expand Down
6 changes: 4 additions & 2 deletions ext/MetalExt.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -346,11 +346,13 @@ MtlArray(H::Dagger.HaloArray) = convert(MtlArray, H)
Base.convert(::Type{C}, H::Dagger.HaloArray) where {C<:MtlArray} =
Dagger.HaloArray(C(H.center),
C.(H.halos),
H.halo_width)
H.halo_width;
own_center=H.own_center)
Adapt.adapt_structure(to::Metal.Adaptor, H::Dagger.HaloArray) =
Dagger.HaloArray(adapt(to, H.center),
adapt.(Ref(to), H.halos),
H.halo_width)
H.halo_width;
own_center=H.own_center)
function Dagger.inner_stencil_proc!(::MtlArrayDeviceProc, f, output, read_vars)
Dagger.Kernel(_inner_stencil!)(f, output, read_vars; ndrange=size(output))
return
Expand Down
6 changes: 4 additions & 2 deletions ext/OpenCLExt.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -320,11 +320,13 @@ CLArray(H::Dagger.HaloArray) = convert(CLArray, H)
Base.convert(::Type{C}, H::Dagger.HaloArray) where {C<:CLArray} =
Dagger.HaloArray(C(H.center),
C.(H.halos),
H.halo_width)
H.halo_width;
own_center=H.own_center)
Adapt.adapt_structure(to::OpenCL.KernelAdaptor, H::Dagger.HaloArray) =
Dagger.HaloArray(adapt(to, H.center),
adapt.(Ref(to), H.halos),
H.halo_width)
H.halo_width;
own_center=H.own_center)
function Dagger.inner_stencil_proc!(::CLArrayDeviceProc, f, output, read_vars)
Dagger.Kernel(_inner_stencil!)(f, output, read_vars; ndrange=size(output))
return
Expand Down
6 changes: 4 additions & 2 deletions ext/ROCExt.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -343,11 +343,13 @@ ROCArray(H::Dagger.HaloArray) = convert(ROCArray, H)
Base.convert(::Type{C}, H::Dagger.HaloArray) where {C<:ROCArray} =
Dagger.HaloArray(C(H.center),
C.(H.halos),
H.halo_width)
H.halo_width;
own_center=H.own_center)
Adapt.adapt_structure(to::AMDGPU.Runtime.Adaptor, H::Dagger.HaloArray) =
Dagger.HaloArray(adapt(to, H.center),
adapt.(Ref(to), H.halos),
H.halo_width)
H.halo_width;
own_center=H.own_center)
function Dagger.inner_stencil_proc!(::ROCArrayDeviceProc, f, output, read_vars)
Dagger.Kernel(_inner_stencil!)(f, output, read_vars; ndrange=size(output))
return
Expand Down
7 changes: 6 additions & 1 deletion src/Dagger.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -162,7 +162,12 @@ import PrecompileTools: @compile_workload
include("precompile.jl")

function __init__()
# Initialize system UUID
# Clear any precompile-cached UUID for this process: the precompile workload
# runs system_uuid() and the resulting SYSTEM_UUIDS entry gets baked into
# the compiled image. Without clearing it here, get!() would return that
# stale build-time UUID instead of reading the actual runtime UUID file,
# causing mismatches between process 1 and workers.
delete!(SYSTEM_UUIDS, myid())
system_uuid()

@static if !isdefined(Base, :get_extension)
Expand Down
Loading
Loading