Skip to content

Precompile common operations in the CUDA extension #196

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
Nov 5, 2026
Merged

Precompile common operations in the CUDA extension #196

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
2 changes: 2 additions & 0 deletions Project.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@ ArgCheck = "dce04be8-c92d-5529-be00-80e4d2c0e197"
Atomix = "a9b6321e-bd34-4604-b9c9-b65b8de01458"
GPUArraysCore = "46192b85-c4d5-4398-a991-12ede77f4527"
KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
PrecompileTools = "aea7be01-6a6a-4083-8856-8a6e6704d82a"

[weakdeps]
CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
Expand All @@ -33,5 +34,6 @@ GPUArraysCore = "0.2.1"
KernelAbstractions = "0.9.43, 0.10"
Metal = "1.10.1"
OpenCL = "0.10, 1"
PrecompileTools = "1"
julia = "1.10"
oneAPI = "1, 2"
28 changes: 28 additions & 0 deletions ext/CUDACoreExt.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@ module CUDACoreExt

using CUDACore
import AcceleratedKernels as AK
using PrecompileTools: @setup_workload, @compile_workload

# Groups of threads read runs of up to 4096 Float32s: on an RTX 5080, summing a 4096×4096 matrix
# along `dims=1` takes 17% less time than with a block per output.
Expand All @@ -21,4 +22,31 @@ AK.sort_tuning(::CUDABackend, ::Type{T}) where {T} =
radix_min_slice_len=1 << 19) :
AK.SortTuning(bitonic_max_len=512, radix_min_len=1 << 23, bitonic_max_slice_len=256)


# CUDACore's own workload can't cache these operations: it doesn't use the GPU, and this
# extension isn't loaded while CUDACore precompiles, so its kernels would use the default
# tunings rather than the ones above. The arrays are just large enough to select the kernels
# that a vector of 10^6 elements and a 1000×1000 matrix do. Without a functional GPU while
# precompiling, these operations are only compiled when first used. Like CUDACore's own
# workload, this needs 1.12: before, GPU code compiled while precompiling leaks into the host
# code of the image.
@static if VERSION >= v"1.12-"
@setup_workload begin
if CUDACore.functional()
@compile_workload begin
m = CuArray(ones(Float32, 256, 32))
v = vec(m)
sum(v)
sum(m; dims=1)
sum(m; dims=2)
cumsum(v)
any(iszero, v)
all(iszero, v)
sum(CuArray(ones(Int, 8192)))
synchronize()
end
end
end
end

end
Loading