Skip to content

Add a dispatch for LinearAlgebra.norm2 #2302

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
May 16, 2026
+17 −4
Merged

Add a dispatch for LinearAlgebra.norm2 #2302

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: 1 addition & 1 deletion CUDACore/Project.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,7 +52,7 @@ CUDA_Runtime_jll = "0.22"
ChainRulesCore = "1"
EnzymeCore = "0.8.2"
ExprTools = "0.1"
GPUArrays = "11.5"
GPUArrays = "11.5.4"
GPUCompiler = "1.10"
GPUToolbox = "1.1"
KernelAbstractions = "0.9.38"
Expand Down
6 changes: 3 additions & 3 deletions lib/cublas/src/linalg.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -209,16 +209,16 @@ function LinearAlgebra.:(*)(transx::Transpose{<:Any,<:StridedCuVector{T}},
return dotu(n, x, y)
end

function LinearAlgebra.norm(x::DenseCuArray{<:Union{Float16, ComplexF16, CublasFloat}},
function LinearAlgebra.norm(x::StridedCuVecOrDenseMat{<:Union{Float16, ComplexF16, CublasFloat}},
p::Real=2)
if p == 2
return nrm2(x)
else
return invoke(norm, Tuple{AbstractGPUArray, Real}, x, p)
return invoke(norm, Tuple{AnyGPUArray, Real}, x, p)
end
end
LinearAlgebra.norm(x::Diagonal{T, <:StridedCuVector{T}}, p::Real=2) where {T<:Union{Float16, ComplexF16, CublasFloat}} = norm(x.diag, p)
LinearAlgebra.norm2(x::DenseCuArray{<:Union{Float16, ComplexF16, CublasFloat}}) = nrm2(x)
LinearAlgebra.norm2(x::StridedCuVecOrDenseMat{<:Union{Float16, ComplexF16, CublasFloat}}) = nrm2(x)

LinearAlgebra.BLAS.asum(x::StridedCuArray{<:CublasFloat}) = asum(length(x), x)

Expand Down
13 changes: 13 additions & 0 deletions lib/cublas/test/level1/core.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -76,6 +76,19 @@ using LinearAlgebra
@test norm(dDx, 2) ≈ norm(Dx, 2)
@test norm(dDx, Inf) ≈ norm(Dx, Inf)
end

@testset "norm of strided views" begin # JuliaGPU/CUDA.jl#2280
# 1D contiguous view: should hit the cuBLAS nrm2 fast path.
x = rand(T, m)
dx = CuArray(x)
@test norm(@view(dx[2:end-1]), 2) ≈ norm(@view(x[2:end-1]), 2)
# Multi-dim non-contiguous view: must avoid scalar iteration.
y = rand(T, 10, 10)
dy = CuArray(y)
@test norm(@view(dy[2:end-1, 2:end-1]), 1) ≈ norm(@view(y[2:end-1, 2:end-1]), 1)
@test norm(@view(dy[2:end-1, 2:end-1]), 2) ≈ norm(@view(y[2:end-1, 2:end-1]), 2)
@test norm(@view(dy[2:end-1, 2:end-1]), Inf) ≈ norm(@view(y[2:end-1, 2:end-1]), Inf)
end
end

@testset for T in [Float16, ComplexF16]
Expand Down