Skip to content

Specialize transpose! for CuMatrix #3015

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
kshyatt merged 1 commit into from
Jan 12, 2026
Merged

Specialize transpose! for CuMatrix #3015

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
8 changes: 8 additions & 0 deletions src/array.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -946,3 +946,11 @@ function Base.resize!(A::CuVector{T}, n::Integer) where T
A.dims = (n,)
return A
end


# CUBLAS.geam! is much faster than the generic implementation of transpose! in GPUArrays:
function LinearAlgebra.transpose!(dest::CuMatrix{T}, src::CuMatrix{T}) where {T <: Union{Float32, Float64, ComplexF32, ComplexF64}}
axes(dest) == reverse(axes(src)) || throw(DimensionMismatch("axes of the destination are incompatible with that of the source"))
CUDA.CUBLAS.geam!('T', 'T', one(T), src, zero(T), src, dest)
return dest
end
10 changes: 10 additions & 0 deletions test/base/array.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -979,6 +979,16 @@ end
@test c === a
end

@testset "transpose!" begin
for T in [Float32, Float64, ComplexF32, ComplexF64]
a = CUDA.rand(T, 10, 20)
b = similar(a, reverse(size(a)))
c = similar(a)
@test Array(transpose!(b, a)) == transpose(Array(a))
@test_throws DimensionMismatch transpose!(c, a)
end
end

@testset "issue 2595" begin
# mixed-type reductions resulted in a deadlock because of union splitting over shfl
a = CUDA.zeros(Float32, 1)
Expand Down