JuliaGPU / Adapt.jl

Other
86 stars 24 forks source link

Reason for missing LinearAlgebra wrappers? #46

Closed danielwe closed 2 years ago

danielwe commented 2 years ago

Is there a reason the WrappedArray union omits some of the basic array wrappers from LinearAlgebra, specifically Symmetric, Hermitian, UpperHessenberg, SymTridiagonal, and Bidiagonal? I know at least symmetric and hermitian matrices have good CUSOLVER support (even if CUDA.jl doesn't wrap everything yet), so at least from the GPU perspective, I don't see an obvious reason why these wouldn't be treated the same as other wrappers.

maleadt commented 2 years ago

No reason, PRs welcome :slightly_smiling_face: