Sheehan Olver
Sheehan Olver
This is fixed in v2.0
```julia julia> using LazyArrays, BandedMatrices julia> A = BandedMatrix(0=>ones(10),1 => zeros(9)) 10×10 BandedMatrix{Float64} with bandwidths (0, 1): 1.0 0.0 ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ ⋅ 1.0 0.0...
I meant also for standard types like ` Float64 ` and something accessible in packages not dependent on this…
Aren't they implemented as `KronTrav`? This issue should really be in LazyBandedMatrices.jl, the slow code should be reproducible there
Note `::KronTrav * ::DiagTrav` can be reduced to matrix-matrix which will probably be the fastest. Is this what you actually want? That is, do you want to store your coefficients...
Here's an example, but I just realised its not smart enough to reduce to banded * matrix * banded..... so I think the first step should be to speed up...
I probably won't be able to look at this for a while. But I want you to look into your heart and ask yourself, do you actually want the code...
I don’t actually know how to use AI so that was all natural, baby
How should `mul!(::DiagTrav, ::KronTrav, ::DiagTrav)` work? To recast as a matrix product `Z = B*X*A'` we would need to have a temporary matrix to store `B*X`. Do we allocate this...