prittjam
This would be a very useful feature. Unfortunately, it doesn't seem to be able to find the path of the currently edited file for some reason.
Eigen also has lazy evaluation and has their own concept of views. Why is the compiler able to inline Eigen views but not xtensor views?
Thanks for the response. I think this would be a common gotcha and maybe worth putting in the docs.
This simple example fails but works with einsum, although einsum casts to a dynamic matrix. ```julia d1 = HybridMatrix{3,StaticArrays.Dynamic()}(rand(3,100)) d2 = HybridMatrix{3,StaticArrays.Dynamic()}(rand(3,100)) @tullio d3[i,j] := d1[i,j]*d2[i,j] ``` The error is...
We want HybridMatrix for speed reasons. I was trying Tullio because matrix multiplication also returns a matrix. I'll get rid of LoopVecotrization for now. Seems like there are some corner...
Thanks for the response. Can you give a TL;DR version of when to prefer one package over the other?
> Yes, TensorOperations ought to be good at this. > > I believe it still decides the order of multiplication based only on the syntax, i.e. before knowing the sizes...
I'm in complete agreement with https://github.com/JuliaLang/julia/issues/22129#issuecomment-1209696536. This fact that you can't return a tuple of arrays causes many headaches for functions that compute on tensors. It's a major weak point...
so broadcasting lazystack over unzip_broadcast achieves what is desired: I can generate a tuple of arrays from broadcasting over a function with multiple return values and then stack each array...
I've also encountered this.