Allocation of nested array

[NeuralCoder3]

The alloc2malloc phase seems to have a problem with the allocation of nested arrays:

%mem.alloc («a; «b; .Idx 4294967296»»,0) mem

Example: https://github.com/NeuralCoder3/thorin2/blob/matrix_dialect/lit/mem/nested_alloc.thorin

It might be the case, that my application of alloc is wrong in this case.

Error:

lit/mem/nested_alloc.thorin:9:9-11:23: error: cannot pass argument 
  '(mem_1264932_303941, %core.trait.size «__1264888_303893#0:(.Idx 2); «__1264888_303893#1:(.Idx 2); .Idx 4294967296»»)' of type 
  '[%mem.M, «__1264888_303893#0:(.Idx 2); «__1264888_303893#1:(.Idx 2); .Idx 4294967296»»]' to 
  '%mem.malloc («__1264888_303893#0:(.Idx 2); «__1264888_303893#1:(.Idx 2); .Idx 4294967296»», 0)' of domain 
  '[%mem.M, .Nat]'

[leissa]

So there are two things at place here.

1. There was a bug with the returned type of a core.trait.size (which I've just fixed)
2. Rn, core.trait.size only deals with structural Arr%ays. We need to add sth in its normalizers to do the right thing for parameterized Arr%ays.

[leissa]

So, the bug is fixed. The other issue is more an enhancement to make it work in the llvm backend.

[leissa]

And this is not entirely trivial and clear to me, how to do that properly. Ideas?
E.g., what is the size of this thing here?

<<i: n; <<i as Nat; T>>

[NeuralCoder3]

Is this a triangle array?
I am not sure if we can view it in a holistic way.
For many cases, we could find iteration strategies with tight bounds or at least some upper bounds.
But I think it is generally not computable what the correct size is in one go.
We would basically need to evaluate n and for [0...n-1] evaluate the inner function to get the bounds.
Overall, this would be the same as unrolling it entirely and having a standard deep initialization.

[leissa]

Yes, this is a triangle array. I think this is a case of: "we support this and that and for other stuff you are out of luck".
The example above at least is a parameterized array.

[leissa]

I'll close this for now and open a new enhancement issue.