Merge continuous infrastructure to Main

[willow-ahrens]

We decided to leave z3 out of it. Here's the final list of what's needed:

• ScaleArray
• ProductArray
• singlerlelevels.jl
• singlelistlevels.jl
• continuous usage test cases from PLDI case studies
• simplify_program.jl
• parameterized storage types for arrays of open or closed intervals

[willow-ahrens]

Decided not to merge https://github.com/willow-ahrens/Finch.jl/blob/3b2504bab426e591e89d4beff937f17336366c09/src/looplets/sequences.jl#L77-L86 ?

[willow-ahrens]

Decided not to merge https://github.com/willow-ahrens/Finch.jl/blob/3b2504bab426e591e89d4beff937f17336366c09/src/looplets/lookups.jl#L50?

I'm not quite sure this is always right?

[willow-ahrens]

Decided not to merge

https://github.com/willow-ahrens/Finch.jl/blob/3b2504bab426e591e89d4beff937f17336366c09/src/looplets/switches.jl#L58-L65
This looks good but not sure if it's always right.

[willow-ahrens]

Might need to merge simplify_program.jl and symbolic.jl likely have some rules that should be merged, perhaps we could add test cases for them too?

[willow-ahrens]

• scale.jl and product.jl should be merged, with a restriction that the extent cannot be negative. If we could figure out a syntax to make it clearer whether we want a scalearray in the expression A[2 * i] that would help. We would also include the changes to wrapperize in this PR.

[willow-ahrens]

• openrlelevels.jl, singlelistlevels.jl, and singlerlelevels.jl should get merged, I suspect singlerlelevel is Single run level #329 and singlelistlevel is Singleton level #331, and I also wonder whether there are some tricks we can play with arrays to get the rlellevels to handle openness more parametrically. Like, if all the right endpoints are closed, maybe we should add a ClosedVector type to wrap the elements of a vector with a closed limit.

[willow-ahrens]

Might need to merge continuous extent changes in dimensionalize.jl

[willow-ahrens]

@nullplay, this is all I have after looking at the diffs between branches. Let me know if I missed anything! I'm prioritizing this because I'd like to merge these before refactoring Finch any further.

[willow-ahrens]

Here's the rules that might need merging:

        # handles counting query 
        (@rule block(~s1..., declare(~a::isvariable, ~v::isconstant), ~s2..., freeze(~a), ~s3...) => begin
            if ortho(a, s2) 
                block(s1..., Postwalk(@rule a => v)(block(s3...)))
            end
        end),
        (@rule access(~a::isliteral, reader(), ~j...) => getval(a)), 


        (@rule block(~s1..., thaw(~a::isvariable), ~s2..., freeze(~a), ~s3...) => if ortho(a, s2)
        (@rule block(~s1..., thaw(~a::isvariable), ~s2..., freeze(~a), ~s3...) => if ortho(a, s2)
            block(s1..., s2..., s3...)
            block(s1..., s2..., s3...)
        end),
        end),

        # For better discordant traversal 
        (@rule loop(~idx, ~ext::isvirtual, ~body) => begin
            body_2 = Postwalk(@rule block(define(~a::isvariable, idx), loop(~idx2, ~ext2::isvirtual, ~body2)) => begin
                if idx ∉ getunbound(body2)
                    if ext2.val.start == a && ext2.val.stop == a
                        loop(idx2, ext, body2)
                    end
                end
            end)(body)

            if body_2 !== nothing
               body_2
            end
        end),

    ]
    ]
end
end


@@ -216,8 +248,18 @@ function visit_simplify(node::FinchNode)
end
end
    root = Rewrite(Prewalk((x) -> if x.kind === literal 
        if x.val isa Finch.Limit
            if x.val == drop_eps(x.val)
                literal(drop_eps(x.val))
            end
        end
    end))(root)