1: \begin{abstract}

2: $\Name$ is an extension of $\Datalog$, where  instead of a program being a collection of

3: union of conjunctive queries over the standard Boolean semiring, a program may now be a

4: collection of sum-sum-product queries over an arbitrary commutative partially ordered

5: pre-semiring. $\Name$ is more powerful than $\Datalog$ in that its additional algebraic

6: structure alows for supporting recursion with aggregation. At the same time, $\Name$ retains

7: the syntactic and semantic simplicity of $\Datalog$: $\Name$ has declarative least fixpoint

8: semantics. The least fixpoint can be found via the na\"ive evaluation algorithm that

9: repeatedly applies the immediate sequence opeator until no further change is possible.

10: 

11: It was shown in~\cite{Khamis0PSW22} that, when the underlying semiring is $p$-stable,

12: then the na\"ive evaluation of any $\Name$ program over the semiring converges in a finite

13: number of steps. However, the upper bounds on the rate of convergence  were exponential in

14: the number of ground IDB atoms.

15: 

16: This paper establishes polynomial upper bounds on the convergence rate  of the na\"ive

17: algorithm on {\bf linear} $\Name$ programs, which is quite common in practice. In

18: particular, the main result of this paper is that the convergence rate of linear $\Name$

19: programs under  any $p$-stable semiring is  $O(pn^3)$. Furthermore, we show a matching lower

20: bound by constructing a $p$-stable semiring and a linear $\Name$ program that requires

21: $\Omega(pn^3)$ iterations for the na\"ive iteration algorithm to converge. Next, we study

22: the convergence rate in terms of the number of elements in the semiring for linear $\Name$

23: programs.  When $L$ is the number of elements, the convergence rate is bounded by $O(pn \log

24: L)$.  This significantly improves the convergence rate for small $L$.  We show a nearly

25: matching lower bound as well.

26: \end{abstract}

27: