PTIME Computation of Transitive Closures of Octagonal Relations

Filip Konecny

DOI: https://doi.org/10.48550/arXiv.1402.2102

2014-02-10

Abstract:Computing transitive closures of integer relations is the key to finding precise invariants of integer programs. In this paper, we study difference bounds and octagonal relations and prove that their transitive closure is a PTIME-computable formula in the existential fragment of Presburger arithmetic. This result marks a significant complexity improvement, as the known algorithms have EXPTIME worst case complexity.

Logic in Computer Science

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to calculate the transitive closure of integer relations within polynomial time, especially the transitive closures of difference bounds (DB) and octagonal relations. The key to this problem lies in finding accurate invariants to verify the correctness of integer programs.

### Specific Problem Description

1. **Background**:
- The calculation of the transitive closure is crucial for finding accurate invariants of integer programs.
- Difference bounds (DB) relations and octagonal relations are widely used in software and hardware verification.
- Currently, known algorithms have exponential - time complexity (EXPTIME) in the worst - case.

2. **Research Objectives**:
- Prove that the transitive closures of difference bounds and octagonal relations can be calculated within polynomial time, and the results can be represented by the existential fragment of Presburger arithmetic.
- Propose a new polynomial - time algorithm to calculate the closed forms of these relations.

3. **Main Contributions**:
- Propose the first polynomial - time algorithm to calculate the closed forms of difference bounds and octagonal relations.
- This result significantly improves the complexity of existing algorithms, since known algorithms have exponential - time complexity in the worst - case.

### Technical Details

- **Definition of Closed Form**:
- Given a binary relation \( R \), its closed form \( \hat{R}(k) \) is a formula that contains unprimed and primed variables as well as a parameter variable \( k \), such that replacing the parameter \( k \) with any integer \( n\geq1 \) can accurately describe \( R^n \).

- **Acceleration Algorithm**:
- The method of calculating the exact closed form of iterative relation combinations is called the acceleration algorithm.
- Known acceleration algorithms are based on the concept of periodicity, but these methods may lead to exponential growth in some cases.

- **Core Idea of the New Algorithm**:
- Reduce the problem of calculating the closed form to calculating two difference bounds relations \( R_{fw} \) and \( R_{bw} \) that are computable within polynomial time.
- Utilize path schemes to capture all the minimum - weight paths in the unfolded constraint graph.
- Ensure the efficiency and correctness of the algorithm through the normalization of paths.

### Summary

The main objective of this paper is to propose a polynomial - time algorithm to calculate the transitive closures of difference bounds and octagonal relations, thereby significantly improving the time complexity of existing algorithms. This result is of great significance for the verification and analysis of integer programs.