[Paper Review] Global Clock, Physical Time Order and Pending Period Analysis in Multiprocessor Systems

Yunji Chen, Tianshi Chen+1 more|ArXiv.org|Mar 29, 2009

Parallel Computing and Optimization Techniques42 references9 citations

TL;DR

This paper introduces physical time order and pending period analysis in multiprocessor systems using a global clock to provide a consistent, real-time-based partial ordering of operations. By leveraging pending periods—time intervals during which operations are active—it reduces complex verification problems like memory consistency and event ordering to polynomial-time solutions, achieving $O(n^2C^p p)$ for complete memory consistency verification and $O(nC^p p)$ for event ordering problems.

ABSTRACT

In multiprocessor systems, various problems are treated with Lamport's logical clock and the resultant logical time orders between operations. However, one often needs to face the high complexities caused by the lack of logical time order information in practice. In this paper, we utilize the \emph{global clock} to infuse the so-called \emph{pending period} to each operation in a multiprocessor system, where the pending period is a time interval that contains the performed time of the operation. Further, we define the \emph{physical time order} for any two operations with disjoint pending periods. The physical time order is obeyed by any real execution in multiprocessor systems due to that it is part of the truly happened operation orders restricted by global clock, and it is then proven to be independent and consistent with traditional logical time orders. The above novel yet fundamental concepts enables new effective approaches for analyzing multiprocessor systems, which are named \emph{pending period analysis} as a whole. As a consequence of pending period analysis, many important problems of multiprocessor systems can be tackled effectively. As a significant application example, complete memory consistency verification, which was known as an NP-hard problem, can be solved with the complexity of $O(n^2)$ (where $n$ is the number of operations). Moreover, the two event ordering problems, which were proven to be Co-NP-Hard and NP-hard respectively, can both be solved with the time complexity of O(n) if restricted by pending period information.

Motivation & Objective

• Address the high computational complexity of verifying memory consistency and event ordering in multiprocessor systems due to incomplete logical time order information.
• Overcome the limitations of Lamport’s logical clock by introducing a physical time order based on a global clock to reflect real execution timing.
• Develop a practical framework—pending period analysis—that enables efficient inference of operation orderings using time intervals instead of precise timestamps.
• Reduce the complexity of known hard problems in multiprocessor systems by exploiting physical time order constraints derived from global clock synchronization.
• Enable scalable and complete verification of memory consistency and event ordering in real-world multiprocessor systems with industrial applicability.

Proposed method

• Define the pending period as a time interval during which an operation is active, serving as a relaxation of the precise performed time.
• Introduce physical time order as a partial order between operations with disjoint pending periods, based on their global clock timestamps.
• Prove that physical time order is independent and consistent with traditional logical time orders (e.g., processor order, execution order).
• Construct a frontier graph to model possible execution paths and use it to efficiently compute must-have and could-have happened-before relations.
• Apply assignment analysis to infer pending periods of all operations from only a subset of observed operations, reducing observation overhead.
• Integrate physical time order with logical time orders to constrain and prune the space of candidate executions during inference.

Experimental results

Research questions

• RQ1 Can a global clock in multiprocessor systems be used to define a consistent, real-time-based partial ordering of operations beyond logical clocks?
• RQ2 How can pending periods—flexible time intervals for operations—be used to simplify the analysis of operation ordering in real executions?
• RQ3 To what extent can physical time order reduce the complexity of NP-hard problems like complete memory consistency verification?
• RQ4 Can event ordering problems that are Co-NP-Hard and NP-hard be solved efficiently under the constraint of pending period information?
• RQ5 What is the impact of combining physical time order with logical time orders on the scalability and accuracy of execution analysis?

Key findings

• Physical time order is proven to be independent and consistent with traditional logical time orders, providing a reliable foundation for real-time execution analysis.
• Pending period analysis enables complete memory consistency verification with a time complexity of $O(n^2C^p p)$, significantly reducing the NP-hard complexity.
• Two event ordering problems—previously Co-NP-Hard and NP-hard—can now be solved in $O(nC^p p)$ time when restricted by pending period information.
• The frontier graph model allows efficient computation of must-have and could-have happened-before relations in $O(n_f + e_f)$ time, where $n_f$ and $e_f$ are nodes and edges in the frontier graph.
• Assignment analysis shows that observing pending periods of only a subset of operations is sufficient to infer the pending periods of all operations.
• The proposed framework has been successfully applied in industrial settings, demonstrating practical feasibility and performance gains in real multiprocessor verification tasks.