Software systems evolve throughout their life cycles. Many revisions are produced over time. Verifying each revision of the software is impractical. Regression verification suggests reusing intermediate results from the previous verification runs. This paper studies regression verification via summary reuse. Not only procedure summaries, but also loop summaries are proposed to be reused. This paper proposes a fully automatic regression verification technique in the context of CEGAR. A lazy counterexample analysis technique is developed to improve the efficiency of summary reuse. We performed extensive experiments on two large sets of industrial programs (3,675 revisions of 488 Linux kernel device drivers). Results show that our summary reuse technique saves 84 to 93 percent analysis time of the regression verification.

Software products are evolving during their life cycles. Ideally, every revision need be formally verified to ensure software quality. Yet repeated formal verification requires significant computing resources. Verifying each and every revision can be very challenging. It is desirable to ameliorate regression verification for practical purposes. In this paper, we regard predicate analysis as a process of assertion annotation. Assertion annotations can be used as a certificate for the verification results. It is thus a waste of resources to throw them away after each verification. We propose to reuse the previously-yielded assertion annotation in regression verification. A light-weight impact-analysis technique is proposed to analyze the reusability of assertions. A novel assertion strengthening technique is furthermore developed to improve reusability of annotation. With these techniques, we present an incremental predicate analysis technique for regression verification. Correctness of our incremental technique is formally proved. We performed comprehensive experiments on revisions of Linux kernel device drivers. Our technique outperforms the state-of-the-art program verification tool CPAchecker by getting 2.8x speedup in total time and solving additional 393 tasks.

Efficient organization and analysis of academic information has many advantages. Most scholar retrieval systems appeared these years can perform keyword-based paper search. However, performing large-scale expert and paper retrieval is an intractable problem. Here we present a platform that can not only reduce the workload of researchers when searching academic literature, but also promote academic communication. In this paper, we introduced a novel community partition algorithm specific to deal with large-scale citation network, the Large-scale Citation Network Partition Algorithm (LCNPA). We demonstrate the construction of the platform, and illustrate the implemented functions and instructions.