Title: Exploiting partial orders and symmetries in efficient analysis of message-passing concurrency

Speaker: Subodh Sharma    (bio) (bio)

Subodh Sharma is a faculty member in the Department of Computer Science and Engineering at IIT Delhi. His research interests lie in the area of software engineering and formal methods, particularly in ensuring the reliability of parallel software via static and dynamic program analyses, model checking, and PL solutions, and employing HPC towards the creation of scalable verification technology. Lately, his research investigations have also spanned the areas of systems security, data privacy, and Blockchain.

When: Tuesday, 06 September 2022 at 1900 hrs (IST)       

Abstract:
The message passing paradigm is the lingua franca for developing large distributed-memory programs, such as high-performance scientific computing and event-driven web applications. Message-passing applications are often found to be using communication nondeterminism (used primarily to obtain efficiency by masking network latencies) and symmetric communication among processes (which keeps programming simple). Under communication nondeterminism, a process can post (possibly asynchronous) receive calls that can potentially match any of the messages sent to the process. Under symmetric communication, a process's communication structure may be partly or completely symmetric with the communication structure of other processes. Interestingly, communication nondeterminism is one of the important sources of analysis complexity and detecting symmetries is, in general, hard. This talk will present practical techniques to efficiently analyse message passing programs by (i) exploiting partial order among the communication dependencies and (ii) detecting symmetries in process-communication. The work in this talk has been published in TOPLAS 2017, FM 2018, ICST 2021, and ASE 2022.