Accelerating multi-party scheduling for transaction-level modeling

Authors:
Di Wang;Vyas Venkataraman;Zhen Wang;Wei Qin;Hangsheng Wang;Mrinal Bose;Jayanta Bhadra
Affiliations:
Boston University, Boston, MA, USA;Boston University, Boston, MA, USA;Boston University, Boston, MA, USA;Boston University, Boston, MA, USA;Freescale Semiconductor Inc., Austin, TX, USA;Freescale Semiconductor Inc., Austin, TX, USA;Freescale Semiconductor Inc., Austin, TX, USA
Venue:
Proceedings of the 19th ACM Great Lakes symposium on VLSI
Year:
2009

Citing 7
Cited 0

Communicating sequential processes

Communicating sequential processes
Introduction to the ISO specification language LOTOS

Computer Networks and ISDN Systems - Special Issue: Protocol Specification and Testing
A comprehensive study of the complexity of multiparty interaction

Journal of the ACM (JACM)
Logic-based 0-1 constraint programming

Logic-based 0-1 constraint programming
Communication and Concurrency

Communication and Concurrency
PCI Express System Architecture

PCI Express System Architecture
SHIM: a deterministic model for heterogeneous embedded systems

IEEE Transactions on Very Large Scale Integration (VLSI) Systems

Quantified Score

Hi-index	0.03

Visualization

Abstract

Transaction-level modeling is an essential component of system-level design. This paper advocates using rendezvous, a construct common to concurrent programming languages, as a theoretical foundation for transactions. Compared to regular function calls, rendezvous are atomic and support multipartiness and parallel composition. However, scheduling multiparty rendezvous is a challenging task due to its NP-hard complexity. This paper describes a heuristic algorithm that significantly reduces the scheduling complexity in practice. It first constructs a relationship graph among rendezvous. It then simplifies the graph and translates it into a decision tree, which assists the scheduler in partitioning and pruning the search space. Our experimental results show that the algorithm is able to improve the efficiency of the scheduler significantly.