Akash Kumar 







Predictable Mapping of Design on Networked Architecture

2005 - 2009

The objective of this project is to design predictable multi-processor systems. This predictability is in terms of both the architecture and the applications.


New applications for embedded systems demand complex multi-processor designs to meet real-time deadlines while achieving other critical design constraints like low power consumption and low area. With high consumer demand the time-to-market has become significantly lower. Multi-Processor Systems-on-Chips (MPSoCs) have been proposed as a promising solution for all these previous problems.

Further, these systems are becoming increasingly heterogeneous with the use of dedicated IP blocks and application domain specific processors. To achieve high performance in such systems, the limited computational resources must be shared. The concurrent execution of dynamic applications on shared resources is a potential source of interference. Modeling and analyzing this interference is a key to building cost-effective systems which can deliver the desired performance of the applications. A number of issues remain to be solved for such systems to be useful.
  1. How should such systems be designed. With ever increasing transistor-density there is more computing power in a single chip that we had in the first super-computer ever built. Designing such systems in a systematic method is a major challenge.
  2. How should such systems be programmed. With over a billion transistors available for processing, programming them and utilizing in an efficient manner is not easy. A lot of human effort is spent on programming these multiprocessor systems. This indirectly translated to money for companies.
  3. How to analyze the performance of applications. With multiple applications sharing resources when they are run, predicting their performance is non-trivial. Even when one application is mapped on a mutliprocessor platform, issues like communication delay, memory limitation, etc make the system very hard to analyze. When the system has to support multiple applications, the problem becomes even harder.
This project aims to alleviate some of the above-mentioned problems and design predictable systems.

Partners involved

People involved from TU/e – ICS-ES

People involved from NUS – ECE