You are on page 1of 8

Design Optimization of System-on-Chip Platforms Authored by Amirhosein Mosavi , Supervised by Professor Oniga Istvn, Reports in Computer Science, University

of Debrecen, Faculty of Informatics, 2011.

Design Optimization of System-on-Chip Platforms


A.Mosavi In the following the answers to the questions of Professor Oniga Istvan, on the subjects of Information Systems and his preferred bridging disciplines and branches of computer science including applied Information Technology and applied Mathematical Information Technology, are briefly documented. It is interested to briefly differ Information Technologies from the Information Systems and focus on the applied Information Technologies, preferably in the field of engineering design. It is questioned to study the optimization algorithms as a sub-branch of the Mathematical Information Technologies and explain the potential applications and innovative tools of Information Technologies in this regards. More precisely it should clear ly be defined how we could facilitate and also benefit from the optimization algorithms, with the aid of IT, for dealing with the multidisciplinary and multicriteria general optimization design tasks. It was preferably suggested to reference the statement with the up-to-date researches and very recently industrial study cases, focused in the field of electronic engineering design optimization and/or embedded system design and applications and explain the suggested solutions and methodologies.

Introduction There are various types of information systems (IS), including but not limited to decision support systems, knowledge management systems, database management systems, and office information systems [1]. An information system usually has an information technology component developed which interacts with the processes components. In this way information systems are distinct from information technology (IT). A series of IT methodologies and processes are used in order to develop and use an information system [3]. In most types of IS, the information technologies are most critical field of research. In other words IT is typically designed to enable humans to perform the most demanding and complicated tasks which would be difficult for human to easily deal with e.g., handling large amounts of information, performing complex calculations, and controlling many simultaneous processes, multiple criteria decision making and multiobjective optimization. IT is essential for executing the demanding computer science related tasks such as design optimization.

Design Optimization of System-on-Chip Platforms Authored by Amirhosein Mosavi , Supervised by Professor Oniga Istvn, Reports in Computer Science, University of Debrecen, Faculty of Informatics, 2011.

Solving multiobjective optimization problems is one of those complicated tasks which need the assistance of IT to be solved. In this regards Mathematical IT has been very active field of research, developing different types of optimization algorithms. Real- life industrial and engineering design problems, including problems of electronic engineering design optimization and embedded system design and applications, typically need to be considered from very different perspectives. For instance in designing a PCB for an embedded system or an smart sensor the perspectives of design could be cost, size of board, material, perfo rmance, cooling, utilized technology etc. This leads to the need of optimizing several conflicting

objectives simultaneously. In the light of this, it is quite natural that one of the main driving forces behind the utilizing mathematical IT is the difficulty related to the implementation of multiobjective optimization. Benefits of multiobjective optimization include that the conflicting objectives are taken into account simultaneously leading to an overall insight of the problem. Therefore, multiobjective optimization can bring about a significant competitive advantage when compared to widely used simplistic approaches where only one objective is optimized and other, although important, objectives are left without a special attention. In different fields of industry, there is a lot of need for multiobjective treatment but not yet enough awareness about it.

Standard form of an optimization problem Optimization means to find a set of system configurations which meet the objectives and satisfy the constraints. Many software packages can be used to describe the behavior of the system during the optimization. Software packages simulate the behavior of the systems in the different criteria of preferred design.

Standard form of an optimization problem [modefrontier user meeting 2010]

Design Optimization of System-on-Chip Platforms Authored by Amirhosein Mosavi , Supervised by Professor Oniga Istvn, Reports in Computer Science, University of Debrecen, Faculty of Informatics, 2011.

Here variables are the free parameters, i.e. the quantities that the designer can vary or the choices the designer can make. The Continuous variables could be point coordinates or process variables and the discrete variables are components from a catalogue or number of components. The black box could be a set of solvers that models and solves in a numerical manner the design problem (e.g. CAD/CAE tools) or a set of experiments that produces some data. For solving the problems in engineering design such as problems related to electronic engineering and embedded systems design the aim is to find the global optimal solutions. That is why the entire design space should be explored during the optimization process.

An example of optimization work flow

Problem Optimization work flow positioning the constraints [modefrontiers user meeting 2010]

Challenges arising while finding design optima There have been more than thousands of optimization algorithms developed so far for solving the industrial problems. As it has also reflected to my thesis too there are challenges rising when starting modeling and solving the problems for finding the optima. Problems include high number of variables, needs for global and multiobjective algorithms, complexity of workflow for engineers, manual workflow, lack of knowledge about the problem, lack of knowledge about the ability of algorithm, difficulties regarding choosing the appropriate algorithm for the modeled problem.

Response to the challenges As a response to the above challenges we suggest an implementation of optimization algorithms as a software package which answer the above mentioned challenges by integrating algorithms,

Design Optimization of System-on-Chip Platforms Authored by Amirhosein Mosavi , Supervised by Professor Oniga Istvn, Reports in Computer Science, University of Debrecen, Faculty of Informatics, 2011.

providing information, providing the ability of coupling CAD, ECAD and CAE packages and advanced visualization techniques. Software for multiobjective optimization and more complex integrated environments such as process integration and design optimization (PIDO) and problem solving environment (PSE) have become popular in the last few years and several new packages are about coming to the market. The importance of multiobjective optimization for the commercial world can be readily seen by the fact that most of the industrial companies eager to support developing one or more of the available packages. It is evidenced that research, commercial and academic communities are becoming increasingly interested in multiobjective optimization software packages.

State-of-the-art study cases; application of multiobjective optimization in real-life industrial proble ms related to electronic engineering and embedded systems

1. Micro-electro- mechanical systems (MEMS) Design Optimization [6] 2. Multiojective Optimization Applied to Mechatronics [6] 3. Multiojective Optimization Multi-Processor System-on-Chip (MPSoC) embedded systems[6] 4. Optimization that involve direct control on hardware, calibration[6] 5. Optimization of System-on-Chip Platform using multiobjective optimization [6]

In the following the first study case is presented with some figures and the rest are explained at the request of the Professor Oniga Istvan.

1. MEMS Design Optimization; a multiobjective optimization design study case MEMS is the ultimate enabling technology for the integration of almost any physical, chemical and biological phenomena that include motion, light, sound, chemistry, biochemistry, radio waves and computation, but all on a single chip.[4] Micro-electro- mechanical systems (MEMS) are very small devices in the dimensions of few micrometers whose behavior is governed both by electronics and mechanics. They are usually composed by several fixed and movable parts which have been obtained after a working process, usually an etching, of a unique silicon wafer opportunely prepared and eventually a subsequent packaging of some wafers. MEMS have wide applications in actuators and smart sensors.

Design Optimization of System-on-Chip Platforms Authored by Amirhosein Mosavi , Supervised by Professor Oniga Istvn, Reports in Computer Science, University of Debrecen, Faculty of Informatics, 2011.

1.1 MEMS Design and improvement Design revolves around a Build and Break iterative. By utilizing traditional numerical methods such as Gradient based search through parameterised model we face numerous problems. Problems are: Lack of expertise and knowledge Solutions are not global optimal Expensive solutions Most non-linear methods (gradient descent, Newtons) are suitable for single parameter optimization. As a conclusion to MEMS design we should admit that available tools are not efficient to detect potential fault and defects within the manufacturing process, and also simulation tools existing for MEMS technology cannot predict the effect of complex external or internal environments parameters [5]. That is why utilizing a multiobjective optimization tool via so called MEMS Automated Design Optimization seems necessary. The work flow of optimization is explained by the applicant at the presentation with more details.

MEMS Design Optimization Framework [modefrontiers user meeting 2010]]

Design Optimization of System-on-Chip Platforms Authored by Amirhosein Mosavi , Supervised by Professor Oniga Istvn, Reports in Computer Science, University of Debrecen, Faculty of Informatics, 2011.

MEMS Design Optimization Framework; involving MATLAB, Simulink, SUGAR, Intellisuite FEA, ANSYS and SolidWorks. [modefrontiers user meeting 2010]

Visualization the results on a Pareto optimal solutions diagram [modefrontiers user meeting 2010]

Design Optimization of System-on-Chip Platforms Authored by Amirhosein Mosavi , Supervised by Professor Oniga Istvn, Reports in Computer Science, University of Debrecen, Faculty of Informatics, 2011.

Visualization of the Response Surface Modeling (an approximation) results on a Pareto optimal solutions diagram [modefrontiers user meeting 2010]

References [1] Mark S. Silver, M. Lynne Markus, Cynthia Mathis Beath (1995) The Information Technology Interaction Model: A Foundation for the MBA Core Co urse, MIS Quarterly, Vol. 19, No. 3, Special Issue on IS Curricula and Pedagogy (Sep., 1995), pp. 361-390. [2] [3] Rockart et al. (1996) Eight imperatives for the new IT organization Sloan Management review. Hayes, Helen; Onkar Sharma (February 2003). "A decade of experience with a common first year program for computer science, information systems and information technology majors". Journal of Computing Sciences in Colleges 18 (3): 217227 [4] [5] MEMS/MOEMS Packaging, McGraw-Hill, NY, published July 2005 Elhadj Benkhelifa et al. (2010). Design and optimisation of microelectromechanical systems: a review of the state-of-the-art, International Journal of Design Engineering, Volume 3, Number 1 [6] Proceedings of the modefrontiers user meetingand conference 2010

Design Optimization of System-on-Chip Platforms Authored by Amirhosein Mosavi , Supervised by Professor Oniga Istvn, Reports in Computer Science, University of Debrecen, Faculty of Informatics, 2011.

Bibliography 1. John L. Hennessy, David A. Patterson: Computer organisation and design: the hardware/software interface Morgan Kaufman Publ., 1998, 2nd. ed., San Francisco 2. Rob Williams, Computer Systems Architecture A Networking Approach, Pearson/AddisonWesley, Harlow-London, 2001. 3. Pokordi, Lszl, Szabolcsi, Rbert, Mathematical Models Applied to Investigate Aircraft Systems, nomogrfia, Monographical Booklets in Applied and Computer Mathematics, MB-12, PAMM, Megyetemi Kiad, Budapest, 1999.

You might also like