الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 138 |  |  |
| | | from | 138 | | |
Abstract
A wide variety of embedded instruments are increasingly integrated within modern System-on-Chips (SoCs) for the purpose of testing, debugging, monitoring, and several other aims. To control this growing use of embedded instruments as well as to standardize their test and access protocols, the IEEE 1687 standard has defined a flexible network infrastructure. In such networks, accessing a certain instrument can be realized in many ways using different sets of network configurations that are generated in a process called pattern retargeting. To find the shortest access pattern among all the possible network configurations, it is necessary to (1) use an algorithm that performs an exhaustive search over a finite solution space to find the optimal solution, and (2) bind the solution space to a limited number of configuration cycles, as it determines when to terminate the search. By fulfilling both challenges, the ability to find the shortest path with minimum access time is ensured.
This thesis contributes an optimal pattern retargeting model based on a Boolean Satisfiability (SAT) problem. The proposed modeling has the advantage of being applicable to any arbitrarily designed IEEE 1687 network, while ensuring minimal access time. Our work also presents a new methodology for computing the right upper-bound on the number of configuration cycles to reduce the time spent searching for the optimal configuration vectors. The feasibility of the computational method has been demonstrated using a number of usage models, so that the algorithm can be reproduced in practice.
To assess the effectiveness of our proposed work, the developed approach was tested on a set of regular and irregular benchmarks that are neither constrained by a specific network design nor a limited size. It was further evaluated against state-of-the-art retargeting models to provide a fair comparison. Results prove the success of our work in finding the optimal retargeting vectors in terms of network access times. Additionally, results show a significant reduction in the computational resources (CPU and memory) used by our algorithm compared to relevant work. These reductions are around 70% in terms of memory used to model the retargeting problem and 60% in terms of CPU time spent to find the optimal solution.