![Institutional Repository [SANDBOX]](images/dias.logo.png "Institutional Repository [SANDBOX]"){kind=logo}
Institutional Repository [SANDBOX]
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/8F49F8E5-94DB-4254-9C4C-01665460791E | ||
| Year | 2019 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Georgios Foteinopoulos, "Design of a distributed producer-consumer system for Golomb ruler computation on a reconfigurable supercomputer", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2019 https://doi.org/10.26233/heallink.tuc.84051 | ||
| Appears in Collections |
An optimal Golomb ruler is a set of distinct positive integers (marks) such that the differences, computed over all different pairs, are distinct, and they occupy less space than any other ruler with the same number of marks. Optimal Golomb rulers are utilized in a great variety of applications over many scientific fields, with their derivation being computationally expensive and extremmelly time-consuming, despite the continuous growth of technology. However, finding and proving a Golomb ruler to be optimal is a problem that can be parallelized. Therefore there is a great incentive to design a custom solution based on FPGA technology. Our design utilizes Convey HC-2ex, a heterogeneous multi-fpga computing platform, that allows us to effectively use many levels of parallelism, in order to improve previous architectures focused on solving this problem. Following the evaluation of the HC-2ex results, we relocated our design to two contemporary FPGAs, capitalizing on the evolution of HDL designing tools. Despite the greater scale of the HC-2ex platform, we confronted spatial limitations, due to the low performance of the older designing tool. On the other hand, the results from the two FPGAs are significantly more promising, as we achieve close to perfect utilization of their available resources, with a high clock, resulting to a potential speedup of 3X over the HC-2ex, down to 2X in the case of larger OGRs.
Copyright © DIAS 2013
