Back to results list
Please use this identifier to cite or link to this item:
|Title:||Efficient C-based SoC architectures and design methodologies||Authors:||Liu, Yidi||Advisors:||Carrion Schafer, Benjamin (EIE)
Lau, Francis C. M. (EIE)
|Keywords:||Systems on a chip -- Design and construction.
|Issue Date:||2016||Publisher:||The Hong Kong Polytechnic University||Abstract:||ITRS  suggest that by 2020 a 10x productivity increase for designing complex SoCs is needed. Two main factors are predicted to help achieving this goal. The first is the re-use of components. ITRS estimates that around 90% of the SoCs will be composed of re-used components. Secondly, the use of new design methodologies to raise the level of abstraction, i.e. High-Level Synthesis (HLS).Companies have started to rely on High-Level Synthesis (HLS) to increase their design productivity and making use of third party behavioral IPs (3PBIPs) to meet their tight schedules.C-Based design has many advantages compared to traditional RTL design. The most salient ones include, the increase in design productivity, which allows design teams to meet the increasingly stringent time-to-market requirements, the ability to create smaller designs compared to hand-coded RTL due to its ability to maximize resource sharing and the possibility of generating a set of different micro-architectures with different area vs. performance trade-offs without having to modify the original behavioural description, also called Design Space Exploration (DSE). In traditional RTL design it is virtually impossible to do DSE as it would involve having to re-write the RTL description completely in order to create the new micro-architecture. Moreover it is common practice not to modify any hardware (HW) block that has been fully verified, even if a more efficient architecture could be achieved in subsequent designs, due to the cost of having to re-verify the new implementation.
HLS is a single process synthesis method, which takes individual behavioral descriptions as inputs and performs resource allocation, scheduling and binding on each of them to obtain an RTL description, which can efficiently execute it. As mentioned previously, HLS allows designers to generate micro-architectures with different area vs. performance trade-offs. Typically, high-performance designs will consume more HW resources as loops are fully unrolled and functions inlined, while low performance designs tend to be much smaller as resources can be shared, functions do not need to be inlined and loops executed sequentially.This thesis investigates effective design methods to automatically generate SoC architectures with unique area vs. performance trade-offs when these are fully described at the behavioral level. State-of-the-art HLS tools now include system level design capabilities which allow this. Two main cases are studied: Static schedule and dynamic tasks schedule architectures. Scheduling and mapping approaches can be classified as online and off-line algorithms. Off-line algorithms have shown to be able to obtain superior results by exploring a larger portion of the design space. These static/off-line methods schedule multiple applications on a system and can be used in order to reduce the complexity of the HW and hence reduce its area and power overheads. On the other side, online methods are much more flexible. It is therefore important to study both approaches.
|Description:||PolyU Library Call No.: [THS] LG51 .H577M EIE 2016 Liu
xx, 69 pages :illustrations
|URI:||http://hdl.handle.net/10397/55265||Rights:||All rights reserved.|
|Appears in Collections:||Thesis|
Show full item record
Files in This Item:
|b28945438_link.htm||For PolyU Users||208 B||HTML||View/Open|
|b28945438_ira.pdf||For All Users (Non-printable)||1.95 MB||Adobe PDF||View/Open|
Citations as of Mar 19, 2018
Citations as of Mar 19, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.