source: anr-2010/section-2.tex @ 277

Last change on this file since 277 was 237, checked in by coach, 15 years ago

Paul coquilles

File size: 6.9 KB
Line 
1Embedded systems (SoC and MPSoC) became an inevitable evolution in the microelectronic industry.
2Due to the exploding fabrication costs, the ASIC technology (Application Specific Integrated Circuit)
3is not an option for SMEs (Small and Medium Enterprises).
4Fortunately, the new FPGA (Field Programmable Gate Array) components,
5such as the Virtex5 family from \xilinx, or the Stratix4 family from \altera can implement a complete
6multi-processor architecture on a single device.
7But the design of embedded system is a long and complex task that requires expertise in software,
8software/hardware partionning, operating system, hardware design, VHDL/Verilog modeling.
9Only very few SMEs have these multiple expertises and are present on the embedded system market.
10\begin{center}\begin{minipage}{.8\linewidth}\textit{
11The major objective of COACH is to provide to SMEs an open-source framework to design
12embedded systems on FPGA devices by system designers.
13}\end{minipage}\end{center}
14%Current design methodologies provide quite low-level abstraction capabilities, and
15%there is an urgent need to leverage system level exploration through the use of a high-level
16%specification of the application and  design space exploration tools.
17%The first system oriented approaches are appearing, among which those
18%based on C/C++ and SystemC are the most popular, but few of them are specifically targetting FPGAs.
19%%%
20\parlf
21The COACH project will leverage on the expertise gained in the field of virtual prototyping
22with the SoCLib platform, to propose a new design flow based on a small number of architectural templates.
23An architectural template is a generic, parameterized architecture, relying on a predefined library
24of IP cores.
25Besides using a specific collection of general purpose IP cores (such as processors cores,
26embedded memory controllers, system bus controllers, I/O and peripheral controllers), each architectural
27template can be enriched by dedicated hardware coprocessors, obtained by high level synthesis (HLS) tools.
28During this project, the COACH partners will develop three different architectural templates:
29\begin{enumerate}
30\item An \altera architectural template based on the \altera IP core library, the AVALON system bus and the NIOS processor.
31\item A \xilinx architectural template based on the \xilinx IP core library, the PLB system bus and the Microblaze processor.
32\item A Neutral architectural template based on the SoCLib IP core library and the VCI/OCP
33      communication infrastructure.
34\end{enumerate}
35The proposed design flow starts from a high level description of the application, specified as a set of
36parallel tasks written in C, without any assumption on the hardware or software implementation
37of these tasks. It lets the system
38designer in charge of expressing the coarse grain parallelism of the application, gives the designer
39the possibility to explore various mapping of the application on the selected template architecture,
40and offers a high predictability of results with respect to cost and performance objectives.
41\\
42When this interactive, system level, design space exploration is completed (converging to
43a specific mapping on a specific version of the selected architectural template), the rest of the flow
44is fully automated: The synthesisable VHDL models for the various hardware components, as well as the binary
45code for the software running on the embedded processors, and the bit-stream to program the the target FPGA
46will be automatically generated by the COACH tools.
47%
48\parlf
49The strength of the COACH approach is the strong integration of the high-level synthesis tools
50in a platform based design flow supporting virtual prototyping and design space exploration.
51Most building blocks already exist (resulting from previous projects): the GAUT
52or UGH synthesis tools, the MUTEKH or DNA embedded operating systems, the ASIP technology,
53the DSX exploration tool, the MWMR hardware/software communication middleware, the BEE parallelisation tool,
54as well as the SoCLib library of systemC simulation models. They must now be enhanced and integrated in
55a consistent design flow.
56%The five academic laboratories worked very closely during more than one year (one monthly meeting
57%in Paris from january 2009 to february 2010, to analyse the issues of interfacing and integrating
58%those various technologies, and to define the detailed architecture of the proposed design flow.
59%%%
60\parlf
61In HPC (High Performance Computing), the targeted application is an existing application
62running on a PC.
63The COACH framework helps designer to accelerate it by migrating critical parts into a
64SoC embedded into an FPGA device plugged to the PC PCI/X bus.
65\begin{center}\begin{minipage}{.8\linewidth}\textit{
66The second objective of COACH is to extend the framework to HPC.
67}\end{minipage}\end{center}
68This will allow SMEs to enter HPC market for the applications that are
69unadapted to the current GPU based solutions.
70%%%
71\parlf
72In summary, the COACH project is clearly oriented toward industry, even if most technology building blocks
73have been previously developed by academic laboratories.
74
75
76%Finally, the key points of the proposed design flow are :
77%\begin{itemize}
78%\item
79%\textbf{System level exploration}: The application coarse grain parallelism
80%is explicitely described as a Tasks and Communication Graph (TCG).
81%A template architecture is selected, and the performances are evaluated
82%on various variant of this architecture using the SoCLib virtual protyping
83%environment. This result in a specific hardware/software partitioning. 
84%This system level exploration is fully controlled by the system designer, and is driven
85%by cost, throughput, latency and power consumption criteria.
86%
87%\item
88%\textbf{High Level Synthesis}: When dedicated hardware coprocessors have been
89%identified as mandatory, they will be generated by the high level synthesis (HLS) tools.
90%The COACH framework will integrate various HLS tools, supporting the micro-architectural space
91%design exploration. Here again, the exploration criteria are cost, throughput, latency
92%and power consumption.
93%At this stage, preliminary source-level transformations and optimisations by front-end
94%tools will be required to improve the efficiency of the back-end HLS tools.
95%
96%\item
97%\textbf{Early performance evaluation}: For each point in the design space,
98%figures of merit must be available such as throughput, latency, power
99%consumption, area, memory allocation and data locality. They are evaluated
100%by reliable estimators obtained by running the actual multi-task software
101%application on the virtual prototype.
102%
103%\item
104%\textbf{Independance from the Target FPGA}: The COACH description of the system
105%(both hardware and software) should be independent of the FPGA family. 
106%Every point of the design space can be implemented on any FPGA component,
107%as long as it contains the hardware ressources required by the selected architectural template.
108%Basically, COACH will support both \altera and \xilinx FPGA families.
109%\end{itemize}
110%
111
112
113
Note: See TracBrowser for help on using the repository browser.