source: anr/section-2.1.tex @ 246

Last change on this file since 246 was 237, checked in by coach, 15 years ago

Paul coquilles

File size: 7.5 KB
Line 
1\begin{table}\leavevmode\center
2\begin{small}\begin{tabular}{|l|l|l|l|}\hline
3Segment                 & 2010   & 2011    & 2012 \\\hline\hline
4Communications          & 1,867  & 1,946   & 2,096 \\
5High end                & 467    & 511     & 550 \\\hline
6Consumer                & 550    & 592     & 672 \\
7High end                & 53     & 62      & 75 \\\hline
8Automotive              & 243    & 286     & 358 \\
9High end                & -      & -       & - \\\hline
10Industrial              & 1,102  & 1,228   & 1,406 \\
11High end                & 177    & 188     & 207 \\\hline
12Military/Aereo          & 566    & 636     & 717 \\
13High end                & 56     & 65      & 82 \\\hline\hline
14Total FPGA/PLD          & 4,659  & 5,015   & 5,583 \\
15Total High-End  FPGA    & 753    & 826     & 914 \\\hline
16\end{tabular}\end{small}
17\caption{\label{fpga_market} Gartner estimation of worldwide FPGA/PLD consumption (Millions \$)}
18\end{table}
19%
20Microelectronic components allow the integration of complex functions into products, increases
21commercial attractivity of these products and improves their competitivity.
22Multimedia and tele-communication sectors have taken advantage from microelectronics facilities
23thanks to the developpment of design methodologies and tools for embedded systems.
24Unfortunately, the Non Recurring Engineering (NRE) costs involded in the design
25and manufacturing ASICs is very high.
26An IC foundry costs several billions of euros and the fabrication of a specific circuit
27costs several millions. For example a conservative estimate for a 65nm ASIC project is 10
28million USD.
29Consequently, it is more and more unaffordable to design and fabricate ASICs for low and medium
30volume markets.
31\parlf
32Today, FPGAs become important actors in the computational domain that was originally dominated
33by microprocessors and ASICs. Just like microprocessors, FPGA based systems can be reprogrammed
34on a per-application basis. For many applications, FPGAs offer significant performance benefits over
35microprocessors implementation. There is still a performance degradation of one order
36of magnitude versus an equivalent ASIC implementations, but low cost
37(500 euros to 10K euros), fast time-to-market and flexibility of FPGAs make them an attractive
38choice for low-to-medium volume applications.
39Since their introduction in the mid eighties, FPGAs evolved from a simple,
40low-capacity gate array to devices (\altera STRATIX III, \xilinx Virtex V) that
41provide a mix of coarse-grained data path units, memory blocks, microprocessor cores,
42on chip A/D conversion, and gate counts by millions. This high logic capacity allows to implement
43complex systems like multi-processors platform with application dedicated coprocessors.
44Table~\ref{fpga_market} shows the estimation of the FPGA worldwide market in the next years in
45various application domains. The ``high end'' lines concern only FPGA with high logic
46capacity for complex system implementations.
47This market is in significant expansion and is estimated to 914\,M\$ in 2012.
48The HPC market size is estimated today by FPGA providers at 214\,M\$.
49Using FPGA limits the NRE costs to the design cost.
50This boosts the developpment of automatic design tools and methodologies.
51%
52%Today, several companies (atipa, blue-arc, Bull, Chelsio, Convey, CRAY, DataDirect, DELL, hp,
53%Wild Systems, IBM, Intel, Microsoft, Myricom, NEC, nvidia etc) are making systems where demand
54%for very high performance (HPC) primes over other requirements. They tend to use the highest
55%performing devices like Multi-core CPUs, GPUs, large FPGAs, custom ICs and the most innovative
56%architectures and algorithms. These companies show up in different "traditional" applications and market
57%segments like computing clusters (ad-hoc), servers and storage, networking and Telecom, ASIC
58%emulation and prototyping, military/aereo etc. The HPC market size is estimated today by FPGA providers
59%at 214\,M\$.
60%%%
61\parlf
62This market is dominated by Multi-core CPUs and GPUs based solutions and the expansion
63of FPGA-based solutions is limited by the lack of design automation.
64Nowadays, there are neither commercial nor academic tools covering the whole design process
65from the system level specification to the bit stream generation.
66% IA to Alain: J'ai remis (et ameliore un peu) ca car sinon le Consequently 20 lignes
67%              au dessous n'a pas de sens.
68% Deplus dans les demandes ANR de la section, il est demande: analyse de la concurrence
69By using SOPC Builder~\cite{spoc-builder} from \altera, designers can select and
70parameterize components from an extensive drop-down list of IP cores (I/O core, DSP,
71processor,  bus core, ...) as well as incorporate their own IP.
72Designers can then generate a synthesized netlist, simulation test bench and custom
73software library that reflect the hardware configuration.
74%% Steven disagree : the C2H compiler bundled with SOPCBuilder does a pretty good job at this.
75%% IA: ces lignes ont ete verifiees et corrigée pa \altera. De plus C2H est plutot limite.
76Nevertheless, SOPC Builder does not provide any facilities to synthesize coprocessors and to
77simulate the platform at a high design level (systemC).
78In addition, SOPC Builder is proprietary and only works together with \altera's Quartus compilation
79tool to implement designs on \altera devices (Stratix, Arria, Cyclone).
80PICO~\cite{pico} and CATAPULT-C~\cite{catapult-c} allow to synthesize
81coprocessors from a C++ description.
82Nevertheless, they can only deal with data dominated applications and they do not handle
83the platform level.
84Similarly, the System Generator for DSP~\cite{system-generateur-for-dsp} is a plug-in to
85Simulink that enables designers to develop high-performance DSP systems for \xilinx FPGAs.
86Designers can design and simulate a system using MATLAB and Simulink. The tool will then
87automatically generate synthesizable Hardware Description Language (HDL) code mapped to
88\xilinx pre-optimized macro-cells.
89However, this tool targets only DSP based algorithms.
90\\
91Consequently, a designer developping an embedded system needs to master four different
92design environments:
93\begin{enumerate}
94  \item a virtual prototyping environment such as SoCLib for system level exploration,
95  \item an architecture compiler (such as SOPC Builder from \altera, or System generator
96  from \xilinx) to define the hardware architecture,
97  \item one or several HLS tools (such as PICO~\cite{pico} or CATAPULT-C~\cite{catapult-c}) for
98        coprocessor synthesis,
99  \item and finally backend synthesis tools (such as Quartus or Synopsys) for the bit-stream generation.
100\end{enumerate}
101Furthermore, mixing these tools requires an important interfacing effort and this makes
102the design process very complex and achievable only by designers skilled in many domains.
103\begin{center}\begin{minipage}{.8\linewidth}\textit{
104The aim of the COACH project is to integrate all these design steps into a single design framework
105and to allow \textbf{pure software} developpers to develop embedded systems.
106}\end{minipage}\end{center}
107\parlf
108We believe that the combination of a design environment dedicated to software developpers
109and FPGA targets,
110will allow small and even very small companies to propose embedded system and accelerating solutions
111for standard software applications with attractive and competitive prices.
112This new market may explode in the same way as the micro-computer market in the eighties,
113whose success was due to the low cost of the first micro-processors (compared to main frames)
114and the advent of high level programming languages which allowed a high number of programmers
115to launch start-ups in software engineering.
Note: See TracBrowser for help on using the repository browser.