[12] | 1 | \begin{figure}\leavevmode\center |
---|
| 2 | \includegraphics[width=.8\linewidth]{architecture-csg} |
---|
[21] | 3 | \caption{\label{archi-csg} software architecture for digital system generation} |
---|
[12] | 4 | %\end{figure}\begin{figure}\leavevmode\center |
---|
| 5 | \mbox{}\vspace*{1ex}\\ |
---|
[21] | 6 | \includegraphics[width=1.0\linewidth]{architecture-hls} |
---|
| 7 | \caption{\label{archi-hls} software architecture of hardware accellerator synthesis} |
---|
[12] | 8 | %\end{figure}\begin{figure}\leavevmode\center |
---|
| 9 | \mbox{}\vspace*{1ex}\\ |
---|
| 10 | \includegraphics[width=.8\linewidth]{architecture-hpc} |
---|
| 11 | \caption{\label{archi-hpc} software architecture of HPC} |
---|
| 12 | \end{figure} |
---|
[56] | 13 | %FIXME: la figure ne montre que l'aspect simulation. Intégrer la partie génération (PC API, PCIX, FPGA-IP, bridge vers VCI, SoC API) serait un plus, non ? |
---|
[12] | 14 | % |
---|
[33] | 15 | Figures~\ref{archi-csg}, \ref{archi-hls} and \ref{archi-hpc} |
---|
| 16 | summarize the software architecture of the COACH framework we plan to develop. |
---|
[12] | 17 | In figures, the dotted boxes are the softwares or formats that COACH |
---|
[21] | 18 | has to provide. |
---|
[12] | 19 | \vspace*{.75ex}\par |
---|
[56] | 20 | For the system generation presented in figure~\ref{archi-csg}, the conductor |
---|
[21] | 21 | is the tool \verb!CSG! (COACH System Generator). Its inputs are a process |
---|
| 22 | network describing the application to design and the synthesis parameters. |
---|
| 23 | The main parameters are the target hardware architectural template |
---|
[12] | 24 | with its instanciation parameters, the hardware/software mapping of the |
---|
[21] | 25 | tasks, the FPGA device and design constraints. |
---|
| 26 | \verb+CSG+ thus requires an architectural template library, a operating system |
---|
| 27 | library, two system hardware component (CPU, memories, BUS...) libraries |
---|
| 28 | (one for synthesis, one for simulation). |
---|
| 29 | For generating the coprocessor of a task mapped as hardware, \verb+CSG+ |
---|
| 30 | controls the HAS tools described below. |
---|
| 31 | From these inputs \verb!CSG! can generate the entire system (both software \& |
---|
| 32 | hardware) either as a SystemC simulator to prototype and explore quickly the |
---|
| 33 | design space or as a bitstream\footnote{COACH generates synthesizable VHDL, and |
---|
| 34 | launch the Xilinx or Altera RTL synthesis tools.} directly downloadable on the |
---|
[56] | 35 | FPGA device\footnote{Additional partial bitstreams are generated in case of |
---|
| 36 | dynamic partial reconfiguration}. |
---|
[12] | 37 | \\ |
---|
[21] | 38 | %To proove CSG that COACH is open and CSG is really configurable, COACH will |
---|
| 39 | %basically support 3 architecture template (the COACH template based on a |
---|
| 40 | %MIPS processors and a VCI token ring, the Altera template based on the NIOS |
---|
| 41 | %and AVALON bus, the Xilinx template based on the MICROBLAZE and OPB bus) |
---|
| 42 | %and 2 operating systems (DNA/OS and MUTEK). Furthermore, thus is enforced |
---|
| 43 | %by the \mustbecompleted{FIXME:zied} contribution that consists in |
---|
| 44 | %implementing an other hardware target. |
---|
| 45 | %\\ |
---|
| 46 | %Finally, it is important to notice that this work is a strong |
---|
| 47 | %enhancement of the SocLib software. |
---|
[12] | 48 | \vspace*{.75ex}\par |
---|
[21] | 49 | The software architecture for HAS is presented in figure~\ref{archi-hls}. |
---|
| 50 | The input is a single task of the process network. The HAS tools do not work |
---|
[12] | 51 | directly on the C++ task description but on an internal format called |
---|
[38] | 52 | \xcoach generated by a plugin into the GNU C compiler (GCC). |
---|
| 53 | This allows on the one hand to insure that all the tools will |
---|
[12] | 54 | accept the same C++ description and on the other hand to make possible |
---|
[21] | 55 | their chaining. The front-end tools read a \xcoach description and generate |
---|
| 56 | a new \xcoach description that exibits more parallelism or implement |
---|
| 57 | specific instructions for ASIP. The back-end tools read a \xcoach |
---|
| 58 | description and generate a \xcoachplus description. This is a \xcoach |
---|
| 59 | description anotated with hardware information (scheduling, binding) required by |
---|
| 60 | the VHDL and systemC drivers. |
---|
| 61 | Furthermore, the back-end tools uses a macro-cell library (functional and memory |
---|
| 62 | unit). |
---|
[12] | 63 | \vspace*{.75ex}\par |
---|
[21] | 64 | In addition to digital system design, HPC requires a supplementary |
---|
| 65 | partitioning step presented in figure~\ref{archi-hpc}. The designer |
---|
| 66 | splits the initial application (tag 1) in two parts: one still on the PC and the |
---|
| 67 | other running in a FPGA plugged on the PCI/X PC bus. The two parts exchange data |
---|
| 68 | through communication primitives (tag 2) implemented in a library. |
---|
[56] | 69 | The relevance of the partitioning is evaluated through a |
---|
[21] | 70 | simulator. Once the partitioning is validated, the design of the FPGA part |
---|
| 71 | is done through \verb!CSG! (figure~\ref{archi-csg}). |
---|
[56] | 72 | |
---|
| 73 | |
---|
[12] | 74 | \vspace*{.75ex}\par |
---|
[46] | 75 | \mustbecompleted{FIXME == MODIFICATION DE LA FIGURE} |
---|
[33] | 76 | The project is split into 8 tasks numbered from 0 to 7. |
---|
[12] | 77 | The first task (task 0) is the project management, the last one (task 7) is |
---|
| 78 | the dissemination the other task are listed below: |
---|
| 79 | \begin{enumerate} |
---|
[46] | 80 | \item\textbf{\Backbone:} This task tackles the fundamental points of the |
---|
[21] | 81 | project such as the defintion of the COACH inputs and outputs, |
---|
| 82 | the internal formats (e.g. \xcoach), the architectural templates and |
---|
| 83 | the design flow. |
---|
[46] | 84 | \item\textbf{System generation:} This task addresses the prototyping and |
---|
[21] | 85 | the generation of digital system. Apart from HAS that belong to the task 3 |
---|
| 86 | and 4, its components are those presented figure~\ref{archi-csg} |
---|
| 87 | (e.g. \verb!CSG!, operating systems). |
---|
| 88 | \item\textbf{HAS front-end:} This task mainly focusses on four functionalities: |
---|
| 89 | optimization of the memory usage, parallelism enhancement through loop |
---|
| 90 | transformations, coarse grain parallelization and ASIP generation. |
---|
| 91 | \item\textbf{HAS back-end:} This task groups two functionalities: |
---|
| 92 | High-Level Synthesis of data dominated description and HLS of control |
---|
| 93 | dominated description. |
---|
| 94 | This task contains also the development of a frequency adaptator |
---|
| 95 | that will allow the coprocessors to respect the processor \& the bus |
---|
| 96 | frequency. |
---|
| 97 | \item\textbf{Communication between PC \& FPGA-SoC:} |
---|
| 98 | This task pools the features dedicated to HPC. The main are the |
---|
[56] | 99 | partitioning validation (see figure~\ref{archi-hpc}), the sytem drivers for |
---|
| 100 | both PC and FPGA-SoC sides, the hardware communication components and |
---|
| 101 | support for dynamic partial reconfiguration. |
---|
[23] | 102 | \item\textbf{Demonstrators:} |
---|
[33] | 103 | This task groups the demonstrators of the COACH project. |
---|
[21] | 104 | \mustbecompleted{FIXME} |
---|
[12] | 105 | \end{enumerate} |
---|
[21] | 106 | % |
---|
[12] | 107 | \begin{figure}\leavevmode\center |
---|
[21] | 108 | %\includegraphics[width=.4\linewidth]{dependence-task} |
---|
| 109 | \includegraphics[width=0.70\linewidth]{dependence-task-h} |
---|
| 110 | \caption{\label{dependence-task}Task dependencies} |
---|
[12] | 111 | \end{figure} |
---|
[21] | 112 | Figure~\ref{dependence-task} presents the dependencies between the tasks. |
---|
| 113 | "$task-N \longrightarrow task-M$" means that $task-N$ requires $task-M$ |
---|
| 114 | to work and be demonstrated. The more bold is the arrow, the more important is |
---|
| 115 | the dependency. |
---|
| 116 | The graph shows: |
---|
| 117 | \begin{itemize} |
---|
| 118 | \item Even that $T3$ and $T4$ functionalities are complementary, their |
---|
| 119 | developments are independent (thanks to \xcoach internal format). |
---|
| 120 | \item $T2$ depends slightly from $T3$ and $T4$. Indeed, $T2$ may works |
---|
| 121 | without $T3$ and $T4$ if we limit to digital systems without hardware |
---|
| 122 | accellerators. |
---|
| 123 | \item $T5$ strongly depends on $T2$ but, $T2$ does not depend at all on |
---|
| 124 | $T5$. So demonstrators ($T6$) of embedded system would not be impacted if |
---|
| 125 | $T5$ would fail. |
---|
| 126 | \item $T1$ drives all the tasks ($T2$, $T3$, $T4$, $T5$) at the heart of |
---|
| 127 | the COACH project. |
---|
| 128 | \item $T7$ and $T0$ respectively depends on and impacts all the other tasks. |
---|
| 129 | \end{itemize} |
---|
[33] | 130 | This organisation offers enough robustness to insure the success of the |
---|
| 131 | project except for the specification task $T1$. |
---|
| 132 | |
---|
[37] | 133 | The only critical task in this chart is T1. \label{xcoach-problem} |
---|
[33] | 134 | However, the partners met |
---|
| 135 | 10 times (a one day meeting per month) during the last year to prepare the |
---|
| 136 | specification and the project proposal. This gives us a degree of confidence |
---|
| 137 | that T1 will be completed in time. |
---|