Subversion Repositories aes_pipe

\documentclass[a4paper]{article}

\usepackage{amsmath}

\usepackage{amssymb}

\usepackage{amsfonts}

\usepackage{graphicx}

\DeclareGraphicsExtensions{.pdf,.eps,.png,.jpg}

\usepackage{color}

\usepackage{psfig}

\usepackage{float}

\usepackage{subfigure}

\setlength\topmargin{0 in}

\setlength\oddsidemargin{0 in}

\setlength\textwidth{6.5 in}

\title{Fully Pipelined AES Core}

\author{Subhasis Das}

\date{}

\begin{document}

\maketitle

\section*{Basic Architecture}

This core meets the NIST FIPS-197 specifications. The basic block diagram is given in Figure \ref{arch}.

\begin{figure}[h]

\centering

\includegraphics[scale=0.7]{arch}

\caption{Basic Architecture}

\label{arch}

\end{figure}

I have generated each of the roundkeys in two steps. Let us call

\[

\text{RotWord}(\text{ Sbox}(C_3)\;)\text{ xor RCon}\; = \;f(C_3)

\]

Then, we can see that

\begin{equation*}

\begin{aligned}

C_0^\prime &= f(C_3) \;\text{xor}\; C_0 \\

C_1^\prime &= f(C_3) \;\text{xor}\; C_0 \;\text{xor}\; C_1 \\

C_2^\prime &= f(C_3) \;\text{xor}\; C_0 \;\text{xor}\; C_1 \;\text{xor}\; C_2 \\

C_3^\prime &= f(C_3) \;\text{xor}\; C_0 \;\text{xor}\; C_1 \;\text{xor}\; C_2 \;\text{xor}\; C_3

\end{aligned}

\end{equation*}

where $C_i$ is the column i of the current roundkey and $C_i^\prime$ is the column i of the next roundkey.

This first step of generating $f(C_3)$ is done alongwith the addkey step of the previous cycle and the second step is done in the combined S-Box and ShiftRows step.

The inputs to the overall processor are as follows:

\begin{itemize}

\item clk\_i: System Clock, Data I/O at rising edge

\item rst\_i: Asynchronous Reset, active high, initializes all inputs to all stages and the final output to zero.

\item plaintext\_i: 16$\times$8 bits plaintext input

\item keyblock\_i: 16$\times$8 bits keyblock input

\end{itemize}

The output is

\begin{itemize}

\item ciphertext\_o: 16$\times$8 bits ciphertext output

\end{itemize}

The timing diagram is shown in Figure \ref{clock}.

\begin{figure}[H]

\centering

\includegraphics[scale=0.7]{clock}

\caption{Timing Diagram}

\label{clock}

\end{figure}

The \texttt{trunk/rtl/vhdl} directory contains the whole source code.

The sample testbench is in \texttt{trunk/bench/vhdl}.

For compiling and running the testbench, the script \texttt{sim\_isim.sh} in \texttt{trunk/sim/rtl\_sim/run} directory can be used for Xilinx ISim simulator and \texttt{sim\_ghdl.sh} for GHDL. The testbench takes in plaintext and key data from \texttt{vectors.dat} in \texttt{trunk/sim/rtl\_sim/src} directory. The expected ciphertext data should be present in \texttt{cipher.dat} in \texttt{trunk/sim/rtl\_sim/src} directory. The results are written to \texttt{output.log} in \texttt{trunk/sim/rtl\_sim/log} directory. The final line is 'OK' if all tests pass, else it is 'FAIL'. This can be used to automate checkings over large test datasets.

The \texttt{trunk/syn/Xilinx/run} directory contains the \texttt{synth.sh} shell script, which will synthesize the design when run using Xilinx ISE WebPack tools.

The speed optimized synthesis results with timing driven map on a Xilinx 5VLX50T device is shown in Table \ref{stats}.

\begin{table}[h]

\centering

        \begin{tabular}{|l|l|}

        \hline

        $f_{max}$ & $\approx$ 330 MHz  \\

        \hline

        Max throughput & $\approx$ 42 Gbps  \\

        \hline

        Slice Registers's & 7873 (27\%) \\

        \hline

        Slice LUT's & 14724 (51\%) \\

        \hline

        Bonded IOB's & 386 (80\%) \\

        \hline

        \end{tabular}

        \caption{Design Statistics}

        \label{stats}

\end{table}

All the synthesis, map and place and route logs are available in \texttt{trunk/syn/Xilinx/log} directory.

\end{document}