OpenCores
URL https://opencores.org/ocsvn/mod_sim_exp/mod_sim_exp/trunk

Subversion Repositories mod_sim_exp

[/] [mod_sim_exp/] [trunk/] [doc/] [src/] [architecture.tex] - Blame information for rev 87

Go to most recent revision | Details | Compare with Previous | View Log

Line No. Rev Author Line
1 47 JonasDC
\chapter{Architecture}
2
 
3
\section{Block diagram}
4
The architecture for the full IP core is shown in the Figure~\ref{blockdiagram}. It consists of 2 major parts, the actual
5
exponentiation core (\verb|mod_sim_exp_core| entity) with a bus interface wrapped around it. In the following sections these
6
different blocks are described in detail.\\
7
\begin{figure}[H]
8
\centering
9
\includegraphics[trim=1.2cm 1.2cm 1.2cm 1.2cm, width=10cm]{pictures/block_diagram.pdf}
10
\caption{Block diagram of the Modular Simultaneous Exponentiation IP core}
11
\label{blockdiagram}
12
\end{figure}
13
\newpage
14
 
15
\section{Exponentiation core}
16
The exponentiation core (\verb|mod_sim_exp_core| entity) is the top level of the modular simultaneous exponentiation
17
core. It is made up by 4 main blocks (Figure~\ref{msec_structure}):\\
18
 
19
\begin{itemize}
20
        \item a pipelined Montgomery multiplier as the main processing unit
21
        \item RAM to store the operands and the modulus
22
        \item a FIFO to store the exponents
23
        \item a control unit which controls the multiplier for the exponentiation and multiplication operations
24
\end{itemize}
25
 
26
\begin{figure}[H]
27
\centering
28
\includegraphics[trim=1.2cm 1.2cm 1.2cm 1.2cm, width=10cm]{pictures/mod_sim_exp_core.pdf}
29
\cprotect\caption{\verb|mod_sim_exp_core| structure}
30
\label{msec_structure}
31
\end{figure}
32
 
33
\subsection{Multiplier}
34
The kernel of this design is a pipelined Montgomery multiplier. A Montgomery multiplication\cite{MontModMul} allows efficient implementation of a
35
modular multiplication without explicitly carrying out the classical modular reduction step. Right-shift operations ensure that the length of the (intermediate) results does not exceed $n+1$ bits. The result of a Montgomery multiplication is given by~(\ref{eq:mont}):
36
\begin{align}\label{eq:mont}
37
r = x \cdot y \cdot R^{-1} \bmod m \hspace{1.5cm}\text{with } R = 2^{n}
38
\end{align}
39
For the structure of the multiplier, the work of \textit{Nedjah and Mourelle}\cite{NedMour} is used as a basis. They show that for large operands ($>$512 bits) the $time\times area$ product is minimal when a systolic implementation is used. This construction is composed of cells that each compute a bit of the (intermediate) result.
40
 
41
Because a fully unrolled two-dimensional systolic implementation would require too many resources, a systolic array (one-dimensional) implementation is chosen. This implies that the intermediate results are fed back to the same same array of cells through a register. A shift register will shift-in a bit of the $x$ operand for every step in the calculation (figure~\ref{mult_structure}). When multiplication is completed, a final check is made to ensure the result is smaller than the modulus. If not, a final reduction with $m$ is necessary.
42
 
43
\textbf{Note:} For this implementation the modulus $m$ has to be uneven to obtain a correct result. However, we can assume that for cryptographic applications, this is the case.
44
 
45
 
46
\begin{figure}[H]
47
\centering
48
\includegraphics[trim=1.2cm 1.2cm 1.2cm 1.2cm, width=15cm]{pictures/mult_structure.pdf}
49
\caption{Multiplier structure. For clarification the $my$ adder and reduction logic are depicted separately, whereas in practice they are internal parts of the stages. (See Figure~\ref{stage_structure})}
50
\label{mult_structure}
51
\end{figure}
52
 
53
\subsubsection{Stage and pipeline structure}
54 78 JonasDC
The Montgomery algorithm uses a series of additions and right shifts to obtain the desired result. The main disadvantage
55
is the carry propagation in the adder, and therefore a pipelined version is used. The length of the operands ($n$) and
56
the number of pipeline stages can be chosen before synthesis. The user has the option to split the pipeline into 2
57
smaller parts so there are 3 operand lengths available during runtime\footnote{e.g. a total pipeline length of 1536 bit
58
split into a part of 512 bit and a part of 1024 bit}.
59 47 JonasDC
 
60 78 JonasDC
The stages and first and last cell logic design are presented in Figure~\ref{stage_structure}. Each stage takes in a
61
part of the modulus $m$ and $y$ operand and for each step of the multiplication, a bit of the $x$ operand is fed to the
62
pipeline (together with the generated $q$ signal), starting with the Least Significant Bit. The systolic array cells
63
need the modulus $m$, the operand $y$ and the sum $m+y$ as an input. The result from the cells is latched into a
64
register, and then passed back to the systolic cells for the next bit of $x$. During this pass the right shift operation
65
is implemented. Each stage thus needs the least significant bit from the next stage to calculate the next step. Final
66
reduction logic is also present in the stages for when the multiplication is complete.
67 47 JonasDC
 
68 78 JonasDC
An example of the standard pipeline structure is presented in Figure~\ref{pipeline_structure}. It is constructed using
69
stages with a predefined width. The first cell logic processes the first bit of the $m$ and $y$ operand and generates
70
the $q$ signal. The last cell logic finishes the reduction and selects the correct result. For operation of this
71
pipeline, it is clear that each stage can only compute a step every 2 clock cycles. This is because the stages rely on
72
the result of the next stage.
73 47 JonasDC
 
74
In Figure~\ref{pipeline_structure_split} an example pipeline design is drawn for a split pipeline. All multiplexers on
75
this figure are controlled by the pipeline select signal (\verb|p_sel|). During runtime the user can choose which part
76
of the pipeline is used, the lower or higher part or the full pipeline.
77
 
78
\newpage
79
\begin{figure}[H]
80
\centering
81
\includegraphics[trim=1.2cm 1.2cm 1.2cm 1.2cm, width=25cm, angle=90]{pictures/sys_stage.pdf}
82
\caption{Pipeline stage and first and last cell logic}
83
\label{stage_structure}
84
\end{figure}
85
\newpage
86
 
87
\newpage
88
\begin{figure}[H]
89
\centering
90
\includegraphics[trim=1.2cm 1.2cm 1.2cm 1.2cm, width=25cm, angle=90]{pictures/sys_pipeline_notsplit.pdf}
91
\caption{Example of the pipeline structure (3 stages)}
92
\label{pipeline_structure}
93
\end{figure}
94
\newpage
95
 
96
\newpage
97
\begin{figure}[H]
98
\centering
99
\includegraphics[trim=1.2cm 1.2cm 1.2cm 1.2cm, width=22cm, angle=90]{pictures/sys_pipeline.pdf}
100
\caption{Example of a split pipeline (1+2 stages)}
101
\label{pipeline_structure_split}
102
\end{figure}
103
\newpage
104
 
105
 
106 78 JonasDC
\subsection{Operand RAM and exponent FIFO} \label{subsec:RAM_and_FIFO}
107
The core's RAM is designed to store 4 operands and a modulus. \footnote{This is the default configuration. The number of operands can be increased, but the control logic is only designed to work with the default configuration.} Three (3) options are available for the implementation of the RAM. Setting the parameter \verb|C_MEM_STYLE|, will change the implementation style. All styles try to use the RAM resources available on the FPGA.
108 47 JonasDC
 
109 87 JonasDC
If the FPGA supports asymmetric RAMs, i.e. with a different read and write width, we suggest that the option \verb|"asym"| is selected. Since the (device specific) RAM blocks are inferred through code, it is imperative to select the right device (\verb|C_FPGA_MAN|), as this inference is different between manufacturers. Currently, only Altera and Xilinx are supported.
110 78 JonasDC
 
111
If there's no asymmetric RAM support, the option \verb|"generic"| should be selected. This option will work for most FPGAs, but the disadvantage is that it will use more resources than the \verb|"asym"| option. This is because a significant number of LUTs will be used to construct an asymmetric RAM.
112
 
113
For both options the size of the RAM adapts dynamically to the chosen pipeline width (\verb|C_NR_BITS_TOTAL|).
114
 
115
Finally, the option \verb|"xil_prim"| is targeted specifically to Xilinx devices. It uses blocks of RAM generated with CoreGen. These blocks are of a fixed width and this results in a fixed RAM of 4x1536 bit for the operands and 1536 bit for the modulus. This option is deprecated in favor of \verb|"asym"|.
116
 
117
Reading and writing (from the bus side) to the operands and modulus is done one 32-bit word at a time. If using a split pipeline, it is important that operands for the higher part of the pipeline are loaded into the RAM with preceding zero's for the lower bits of the pipeline. As a rule of thumb, the number of FPGA RAM blocks that will be used is given by (\ref{eq:ramblocks}):
118
\begin{align}
119
        2 \cdot \mathtt{C\_NR\_BITS\_TOTAL} / 32\label{eq:ramblocks}
120
\end{align}
121
\newline
122
 
123
To store the exponents, there is a FIFO of 32 bit wide. Every 32 bit entry has to be formatted as 16 bit of $e_0$ for the
124
lower part [15:0] and 16 bit of $e_1$ for the higher part [31:16]. Entries have to be pushed in the FIFO starting with the least significant word and ending with the most significant word of the exponents.
125
 
126
For the FIFO there are 2 styles available. The implementation style depends on the style of the operand memory and it can not be set directly. When  the RAM option \verb|"xil_prim"| is chosen, the resulting FIFO will use the FIFO18E1 primitive. It is able to store 512 entries, meaning 2 exponents of each 8192 bit long.
127
 
128
When the RAM options \verb|"generic"| or \verb|"asym"| are chosen, a generic FIFO will be implemented. This consist of a symmetric RAM with the control logic for a FIFO. The depth of this generic FIFO is adjustable with the parameter \verb|C_FIFO_DEPTH|.
129
The number of RAM blocks for the FIFO is given by (\ref{eq:fifoblocks}), where \verb|RAMBLOCK_SIZE| is the size [bits] of the FPGA's RAM primitive.
130
\begin{align}
131
        \left[\left(\mathtt{C\_FIFO\_DEPTH}+1\right) \cdot 32 \right]/ \mathtt{RAMBLOCK\_SIZE} \label{eq:fifoblocks}
132
\end{align}
133
 
134 47 JonasDC
\subsection{Control unit}
135
The control unit loads in the operands and has full control over the multiplier. For single multiplications, it latches in
136
the $x$ operand, then places the $y$ operand on the bus and starts the multiplier. In case of an exponentiation, the FIFO is
137
emptied while the necessary single multiplications are performed. When the computation is done, the ready signal is
138
asserted to notify the system.
139
 
140
\newpage
141
\subsection{IO ports and memory map}
142
The \verb|mod_sim_exp_core| IO ports\\
143
\newline
144
% Table generated by Excel2LaTeX
145
\begin{tabular}{|l|c|c|p{8cm}|}
146
\hline
147
\rowcolor{Gray}
148
\textbf{Port} & \textbf{Width} & \textbf{Direction} & \textbf{Description} \bigstrut\\
149
\hline
150
\verb|clk|   & 1     & in    & core clock input \bigstrut\\
151
\hline
152
\verb|reset| & 1     & in    & reset signal (active high) resets the pipeline, fifo and control logic \bigstrut\\
153
\hline
154
\multicolumn{4}{|l|}{\textbf{\textit{operand memory interface}}} \bigstrut\\
155
\hline
156
\verb|rw_address| & 9     & in    & operand memory read/write address (structure descibed below) \bigstrut\\
157
\hline
158
\verb|data_out| & 32    & out   & operand data out (0 is lsb) \bigstrut\\
159
\hline
160
\verb|data_in| & 32    & in    & operand data in (0 is lsb) \bigstrut\\
161
\hline
162
\verb|write_enable| & 1     & in    & write enable signal, latches \verb|data_in| to operand RAM \bigstrut\\
163
\hline
164
\verb|collision| & 1     & out   & collision output, asserts on a write error \bigstrut\\
165
\hline
166
\multicolumn{4}{|l|}{\textbf{\textit{exponent FIFO interface}}} \bigstrut\\
167
\hline
168
\verb|fifo_din| & 32    & in    & FIFO data in, bits [31:16] for $e_1$ operand and bits [15:0] for $e_0$ operand \bigstrut\\
169
\hline
170
\verb|fifo_push| & 1     & in    & push \verb|fifo_din| into the FIFO \bigstrut\\
171
\hline
172
\verb|fifo_nopush| & 1     & out   & flag to indicate if there was an error pushing the word to the FIFO \bigstrut\\
173
\hline
174
\verb|fifo_full| & 1     & out   & flag to indicate the FIFO is full \bigstrut\\
175
\hline
176
\multicolumn{4}{|l|}{\textbf{\textit{control signals}}} \bigstrut\\
177
\hline
178
\verb|x_sel_single| & 2     & in    & selection for x operand source during single multiplication \bigstrut\\
179
\hline
180
\verb|y_sel_single| & 2     & in    & selection for y operand source during single multiplication \bigstrut\\
181
\hline
182
\verb|dest_op_single| & 2     & in    & selection for the result destination operand for single multiplication \bigstrut\\
183
\hline
184
\verb|p_sel| & 2     & in    & specifies which pipeline part to use for exponentiation / multiplication. \bigstrut[t]\\
185
      &       &       & ``01'' : use lower pipeline part \\
186
      &       &       & ``10'' : use higher pipeline part \\
187
      &       &       & ``11'' : use full pipeline \bigstrut[b]\\
188
\hline
189 78 JonasDC
\verb|modulus_sel| & 1     & in    & selection for which modulus to use for the calculations (only available if \verb|C_MEM_STYLE| = \verb|"generic"| or \verb|"asym"|). Otherwise set to 0 \bigstrut\\
190
\hline
191 47 JonasDC
\verb|exp_m| & 1     & in    & core operation mode. ``0'' for single multiplications and ``1'' for exponentiations \bigstrut\\
192
\hline
193
\verb|start| & 1     & in    & start the calculation for current mode \bigstrut\\
194
\hline
195
\verb|ready| & 1     & out   & indicates the multiplication/exponentiation is done \bigstrut\\
196
\hline
197
\verb|calc_time| & 1     & out   & is high during a multiplication, indicator for used calculation time \bigstrut\\
198
\hline
199
\end{tabular}%
200
\newpage
201 78 JonasDC
The \verb|mod_sim_exp_core| parameters\\
202
\begin{center}
203
        \begin{tabular}{|l|p{6.5cm}|c|l|}
204
                \hline
205
                \rowcolor{Gray}
206
                \textbf{Name} & \textbf{Description} & \textbf{VHDL Type} &\textbf{Default Value} \bigstrut\\
207
                \hline
208
                \verb|C_NR_BITS_TOTAL| & total width of the multiplier in bits & integer & 1536\bigstrut\\
209
                \hline
210
                \verb|C_NR_STAGES_TOTAL| & total number of stages in the pipeline & integer & 96\bigstrut\\
211
                \hline
212
                \verb|C_NR_STAGES_LOW| & number of lower stages in the pipeline, defines the bit-width of the lower pipeline part & integer & 32 \bigstrut\\
213
                \hline
214
                \verb|C_SPLIT_PIPELINE| & option to split the pipeline in 2 parts & boolean & true \bigstrut\\
215
                \hline
216
                \verb|C_FIFO_DEPTH| & depth of the generic FIFO, only applicable if \verb|C_MEM_STYLE| = \verb|"generic"| or \verb|"asym"|  & integer & 32 \bigstrut\\
217
                \hline
218
                \verb|C_MEM_STYLE| & select the RAM memory style (3 options): & string & \verb|"generic"| \bigstrut\\
219
                                                        & \verb|"generic"| : use general 32-bit RAMs & & \\
220
                                                & \verb|"asym"| : use asymmetric RAMs & & \\
221
                                                & (For more information see \ref{subsec:RAM_and_FIFO}) & & \\
222
                                                & \verb|"xil_prim"| : use xilinx primitives & &\\
223
                                                & (deprecated) & & \bigstrut[b] \\
224
                \hline
225 87 JonasDC
                \verb|C_FPGA_MAN| & device manufacturer: & & \\
226 78 JonasDC
                                                & \verb|"xilinx"| or \verb|"altera"| & string & \verb|"xilinx"| \bigstrut\\
227
                \hline
228
        \end{tabular}%
229
\end{center}
230
 
231 47 JonasDC
The following figure describes the structure of the Operand RAM memory, for every operand there is a space of 2048 bits
232 78 JonasDC
reserved. So operand widths up to 2048 bits are supported.\\
233
\newline \\
234 47 JonasDC
\begin{figure}[H]
235
\centering
236
\includegraphics[trim=1.2cm 1.2cm 1.2cm 1.2cm, width=15cm]{pictures/msec_memory.pdf}
237 78 JonasDC
\caption{Address structure of the exponentiation core}
238
\label{Address_structure}
239 47 JonasDC
\end{figure}
240
 
241
\section{Bus interface}
242
The bus interface implements the register necessary for the control unit inputs to the \verb|mod_sim_exp_core| entity.
243
It also maps the memory to the required bus and connects the interrupt signals. The embedded processor then has full control
244
over the core.

powered by: WebSVN 2.1.0

© copyright 1999-2022 OpenCores.org, equivalent to Oliscience, all rights reserved. OpenCores®, registered trademark.