URL
https://opencores.org/ocsvn/cpu_lecture/cpu_lecture/trunk
Subversion Repositories cpu_lecture
[/] [cpu_lecture/] [trunk/] [html/] [09_Toolchain_Setup.html] - Rev 4
Go to most recent revision | Compare with Previous | Blame | View Log
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
"http://www.w3.org/TR/html4/strict.dtd">
<HTML>
<HEAD>
<TITLE>html/Toolchain_Setup</TITLE>
<META NAME="generator" CONTENT="HTML::TextToHTML v2.46">
<LINK REL="stylesheet" TYPE="text/css" HREF="lecture.css">
</HEAD>
<BODY>
<P><table class="ttop"><th class="tpre"><a href="08_IO.html">Previous Lesson</a></th><th class="ttop"><a href="toc.html">Table of Content</a></th><th class="tnxt"><a href="10_Listing_of_alu_vhd.html">Next Lesson</a></th></table>
<hr>
<H1><A NAME="section_1">9 TOOLCHAIN SETUP</A></H1>
<P>In this lesson we will learn how to set up a toolchain on a Linux box.
We will not describe, however, how the tools are downloaded and installed.
The installation of a tools is normally described in the documentation
that comes with the tool.
<P>Places from where tools can be downloaded were already presented in the
first lecture.
<P>The following figure gives an overview of the entire flow. We show source files
in yellow, temporary files in white and tools in green.
<P><br>
<P><img src="toolchain_1.png">
<P><br>
<P>We start with a C source file <STRONG>hello.c</STRONG>. This file is compiled with <STRONG>avr-gcc</STRONG>,
a <STRONG>gcc</STRONG> variant that generates opcodes for the AVR CPU. The compilation
produces 2 output files: <STRONG>hello.lss</STRONG> and <STRONG>hello.hex</STRONG>.
<P><STRONG>hello.lss</STRONG> is a listing file and may optionally be post-processed by the
tool <STRONG>end_conv</STRONG> which converts the little-endian format of <STRONG>hello.lss</STRONG> into
a slightly different format that is more in line with the way gtkwave shows
the hex values of signals.
<P>The main purpose of the compilation is to produce <STRONG>hello.hex</STRONG>. <STRONG>hello.hex</STRONG>
contains the opcodes produced from <STRONG>hello.c</STRONG> in Intel-Hex format.
<P><STRONG>hello.hex</STRONG> is then fed into <STRONG>make_mem</STRONG>. <STRONG>make_mem</STRONG> is a tool that converts
the Intel-Hex format into VHDL constants. These constants are used to
initialize the block RAM modules of the program memory. The output of
<STRONG>make_mem</STRONG> is <STRONG>memory_content.vhd</STRONG> (which, as you certainly remember,
was included by <STRONG>prog_mem.vhd</STRONG>).
<P>At this point, there are two possible ways to proceed. You could do a
functional simulation or a timing simulation.
<H2><A NAME="section_1_1">8.1 Functional Simulation</A></H2>
<P>Initially you will be concerned mostly with the functional simulation.
On this branch you debug the VHDL code until it looks functionally OK.
In order to perform the functional simulation, you need 3 sorts of VHDL files:
<OL>
<LI>The VHLD source files that were discussed in the previous lessons,
<LI>the <STRONG>memory_content.vhd</STRONG> just described, and
<LI>a testbench that mimics the board containing the FPGA to be (<STRONG>test_tb.vhd</STRONG>,
and a VHDL implementation of device specific components used (in our case
this is only <STRONG>RAMB4_S4_S4.vhd</STRONG>). Both files are provided in the directory
called <STRONG>test</STRONG>.
</OL>
<P>All these VHDL files are then processed by <STRONG>ghdl</STRONG>. <STRONG>ghdl</STRONG> produces a single
output file <STRONG>testbench</STRONG> in directory <STRONG>simu</STRONG>. <STRONG>testbench</STRONG> is an executable
file. <STRONG>testbench</STRONG> is then run in order to produces a gzip'ed <STRONG>vcd</STRONG> (value change
dump) file called <STRONG>testbench.vcdgz</STRONG>.
<P>The last step is visualize <STRONG>testbench.vcdgz</STRONG> by means of the tool <STRONG>gtkwave</STRONG>.
<STRONG>gtkwave</STRONG> is similar to the <STRONG>ModelSim</STRONG> provided by Xilinx, but it has
two advantages: it does not bother the user with licence installations
(even in the "free" versions provided by Xilinx) and it runs under Linux.
There are actually more advantages of the <STRONG>ghdl</STRONG>/<STRONG>gtkwave</STRONG> combination;
after having used both tools in the past the author definitely prefers
<STRONG>ghdl</STRONG>/<STRONG>gtkwave</STRONG>.
<P>An example output of the functional simulation that shows the operation
our CPU:
<P><br>
<P><img src="GTKWave.png">
<P><br>
<P>We can compare the CPU signals shown with the assembler code being executed.
The CPU is executing inside the C function <STRONG>uart_puts()</STRONG>:
<P><br>
<pre class="vhdl">
156 00000095: (uart_puts):
157 95: 01AC movw r20, r24
158 96: C003 rjmp 0x9A ; 0x134 <uart_puts+0xa>
159 97: 9B5D sbis 0x0b, 5 ; 11
160 98: CFFE rjmp 0x97 ; 0x12e <uart_puts+0x4>
161 99: B92C out 0x0c, r18 ; 12
162 9A: 01FC movw r30, r24
163 9B: 9601 adiw r24, 0x01 ; 1
164 9C: 9124 lpm r18, Z
165 9D: 2322 and r18, r18
166 9E: F7C1 brne 0x97 ; 0x12e <uart_puts+0x4>
167 9F: 1B84 sub r24, r20
168 A0: 0B95 sbc r25, r21
169 A1: 9701 sbiw r24, 0x01 ; 1
170 A2: 9508 ret
<pre class="filename">
app/hello.lss1
</pre></pre>
<P>
<P><br>
<H2><A NAME="section_1_2">8.2 Timing Simulation and FPGA Configuration</A></H2>
<P>After the CPU functions correctly, the design can be fed into the Xilinx
toolchain. This toolchain is better described in the documentation that
comes with it, so we don't go to too much detail here.
<P>We used Webpack 10.1, which can be downloaded from Xilinx.
<P>The first step is to set up a project in the ISE project navigator with
the proper target device. Then the VHDL files in the <STRONG>src</STRONG> directory are
added to the project. Next the <STRONG>Synthesize</STRONG> and <STRONG>Implementation</STRONG> steps
of the design flow are run.
<P>If this is successful, then we can generate a programming file. There are
a number of ways to configure Xilinx FPGAs, and the type of programming file
needed depends on the particular way of configuring the device. The board
we used for testing the CPU had a serial PROM and therefore we generated
a programming file for the serial PROM on the board. The FPGA would then
load from the PROM on start-up. Other ways of configuring the device are
via JTAG, which is also quite handy during debugging.
<P>The entire build process is a little lengthy (and the devil is known to
hide in the details). We therefore go through the entire design flow in
a step-by-step fashion.
<H2><A NAME="section_1_3">8.3 Downloading and Building the Tools </A></H2>
<UL>
<LI>Download and install <STRONG>ghdl</STRONG>.
<LI>Download and install <STRONG>gtkwave</STRONG>.
<LI>Download and install the Xilinx toolchain.
<LI>Build the <STRONG>make_mem</STRONG> tool. The source is this:
</UL>
<P><br>
<pre class="vhdl">
1 #include "assert.h"
2 #include "stdio.h"
3 #include "stdint.h"
4 #include "string.h"
5
6 const char * hex_file = 0;
7 const char * vhdl_file = 0;
8
9 uint8_t buffer[0x10000];
10
11 //-----------------------------------------------------------------------------
12 uint32_t
13 get_byte(const char * cp)
14 {
15 uint32_t value;
16 const char cc[3] = { cp[0], cp[1], 0 };
17 const int cnt = sscanf(cc, "%X", &value);
18 assert(cnt == 1);
19 return value;
20 }
21 //-----------------------------------------------------------------------------
22 void
23 read_file(FILE * in)
24 {
25 memset(buffer, 0xFF, sizeof(buffer));
26 char line[200];
27 for (;;)
28 {
29 const char * s = fgets(line, sizeof(line) - 2, in);
30 if (s == 0) return;
31 assert(*s++ == ':');
32 const uint32_t len = get_byte(s);
33 const uint32_t ah = get_byte(s + 2);
34 const uint32_t al = get_byte(s + 4);
35 const uint32_t rectype = get_byte(s + 6);
36 const char * d = s + 8;
37 const uint32_t addr = ah << 8 | al;
38
39 uint32_t csum = len + ah + al + rectype;
40 assert((addr + len) <= 0x10000);
41 for (uint32_t l = 0; l < len; ++l)
42 {
43 const uint32_t byte = get_byte(d);
44 d += 2;
45 buffer[addr + l] = byte;
46 csum += byte;
47 }
48
49 csum = 0xFF & -csum;
50 const uint32_t sum = get_byte(d);
51 assert(sum == csum);
52 }
53 }
54 //-----------------------------------------------------------------------------
55 void
56 write_vector(FILE * out, bool odd, uint32_t mem, uint32_t v)
57 {
58 const uint8_t * base = buffer;
59
60 // total memory is 2 even bytes, 2 odd bytes, 2 even bytes, ...
61 //
62 if (odd) base += 2;
63
64 // total memory is 4 kByte organized into 8 memories.
65 // thus each of the 16 vectors covers 256 bytes.
66 //
67 base += v*256;
68
69 // memories 0 and 1 are the low byte of the opcode while
70 // memories 2 and 3 are the high byte.
71 //
72 if (mem >= 2) ++base;
73
74 const char * px = odd ? "po" : "pe";
75 fprintf(out, "constant %s_%u_%2.2X : BIT_VECTOR := X\"", px, mem, v);
76 for (int32_t d = 63; d >= 0; --d)
77 {
78 uint32_t q = base[4*d];
79 if (mem & 1) q >>= 4; // high nibble
80 else q &= 0x0F; // low nibble
81 fprintf(out, "%X", q);
82 }
83
84 fprintf(out, "\";\r\n");
85 }
86 //-----------------------------------------------------------------------------
87 void
88 write_mem(FILE * out, bool odd, uint32_t mem)
89 {
90 const char * px = odd ? "po" : "pe";
91
92 fprintf(out, "-- content of %s_%u --------------------------------------"
93 "--------------------------------------------\r\n", px, mem);
94
95 for (uint32_t v = 0; v < 16; ++v)
96 write_vector(out, odd, mem, v);
97
98 fprintf(out, "\r\n");
99 }
100 //-----------------------------------------------------------------------------
101 void
102 write_file(FILE * out)
103 {
104 fprintf(out,
105 "\r\n"
106 "library IEEE;\r\n"
107 "use IEEE.STD_LOGIC_1164.all;\r\n"
108 "\r\n"
109 "package prog_mem_content is\r\n"
110 "\r\n");
111
112 for (uint32_t m = 0; m < 4; ++m)
113 write_mem(out, false, m);
114
115 for (uint32_t m = 0; m < 4; ++m)
116 write_mem(out, true, m);
117
118 fprintf(out,
119 "end prog_mem_content;\r\n"
120 "\r\n");
121 }
122 //-----------------------------------------------------------------------------
123 int
124 main(int argc, char * argv[])
125 {
126 if (argc > 1) hex_file = argv[1];
127 if (argc > 2) vhdl_file = argv[2];
128
129 FILE * in = stdin;
130 if (hex_file) in = fopen(hex_file, "r");
131 assert(in);
132 read_file(in);
133 fclose(in);
134
135 FILE * out = stdout;
136 if (vhdl_file) out = fopen(vhdl_file, "w");
137 write_file(out);
138 assert(out);
139 }
140 //-----------------------------------------------------------------------------
<pre class="filename">
tools/make_mem.cc
</pre></pre>
<P>
<P><br>
<P>The command to build the tool is:
<P><br>
<pre class="cmd">
# Build makemem.
g++ -o make_mem make_mem.cc
</pre>
<P>
<P><br>
<UL>
<LI>Build the <STRONG>end_conv</STRONG> tool. The source is this:
</UL>
<P><br>
<pre class="vhdl">
1 #include "assert.h"
2 #include "ctype.h"
3 #include "stdio.h"
4 #include "string.h"
5
6 //-----------------------------------------------------------------------------
7 int
8 main(int argc, const char * argv)
9 {
10 char buffer[2000];
11 int pc, val, val2;
12
13 for (;;)
14 {
15 char * s = fgets(buffer, sizeof(buffer) - 2, stdin);
16 if (s == 0) return 0;
17
18 // map lines ' xx:' and 'xxxxxxxx; to 2* the hex value.
19 //
20 if (
21 (isxdigit(s[0]) || s[0] == ' ') &&
22 (isxdigit(s[1]) || s[1] == ' ') &&
23 (isxdigit(s[2]) || s[2] == ' ') &&
24 isxdigit(s[3]) && s[4] == ':') // ' xx:'
25 {
26 assert(1 == sscanf(s, " %x:", &pc));
27 if (pc & 1) printf("%4X+:", pc/2);
28 else printf("%4X:", pc/2);
29 s += 5;
30 }
31 else if (isxdigit(s[0]) && isxdigit(s[1]) && isxdigit(s[2]) &&
32 isxdigit(s[3]) && isxdigit(s[4]) && isxdigit(s[5]) &&
33 isxdigit(s[6]) && isxdigit(s[7])) // 'xxxxxxxx'
34 {
35 assert(1 == sscanf(s, "%x", &pc));
36 if (pc & 1) printf("%8.8X+:", pc/2);
37 else printf("%8.8X:", pc/2);
38 s += 8;
39 }
40 else // other: copy verbatim
41 {
42 printf("%s", s);
43 continue;
44 }
45
46 while (isblank(*s)) printf("%c", *s++);
47
48 // endian swap.
49 //
50 while (isxdigit(s[0]) &&
51 isxdigit(s[1]) &&
52 s[2] == ' ' &&
53 isxdigit(s[3]) &&
54 isxdigit(s[4]) &&
55 s[5] == ' ')
56 {
57 assert(2 == sscanf(s, "%x %x ", &val, &val2));
58 printf("%2.2X%2.2X ", val2, val);
59 s += 6;
60 }
61
62 char * s1 = strstr(s, ".+");
63 char * s2 = strstr(s, ".-");
64 if (s1)
65 {
66 assert(1 == sscanf(s1 + 2, "%d", &val));
67 assert((val & 1) == 0);
68 sprintf(s1, " 0x%X", (pc + val)/2 + 1);
69 printf(s);
70 s = s1 + strlen(s1) + 1;
71 }
72 else if (s2)
73 {
74 assert(1 == sscanf(s2 + 2, "%d", &val));
75 assert((val & 1) == 0);
76 sprintf(s2, " 0x%X", (pc - val)/2 + 1);
77 printf(s);
78 s = s2 + strlen(s2) + 1;
79 }
80
81 printf("%s", s);
82 }
83 }
84 //-----------------------------------------------------------------------------
<pre class="filename">
tools/end_conv.cc
</pre></pre>
<P>
<P><br>
<P>The command to build the tool is:
<P><br>
<pre class="cmd">
# Build end_conv.
g++ -o end_conv end_conv.cc
</pre>
<P>
<P><br>
<H2><A NAME="section_1_4">8.4 Preparing the Memory Content</A></H2>
<P>We write a program <STRONG>hello.c</STRONG> that prints "Hello World" to the serial line.
<P>The source is this:
<P><br>
<pre class="vhdl">
1 #include "stdint.h"
2 #include "avr/io.h"
3 #include "avr/pgmspace.h"
4
5 #undef F_CPU
6 #define F_CPU 25000000UL
7 #include "util/delay.h"
8
9
10 //----------------------------------------------------------------------//
11 // //
12 // print char cc on UART. //
13 // return number of chars printed (i.e. 1). //
14 // //
15 //----------------------------------------------------------------------//
16 uint8_t
17 uart_putc(uint8_t cc)
18 {
19 while ((UCSRA & (1 << UDRE)) == 0) ;
20 UDR = cc;
21 return 1;
22 }
23
24 //----------------------------------------------------------------------//
25 // //
26 // print char cc on 7 segment display. //
27 // return number of chars printed (i.e. 1). //
28 // //
29 //----------------------------------------------------------------------//
30 // The segments of the display are encoded like this:
31 //
32 //
33 // segment PORT B
34 // name Bit number
35 // ----A---- ----0----
36 // | | | |
37 // F B 5 1
38 // | | | |
39 // ----G---- ----6----
40 // | | | |
41 // E C 4 2
42 // | | | |
43 // ----D---- ----3----
44 //
45 //-----------------------------------------------------------------------------
46
47 #define SEG7(G, F, E, D, C, B, A) (~(G<<6|F<<5|E<<4|D<<3|C<<2|B<<1|A))
48
49 uint8_t
50 seg7_putc(uint8_t cc)
51 {
52 uint16_t t;
53
54 switch(cc)
55 { // G F E D C B A
56 case ' ': PORTB = SEG7(0,0,0,0,0,0,0); break;
57 case 'E': PORTB = SEG7(1,1,1,1,0,0,1); break;
58 case 'H': PORTB = SEG7(1,1,1,0,1,1,0); break;
59 case 'L': PORTB = SEG7(0,1,1,1,0,0,0); break;
60 case 'O': PORTB = SEG7(0,1,1,1,1,1,1); break;
61 default: PORTB = SEG7(1,0,0,1,0,0,1); break;
62 }
63
64 // wait 800 + 200 ms. This can be quite boring in simulations,
65 // so we wait only if DIP switch 6 is closed.
66 //
67 if (!(PINB & 0x20)) for (t = 0; t < 800; ++t) _delay_ms(1);
68 PORTB = SEG7(0,0,0,0,0,0,0);
69 if (!(PINB & 0x20)) for (t = 0; t < 200; ++t) _delay_ms(1);
70
71 return 1;
72 }
73
74 //----------------------------------------------------------------------//
75 // //
76 // print string s on UART. //
77 // return number of chars printed. //
78 // //
79 //----------------------------------------------------------------------//
80 uint16_t
81 uart_puts(const char * s)
82 {
83 const char * from = s;
84 uint8_t cc;
85 while ((cc = pgm_read_byte(s++))) uart_putc(cc);
86 return s - from - 1;
87 }
88
89 //----------------------------------------------------------------------//
90 // //
91 // print string s on 7 segment display. //
92 // return number of chars printed. //
93 // //
94 //----------------------------------------------------------------------//
95 uint16_t
96 seg7_puts(const char * s)
97 {
98 const char * from = s;
99 uint8_t cc;
100 while ((cc = pgm_read_byte(s++))) seg7_putc(cc);
101 return s - from - 1;
102 }
103
104 //-----------------------------------------------------------------------------
105 int
106 main(int argc, char * argv[])
107 {
108 for (;;)
109 {
110 if (PINB & 0x40) // DIP switch 7 open.
111 {
112 // print 'Hello world' on UART.
113 uart_puts(PSTR("Hello, World!\r\n"));
114 }
115 else // DIP switch 7 closed.
116 {
117 // print 'HELLO' on 7-segment display
118 seg7_puts(PSTR("HELLO "));
119 }
120 }
121 }
122 //-----------------------------------------------------------------------------
<pre class="filename">
app/hello.c
</pre></pre>
<P>
<P><br>
<P>The commands to create <STRONG>hello.hex</STRONG> and <STRONG>hello.css</STRONG> are:
<P><br>
<pre class="cmd">
# Compile and link hello.c.
avr-gcc -Wall -Os -fpack-struct -fshort-enums -funsigned-char -funsigned-bitfields -mmcu=atmega8 \
-DF_CPU=33333333UL -MMD -MP -MF"main.d" -MT"main.d" -c -o"main.o" "main.c"
avr-gcc -Wl,-Map,AVR_FPGA.map -mmcu=atmega8 -o"AVR_FPGA.elf" ./main.o
# Create an opcode listing.
avr-objdump -h -S AVR_FPGA.elf >"AVR_FPGA.lss"
# Create intel hex file.
avr-objcopy -R .eeprom -O ihex AVR_FPGA.elf "AVR_FPGA.hex"
</pre>
<P>
<P><br>
<P>Create <STRONG>hello.css1</STRONG>, a better readable from of <STRONG>hello.css</STRONG>:
<P><br>
<pre class="cmd">
# Create hello.css1.
./end_conv < hello.css > hello.css1
</pre>
<P>
<P><br>
<P>Create <STRONG>prog_mem_content.vhd</STRONG>.
<P><br>
<pre class="cmd">
# Create prog_mem_content.vhd.
./make_mem < hello.hex > src/prog_mem_content.vhd
</pre>
<P>
<P><br>
<H2><A NAME="section_1_5">8.5 Performing the Functional Simulation</A></H2>
<H3><A NAME="section_1_5_1">8.5.1 Preparing a Testbench</A></H3>
<P>We prepare a testbench in which we instantiate the top-level FPGA design
of the CPU. The test bench provides a clock signal and a reset signal
for the CPU:
<P><br>
<pre class="vhdl">
1 -------------------------------------------------------------------------------
2 --
3 -- Copyright (C) 2009, 2010 Dr. Juergen Sauermann
4 --
5 -- This code is free software: you can redistribute it and/or modify
6 -- it under the terms of the GNU General Public License as published by
7 -- the Free Software Foundation, either version 3 of the License, or
8 -- (at your option) any later version.
9 --
10 -- This code is distributed in the hope that it will be useful,
11 -- but WITHOUT ANY WARRANTY; without even the implied warranty of
12 -- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 -- GNU General Public License for more details.
14 --
15 -- You should have received a copy of the GNU General Public License
16 -- along with this code (see the file named COPYING).
17 -- If not, see http://www.gnu.org/licenses/.
18 --
19 -------------------------------------------------------------------------------
20 -------------------------------------------------------------------------------
21 --
22 -- Module Name: alu - Behavioral
23 -- Create Date: 16:47:24 12/29/2009
24 -- Description: arithmetic logic unit of a CPU
25 --
26 -------------------------------------------------------------------------------
27 --
28 library IEEE;
29 use IEEE.STD_LOGIC_1164.ALL;
30 use IEEE.STD_LOGIC_ARITH.ALL;
31 use IEEE.STD_LOGIC_UNSIGNED.ALL;
32
33 entity testbench is
34 end testbench;
35
36 architecture Behavioral of testbench is
37
38 component avr_fpga
39 port ( I_CLK_100 : in std_logic;
40 I_SWITCH : in std_logic_vector(9 downto 0);
41 I_RX : in std_logic;
42
43 Q_7_SEGMENT : out std_logic_vector(6 downto 0);
44 Q_LEDS : out std_logic_vector(3 downto 0);
45 Q_TX : out std_logic);
46 end component;
47
48 signal L_CLK_100 : std_logic;
49 signal L_LEDS : std_logic_vector(3 downto 0);
50 signal L_7_SEGMENT : std_logic_vector(6 downto 0);
51 signal L_RX : std_logic;
52 signal L_SWITCH : std_logic_vector(9 downto 0);
53 signal L_TX : std_logic;
54
55 signal L_CLK_COUNT : integer := 0;
56
57 begin
58
59 fpga: avr_fpga
60 port map( I_CLK_100 => L_CLK_100,
61 I_SWITCH => L_SWITCH,
62 I_RX => L_RX,
63
64 Q_LEDS => L_LEDS,
65 Q_7_SEGMENT => L_7_SEGMENT,
66 Q_TX => L_TX);
67
68 process -- clock process for CLK_100,
69 begin
70 clock_loop : loop
71 L_CLK_100 <= transport '0';
72 wait for 5 ns;
73
74 L_CLK_100 <= transport '1';
75 wait for 5 ns;
76 end loop clock_loop;
77 end process;
78
79 process(L_CLK_100)
80 begin
81 if (rising_edge(L_CLK_100)) then
82 case L_CLK_COUNT is
83 when 0 => L_SWITCH <= "0011100000"; L_RX <= '0';
84 when 2 => L_SWITCH(9 downto 8) <= "11";
85 when others =>
86 end case;
87 L_CLK_COUNT <= L_CLK_COUNT + 1;
88 end if;
89 end process;
90 end Behavioral;
91
<pre class="filename">
test/test_tb.vhd
</pre></pre>
<P>
<P><br>
<H3><A NAME="section_1_5_2">8.5.2 Defining Memory Modules</A></H3>
<P>We also need a VHDL file that implements the Xilinx primitives that
we use. This is only one: the memory module RAMB4_S4_S4:
<P><br>
<pre class="vhdl">
1 -------------------------------------------------------------------------------
2 --
3 -- Copyright (C) 2009, 2010 Dr. Juergen Sauermann
4 --
5 -- This code is free software: you can redistribute it and/or modify
6 -- it under the terms of the GNU General Public License as published by
7 -- the Free Software Foundation, either version 3 of the License, or
8 -- (at your option) any later version.
9 --
10 -- This code is distributed in the hope that it will be useful,
11 -- but WITHOUT ANY WARRANTY; without even the implied warranty of
12 -- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 -- GNU General Public License for more details.
14 --
15 -- You should have received a copy of the GNU General Public License
16 -- along with this code (see the file named COPYING).
17 -- If not, see http://www.gnu.org/licenses/.
18 --
19 -------------------------------------------------------------------------------
20 -------------------------------------------------------------------------------
21 --
22 -- Module Name: prog_mem - Behavioral
23 -- Create Date: 14:09:04 10/30/2009
24 -- Description: a block memory module
25 --
26 -------------------------------------------------------------------------------
27
28 library IEEE;
29 use IEEE.STD_LOGIC_1164.ALL;
30 use IEEE.STD_LOGIC_ARITH.ALL;
31 use IEEE.STD_LOGIC_UNSIGNED.ALL;
32
33 entity RAMB4_S4_S4 is
34 generic(INIT_00 : bit_vector := X"00000000000000000000000000000000"
35 & "00000000000000000000000000000000";
36 INIT_01 : bit_vector := X"00000000000000000000000000000000"
37 & X"00000000000000000000000000000000";
38 INIT_02 : bit_vector := X"00000000000000000000000000000000"
39 & X"00000000000000000000000000000000";
40 INIT_03 : bit_vector := X"00000000000000000000000000000000"
41 & X"00000000000000000000000000000000";
42 INIT_04 : bit_vector := X"00000000000000000000000000000000"
43 & X"00000000000000000000000000000000";
44 INIT_05 : bit_vector := X"00000000000000000000000000000000"
45 & X"00000000000000000000000000000000";
46 INIT_06 : bit_vector := X"00000000000000000000000000000000"
47 & X"00000000000000000000000000000000";
48 INIT_07 : bit_vector := X"00000000000000000000000000000000"
49 & X"00000000000000000000000000000000";
50 INIT_08 : bit_vector := X"00000000000000000000000000000000"
51 & X"00000000000000000000000000000000";
52 INIT_09 : bit_vector := X"00000000000000000000000000000000"
53 & X"00000000000000000000000000000000";
54 INIT_0A : bit_vector := X"00000000000000000000000000000000"
55 & X"00000000000000000000000000000000";
56 INIT_0B : bit_vector := X"00000000000000000000000000000000"
57 & X"00000000000000000000000000000000";
58 INIT_0C : bit_vector := X"00000000000000000000000000000000"
59 & X"00000000000000000000000000000000";
60 INIT_0D : bit_vector := X"00000000000000000000000000000000"
61 & X"00000000000000000000000000000000";
62 INIT_0E : bit_vector := X"00000000000000000000000000000000"
63 & X"00000000000000000000000000000000";
64 INIT_0F : bit_vector := X"00000000000000000000000000000000"
65 & X"00000000000000000000000000000000");
66
67 port( ADDRA : in std_logic_vector(9 downto 0);
68 ADDRB : in std_logic_vector(9 downto 0);
69 CLKA : in std_ulogic;
70 CLKB : in std_ulogic;
71 DIA : in std_logic_vector(3 downto 0);
72 DIB : in std_logic_vector(3 downto 0);
73 ENA : in std_ulogic;
74 ENB : in std_ulogic;
75 RSTA : in std_ulogic;
76 RSTB : in std_ulogic;
77 WEA : in std_ulogic;
78 WEB : in std_ulogic;
79
80 DOA : out std_logic_vector(3 downto 0);
81 DOB : out std_logic_vector(3 downto 0));
82 end RAMB4_S4_S4;
83
84 architecture Behavioral of RAMB4_S4_S4 is
85
86 function cv(A : bit) return std_logic is
87 begin
88 if (A = '1') then return '1';
89 else return '0';
90 end if;
91 end;
92
93 function cv1(A : std_logic) return bit is
94 begin
95 if (A = '1') then return '1';
96 else return '0';
97 end if;
98 end;
99
100 signal DATA : bit_vector(4095 downto 0) :=
101 INIT_0F & INIT_0E & INIT_0D & INIT_0C & INIT_0B & INIT_0A & INIT_09 & INIT_08 &
102 INIT_07 & INIT_06 & INIT_05 & INIT_04 & INIT_03 & INIT_02 & INIT_01 & INIT_00;
103
104 begin
105
106 process(CLKA, CLKB)
107 begin
108 if (rising_edge(CLKA)) then
109 if (ENA = '1') then
110 DOA(3) <= cv(DATA(conv_integer(ADDRA & "11")));
111 DOA(2) <= cv(DATA(conv_integer(ADDRA & "10")));
112 DOA(1) <= cv(DATA(conv_integer(ADDRA & "01")));
113 DOA(0) <= cv(DATA(conv_integer(ADDRA & "00")));
114 if (WEA = '1') then
115 DATA(conv_integer(ADDRA & "11")) <= cv1(DIA(3));
116 DATA(conv_integer(ADDRA & "10")) <= cv1(DIA(2));
117 DATA(conv_integer(ADDRA & "01")) <= cv1(DIA(1));
118 DATA(conv_integer(ADDRA & "00")) <= cv1(DIA(0));
119 end if;
120 end if;
121 end if;
122
123 if (rising_edge(CLKB)) then
124 if (ENB = '1') then
125 DOB(3) <= cv(DATA(conv_integer(ADDRB & "11")));
126 DOB(2) <= cv(DATA(conv_integer(ADDRB & "10")));
127 DOB(1) <= cv(DATA(conv_integer(ADDRB & "01")));
128 DOB(0) <= cv(DATA(conv_integer(ADDRB & "00")));
129 if (WEB = '1') then
130 DATA(conv_integer(ADDRB & "11")) <= cv1(DIB(3));
131 DATA(conv_integer(ADDRB & "10")) <= cv1(DIB(2));
132 DATA(conv_integer(ADDRB & "01")) <= cv1(DIB(1));
133 DATA(conv_integer(ADDRB & "00")) <= cv1(DIB(0));
134 end if;
135 end if;
136 end if;
137 end process;
138
139 end Behavioral;
140
<pre class="filename">
test/RAMB4_S4_S4.vhd
</pre></pre>
<P>
<P><br>
<H3><A NAME="section_1_5_3">8.5.3 Creating the testbench executable</A></H3>
<P>We assume the following file structure:
<UL>
<LI>a <STRONG>test</STRONG> directory that contains the testbench (<STRONG>test_tb.vhd</STRONG>) and the
memory module (<STRONG>RAMB4_S4_S4.vhd</STRONG>).
<LI>a <STRONG>src</STRONG> directory that contains all other VHDL files.
<LI>a <STRONG>simu</STRONG> directory (empty).
<LI>A <STRONG>Makefile</STRONG> like this:
</UL>
<P><br>
<pre class="vhdl">
1 PROJECT=avr_core
2
3 # the vhdl source files (except testbench)
4 #
5 FILES += src/*.vhd
6
7 # the testbench sources and binary.
8 #
9 SIMFILES = test/test_tb.vhd test/RAMB4_S4_S4.vhd
10 SIMTOP = testbench
11
12 # When to stop the simulation
13 #
14 # GHDL_SIM_OPT = --assert-level=error
15 GHDL_SIM_OPT = --stop-time=40us
16
17 SIMDIR = simu
18
19 FLAGS = --ieee=synopsys --warn-no-vital-generic -fexplicit --std=93c
20
21 all:
22 make compile
23 make run 2>& 1 | grep -v std_logic_arith
24 make view
25
26 compile:
27 @mkdir -p simu
28 @echo -----------------------------------------------------------------
29 ghdl -i $(FLAGS) --workdir=simu --work=work $(SIMFILES) $(FILES)
30 @echo
31 @echo -----------------------------------------------------------------
32 ghdl -m $(FLAGS) --workdir=simu --work=work $(SIMTOP)
33 @echo
34 @mv $(SIMTOP) simu/$(SIMTOP)
35
36 run:
37 @$(SIMDIR)/$(SIMTOP) $(GHDL_SIM_OPT) --vcdgz=$(SIMDIR)/$(SIMTOP).vcdgz
38
39 view:
40 gunzip --stdout $(SIMDIR)/$(SIMTOP).vcdgz | gtkwave --vcd gtkwave.save
41
42 clean:
43 ghdl --clean --workdir=simu
44
<pre class="filename">
Makefile
</pre></pre>
<P>
<P><br>
<DL>
<DT>Then</DT>
<DD>
</DL>
<P><br>
<pre class="cmd">
# Run the functional simulation.
make
</pre>
<P>
<P><br>
<P>It will take a moment, but then a <STRONG>gtkwave</STRONG> window like the one shown
earlier in this lesson will appear. It my look a little different due
due to different default settings (like background color). In that
window you can add new signals from the design that you would like
to investigate, remove signals you are not interested in, and so on.
At the first time, no signals will be shown; you can add some by selecting
a component instance at the right, selecting a signal in that component,
and then pushing the <STRONG>append</STRONG> button on the right.
<P>The <STRONG>make</STRONG> command has actually made 3 things:
<UL>
<LI>make compile (compile the VHLD files)
<LI>make run (run the simulation), and
<LI>make view
</UL>
<P>The first two steps (which took most of the total time) need only be run
after changes to the VHDL files.
<H2><A NAME="section_1_6">8.6 Building the Design</A></H2>
<P>When the functional simulation looks OK, it is time to implement the design
and check the timing. We describe this only briefly, since the Xilinx
documentation of the Xilinx toolchain is a much better source of
information.
<H3><A NAME="section_1_6_1">8.6.1 Creating an UCF file</A></H3>
<P>Before implementing the design, we need an <STRONG>UCF</STRONG> file. That file
describes timing requirements, pin properties (like pull-ups for our
DIP switch), and pin-to-signal mappings:
<P><br>
<pre class="vhdl">
1 NET I_CLK_100 PERIOD = 10 ns;
2 NET L_CLK PERIOD = 35 ns;
3
4 NET I_CLK_100 TNM_NET = I_CLK_100;
5 NET L_CLK TNM_NET = L_CLK;
6
7 NET I_CLK_100 LOC = AA12;
8 NET I_RX LOC = M3;
9 NET Q_TX LOC = M4;
10
11 # 7 segment LED display
12 #
13 NET Q_7_SEGMENT<0> LOC = V3;
14 NET Q_7_SEGMENT<1> LOC = V4;
15 NET Q_7_SEGMENT<2> LOC = W3;
16 NET Q_7_SEGMENT<3> LOC = T4;
17 NET Q_7_SEGMENT<4> LOC = T3;
18 NET Q_7_SEGMENT<5> LOC = U3;
19 NET Q_7_SEGMENT<6> LOC = U4;
20
21 # single LEDs
22 #
23 NET Q_LEDS<0> LOC = N1;
24 NET Q_LEDS<1> LOC = N2;
25 NET Q_LEDS<2> LOC = P1;
26 NET Q_LEDS<3> LOC = P2;
27
28 # DIP switch(0 ... 7) and two pushbuttons (8, 9)
29 #
30 NET I_SWITCH<0> LOC = H2;
31 NET I_SWITCH<1> LOC = H1;
32 NET I_SWITCH<2> LOC = J2;
33 NET I_SWITCH<3> LOC = J1;
34 NET I_SWITCH<4> LOC = K2;
35 NET I_SWITCH<5> LOC = K1;
36 NET I_SWITCH<6> LOC = L2;
37 NET I_SWITCH<7> LOC = L1;
38 NET I_SWITCH<8> LOC = R1;
39 NET I_SWITCH<9> LOC = R2;
40
41 NET I_SWITCH<*> PULLUP;
42
<pre class="filename">
src/avr_fpga.ucf
</pre></pre>
<P>
<P><br>
<H3><A NAME="section_1_6_2">8.6.2 Synthesis and Implementation</A></H3>
<UL>
<LI>Start the ISE project manager and open a new project with the desired
FPGA device.
<LI>Add the VHDL files and the <STRONG>UCF</STRONG> file in the <STRONG>src</STRONG> directory to the
project (Project->Add Source).
<LI>Synthesize and implement the design (Process->Implement top Module).
</UL>
<P>This generates a number of reports, netlists, and other files.
There should be no errors. There will be warnings though, including
timing constraints that are not met.
<P>It is important to understand the reason for each warning. Warnings often
point to faults in the design.
<P>The next thing to check is the timing reports. We were lucky:
<P><br>
<pre class="cmd">
#Timing report fragment:
================================================================================
Timing constraint: NET "L_CLK" PERIOD = 35 ns HIGH 50%;
676756190 paths analyzed, 2342 endpoints analyzed, 0 failing endpoints
0 timing errors detected. (0 setup errors, 0 hold errors)
Minimum period is 34.981ns.
--------------------------------------------------------------------------------
================================================================================
Timing constraint: NET "I_CLK_100_BUFGP/IBUFG" PERIOD = 10 ns HIGH 50%;
19 paths analyzed, 11 endpoints analyzed, 0 failing endpoints
0 timing errors detected. (0 setup errors, 0 hold errors)
Minimum period is 3.751ns.
--------------------------------------------------------------------------------
All constraints were met.
</pre>
<P>
<P><br>
<P>This tells us that we have enough slack on the crystal CLK_100 signal
(8.048ns would allow for up to 124 MHz). We had specified a period
of 35 ns irequirement for the CPU clock:
<P><br>
<pre class="vhdl">
2 NET L_CLK PERIOD = 35 ns;
<pre class="filename">
src/avr_fpga.ucf
</pre></pre>
<P>
<P><br>
<P>The CPU runs at 25 MHz, or 40 ns. The 35 ns come from the 40 ms minus
a slack of 5 ns. With some tweaking of optimization options, we could
have reached 33 MHz, but then the slack would have been pretty small.
<P>However, we rather stay on th safe side.
<H2><A NAME="section_1_7">8.7 Creating a Programming File</A></H2>
<P>Next we double-click "Generate Programming file" in the ISE project navigator.
This generates a file <STRONG>avr_fpga.bit</STRONG> in the project directory. This can also
be run from a Makefile or from the command line (the command is <STRONG>bitgen</STRONG>).
<H2><A NAME="section_1_8">8.8 Configuring the FPGA</A></H2>
<P>At this point, we have the choice between configuring the FPGA directly
via JTAG, or flashing an EEPROM and then loading the FPGA from the EEPROM.
<H3><A NAME="section_1_8_1">8.8.1 Configuring the FPGA via JTAG Boundary Scan</A></H3>
<P>Configuring the FPGA can be done with the Xilinx tool called <STRONG>impact</STRONG>.
The file needed by <STRONG>impact</STRONG> is <STRONG>avr_fpga.bit</STRONG> from above. The configuration
loaded via JTAG will be lost when the FPGA looses power.
<P>Choose "Boundary Scan" in <STRONG>impact</STRONG>, select the FPGA and follow the instructions.
<H3><A NAME="section_1_8_2">8.8.2 Flashing PROMs</A></H3>
<P>In theory this can also be done from ISE. In practice it could (and actually
did) happen that the programming cable (I use an old parallel 3 cable)
is not detected by impact.
<P>Before flashing the PROM, the <STRONG>avr_fpga.bit</STRONG> from the previous step needs to
translated into a format suitable for the PROM. My PROM is of the serial
variety, so I start <STRONG>impact</STRONG>, choose "PROM File Formatter" and follow the
instructions.
<P>After converting <STRONG>avr_fpga.bit</STRONG> into, for example, <STRONG>avr_fpga.mcs</STRONG>, the
PROM can be flashed. Like before choose "Boundary Scan" in #impact. This
time, however, you select the PROM and not the FPGA, and follow the
instructions.
<P>This concludes the description of the design flow and also of the CPU.
The remaining lessons contain the complete listings of all sources files
discussed in this lectures.
<P>Thank you very much for your attention.
<P><hr><BR>
<table class="ttop"><th class="tpre"><a href="08_IO.html">Previous Lesson</a></th><th class="ttop"><a href="toc.html">Table of Content</a></th><th class="tnxt"><a href="10_Listing_of_alu_vhd.html">Next Lesson</a></th></table>
</BODY>
</HTML>
Go to most recent revision | Compare with Previous | Blame | View Log