Labs 3 and 4 16- bit Processor

Objectives
Work in Xilinx Vivado and design a simplified processor in Verilog. The full architecture to be
implemented is described at the end of this document. The steps for the lab will guide you through the
design of the instruction decode, ALU, register file, branch comparator, and program counter blocks.
Several test programs will be provided.
Inputs/Outputs
The individual inputs and outputs ofthe top processor module are already connected and are described
below. These are the same inputs and outputs as the previous labs, though they are used differently;
besides the clock, all of the inputs and outputs are part of the data memory map for the processor and
are accessible with load and store operations. For more details, see the Appendix: Memory Map below.
Inputs to the Processor
• CLK – the onboard 100Mhz clock
o As a non-pipelined processor, your code may notrun at 100MHz. A clock enable signal is
provided to each ofthe sequential logic blocks to reduce the system clock frequency.
o The clk_en signal divides the system clock frequency down to about 24 kHz –this is
fast enough for the lab and all test programs provided.
• SW[7:0] – the eight board switches
o The switches are accessible as a memory mapped register, by reading from data address
0x8000
• BTN[4:0] – the five board push buttons
• CPU_RESETN – CPU Reset N (Active Low)
o Resets all CPU state (registers, PC, etc.) and begins code execution again from program
address 0x0
Outputs of the Processor
• LED[7:0] – the eight board LEDs
o The LED bits are accessible as a memory mapped register, exercised by writing to data
address 0x8000.
§ Yes, this is the same data address as the switches. As one is accessible when
reading from 0x8000 and one when writing to 0x8000,there is no conflict.
• SEG[6:0], DP, AN[3:0] – the seven segment display
o Accessible as a memory mapped register by writing to 0x9000. The seven-segment
display formatting is already implemented, as in Lab 2.
Provided files
Download processor.v, data_flow.v, instruction_mem.v, data_mem.v, and Nexys4DDR_Master.xdc
from the course webpage.
In this lab, the top-level processor.v is already complete. Please read and become familiar with the
modules.
Additionally, data_mem.v is already complete. The instruction_mem.v file has several example
programs written into it that are commented out –these programs can be used at different steps in the
lab below to test that your code is working
The data_flow.v file is where the bulk of the work in this lab will be written. All of these modules are
instantiated and connected in the processor.v file; your task will be completing the five data flow
modules.Eachofthesemodules,exceptfortheinstruction_decodemodule(Step1),defaults todriving
useful values on its outputs so that the system may be tested at the end of each step – e.g.,the register
file module, before Step 2, drives a constant 0x9000 as the read value for all registers, allowing the
seven segment display to be written afterimplementing only the instruction_decode submodule.
Besides the provided Verilog and constraint files, an assembler is also provided. It is java based, so
should run on your own computers or the lab machines, and three test programs are provided,
corresponding to the three test programs coded into the instruction_mem.v file. The assembler is
provided if you wish to write your own test programs.
Step 1: Instruction Decode
The instruction_decode submodule takes one input (the 16 bit instruction) and drives twenty-eight
outputs. All logic in the instruction decode block should be combinational;there is no clock inputto the
module. The full description of the instructions is at the end of the document in the “Appendix:
Instruction Set”, but here is a brief description of some of the output signals:
• alu_func: the FUNC bits of the 1-input and 2-input arithmetic operations
• destination_reg:the Rd bits for all operations. This should be the same set of bits regardless of
op-code; all operations thatwrite to the registerfile use this as the destination register, and the
STORE command uses these bits as the data register (so both Load and Store use Rd as the
register to load from or store to the data memory)
• source_reg1/2: These bits are consistent across most op-codes. The only exceptions are the
branch op-codes – make sure that when assigning the bits for the two source_reg outputs
that you pay attention to the alignment in the arithmetic op-codes vs the branch op-codes.
• Immediate: These bits represent an integer to be used directly rather than a reference to a
register (e.g. ADD sums the values in two registers, ADDI sums the value in a register with a 6-
bit immediate). They are always the least significant bits of any instruction. Sign extension
should not be performed in the instruction_decode module, but should be left to other blocks
For instance, the branch/jump immediate values are sign extended, but it is expected that
this will occur in the program counter block. In fact, the ONLY instructions for which the
immediate is sign extended are branches and jumps.
All other outputs should be self-explanatory, and are Boolean values based upon a single op-code or opcode+function. Referto the instruction set Appendix forfurther details. For decoding purposes, it may
be usefulfor you to use the `define constants provided atthe beginning ofthe data_flow.v file.
When you have finishedyourinstruction_decodeblock, you should be able to execute a simple program
thatdoes notrequire theALU,registerfile, or any branches.This is the first program initially provided to
you in the instruction_mem file. After programming your board, “CAFE” should be displayed on
the seven-segment display.
At this stage you will see quite a few warnings about unused signals or logic removed during synthesis.
This is expected, as much of the logic for the processor is not being exercised yet.
Step 2: Register File
The register file module, reg_file, is a block of eight 16-bit registers.
A description of its inputs and outputs follows:
• clk & clk_en: the 100 MHz clock and a clock enable signal. The data written into a register
should only occur on the clock edge when the clk_en signal is active
o Effectively,this is providing a slower clock to this and the other sequential blocks in the
processor. This is a method of dividing down a clock to run logic at a slowerfrequency.
• reset: Active high. When this input is a 1, reset the registers to all zeros.
• source_reg1: A 3-bit identifier for a register.
• reg1_data: The data corresponding to the register # provided by the source_reg1 input.
This should NOT be registered as an output, and should be combinational based on the
input source_reg1 value.
• source_reg2/reg2_data: Analogousto source_reg1/reg1_data
• destination_reg: A 3-bit identifier for a register
• wr_destination_reg: When true, an instruction is writing to a register.ONLY when this signal is
true should the values in any of the registers change.
• dest_result_data:The 16-bitresult of an instruction (ALUoutput or data_mem output) that is
tobewrittentotheregister specifiedbydestination_reg if wr_destination_reg is true.
• regD_data: Analogous to reg1_data or reg2_data, but for the destination_reg input. This is
the value to be written by a STORE instruction.
• Immediate: An 8-bit immediate value. Only used for a MOVIL/MOVIH instruction.
• movi_lower: The MOVIL instruction. Copies the 8-bit immediate input into the lower byte of
• the destination register.
• movi_higher: The MOVIH instruction. Copies the 8-bit immediate input into the upper byte of
the destination register.
It is advisable for this module to have two sections of code – combinational logic that reads the
values for the data outputs (reg1_data/reg2_data/regD_data), and sequential logic that
handles the register write-back logic. A write-back to a register should only occur if clk_en is true and
wr_destination_reg is TRUE. The movi_lower and movi_higher commands take precedence over
dest_result_data (if movi_lower or movi_higher is true, the dest_result_data is not valid).
When you have finished your reg_file block, you should be able to execute the first simple test program
again.Afterprogramming yourboard, “BEEF” should be displayed on the seven-segmentdisplay.Itis the
same test program as Step 1, but with the additional functionality of the register file, it will no longer
display the previous tie-off value of “CAFÉ”, but the values copied into the registers with the
MOVIH/L instructions.
Step 3: ALU
The arithmetic logic unit (ALU) is primarily combinational logic. There are two registered logic values,
however – the borrow bit and carry bit.
The carry and borrow bits are used in 16-bit systems like the one you are designing to allow for 32-bit
logic; successive ADDC commands allow the carry bit from one addition to be propagated to the next
one. The carry bit can be set by any addition operation (ADD, ADDC, ADDI), and the borrow bit can
be set by any subtraction operation (SUB, SUBB, SUBI), butthe bits are only used during ADDC or SUBB
instructions.
A description of the inputs and outputs:
• clk/clk_en/reset: The same function they served in the reg_file block. The clk and clk_en
should be used when registering the carry and borrow bits, and the reset input should be
used to set both of those state bits to zero.
• alu_result: The main output of the ALU. This 16-bit value should be the computational result
for the currentinstruction.
• arith_1op/arith_2op/addi/subi: Each ofthese 1-bit signals is TRUE if and only if it holds for
the currentinstruction op-code
• load_or_store: True if the current instruction op-code is either LOAD or STORE. In this
case, the alu_result should be the same as addi, except that the carry bit should NOT be
modified. The ALU is used for load/store instructions to calculate the address in the data
memory; these instructions allow an immediate offset from a register so that data can be
accessed by specifying an offset from a fixed base register.
• reg1_data: The first operand for a 2op instruction, or the only operand for a 1op instruction,
add/subtract immediate, or LOAD/STORE.
• reg2_data: The second operand for a 2op instruction.
• Immediate: The six bit immediate for add/subtract immediate, or for the LOAD/STORE
address offset.
• stc_cmd/stb_cmd: Command instruction to set carry bit or borrow bit, respectively.
• alu_carry_bit: The carry bit ofthe ALU, output so it may be used in the branch comparator.
Most of the code for this block should be combinational. You should have a separate sequential block
for the carry/borrow bit logic. Once you have finished the ALU implementation, comment out the
first test program in instruction_mem.v, and uncommentthe test program labeled as “Test Program for
Step 3”. The display outputforthe step 3 test program should read 0xDFF8 on the seven-segment display.
Step 4: Branch Comparator and Program Counter
The final step to complete the basic processor involves two modules: the branch comparator and the
program counter. The two cannot be tested separately, as you are implementing two things that are
interrelated: determining whether a branch is taken, and updating the PC so that a branch or jump can
have effect.
The first partis fairly simple –the branch comparator has fourinputs forthe four branch op-codes,input
data to determine branch outcome, and a single output (branch_taken) that should be true if a
branch op-code is active and its conditions are true.
The four branch instructions are:
• branch_eq: Branch if the two reg_data inputs are equal
• branch_ge: Branch if reg1 is greater than or equal to reg2
• branch_le: Branch if reg1 input is less than or equal to reg2
• branch_carry: Branch ifthe ALU carry bitis set. This is particularly usefulfor overflow checks.
There is no way to directly test that this code is working until you have also updated the program
counter module, as the branch_taken flag is an input to the program_counter module that you will
need to handle before changes to the branch comparator module have any effect on the system.
The sequential logic forthe program_counter module is already presentin the module, and in the code
provided to you, pc_next is always set to the current pc + 2 (2, because the instructions are 2
bytes wide).
You will need to do two things to finish the program_counter module. First, you will need to sign extend
the branch and jump immediate signals to 16-bits. Then you will need to modify the code that
determines pc_next based upon the jump_taken and branch_taken inputs.
IMPORTANT NOTE: even when a branch or jump is taken, the pc still increments by 2. This is fairly
standard across architectures, usually to aid in pipelining. Itis expected here as well, and the finaltest
code will not work correctly if you do not do this. Ensure that no matter what you add to the pc to get
pc_next, you also add 2. (e.g. “BEQ $1 $1 0” should advance to the next command, but “BEQ $1 $1 -2”
should be an infinite loop)
There is a finaltest program you can execute atthis pointto demonstrate your working processorin the
lab. Comment out the test code for the earlier lab steps and uncomment the code below the comment
“Final Test Code.”
When running the final test program, you should observe the following:
• The switch values should be displayed on the LEDs - there is a loop that reads the switches and
writes the values back to the LEDs.
• The 7-segment display should output a Fibonacci sequence that increments every second.
• When the Fibbonacci sequence exceeds what can be shown on the 7-segment display, the
CPU halts (so the switches will no longer have any effect on the LEDs).
o The value when the CPU halts should be 0xB520
o Immediately before halting,the program sets the ‘blink’flag at data address 0x9002.
Once the 7-segment display starts blinking 0xB520,the switches should have no
effect.
o Once halted (or any point, really), CPU_RESET can be pressed to reset the system to
start all overagain.
Step 5: Processor Additions and Assembly
Choose one ofthe options below. In yourreport, describe the tradeoffs ofimplementing in Assembly on
this custom processor instead of a pure Verilog based implementation for your chosen option.
1. Assembly. Implement multiplication as a software program in assembly on your new processor.
Read values from the sliding switches SW[7:0]. Then perform a 4-bit by 4-bit multiplication
using operands from the sliding switches values as a simple software addition loop. Show the
result on the right set of seven-segment displays. You should allow for a new set of inputs to
be set by changing the switches and re-running your multiply program.
2. Buttons. Add support to enable reading the values ofthe push buttons. Write a short assembly
program to (i) perform addition and subtraction for two 4-bit operands read from the
sliding switches SW[7:0] (memory map 0x9000) and (ii) read the system clock (memory map
0xF000). Use buttons BTN[2:0] to select the operation +,-, or show system clock. Simply
enter a continuous loop, reading the button and switch values and showing the results of the
computation on the on the right set of seven-segment displays. To read values from the
buttons, alter data_mem.v so that processor memory address 0x9006 reads from the BTN
values.
3. Extended I/O. Add support to the enable the set of the left four seven-segment displays, left 8
sliders and LEDs. Write an assembly program that demonstrates them working such as reading
switches values and displaying them on the seven segment displays while looping overthe LEDs
or showing the NOT of the switches on the LEDs. To read/write additional I/O, alter
data_mem.v as outlined in the Appendix Memory Map, Extended I/O. This will require
altering the data_mem module to map the additional addresses and extend the data bus,
as well as modifying the processor.v to also extend the data. You must also uncomment the
associated lines in the design constraintfile (Nexys4DDR_Master.xdc)to enable the additional
switchesand LEDs.
4. Colors, Tri-Color LED, Bonus +5. The Nexys 4 board contains two tri-color LED's. Read sections
10.2 and 16.1 in the Nexys 4 DDR Manual. Construct a new Verilog module from scratch that
receives CLK and 8-bit Red, Green, and Blue inputs, and outputs three values to one ofthe tri-
color LED's pins using pulse width modulation (PWM). Create a simple Assembly program
thatrotates through various color combinations or changes colors based on the sliding switches.
Do not simply supply constant power to the RGB LED in your custom module. The module
should take each component color value and pulse each output corresponding to the eight-bit
value - make the value 255 pulse 50% of the time to temper LED brightness, so a value of 128
should pulse the component color high for 25% of the time. Enable the RGB LED in the design
constraint file (LED16_R, LED16_G, LED16_B). Alter the data_mem.v and processor.v files
to pass component colors as arguments to the data_mem module. Create a new Verilog
module in processor.v that receives these component colors on a memory map write to
addresses 0x9010, 0x9012 or 0x9014, and saves the color components in a register. Use the
system CLK to count from 0 to 512, for each component color, enable the component color
on only if the value is greater than the current count.
5. Bonus +5. Combine Assembly and Buttons (options 1 and 2) to implement a variant of lab 1 in
software assembly code. Use buttons BTN[3:0].
6. Bonus +10. Combine 1, 2, and 3 into an 8-bit calculator. Implement the Extended I/O (option
3). Then add Button (option 2) support and write an 8-bit calculator in Assembly (option 1) as
above except using 16 bit values from the switches for the two eight bit numbers. The LEDs
should show the values of the switches, the left seven-segment displays should show the two
eight-bit numbers and the right seven-segment display should show the result after pushing one
of the four buttons BTN[3:0].
Lab Report
Follow the report outline for previous labs. Only include code from your Step 4 and your custom
assembly program from Step 5. A lab grader should sign off on Step 4 and Step 5.
In your report, describe your Step 5 and the tradeoffs of implementing in Assembly on this custom
processor instead of a pure Verilog based implementation for your chosen option.
Document the role/contribution of each lab partner.
For this lab capture 4 Screenshots and Implementation Utilization.
Screenshot 1. RTL Analysis - Schematic. Identify the major system components. Relate to the in-class
discussion processor components.
Screenshot 2. Synthesis – Synthesize – Schematic. Make sure you close previous schematics and are
viewing the complete design. Capture a screenshot and comment on the complexity.
Screenshot 3. Synthesis – Synthesize – Schematic – Expand ALU. Comment on the ALU complexity
compared to the calculator in Lab 1.
Screenshot 4. Device Implemented Design. Make sure to close previous schematics and refresh
Implementation and select the top-most module to get the latest design or your changes will not show
up. Compare complexity to other labs (light blue areas utilized).
Utilization. Report the total usage of components on the FPGA based on your design. (Implementation
– ImplementedDesign–ReportUtilization) LookattheSliceLogicintheUtilizationwindow.
How many Slice Lookup Tables (LUT) and Slice Registers are used?
What is the total percentage of utilization?
Could you put 15 more processors on your FPGA?
How could you communicate with the other processors or a co-processor / specialized processor?
Appendix: Instruction Set
Op-codes
The four most significant bits of the instruction code represent the op-code for the instruction. These
op-codes are defined as follows:
Op-code Mnemonic Instruction Format Details
0000 NOP NOP
0001 ARITH_2OP FUNC Rs1, Rs2, Rd Valid FUNC values: ADD, ADDC, SUB, SUBB,
AND, OR, XOR, XNOR
0010 ARITH_1OP FUNC Rs1, Rd Valid FUNC values: SHIFTL, SHIFTR, NOT, CP
0011 MOVI FUNC Rd, #8-bit Valid FUNC values: MOVIL, MOVIH
0100 ADDI ADDI Rs1, Rd, #6-bit
0101 SUBI SUBI Rs1, Rd, #6-bit
0110 LD LD Rs1, Rd, #6-bit
0111 ST ST Rs1, Rd, #6-bit
1000 BEQ BEQ Rs1, Rs2, #6-bit
1001 BGE BGE Rs1, Rs2, #6-bit
1010 BLE BLE Rs1, Rs2, #6-bit
1011 BC BC #6-bit
1100 J J #12-bit
1101 JL JL #12-bit
1110 INT FUNC #4-bit Valid FUNC values: INTE, INTD, INT_TRIG
1111 CONTROL FUNC ValidFUNCvalues:RET, STC, STB,HALT,RESET
In the instruction formats above and the Op-code Detail below, the following conventions are observed:
• Register designations (Rs1, Rs2, Rd, etc)represent Register File identifiers. They are three bits
indices in allcases.
o In the assembler, register ids are written with a dollar sign, e.g. “ADD $0 $1 $2”
• #N-bit – an N bit immediate value. May be signed (sign-extended) or unsigned (literal).
o Example: Jumpusesa sign-extendedimmediate, soJ 0xFF8wouldjumpbackwards,as it
would be interpreted as “Jump -8”. Note that the PC is still incremented, so “J 0x0”
would proceed to the nextinstruction, but “J 0xFFE” would be an infinite loop.
o On that note, “J 0xFFF” is not valid. The LSB of relative jump or branch commands
should always be zero, as instructions are aligned to even addresses.
o In the assembler,immediate values can be any number offormats (anything that Java’s
Integer.decode will parse). Examples include:
§ ADDI $0 $0 1 // Increments Reg0 by 1
§ MOVIH $0 0x80 // MOVE 0x80 into the upper byte of Reg0
• Instruction formats are consistent within a single op-code, but may vary between op-codes. The
destination register location is consistent (all op-codes that store data in the register file do so
with the three bits immediately following the op-code), butthe source registers may be located
in different bits. You will implement this in the instruction decode block.
Op-code Detail
NOP
Op-code Reserved
0000 000000000000
No operation.
ADD
Op-code RD RS1 RS2 Func
0001 Rd Rs1 Rs2 000
Adds Rs1 and Rs2 and stores the sum in Rd, ignoring any previous carry.
ADDC
Op-code RD RS1 RS2 Func
0001 Rd Rs1 Rs2 001
Adds Rs1 and Rs2 and stores the sum in Rd, with previous carry.
SUB
Op-code RD RS1 RS2 Func
0001 Rd Rs1 Rs2 010
Subtracts Rs2 from Rs1 and stores the difference in Rd, ignoring previous borrow.
SUBB
Op-code RD RS1 RS2 Func
0001 Rd Rs1 Rs2 011
Subtracts Rs2 from Rs1 and stores the difference in Rd, with previous borrow.
AND
Op-code RD RS1 RS2 Func
0001 Rd Rs1 Rs2 100
Performs bitwise AND of Rs1 and Rs2 and stores the result in Rd.
OR
Op-code RD RS1 RS2 Func
0001 Rd Rs1 Rs2 101
Performs bitwise OR of Rs1 and Rs2 and stores the result in Rd.
XOR
Op-code RD RS1 RS2 Func
0001 Rd Rs1 Rs2 110
Performs bitwise XOR of Rs1 and Rs2 and stores the result in Rd.
XNOR
Op-code RD RS1 RS2 Func
0001 Rd Rs1 Rs2 111
Performs bitwise XNOR of Rs1 and Rs2 and stores the result in Rd.
NOT
Op-code RD RS1 Reserved Func
0010 Rd Rs1 000 000
Performs bitwise NOT of Rs1 and stores the result in Rd.
SHIFTL
Op-code RD RS1 Reserved Func
0010 Rd Rs1 000 001
Shifts Rs1 by one place to the left and stores the result in Rd.
SHIFTR
Op-code RD RS1 Reserved Func
0010 Rd Rs1 000 010
Shifts Rs1 by one place to the right and stores the result in Rd.
CP
Op-code RD RS1 Reserved Func
0010 Rd Rs1 000 011
Copies Rs1 to Rd.
MOVIL
Op-code RD FUNC Unsigned Immediate
0011 Rd 0 #8-bit
Copies immediate value into lower byte of Rd.
MOVIH
Op-code RD FUNC Unsigned Immediate
0011 Rd 1 #8-bit
Copies immediate value into higher byte of Rd.
ADDI
Op-code RD RS1 Unsigned Immediate
0100 Rd Rs1 #6-bit
Adds a 6-bit unsigned value to Rs1 and stores the sum in Rd.
SUBI
Op-code RD RS1 Unsigned Immediate
0101 Rd Rs1 #6-bit
Subtracts a 6-bit unsigned value from Rs1 and stores the difference in Rd.
LD
Op-code RD RS1 Unsigned Immediate
0110 Rd Rs1 #6-bit
Loads Rd from the address given by [Rs1 + 6 bit unsigned value].
ST
Op-code RD RS1 Unsigned Immediate
0111 Rd Rs1 #6-bit
Stores Rd at the address given by [Rs1 + 6 bit unsigned value].
In spite of the name and position, Rd acts as a source register for this command – the value in register
Rd is the data to be stored in the data memory.
BEQ
Op-code RS1 RS2 Signed Immediate
1000 Rs1 Rs2 #6-bit
Branches to the relative PC address given by the signed 6-bit immediate value if the values in Rs1 and
Rs2 are equal.
BGE
Op-code RS1 RS2 Signed Immediate
1001 Rs1 Rs2 #6-bit
Branches to the relative PC address given by the signed 6-bit immediate value if the value in Rs1 is
greater than or equal to the value in Rs2.
BLE
Op-code RS1 RS2 Signed Immediate
1010 Rs1 Rs2 #6-bit
Branches to the relative PC address given by the signed 6-bit immediate value if the value in Rs1 is less
than or equal to the value in Rs2.
BC
Op-code Reserved Reserved Signed Immediate
1011 000 000 #6-bit
Branches to the relative PC address given by the signed 6-bitimmediate value ifthe carry bitin the ALU
is set.
J
Op-code Signed Immediate
1100 #12-bit
Jumps to the relative PC address given by the signed 12-bit immediate value.
STC
Op-code Func
1111 000000000001
Set the carry bit in the ALU.
STB
Op-code Func
1111 000000000010
Set the borrow bit in the ALU.
RESET
Op-code Func
1111 101010101010
Reset the CPU.
HALT
Op-code Func
1111 111111111111
Halt the CPU.

Appendix: Memory Map
The system has a fairly simple memory map – only four functional addresses and no general purpose
registers. The ones currently mapped provide inputs and outputs to the cpu that are visible externally,
throughtheLEDs, switches,anddisplay.Thedata_mem.vfilealreadyimplements thismapping,andyou
do not need to modify the code. This is provided as a reference if you wish to write your own assembly
code for the system and wish to access any of the memory mapped data.
Address Name R/W Function
0x8000 Switches & LEDs RW Bits[7:0]
Read from SW,the board switches
Write to LED, the board LEDs
Bits[15:8] return0unlessmodifiedtoextendI/Oto16.
0x9000 Right
Seven-segment Disp
W Write a 16-bit number to the seven-segment display,
formatted as in the previous labs.
0x9002 Blink W Writing a 1 to bit[0] enables the seven-segment display
blink functionality from Lab2
0xF000 Uptime R Reads the current real-time clock value in seconds. The
value is approximate, and is the same seconds counter
from Lab2.
In a real system, this would likely be more than a 16-bit
number, as the 2^16 seconds is only about 18 hours.
Also in a real system, the real time clock would be
significantly more accurate, and not from trial-and-error
on a single Nexys 4 board. The boards vary, and your RTC
counter may run faster or slower than 1 Hz.
Extending I/O
If you choose to extend the I/O of your processor in Step 5, then you must implement additional
functionality in data_mem.v and processor.v (to increase data width). Try and map the following
memory locations by modifying the data_mem module inputs to the following memory locations.
Address Name R/W Function
0x8000 Switches & LEDs RW Bits[15:0]
Read from SW,the board switches
Write to LED, the board LEDs
0x9004 Left Seven-segment
Display
W Write a 16-bit number to the seven-segment display,
formatted as in the previous labs.
0x9006 Buttons R Read the value of the buttons BTN[4:0]
0x9010 Tri-Color LED Red W Write an 8-bit component Red color to the tri-color LED
module.
0x9012 Tri-Color LED Green W Write an 8-bit component Green color to the tri-color
LED module.
0x9014 Tri-Color LED Blue W Write an 8-bit component Blue color to the tri-color LED
module.
Note onnon-pipelined systems
While any number ofthings could be added to this memory map,including any ofthe remaining
board I/O,the non-pipelined nature ofthis system actually makes this more complicated.In a
traditional pipelined system, any external access (say,to the board RAM)is part ofthe cpu
pipeline, and occurs before any write back to the register file.
Withanon-pipelined system,thenormalwrite-backto theregisterfileoccurs attheendofthe
singleprocessing cycle;the entire instruction isprocessed before thenext clock edge.This is
problematic when theRAM,or any otherblock,registers its outputs;there isa cycledelay, and
so if instruction 0x4 is a read from a RAM, itis the next cycle when instruction 0x6 is being
processed thatthe RAM data is available. Additional hooks would need to be added to the cpu
to allow register write-backs on subsequent cycles.