Due: 5 pm Friday 8 April 2022
Summary: You will write MIPS assembly to control a speaker and play a song.
Collaboration: You will work during lab in randomly assigned pairs. You must complete the work together in these groups, whether during or after the scheduled lab time.
Submitting: After completing each exercise, show your completed work to the instructor or a course mentor. If you are unable to complete the lab during class time, schedule a time during office hours to demonstrate each of your solutions.
In this lab, you will wire a speaker to the PIC32 and play a song. Along the way, you will write functions in assembly code, work with arrays stored in memory, and use instructions for multiplication and division.
As you might guess, multiplication and division are more complicated than arithmetic (addition and subtraction) and logic. Hence, they happen in a separate part of the chip called the MDU (multiplication and division unit), rather than the ALU. The multiplication and division instructions bring their own set of intricacies and complications that we must understand to use them correctly.
Let us recall that with addition, the only bit of “extra space” (beyond the storage capacity of each register) that we might have to worry about is the carry out—a single bit. However, if you recall your grade-school multiplication algorithm, for every digit (or bit, in our case) in the multiplier, we must shift the product with that digit further and further to the left. Thus, a product could take twice as many bits as the multiplicand and multiplier. To handle this situation, the MIPS ISA stores the resulting product in two special, reserved registers called Hi
and Lo
from which you must copy into a general purpose register using instructions mflo
(for the lower bits) or mfhi
(for the higher bits).
Fortunately, the PIC32 architecture can do this multiplication fairly quickly—though not always in a single “cycle,” the architecture handles the messy details for you, so that you never have to worry about any delay, the way you do with jumps and branches. If you want to read more about efficient multiplication, consult section 3.3 of Patterson and Hennessy.
Unfortunately, division is not quite so clean, both from an algorithmic standpoint and an implementation standpoint. Division takes much longer, but (fortunately) these details are again hidden from you. Much like multiplication, the quotient can be copied from Lo
using instruction mflo
, and the remainder can be copied from Hi
using instruction mfhi
. Unfortunately, we have detected a side-effect in the division instructions that seems to put the quotient into the dividend register, as the following code snippet demonstrates.
li $t0, 60 # t0 = 60
li $t1, 18 # t1 = 18
divu $t0, $t1 # Calculate t0/t1
mflo $t2 # Put the quotient (3) into t2 for later use
mfhi $t3 # Put the remainder (6) into t3 for later use
Contrary to our expectations about the MIPS ISA and the description on the Green Card, at the end of this code sequence we find $t0
also contains 3 with our Microstick PIC32s. For now, this is simply a fact you will have to be aware of and program around.
There is an additional caveat: The two instructions following an mflo
or mfhi
instruction must not modify the HI register. That precludes multiplication and division instructions from immediately following either of these moves.
Recall that our delayloop
function from the previous lab takes as its parameter a number of loop iterations to execute. As a first step to playing musical notes, we’d like to be able to convert units of time into delay loop iterations. For example, consider the following function in C:
#define CLOCK_RATE_HZ 4000000 // Cycles per second for this CPU
#define CPI 1 // Cycles per instruction
#define IC 3 // Instructions per delay loop
#define MICROSEC_PER_SEC 1000000 // Microseconds per second
int
(int microsec) {
time_to_iters return (microsec * CLOCK_RATE_HZ) / (MICROSEC_PER_SEC * IC * CPI);
}
(Note that we use C-style comments in the #define
lines because these are consumed by the preprocessor, which is operating in the C programming language.)
I translated this C code into assembly. To test my implementation, I wrote a simple program that calls time_to_iters
and delayloop
to blink the onboard LED at a rate of 1 Hz. To test my program, I simultaneously pushed the PIC32 reset button and started a 60 second timer. I verified that the blinking stopped after the 60 second timer was complete.
Here is my test program:
# timetest.S
# Test harness for the time_to_iters function
# Written by Janet Davis, 23 October 2013
# Last revised by Janet Davis, 24 October 2013
#define OFF 0x0
#define ON 0x1
#define CLOCK_RATE_HZ 4000000 // Cycles per second for this CPU
#define CPI 1 // Cycles per instruction
#define IC 3 // Instructions per delay loop
#define MICROSEC_PER_SEC 1000000 // Microseconds per second
.set noreorder # Avoid reordering instructions
.text # Start generating instructions
.globl main # The label should be globally known
# The label marks an entry point
.ent main
# Compute number of delay loops in a given period of time
# (measured in microseconds)
time_to_iters:
# $t0 = time * CLOCK_RATE_HZ
li $t2, CLOCK_RATE_HZ # Load CLOCK_RATE_HZ constant
multu $a0, $t2 # Multiply argument with CLOCK_RATE_HZ
mflo $t0 # Copy result from lo to $t0
# $t1 = MICROSEC_PER_SEC * IC * CPI
li $t2, MICROSEC_PER_SEC # Load MICROSEC_PER_SEC constant
li $t3, IC # Load IC constant
# NB: At least two instructions occur between the mflo above and multu below
multu $t2, $t3 # Multiply IC * MICROSEC_PER_SEC
mflo $t1 # Copy result from lo to $t1
li $t4, CPI # Load CPI constant
nop # Ensure delay between mflo and multu (2 instr)
multu $t1, $t4 # Multiply previous result by CPI
mflo $t1 # Copy result from lo to $t1
# result = (time * CLOCK_RATE_HZ) / (MICROSEC_PER_SEC * IC * CPI)
nop # Ensure delay betweeen mflo and divu (2 instr)
nop # [continue delay]
divu $t0, $t1 # Divide $t0 by $t1
mflo $v0 # Copy quotient from lo to return val reg
jr $ra # Return to caller
nop # Delay slot
# Delay for the given number of loop iterations (3 cycles/iteration)
# Precondition: $a0 > 0
delayloop:
addi $a0, $a0, -1
bnez $a0, delayloop
nop
jr $ra
nop
# Main program.
# Test time_to_iters computation.
# The light should blink at a rate of 1 Hz.
main:
# Set up port A for output
la $s0, TRISA # Load the address mapped to TRISA
li $t0, 0x0000 # Output on all pins
sw $t0, 0($s0) # Write to TRISA
la $s0, LATA # Load the address mapped to LATA
# Compute number of delayloop iterations for blinking at 1Hz
li $a0, 500000 # Half a second, in microsec
jal time_to_iters # Call time_to_iters procedure.
nop # Delay slot
add $s1, $v0, $zero # Store result in $s1
# Blink on and off 60 times; should take 60 seconds to finish blinking.
li $s2, 60 # Set countdown to 60.
loop:
li $t0, ON # Turn LED on.
sw $t0, 0($s0)
move $a0, $s1 # Call delayloop on the computed value.
jal delayloop
nop # Delay slot
li $t0, OFF # Turn LED off.
sw $t0, 0($s0)
move $a0, $s1 # Call delayloop on the computed value.
jal delayloop
nop # Delay slot
addi $s2, $s2, -1 # Decrement countdown
bnez $s2, loop # Loop while countdown > 0
nop # Delay slot
forever:
j forever # Infinite loop that does nothing.
nop
# Marks the end of the program .end main
Unfortunately, I made a significant error when writing the time_to_iters
function; the blinking doesn’t even start. It’s your job to identify and fix my bug. (Note: It does not have to do with violating the mflo
delay; however, any purported fix must be sure to respect this condition.)
Note: You may NOT change any of the pre-#define
d constants.
time_to_iters
produce in the given program?timetest.S
time_to_iters
function so it works correctly.When you have completed this task, explain your fix and demonstrate the program to the mentor or instructor.
Next, you will use your debugged time_to_iters
function to play a song. The overall algorithm is as follows:
Get a couple of long wires, one red and one black. Connect the red wire to RA0
(pin 2) and the black wire to VSS
(pin 8). Connect the other end of one wire to the top row of holes in the protoboard’s speaker connection, and the other wire to the bottom row. It does not matter which wire goes in which spot.
If you run the timetest.S
program now, you will hear a series of quiet, dull clicks. This is because 1Hz is much too slow an oscillation for the human mind to perceive as a tone. If you change the time passed in to time_to_iters
so it is much shorter, you will hear a tone.
Use this assembly program as a starting point for your work:
# song.S
# Plays a song
# Written by Janet Davis, 23 October 2013
# Last revised by YOUR NAME(S), THE DATE
#define OFF 0x0
#define ON 0x1
#define CLOCK_RATE_HZ 4000000 // Cycles per sec for this CPU
#define CPI 1 // Cycles per instruction
#define IC 3 // Instructions per loop
#define MICROSEC_PER_SEC 1000000 // Microseconds per second
# C-major scale
# Reference: http://en.wikipedia.org/wiki/Piano_key_frequencies
#define REST 0 // Use 0 to represent rests (no sound played)
#define C 3822 // Period in microseconds for middle C
#define CSHRP 3608 // Period in microseconds for middle C#
#define D 3405 // Period in microseconds for D above middle C
#define DSHRP 3214 // Period in microseconds for D# above middle C
#define E 3034 // Period in microseconds for E above middle C
#define F 2863 // Period in microseconds for F above middle C
#define FSHRP 2703 // Period in microseconds for F# above middle C
#define G 2551 // Period in microseconds for G above middle C
#define GSHRP 2408 // Period in microseconds for G# above middle C
#define A 2273 // Period in microseconds for A above middle C
#define Bee 2025 // Period in microseconds for B above middle C
#define CC 1911 // Period in microseconds for C above middle C
// (Note this is half of middle C!)# Durations
#define WHOLE 2000000 // Whole note (microseconds)
#define DTHALF 1500000 // Dotted half note
#define HALF 1000000 // Half note
#define DTQUART 750000 // Dotted quarter note
#define QUARTER 500000 // Quartner note
#define EIGHTH 250000 // Eighth note
#define SIXTEENTH 125000 // Sixteenth note
#define BREATH 10000 // A breath between phrases
# The song (Rodgers & Hammerstein, 1959)
# notes and durations are arrays stored in memory.
# songlength, also stored in memory, gives the size of the arrays.
# Reference for data segment format:
# http://web.archive.org/web/20091112173310/http://www.cs.umd.edu/class/sum2003/cmsc311/Notes/Mips/dataseg.html
.data
songlength:
.word 65
notes:
.word C, REST, D, E, REST, \
C, E, C, E, REST, \
D, E, F, F, E, D, F, REST, \
E, F, G, E, G, E, G, REST, \
F, G, A, A, G, F, A, REST, \
G, C, D, E, F, G, A, REST, \
A, D, E, FSHRP, G, A, Bee, REST, \
Bee, E, FSHRP, GSHRP, A, Bee, CC, REST, \
F, F, A, F, Bee, G, CC
durations:
.word QUARTER, EIGHTH, EIGHTH, QUARTER, EIGHTH, \
EIGHTH, QUARTER, QUARTER, HALF, BREATH, \
DTQUART, EIGHTH, EIGHTH, EIGHTH, EIGHTH, EIGHTH, WHOLE, BREATH, \
DTQUART, EIGHTH, DTQUART, EIGHTH, QUARTER, QUARTER, HALF, BREATH, \
DTQUART, EIGHTH, EIGHTH, EIGHTH, EIGHTH, EIGHTH, WHOLE, BREATH, \
DTQUART, EIGHTH, EIGHTH, EIGHTH, EIGHTH, EIGHTH, WHOLE, BREATH, \
DTQUART, EIGHTH, EIGHTH, EIGHTH, EIGHTH, EIGHTH, WHOLE, BREATH, \
DTQUART, EIGHTH, EIGHTH, EIGHTH, EIGHTH, EIGHTH, DTHALF, BREATH, \
EIGHTH, EIGHTH, QUARTER, QUARTER, QUARTER, QUARTER, DTHALF
# The program starts here
.set noreorder # Avoid reordering instructions
.text # Start generating instructions
.globl main # The label should be globally known
# The label marks an entry point
.ent main
# Play the specified note for the specified length of time
# $a0: period in microseconds for the note (0 specifies a rest)
# $a1: duration in microseconds to play the note
playnote:
# TODO: YOUR CODE HERE
# Compute number of delay loops in a given period of time
# (measured in microseconds)
time_to_iters:
# TODO: YOUR CODE HERE
# Delay for the given number of loop iterations (3 cycles/iteration)
# Precondition: $a0 > 0
delayloop:
addi $a0, $a0, -1
bnez $a0, delayloop
nop
jr $ra
nop
# Main program
# Plays the first few notes of the song stored in memory.
# TODO: PLAY THE ENTIRE SONG
main:
# Set port A for output.
la $s0, TRISA # Load the address mapped to TRISA
li $t0, 0x0000 # Output on all pins
sw $t0, 0($s0) # Write to TRISA
# Pause for a moment so that the song does not begin playing while
# MPLAB is programming the microprocessor.
li $a0, REST
li $a1, WHOLE
jal playnote
nop
# Load data addresses.
la $s5, songlength
la $s6, notes
la $s7, durations
# Play the first five notes/rests of the song.
# TODO: CHANGE THIS TO A LOOP
lw $a0, 0($s6) # Play the first note or rest.
lw $a1, 0($s7)
jal playnote
nop # Delay slot
lw $a0, 4($s6) # Play the second note or rest.
lw $a1, 4($s7)
jal playnote
nop # Delay slot
lw $a0, 8($s6) # Play the third note or rest.
lw $a1, 8($s7)
jal playnote
nop # Delay slot
lw $a0, 12($s6) # Play the fourth note or rest.
lw $a1, 12($s7)
jal playnote
nop # Delay slot
lw $a0, 16($s6) # Play the fifth note or rest.
lw $a1, 16($s7)
jal playnote
nop # Delay slot
forever:
j forever # Infinite loop that does nothing.
nop # Delay slot
# Marks the end of the program .end main
song.S
.time_to_iters
function with your corrected implementation.Implement the playnote
function. To help you get started, here is a C function that does the same thing:
/*
* Play a musical note for a specified duration.
* period:
* The period of the note (1/frequency), in microseconds.
* 0 indicates a rest (silence).
* duration:
* The duration to play the note, in microseconds.
* Preconditions:
* period >= 0, duration >= 0
* Port A, pin 0 is configured for output, and is connected to a speaker.
* Port A, pin 0 is off (0).
* Postconditions:
* The note (or rest) has been played.
* Port A, pin 0 is off (0).
*/
void
(int period, int duration)
playnote {
int * p_lata = LATA;
int delay, count;
if (period == 0) {
= time_to_iters (duration);
delay (delay);
delayloop } else {
= duration/period;
count = time_to_iters(period/2);
delay while (count > 0) {
*p_lata = ON;
(delay);
delayloop *p_lata = OFF;
(delay);
delayloop --;
count} // while
} // if
} // playnote
Test your program. You should hear the first few notes of a song.
Revise the body of the main program so it will play the entire song instead of just the first few notes. It may be helpful to write the loop in C before translating it to assembly.
If you have time, revise the data segment to play a different song.
So that others are not startled or confused when they connect your MIPS processor, please re-program your chip with a different program that does not do anything useful. While one solution would be to program your chip with a project from an earlier lab, the easiest thing to do might be to change the first few lines of the main
procedure to be an infinite loop, like follows:
main:
j main # Infinitely do nothing
nop
After you make the change to your source code, remember to build the project and program it to your chip.
Copyright © 2018, 2019, 2020, 2022 Marge Coahran, Charlie Curtsinger, Janet Davis, and Jerod Weinman
This work is licensed under a Creative Commons Attribution NonCommercial ShareAlike 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-sa/4.0/ or send a letter to Creative Commons, 543 Howard Street, 5th Floor, San Francisco, California, 94105, USA.