Sale!

Embedded Systems – Homework 3

$30.00

ECE471: Embedded Systems – Homework 3
Linux Assembly and Code Density
1. Use your Raspberry-Pi to work on this project.
• Download the code from:
http://web.eece.maine.edu/~vweaver/classes/ece471/ece471_hw3_code.tar.gz
and copy it to the Raspberry-Pi.
• Uncompress/unpack it with the command tar -xzvf ece471_hw3_code.tar.gz
• Change into the ece471_hw3_code directory cd ece471_hw3_code
• Put all answers to questions into the included text README file. This will automatically be
bundled up with your submission.

Category:
5/5 - (2 votes)

ECE471: Embedded Systems – Homework 3
Linux Assembly and Code Density
1. Use your Raspberry-Pi to work on this project.
• Download the code from:
http://web.eece.maine.edu/~vweaver/classes/ece471/ece471_hw3_code.tar.gz
and copy it to the Raspberry-Pi.
• Uncompress/unpack it with the command tar -xzvf ece471_hw3_code.tar.gz
• Change into the ece471_hw3_code directory cd ece471_hw3_code
• Put all answers to questions into the included text README file. This will automatically be
bundled up with your submission.
2. Modify the exit_asm.s file to return the value 42. (1 point total)
(a) Modify exit_asm.s
(b) Be sure any code comments are accurate!
(c) Run make to generate an updated version
(d) To test, run ./exit_asm followed by echo $? which will show you the last program’s exit
status.
(e) Some reminders about Linux GNU assembler (as) syntax:
.equ IDENTIFIER,value sets a macro replacement, like
#define IDENTIFIER value would in C
You can use @ to specify a comment, like // in C
You prefix a constant value with #
(to move the number 5 into a register you would do mov r0,#5 )
(f) Reminders about the Linux kernel ARM syscall EABI:
Arguments go in r0 to r6
System Call Number goes in r7
Use swi 0x0 to trigger a system call.
3. Investigate the code density of print_integer (3 points total)
(a) Run make if you haven’t already, which should generate the
print_integer and print_integer.thumb executables.
(b) Look at the source code (the same file is used for both): print_integer.c. The algorithm
used was described in class.
(c) Find the size of the print_integer() function in the ARM32 executable and record it in the
README
i. To disassemble the code, run the command
objdump –disassemble-all ./integer_print | less
the | less at the end says to feed the output of the call to the program less, which lets you
scroll backwards and see all the output.
ii. Find the print_integer() function. You could look manually, but the best way is to
use the ‘/’ slash character to start a search and then search for print_integer. The first
location you find is the call from main() to the function, press ‘/’ again to find the actual
function.
iii. You should find something like the following:
0001041c <print_integer>:
1041c: e59fc064 ldr ip, [pc, #100] ; 10488 <print_integer+0x610420: e52de004 push {lr} ; (str lr, [sp, #-4]!)

The first column is the address in memory where this code lives, the next is the raw machine code for the instruction, the next is the decoded assembly language, and the last is the
disassembler giving “helpful” hints about what’s going on.
iv. You can see in this first case all the instructions are 32-bit hex values.
v. Calculate the length of this function and note it in the README. You can calculate length by
scrolling down to where the function ends, (probably at something like __libc_csu_init)
getting the memory address (the first column), then subtract from that the address of the start
of the function.
(d) Now go back and do the same for the THUMB2 print_integer.thumb2 executable too.
Record the size in the README.
(e) Answer in README: Does the machine language look different for THUMB2 than ARM32?
How?
4. C vs Assembly code density: (2 points total)
Put the answer to these in the README.
(a) Compare the size of the ARM32 integer_print executable and the
THUMB2 integer_print.thumb2 executables. (list both sizes)
You can get filesize with ls -l (that’s a lowercase L) You will want to run the strip command
on the executables first (i.e. strip integer_print ) or your results might be unexpected.
(b) When you ran make it also compiled a pure assembly language version of the integer print code:
integer_print_asm. What is the size of that file?
(c) Given these results, Which language might you use in space constrained system? Why?
(d) Which code to you think is easier to write, the C or assembly one?
5. Use gdb to track down the source of a segfault. (2 points total)
(a) When you ran make it should have built a program called crash
(b) Run that program. It should crash with a Segmentation fault error.
(c) Use the gdb debugger to find the source of the error.
(d) Run gdb ./crash
(e) When it comes up to a prompt, type run and press enter to run it until it crashes.
(f) It should tell you it crashed, then tell you what line of code caused the crash. Put the line that
caused the crash in the README
2
(g) You can do various other things here, such as run bt to get a backtrace, which shows you which
functions were called to get you to this error. You can run info regis to see the current
register values.
(h) Run disassem to get a disassembly of the function causing problems. There should be an
arrow pointing to the problem code. Cut and paste this line into the README
(i) In the end, what was the cause of the error in this program? (again, put this in the README
6. Something cool: (1 point total)
Modify integer_print.extra.c (which starts out the same as integer_print.c and do
one of the following:
• Easy: modify so instead of printing a hard-coded value in main(), it reads a value from
the prompt. To read from the console into an integer in C you would use something like
scanf(“%d”,&value);
• Moderate: make the file print the integer in hexadecimal instead of decimal. (modify the existing
algorithm, don’t use printf())
• Hard: modify integer_print_asm.s and make it print hexadecimal
7. Linux Command Line Exploration (1 point total)
Try out the cal program. This prints a calendar, by default the current month. You can also cal
2016 or cal 12 2016. Beware not to do cal 16 as that will give you year 16, not 2016.
(a) Run cal 9 1752 Is there a bug here? Can you explain what is happening?
8. Submitting your work
• Run make submit which will create hw3_submit.tar.gz containing the various files.
You can verify the contents with tar -tzvf hw3_submit.tar.gz
• e-mail the hw3_submit.tar.gz file to me by the homework deadline. Be sure to send the
proper file!
3

Open chat
Need help?
Hello
Can we help you?