This lab should be done with a partner of your choosing.
The setup procedure for this lab will be similar to Labs 2 and 3.
First, both you and your partner should run setup31 to
grab the starting point code for this assignment. Suppose users molly and tejas which to work together. Molly (mdanner1) can start by running
[~]$ setup31 labs/04 tdanner1
Once the script finishes, Tejas (tdanner1) should run
[~]$ setup31 labs/04 mdanner1
For the next step only one partner should copy over the starting code
[~]$ cd ~/cs31/labs/04
[04]$ cp -r ~lammert/public/cs31/labs/04/* ./
[04]$ ls
Makefile loop.s mainloop.c readfile.c small.txt swap.s
large.txt loop_C_goto_version mainswap.c readfile.h stats.c
Now push the changes to your partner
[04]$ git add *
[04]$ git commit -m "lab 4 start"
[04]$ git push
Your partner can now pull the changes. In this case if Tejas wishes to get files Molly pushed, he would run
[~]$ cd ~/cs31/labs/04
[04]$ git pull
Lab 4 Goals:
- Gain experience using pointers and dynamic memory allocation in C.
- Practice using gdb and valgrind to debug C memory bugs.
- Practice converting C code to equivalent IA32 assembly instructions.
Part 1. C Pointers and Memory Allocation
Experimental scientists collect data from experiments and often
want to compute some simple statistical analyses over the set of experimental
data. A useful tool would be a program that could compute these
statistical results for any size data set (i.e. it would work for 10 data
values or for 10 million without re-compilation).
For this program you will implement the program started in stats.c
that takes a single command line argument, which is the name of a file of
data values (floats, one per line), and computes and prints out a set of
statistics about the data values.
The starting point code comes with two input files that you can use to
test your solution:
./stats small.txt
Results:
--------
num values: 16
mean: 1.812
median: 1.500
std dev: 1.031
unused array capacity: 4
./stats large.txt
Results:
--------
num values: 94
mean: 1.161
median: 1.000
std dev: 0.788
unused array capacity: 66
This program includes the readfile library code that it links in as well
as linking in the math library:
use the makefile to compile.
You can see how the executable is built from a .c, a .o, and explicitly
linking in the math library (-lm), by reading the Makefile.
In stats.c is the starting point into which you will put your
code. It contains the prototype for the getvalues function that
you need to implement, and has some code to get the filename
command line argument.
You should structure your program in the following way:
- Make a call to getvalues, passing in the filename of the
file containing the data values, and passing in two values by reference:
the address of an int variable to store the size of the array (number
of values read in); and the address of an int variable to store the total
array capacity.
Note: The function will return an array of float values initialized to the values
read in from the file, or NULL on error (like malloc fails or the file cannot
be opened).
- Compute the mean (average), the median (the middle value) and
the standard deviation of the set of values and print them out.
Note: You will likely need to sort the values prior to computing the median.
- Print out the results, plus information about the number of values in the
data set and the amount of unused capacity in the array storing the values.
Statistic Functions
The statistics you need to compute on the set of values are the following:
- mean: the average of the set of values. For example, if
the set is: 5, 6, 4, 2, 7, the mean is 4.8 (24.0/5).
- median: the middle value in the set of values. For example,
if the set is: 5, 6, 4, 2, 7, the median value is 5 (2 and 4 are smaller
and 6 and 7 are larger).
- stddev: is given by the following formula:
$$s = \sqrt{\frac{1}{N-1}\sum_{i=1}^N(x_i - \bar{x})^2}$$
Where $N$ is the number of data values, $x_i$ is the $i$th data value, and
$\bar{x}$ is the mean value of the $N$ input values.
Feel free to discuss the math on Piazza or in person with your classmates.
While we're using it to ensure that your pointers are working correctly, the
math itself is not the focus of this assignment...
Requirements
- getvalues: The array of values must be dynamically allocated
on the heap by calling malloc. You should start out allocating
an array of 20 float values. As you read in values into the current array,
if you run out of capacity:
- Call malloc to allocate space for a new array that is
twice the size of the current full one.
- Copy values from the old full array to the new array (and make
the new array the current one).
- Free the space allocated by the old array by calling free.
When all of the data values have been read in from the file, the function
should return the filled, dynamically allocated, array to the caller
(the function's return type is float *).
The size and capacity of the array should be "passed" back to the caller
via the pointer parameters that are used for pass-by-reference values.
NOTE: there are other ways to do this
type of alloc and re-alloc in C. However, this is the way I want you to
do it for this assignment: make sure you start out with a dynamically
allocated array of 20 floats, then each time it fills up, allocate a
new array of twice the current size, copy values from the old to new,
and free the old.
- For full credit, your program must be free of valgrind errors.
- Your code should be commented, modular, robust, and use meaningful variable
and function names. This includes having a top-level comment describing your
program and listing your and your partner's names and the date. In addition,
every function should include a brief description of its behavior. You
should not use any global variables for this assignment.
- It should be evident that you applied top-down design when constructing
your submission (e.g., there are multiple functions, each with a specific,
documented role).
- You should not assume that we will test your code with the sample input
files that have been provided.
Hints & Tips
Try getting getvalues to work without the re-allocation and
copying part first (for fewer than 20 values). Once that works, then
go back and get it to work for larger numbers of input values that require
mallocing up new heap space, copying the old values to the new larger space,
and freeing up the old space.
Use doubles to store and compute the mean and the square root.
The C math library (in math.h), has functions to compute the square root:
double sqrt(double val);
Your program must be free of valgrind errors.
See Lab 2 for documentation about using the
readfile library (and also look at the readfile.h comments for how to
use its functions).
Take a look at the weekly lab code and in-class exercises to
remind yourself about malloc, free, pass-by-reference, pointer variables,
dereferencing pointer variables, and dynamically allocated arrays.
Make use of my C programming resources and links for C references (C pointer
references in particular), and the C Style Guide for tips on good
commenting, avoiding line wrapping, and other programming style tips.
Part 2. Writing a Loop in IA32
In the file named loop.s finish the implementation of
the following function in IA32:
// this function computes the sum of the first x positive int values
// x: the top of the range of values to sum
// returns: the sum of the values 0 to x inclusive
int loop(int x) {
int res, i;
res = 0;
for(i=1; i <= x; i++) {
res = res + i;
}
return res;
}
In mainloop.c is a main program that makes a call to this
function.
If you run make, you can build an executable, mainloop,
that links in your loop.s solution.
Use this to test your solution for correctness.
Hints and Requirements
- In the file loop_C_goto_version you should write
your C goto translation of the loop function above (you do not need
to compile this, I just want to set your application of Step 1 in
converting a C for loop to IA32).
- The basic stack set-up and return code is provided for you in loop.s,
including space for the two local variables res and i.
- The parameter value x is in the stack in the caller's
stack frame. It can be accessed relative to the %ebp pointer:
esp ----> # loop's stack frame:
ebp-0x8 # space for one local variable
ebp-0x4 # space for one local variable
ebp ---->
# main's stack frame:
ebp+0x8: # space for the parameter (x)
- The return value must be stored in register %eax right
before the leave, ret instruction sequence.
- Comment your IA32 instructions with what they are doing in terms of
the C code. As an example:
movl 0x8(%ebp), %eax # load x into R[%eax]
addl $1, %eax # x + 1
Help:
- Draw a picture of the stack, draw where parameter and argument
values are and trace through the instruction execution
to help you determine what IA32 instructions you need to use to implement
loop.
- Try implementing and testing incrementally. For example, first see
if you can add code to return the value of the parameter. Next, try
returning some function of the parameter (like x+10).
- You can use gdb (or ddd) to debug your solution. Set a breakpoint in
loop, and use disass, ni, and print $reg to see values as you go. If you
want to see the value referred to by an address, you need to tell gdb what
type the address points to. For example, if you want to see an int value
at address 0x1234, do:
# re-cast address as an int pointer (int *), then dereference the pointer *
(gdb) *(int *)(0x1234)
- look in mainloop.c for the call to your loop function to help you
see how it is called.
Part 3. Writing a Swap Function in IA32
In the file named swap.s finish the implementation of
the a swap function that swaps two int values:
void swap(int *x, int *y);
This function takes two ints passed by reference (the parameters
point to the storage location of their argument variables). After the call,
the two int variables passed to swap should have their values swapped.
A call to swap would look like (see mainswap.c for another example):
int a, b;
a = 10;
b = 8;
printf("%d %d\n", a, b); // prints: 10 8
swap(&a, &b); // swap the values stored in a and b
printf("%d %d\n", a, b); // prints: 8 10
The swap.s file already contains some IA32 instructions
that set up up and tear down the stack frame associated with the swap
function. You will add IA32 instructions to the body of this IA32 function
to perform swapping operation.
In mainswap.c is code to test your swap implementation.
Run make to build an executable, mainswap, that
links in your loop.s solution and calls it.
Use this to test your solution for correctness.
Requirements
- As its name implies, your swap function should swap values stored in the given
input pointers.
- Your code should access the function's parameter values of x and y
on the stack in the caller's stack frame. They can be accessed relative to
the %ebp pointer:
esp ----> # swap's stack frame:
ebp ---->
# main's stack frame:
ebp+0x8: first parameter value
ebp+0xc: second parameter value
- Briefly comment your IA32 instructions with what they are doing in terms of
the C code. As an example:
movl 0x8(%ebp), %eax # load x into R[%eax]
addl $1, %eax # x + 1
Hints & Tips
Handy References
Submit
To submit your code, simply commit your changes locally
using git add and git commit. Then run git push while in the
labs/04 directory. Only one partner needs to run the final
push, but make sure both partners have pulled and merged
each others changes. See the section on using a shared repo
on the git help page.