Due Wednesday, November 04 before 9AM.

Please post all questions to the piazza page: http://www.piazza.com/wisc/fall2015/cs354/

Collaboration Policy

For this assignment, you may work in pairs (2 people). All students (whether working in a pair or not) must individually turn in the assignment. Therefore, 2 copies of the assignment will be turned in for those working as a pair. The grader will choose to grade only one copy of a pair's work, and assign the same score to both students.

If one member of a pair does something unexpected with respect to turning in the assignment (for example: bad copy of the source code, late, or perhaps not turning in portions of the assignment), then both of the pair get the same penalty and the same score.

If working in a pair, the names and sections of both students must appear as comments at the top of all files that will be turned in.

Overview

The purpose of this assignment is to understand more about how caches work, learning a bit about simulation along the way, and picking up introductory skills in generating a Makefile.

Please read this entire assignment specification before starting on the first step.

Description

For this assignment, you will use a program called pin that produces cache performance statistics, given cache parameters and an executable.

pin runs the executable to internally produce a series of address traces. These are the ordered set of addresses that a program generates as it runs. They represent the addresses read from or written to as the program runs. Each address may represent a read for the instruction fetch or a read or write to a variable stored in memory.

The address traces are then used internally with a cache simulator. The cache simulator is a program that acts as if it is a cache, and for each trace, does a look up to determine if that address causes a cache hit or a cache miss. The simulator tallies the hits/misses, and when it has completed simulation of all the traces, it prints these statistics for you.

To run a simulation, use a command line similar to

/p/course/cs354-common/public/cache/pin -t /p/course/cs354-common/public/cache/source/tools/Memory/obj-ia32/allcache.so -is 16384 -ia 1 -ib 64 -ds 16384 -da 1 -db 64 -- yourexe

6 of these command line arguments specify cache parameters to pin. The simulator presumes separate I-cache and D-cache. The I-cache holds only the machine code instructions, as read when doing an instruction fetch. The D-cache holds all other data read or written while a program runs.

For the I-cache, specify

• -ia 1
  This causes the set associativity of the cache to be 1, or direct mapped. This is the only value we will use in this assignment.
• -is N
  Substitute a power of 2 for N. This sets the capacity for this cache. For all our simulations, use a 16KB (16384) size.
• -ib N
  Substitute a power of 2 for N. N is the number of bytes per block. You will be changing this parameter in this assignment.

For the D-cache, specify

• -da 1
  This causes the set associativity of the cache to be 1, or direct mapped. This is the only value we will use in this assignment.
• -ds N
  Substitute a power of 2 for N. This sets the capacity for this cache. For all our simulations, use a 16KB (16384) size.
• -db N
  Substitute a power of 2 for N. N is the number of bytes per block. You will be changing this parameter in this assignment.

As you work your way through this assignment, you will be answering questions that are in the file:

   /p/course/cs354-common/public/html/handouts/a3questions

Copy this file to your own directory, and add your answers to this file. The file will be turned in.

Write a very small C program called cache1D.c that sets each element of an array of 100,000 integers to the value of its index. The statement that sets a single array element will be something like

  arr[i] = i;

• Declare the array as a global variable, so the declaration will be outside of main() (and prior to main() within the source code file). This requirement will cause the array to be within the global data segment, and not on the stack.
• Place a for loop inside main() to set each element of the array. One array element is set during each iteration of the for loop.

Compile your program with

  gcc -o cache1D cache1D.c -Wall -m32

Notice that this command differs slightly from the compilation that we did for assignment 1 this semester. It does NOT have the -O option, so the compiler is not going to do optimizations. It is important that you do not use the -O option. Use these same compiler options for all the programs in this assignment.

Before you go any further, it is time to figure out how to do this compilation using make: a powerful tool that helps us to automate our compilation, as well as automate many other sets of commands.

To help you put together a Makefile, try one or more of

• A brief tutorial about make by Hector Urtubia can start you on your way.
• Within the online textbook written by our own Arpaci-Dusseau professors, see the Laboratory: Tutorial. You'll benefit from pages 1-8 of this tutorial, although pages 5-8 specifically discuss make.
• University of Maryland has a nice introduction to make; especially helpful will be to look at the topic of dummy targets. You will want to add a dummy target for each of your simulations.
• UW-Madison alumni Professor Newhall explains using make in helpful and concise prose.

Make sure that you have a single target to make the cache1D executable. (And for each 2 subsequent programa, add a new target, and then use make. You will be turning in your Makefile. Please name your target with the name of the executable, which should be the same as your source code, but without the .c file name extension.)

Step Two: 1-dimensional array code analysis

Run the 3 simulations needed to answer the questions in the a3questions file. Then, answer the questions. Running the simulations will be amazingly easy, if you put a target for each block size simulation into your Makefile! So, do it!

Repeat what you did for the 1 dimensional array with two programs that set elements of a 2-dimensional array. Many students will not have used 2-dimensional arrays in C before; the K&R book, section 5.7-5.9, starting on page 110 will be a good reference for declaring and using 2D arrays.

First, write a C program called cache2Drows.c that sets each element of a 3000 row by 500 column array of integers to the sum of the row index and the column index. The statement that sets an element will be something like

  arr2d[row][col] = row + col;

Run the cache analysis requested in the a3questions file.

Second, write another C program called cache2Dcols.c that does the same thing as cache2Drows.c did, but in a different order. This program has the inner loop work its way through the elements of a single column of the array, and the outer loop iterates through the columns. If you truly understand your code, it should be close to trivial to copy cache2Drows.c and modify it to become cache2Dcols.c.

Handing In the Assignment

We will use your Makefile to compile. We will compile the 3 source files using following commands:

make cache1D
make cache2Drows
make cache2Dcols

If you are working as part of a pair, you must turn in an extra file. This file will contain the names and sections of both students in the pair. As an example, if Kevin worked with Haseeb on this assignment, the contents of the extra file for both Kevin and Haseeb would be

Kevin Zhang   section 1
Haseeb Tariq  section 2

The name of this file is specialized to help the 354 automated grading tools identify who worked together. This file name is composed of the CS logins of the partners separated by a period. The file name is of the form <login1>.<login2>. Kevin's login is kzhang, and Haseeb's login is haseeb. The file name that both use will be kzhang.haseeb ; please have both partners use the same file name. It does not matter which partner's name is first within this file name.

Requirements

1. Include a comment at the top of the source code of both programs with your name and section (and your partner's name and section, if working in a pair).
2. Use the instructional Linux machines for this assignment! A penalty will be imposed for any program that was obviously edited on a Windows machine and transferred to the Unix machines for turning in. It is annoying to see Windows line endings (^M) at the end of every line. Points will be taken off for such mistakes.
3. Your programs must compile on an instructional Linux machine as indicated in this specification without warnings or errors. When grading, we will use your Makefile, so make sure that it works!

About Copying Code and Academic Misconduct

Don't cheat. Read this link carefully.

Do not post your assignment solutions (or drafts) on any publicly accessible web sites. This specifically includes GitHub. It is academic misconduct to post your solution.

For almost any C program that does something useful, someone has already written this program and further, has posted it for others to use. These programs do not do much that is useful, and are not likely posted anywhere. Still, it is academic misconduct for you to copy or use some or all of a program that has been written by someone else.

The penalty for academic misconduct on this assignment (and all CS/ECE 354 assignments) will be a failing grade in the course. This penalty is significantly more harsh than if you simply do not do the assignment. You will gain much more by doing the assignment than by copying, possibly modifying, and turning in someone else's effort.