Due: Wednesday, October 28, at
Due: Friday, October 30, at 11:59pm
The baseline Nachos implementation has an incomplete thread system. In this project, your job is to complete it, and then use it to solve synchronization problems.
Properly synchronized code should work no matter what order the scheduler chooses to run the threads on the ready list. In other words, we should be able to put a call to KThread.yield (causing the scheduler to choose another thread to run) anywhere in your code where interrupts are enabled, and your code should still be correct. You will be asked to write properly synchronized code as part of the later assignments, so understanding how to do this is crucial to being able to do the project.
To aid you in this, code linked in with Nachos will cause KThread.yield to be called on your behalf in a repeatable (but sometimes unpredictable) way. Nachos code is repeatable in that if you call it repeatedly with the same arguments, it will do exactly the same thing each time. However, if you invoke nachos -s <number> with a different number each time, calls to KThread.yield will be inserted at different places in the code.
You will be modifying source code files in the threads subdirectory, and compiling in the proj1 subdirectory. There should be no busy-waiting in any of your solutions to this assignment. (The initial implementation of Alarm.waitUntil is an example of busy-waiting.)
Browse through the initial thread system implementation, starting with KThread.java. This thread system implements thread fork, thread completion, and semaphores for synchronization. It also provides locks and condition variables built on top of semaphores.
Trace the execution path (by hand) for the startup test case provided. When you trace the execution path, it is helpful to keep track of the state of each thread and which procedures are on each thread's execution stack. You will notice that when one thread calls TCB.contextSwitch, that thread stops executing, and another thread starts running. The first thing the new thread does is to return from TCB.contextSwitch. We realize this will seem cryptic to you at first, but you will understand threads once you understand why the TCB.contextSwitch that gets called is different from the TCB.contextSwitch that returns.
Compile and run the baseline implementation of Nachos in the proj1 directory:
% cd nachos/proj1 % make % nachos
The output should be the same as in project 0 before your changes.
Note that only one instance of Alarm may exist at a time (due to a limitation of Nachos), and Nachos already creates one global alarm that is referenced via ThreadedKernel.alarm.
Testing: Implement tests that verify that a thread waits (approximately) for its requested duration; if the wait parameter is 0 or negative, the thread does not wait; multiple threads waiting on the alarm are woken up at the proper times, and in the proper order. For examples and strategies for implementing tests, see the Testing section below.
we say that thread B joins with thread A. When B calls join on A, there are two possibilities. If Ahas already finished, then B returns immediately from join without waiting. If A has not finished, then B waits inside of join until A finishes; when A finishes, it resumes B. Often thread B is called the "parent" and A is called the "child" since a common pattern is for a thread that creates child threads to join on them to wait for them to finish. However, note that any thread can call KThread.join on another (it does not have to be a parent/child relationship).KThread A = new KThread (...); ... A.join ();
Note that: (1) join does not have to be called on a thread. A thread should be able to finish successfully even if no other thread calls join on it; (2) A thread cannot join to itself. (The initial implementation already checks for this case and invokes Lib.assert when it happens. Keep this Lib.assert call in your code); (3) Join can be called on a thread at most once. If thread B calls join on A, then it is an error for B or any other thread C to call join on A again. Assert on this error.
Testing: Implement tests that verify if B calls join on A and A is still executing, B waits; if B calls join on A and A has finished executing, the B does not block; if a thread calls join on itself, Nachos asserts; if join is called more than once on a thread, Nachos asserts; one thread can join with multiple other threads in succession; independent pairs of threads can join with each other without interference.
A thread must have acquired the lock associated with the condition variable when it invokes methods on the CV. The underlying implementation of the Lock class already has code to assert in these cases, but we recommend writing a test program that causes such an error so that you can see what happens.
Testing: Implement tests that verify that sleep blocks the calling thread; wake wakes up at most one thread, even if multiple threads are waiting; wakeAll wakes up all waiting threads; if a thread calls any of the synchronization methods without holding the lock, Nachos asserts; wake and wakeAll with no waiting threads have no effect, yet future threads that sleep will still block (i.e., the wake/wakeAll is "lost", which is in contrast to the semantics of semaphores).
Testing: Implement tests that verify that a thread that calls sleepFor will timeout and return after x ticks if no other thread calls wake to wake it up; a thread that calls sleepFor will wake up and return if another thread calls wake before the timeout expires; sleepFor handles multiple threads correctly (e.g., different timeouts, all are woken up with wakeAll).
A GameMatch is created specifying the number of players N required for a game match. You may assume that N will always be greater than 0. Each player is a separate thread that invokes the play method to join a match with a parameter indicating their ability level. The parameter ability can have one of three values: abilityBeginner, abilityIntermediate, and abilityExpert. play blocks a calling thread until N player threads of the same ability have called play. At that point, all threads of that ability waiting for a match should continue and return from play (but not threads of other abilities). Subsequent threads of that ability calling play will then wait to form a group for the next match. play returns -1 if it is invoked with an unknown ability.
The return value of play is the match number. The first match returned by play has match number 1, and every subsequent match returned by play increments the match number by one, independent of ability. No two matches should have the same match number, match numbers should be strictly monotonically increasing, and there should be no gaps between match numbers. Different instances of a GameMatch have their own independent match number history.
Tip: Implement GameMatch in stages. First implement correct synchronization behavior for one ability, ignoring tracking match numbers. Then extend your implementation to handle all abilities. Then track match numbers.
Testing: Implement tests that verify that play only returns when N threads of the same ability have called it; play can create matches of threads with different abilities; threads do not return from play unless they are matched; multiple GameMatch instances do not interfere with each other; play returns match numbers as specified; play returns an error if invoked with an unknown ability.
Web programming proliferated an asynchronous programming style, where performing some action triggers an asynchronous operation that takes an unknown amount of time (e.g., downloading an image from a Web server). An early design pattern for a program to know when the operation finishes is to associate a callback with the operation: when the operation finishes, the callback is invoked.
Implement the Future class using a Nachos KThread to execute the function asynchronously. Happily, the Future class now enables us to effectively have KThreads return a result.
It is your responsibility to implement your own tests to thoroughly exercise your code to ensure that it meets the requirements specified for each part of the project. Testing is an important skill to develop, and the Nachos projects will help you to continue to develop that skill. You can add calls to testing code in ThreadedKernel.selfTest, and add class-specific code in selfTest methods of each class.
As a testing strategy, first start with simple tests and then implement more complicated tests. When something goes wrong with a simple test, it is easier to pinpoint what aspect of your implementation has a bug. When something goes wrong with a more complicated test, it is more difficult to determine where the bug may be unless you can rule out all the causes that your simple tests have shown to already be correct. We also strongly recommend implementing tests as separate methods, rather than making changes to just one or a few methods. Rather than making a change to an existing test to evaluate new functionality, copy the test into a new method and make the change. That way, your earlier tests are always there in case you need to use them again. You can comment out calls to previous tests so that you can concentrate on one test at a time.
To help you get jumpstarted on testing, here are a handful of example test programs across the various problems:
When implementing your own tests, note that if the main Nachos thread (the first thread that executes) ever returns or exits, then all of Nachos exits (even if other threads have not finished). This behavior is a peculiarity of how Nachos is implemented. If these semantics interfere with your tests then, if you have join implemented, have the main thread join on all children it creates (thereby guaranteeing that the main thread is the last one to finish). If you have only implemented Alarm, then in your Alarm tests have the main thread be the one that has the longest wait time (or you can have the main thread spin in test code until all children have woken up).
Keep in mind that Nachos has a number of command line arguments, two of which are particularly useful for debugging this project:
Our grader will ignore any output that you add with Lib.debug or System.out.println. You do not need to remove these statements for your submission (and some of them likely will come in handy for later projects).
During the project period, you can also use Gradescope to run a snapshot of your code on the sample tests that we have given. See the following Piazza post describing how to use Gradescope to compile and run the sample tests. Important: We will be using Gradescope as the grading platform. Before the deadline, you must submit your code to Gradescope at least once to initialize the grading system for your project.
As a final step, create a file named README in the proj1 directory. The README file should list the members of your group and provide a short description of what code you wrote, how well it worked, how you tested your code, and how each group member contributed to the project. The goal is to make it easier for us to understand what you did as we grade your project in case there is a problem with your code, not to burden you with a lot more work. Do not agonize over wording. It does not have to be poetic, but it should be informative.
We will use a snapshot of your Nachos implementation in your github repository as it exists at the deadline, and grade that version. (Even if you have made changes to your repo after the deadline, that's ok, we will use a snapshot of your code at the deadline.) Important: We will be using Gradescope as the grading platform. Before the deadline, you must submit your code to Gradescope at least once. See the instructions in the Testing section above.
You can discuss concepts with students in other groups, but do not cheat when implementing your project. Cheating includes copying code from someone else's implementation, or copying code from an implementation found on the Internet. See the main project page for more information.
We will manually check and also run code plagiarism tools on submissions and multiple Internet distributions (if you can find it, so can we).