This specification describes a project which just barely uses a mutex. Rather than concentrating on a difficult synchronization problem, this project is more of a summative challenge in using C++ threads, getopt and writing robust code. It is intended to be relatively straight forward, a gentle start in our learning about concurrency.
You are expected to support and implement all of these command line options.
Prints program usage. Your text should include any defaults and limitations on values, printed nicely.
Specifies the size of a vector of uint32_t. The minimum is (1 << 8). The default for this parameter (and also its maximum allowable value) is (1 << 20).
Specifies the value of the random number generator seed. The default is to use the time-of-day. If this command line argument is given, use its value as the seed instead.
Specifies the number of threads to create. The minimum value (and also the default value) is 4. The maximum value is 30.
You will be expected to ensure various command line option combinations make sense. Specifically, any minimums or maximums specied above must be enforced. Any violation of these should produce an error message and stop the program.
This is kind of like the card game spit.
Some number of threads each have a list of numbers they will respond to. They each are looking at the "pile" initialized to 0. The thread that has the current value of the "pile" in its list, increments the "pile" to the next number.
Clearly, the "pile" is the shared resource that must be protected.
When a thread has "played" every number in its list, it ends.
Your program will create a large vector of uint32_t. The integers will run the range of 0 to the size of the vector - 1. In other words, something like this to start with:
for (uint32_t i = 0; i < size; i++) v[i] = i;
After initializing the vector, you are to shuffle it so that the numbers are no longer consecutive. It is absoltely critical that each value from 0 to size - 1 be in the vector only once.
Think of this large list as a shuffled deck of cards.
Divy the vector up into nearly equal size subranges and give each subrange to one of the threads. By way of example, imagine the following scenario (using a vector size far below the minimum size of the actual program):
size = 13
vector = { 11, 4, 7, 2, 9, 3, 6, 1, 10, 0, 5, 8, 12 }
threads = 4
thread[0] gets { 11, 4 and 7 }
thread[1] gets { 2, 9 and 3 }
thread[3] gets { 6, 1 and 10 }
thread[4] gets { 0, 5, 8 and 12 }
Notice all the threads got a roughly equal hand size for its own work. The last thread gets any strays.
Once the hands have been dealth, launch all the threads. As each thread plays a number, print out that fact. For example:
$ ./a.out -t 8 -z 300 Thread: 5 hit on: 0 Thread: 7 hit on: 1 Thread: 3 hit on: 2 Thread: 0 hit on: 3 Thread: 2 hit on: 4 Thread: 3 hit on: 5 Thread: 0 hit on: 6 Thread: 0 hit on: 7 Thread: 3 hit on: 8 Thread: 5 hit on: 9 ... Thread: 1 hit on: 294 Thread: 0 hit on: 295 Thread: 4 hit on: 296 Thread: 0 hit on: 297 Thread: 2 hit on: 298 Thread: 7 hit on: 299 $
Notice, the output will come out in order and with no string stomping on any other string. Also, use iomanip to space the strings out so as to allow for numbers with different numbers of digits to print lined up nicely.
The main program will simply attempt to join all the threads it launched. If the program is correctly written, all threads will end and become joinable. When the main program has joined all the threads it launched, it will end.
My version is about 130 lines without any meaningful comments. This is stated only to set expectations, not to pose a challenge.
Grading for this project includes your meaningful commenting and a beautiful code design leveraging C++ classes, and vector / iterators.
For example, here's my main():
int main(int argc, char **argv) { uint64_t size; uint32_t nthreads; uint64_t seed; IntVec numbers; HandleOptions(argc, argv, size, seed, nthreads); InitializeNumbers(numbers, seed, size); LaunchThreads(numbers, nthreads, size); JoinThreads(); return 0; }
IntVec is a typedef for vector<uint32_t> or more precisely, the updated C++ version of typedef. You don't have to do this but if you don't, you will find yourself typing vector<uint32_t> over and over again.
using IntVec = vector<uint32_t>;
You don't have to follow this outline above, of course. However, your code must be well designed. Be specific about your types (i.e. use cinttypes).
I will be grading on MacOS command line (not xcode)
There can be no warnings generated by your code. This is the command line I will use to build your program:
g++ -Wall -std=c++11 -O3 yourprog.cpp -lpthread
Rather than deducting 10 points, any warnings will result in a grade of 0.