[SOLVED] Cse 3100 systems programming homework #7

In this exercise we will continue to work on the matrix ADT. An implementation of matrix ADT (abstract data type) is given in matrix.c. The API (application programming interface) functions that operates on the matrices are listed in matrix.h. One of the functions, mulMatrix(), performs matrix multiplication, which is implemented in matrix.c. In this assignment, we implement mulMatrix thread() in mmul.c, which has the same interface as mulMatrix(), but performs matrix multiplication with two threads. We only need to change mmul.c. test-mmul.c is provided to test our implementation.

The program takes the following arguments from the command line: the number of rows in the first matrix, the number of columns in the first matrix, and the number of columns in the second matrix. Then it fills two matrices with random numbers, and compares the result of mulMatrix() and mulMatrix thread(). If no argument is specified, the program works on two matrices of size 6 × 6. In addition, if a command line option -t is present, test-mmul prints the time (in seconds) spent on matrix multiplications, using the average of calls.

Here are some sample sessions running test-mmul. The first command multiplies a matrix of 1000 by 500 with a matrix of 500 by 800. The resulting matrix is 1000 by 800. The second command also shows the timing information, the average of calling each multiplication function 3 times. time1 is the average time on mulMatrix() and time2 is the average time on mulMatrix thread(). The numbers are likely to change in different runs. $./test-mmul 1000 500 800 Good work! $./test-mmul 100 500 300 -t3 Good work! num_runs=3 time1=0.0652 time2=0.0341 speedup=1.9132

Suppose p printers need to get j print jobs done. The print jobs are already placed in a queue. The starter code printing.c defines a type job_queue_t for the queue and provides functions to operate on the queue. A printer performs the following operations in a loop. 1. Call q_num_jobs() to get the number of remaining jobs in the queue. 2. If no job is pending, exit from the loop. 3. Call q fetch job() to get a job from the queue. The function returns an integer indicating how long the job takes. 4. Use macro print_job() to print. The macro simulates the fact that different print jobs take different amounts of time to complete. 1

5. Keep track the number of jobs the printer has done. The function printer_single() in the starter code printing.c shows how a single printer completes all the jobs. The tasks in this problem are to use threads to simulate the process of multiple printers completing the print jobs. Each thread is a printer and performs similar operations as printer_single(). Apparently, threads need to coordinate their operations on the queue, which is shared by all printers. A mutex is defined in the job_queue_t structure for this purpose. The program printing takes optional arguments from the command line. An argument can be one of the following. • -p . Specify the number of printers. The default value is 2. • -j . Specify the number of jobs. The default value is 20. • -d. Call the demo function showing the operations of a single printer and exit. Checking results. A script check-printing.py is provided to check the output of printing. Below is an example of how to use check-printing.py. $./printing -p 5 -j 1000 | python3 ./check-printing.py If you have made check-printing.py executable by command “chmod +x ./check-printing.py”, you can run it directly. $./printing -p 5 -j 1000 | ./check-printing.py Note that it is not guaranteed that a program that passes the check is correct. We should also examine the output manually sometimes.

Some synchronization errors may manifest themselves only for some values of the parameters. And even for the same parameters, errors may happen non-deterministically due to different timing and scheduling orders of the threads. You may run the program multiple times even with the same parameters.

For example, the following bash command runs the above example for 10 times. (Yes, it looks like a loop in C! and it may not work in other shells.) for ((n=0;n<10;n++)); do ./printing -p 5 -j 1000 | python3 check-printing.py; done Debugging. gdb supports multithreading. Run your code in gdb until it stops at a breakpoint or appears to stop making progress.

If threads are not making progress, interrupt the execution with Ctrl-C to get to the gdb prompt. Here are some commonly used thread commands. • info threads See what threads are running. • thread n Switch to thread n, where n is a thread number. • thread apply [threadno] [all] args Apply commands to one or more threads.