Forks

Both programs use shared memory.

Gauss method

Generate a random matrix of size 5 and bring it to the upper triangular form according to the Gauss method:

./fork-gauss -n 5 -c 4 -d

Flags:

• -n: matrix size (default: 5).
• -c: core count (if not specified, the count of cores is taken from the system information).
• -d: debug mode (print some additional information).

Merge sort

Generate a random array of size 10 and sort it using a combination of bubble sort and merge sort:

./fork-merge -n 20 -d

Flags:

• -n: array size (default: 10).
• -d: debug mode (print some additional information).

Threads

Threads work exactly the same way using pthread:

./thread-gauss -n 4 -n 2 -d
./thread-merge -n 20

Program speed

Let's compare forks and threads by speed. Let's run the Gauss method with the same parameters:

yukatan@LEO-DEB:~/Forks$ time ./fork-gauss -n 1000 -c 4
Count of cores: 4
Matrix size: 1000

Done.

real	0m2,460s
user	0m2,188s
sys	0m0,644s

yukatan@LEO-DEB:~/Threads$ time ./thread-gauss -n 1000 -c 4
Count of cores: 4
Matrix size: 1000

Done.

real	0m1,417s
user	0m5,104s
sys	0m0,136s

Let's compare merge sort:

yukatan@LEO-DEB:~/Forks$ time ./fork-merge -n 50000
Array size: 50000
Done.

real	0m4,564s
user	0m9,112s
sys	0m0,000s

yukatan@LEO-DEB:~/Threads$ time ./thread-merge -n 50000
Array size: 50000
Done.

real	0m2,980s
user	0m5,924s
sys	0m0,000s

So the threads are obviously faster.