Divide and Conquer Parallel Patterns

Overview:

Demonstrates how common divide and conquer problems including the Fibonacci sequence, merge sort, and quick sort can be parallelised using a parallel divide and conquer skeleton.

---

Compilation/Execution Instructions

Implemented using C++ 17.

Compilation

1. Open a terminal window in the ’divide-conquer-parallel-patterns/src’ directory

2. Run the command ’make’ to run the Makefile

Execution

1. Open a terminal window in the ’divide-conquer-parallel-patterns/src’ directory

2. Run a command, such as the following, to execute the program:

./dac -p -f -m -q

-p, specifies the number of CPU cores to use

-f, specifies that the Fibonacci benchmark should be run

-m, specifies that the merge sort benchmark should be run

-q, specifies that the quick sort benchmark should be run

For example:

./dac -p 4 -f -m -q

The above command runs the Fibonacci, merge sort, and quick sort benchmarks using 4 CPU cores.

The results of each benchmark are written to a file in the ’divide-conquer-parallel-patterns/data’ directory.

The data files produced are named using the benchmark which was run, followed bythe number of CPU cores used, followed by the current repeat reading performed.

For example, the files produced by running the above command are as follows:

fib_4_1.csv, 
fib_4_2.csv, 
fib_4_3.csv,
merge_4_1.csv, 
merge_4_2.csv, 
merge_4_3.csv,
qsort_4_1.csv, 
qsort_4_2.csv, 
qsort_4_3.csv