Assignment: game of life simulation
Requirements
– This is a group assignment
– Assignments are submitted through Canvas, deadlines are handled by Canvas
– Implement Game of Life as described in https://conwaylife.com/wiki/Conway%27s_Game_of_Life
on a finite grid with a border of permanently dead cells. Parallelize it using OpenMP and/or MPI.
– Code should be of decent quality: documented, commented, proper error handling,
proper use of the C language
– To verify the correctness of your implementation you are encouraged to use
a number of verification patterns, see below.
– For all benchmark executions of the code the board should be of size 3000×3000. You should start
with the ‘grower’ pattern (see below), where cell (0,0) of the pattern should be at cell(1500,1500)
of the board. You should then run the code for 5000 iterations.
– Write a report containing the full code of the implementation, a motivation
of your chosen implementation, data distribution and communication methods, the population
size at the end of a benchmark run, a speedup curve for
at least 1, 2, 4, 8, and 16 processes/thread, and a discussion of this speedup curve.
– A passing grade requires that you achieve at least a speedup of 3 on 16
processes/threads on the benchmark.
– 1 bonus point for achieving a speedup of at least 10 on 16 processes/threads.
– 1 bonus point for achieving a speedup of at least 17 on the lisa cluster.
– You are free to use as many nodes of the lisa cluster as you can get. The use
of GPUs is NOT permitted.
Implementation hints
On Canvas I provide a set of header files with different cell patterns of increasing complexity.
They are used to ensure that your implementation of the Game of Life rules is correct.
You can do a ‘#include’ of the patterns to incorporate them into your code.
– Verify that the ‘beehive’ pattern has a population of exactly 6 in all generations
– Verify that the ‘glider’ pattern has a population of exactly 5 in all generations until
the pattern leaves the board
– Verify that the ‘grower’ pattern has a population of 49 at generation 10, and a population
of 138 at generation 100.
– There are a number of programs and websites to help you visualize the Game of Life. The one I
use is https://golly.sourceforge.net. The file set on Canvas also contains .rle files of
the patterns. Most Game of Life programs, including Golly, can read these .rle files.
NOTE: As mentioned before, code should be of decent quality. Emphatically part of this
is handling of error conditions from MPI function calls, malloc() calls,
file opening, and other functions that can return errors.
In the context of this assignment, error handling can be as simple as
printing a warning and stopping the program. I will deduct points
for missing error handling.
NOTE: You will be running the experiments on the Lisa cluster. Such clusters are heavily used,
so it may take some time before your job is run. Keep this in mind while observing
the deadline. Start early enough to do all the runs you need. Early development can also be
done on your own computer. Hint: the Lisa has a special queue for small test jobs.