Note: For this exercise, you may use the provided scratch pa…
Note: For this exercise, you may use the provided scratch paper to perform calculations. Please write your name and G# on the top of the page and return to the proctor when you have completed the exam. Galactica 3000, is a galaxy simulator developed by the (fictitious) computational astrophysics institute at GMU. It is used to model the evolution of galaxies. A simulation consists of two main parts: Gravitational calculations (80% of the runtime): These are parallelizable and can be sped up by adding more processors. Black hole collision predictions (20% of the runtime): Due to the complexity of interactions, this part is not parallelizable. The institute is considering purchasing a new supercomputer with 16 processors to speed up the simulation. The director asks you to evaluate the expected speedup of Galactica 3000 when run on this new hardware. Question 1: Using Amdahl’s law, calculate the theoretical speedup of the simulation on the 16-processor supercomputer. Question 2: Suppose the engineers manage to improve the code so that 95% of the runtime becomes parallelizable. Recalculate the theoretical speedup for this improved version. Question 3: The director wants the simulation to run 10× faster than the current version. Is it possible with the 16-processor supercomputer? If not, how much of the program must be parallelizable to achieve this goal?
Read Details