Name the two independent dimensions used in Flynn’s taxonomy to classify multi – processor computer architectures.. Provide a brief definition of each category and explain what types of problems would benefi t the most from each implementation.. Which category of Flynn’s taxonomy do Graphical Processing Units (GGPU)) fall in and why?? (225 marks))
Question 2 – Parallel Patterns
Explain the reduce pattern and provide two example combiner functions for it.. What are th e requirements for the combiner functions in the reduce pattern?? Select one combiner function and demonstrate the reduction process on 8 elements {11,2,3,4,5,6,7,8}} in graphical form.. What is the span and work complexity of the reduce pattern?? (225 marks))
Qu estion 3 – Parallel Algorithms
What is the main idea behind the optimisation technique called “pprivatisation”” ? Describe the application of this technique for histogram calculations and provide detail how this algorithm would be implemented on a parallel ha rdware . Summarise the benefits and limitations of this approach.. (225 marks))
Question 4 – Performance
The following graph depicts the theoretical maximum performance gain (vvertical axis)) obtained from parallelising and running the program on a number of pa rallel units (hhorizontal axis)).. Individual lines correspond to an amount of serial (nnon – parallelisable)) code in the program..
Explain the graph by using the Amdahl’s law and describe how this affects the parallelisation techniques.. Using the concept of efficiency,, describe further impact on parallelisation of computer programs.. (225 marks))