Discussion 8: Wavefront Parallelization

  1. For prog1, compare the performance variation for the transformations: (i,j->p,t) = (i,j->j,i+j-1) and (i,j->p,t) = (i,j->i,i+j-1).
  2. For prog2, what were the transformations that you came up with?
  3. For prog2, were you able to account for the memory based dependencies?
  4. What were your performance gain/loss on 1-8 processors for both the programs?
  5. Mention at-least two drawbacks of wavefront parallelization when compared to the parallelizations you did for PA1.
    (Note: Trade-off is the performance gain that you obtain)
Submit individual report, come to a group conclusion and submit a single group report on Piazza.