[General boards] [Fall 2018 courses] [Summer 2018 courses] [Winter 2018 courses] [Older or newer terms]

In Assignment 2, does the reduced system solver from partitioning method 2 need to be parallelized?


#1

If we perform the partitioning and compute the reduced system in parallel, do we also have to implement a parallel solver (i.e., parallel LU, backsub, etc.) for the reduced system itself, or is it OK to solve the final reduced system serially?

Thank you,
Andrei


#2

Only for the solution of the reduced system,
and for the purposes of this assignment,
assume you use total exchange,
to make all of the matrices and vectors needed
for the reduced system available (and common) to all processors,
then duplicate the solution in all processors.

Duplicate means you are doing it sequentially in each processor.

Clearly, this holds only for the step of solving the reduced system,
not for any other step.


#3

Makes perfect sense, thank you for the clarification!