Parallel computation in CFD

• Parallel computation only in vector assignments. Subscript is *.

• Communication between PEs by adding shift to *:

  DIFP(*) = P(* + 1) – P(* - 1)
  

• Conditional vector assignment:

  IF(A(*) .LT. 0) A(*) = -A(*)
  

• PEs can locally access different locations:

  DIAG(*) = RHO(*, X(*))
  

Vector subscript must be in sequential dimension.

Previous slide

Next slide

Back to first slide

View graphic version