Example: Data Distribution

• To avoid the bottle neck resulting from memory allocation on a single node:

  • Perform initialization in parallel, such that each processor initializes data that it is likely to access later for calculation
  • Use explicit data placement compiler directives (e.g. C$ distribute )
  • Enable environment variable for data placement (setenv _DSM_PLACEMENT)
  • Use dplace command (with appropriate options) to start the program execution

REAL*8 A(N),B(N),C(N),D(N)

C$DISTRIBUTE A(BLOCK),B(BLOCK),C(BLOCK),D(BLOCK)

DO I=1,N

A(I) = 0.0

B(I) = I/2

C(I) = I/3

D(I) = I/7

ENDDO

C$OMP PARALLEL DO

DO I=1,N

A(I) = B(I) + C(I) + D(I)

ENDDO

Previous slide

Next slide

Back to first slide

View graphic version