Performance of Loop-Level Model

• Fine Grain Overhead

  • Start a parallel region each time
  • Frequent synchronization

• Fraction of non-parallel work will dominate

  • Amdahl’s law

• Loops may really change layout of data

  • Lots of data motion

• Not scalable to large number of threads

C$OMP PARALLEL DO

do i=1,n

…………

enddo

alpha = xnorm/sum

C$OMP PARALLEL DO

do i=1,n

…………

enddo

C$OMP PARALLEL DO

do i=1,n

…………

enddo

Previous slide

Next slide

Back to first slide

View graphic version