Skip to main content

PWD009: Incorrect privatization in OpenMP parallel region

Issue​

A variable is being incorrectly privatized in the OpenMP datascoping clauses.

Actions​

Change the data scope of the variable from private to shared.

Relevance​

Specifying an invalid scope for a variable may introduce race conditions and produce incorrect results. For instance, when a variable must be shared among threads but it is privatized instead.

Code example​

C​

In the following code, C is incorrectly privatized since the threads operate on non-overlapping positions of the array:

void example(int m, double *A, double *B, double *C) {
  double temp;

  #pragma omp parallel for default(none) private(i, temp, C) shared(A, B, m)
  for (int i = 0; i < m; i++) {
    temp = A[i] * B[i];
    C[i] = C[i] + temp;
  }
}

To fix this, C should be moved to a shared clause:

void example(int m, double *A, double *B, double *C) {
  double temp;

  #pragma omp parallel for default(none) private(i, temp) shared(A, B, C, m)
  for (int i = 0; i < m; i++) {
    temp = A[i] * B[i];
    C[i] = C[i] + temp;
  }
}

Fortran​

In the following code, C is incorrectly privatized since the threads operate on non-overlapping positions of the array:

subroutine example(A, B, C)
  real, intent(in) :: A(:), B(:)
  real, intent(inout) :: C(:)
  real :: temp
  integer :: i

  !$omp parallel do default(none) private(i, temp, C) shared(A, B)
  do i = 1, size(C, 1)
    temp = A(i) * B(i)
    C(i) = C(i) + temp
  end do
end subroutine example

To fix this, C should be moved to a shared clause:

subroutine example(A, B, C)
  real, intent(in) :: A(:), B(:)
  real, intent(inout) :: C(:)
  real :: temp
  integer :: i

  !$omp parallel do default(none) private(i, temp) shared(A, B, C)
  do i = 1, size(C, 1)
    temp = A(i) * B(i)
    C(i) = C(i) + temp
  end do
end subroutine example

References​