[OpenMP] Issues with Work-sharing region errors in ludcmp, fdtd-2d

[4fterlook]

In the ludcmp benchmark, the barrier is inside a parallel for loop which results in a error during compiling. There're other benchmarks that have similar problems.
#pragma scop b[0] = 1.0; #pragma omp parallel { #pragma omp for private (j, k, w) for (i = 0; i < _PB_N; i++) { for (j = i+1; j <= _PB_N; j++) { w = A[j][i]; for (k = 0; k < i; k++) w = w- A[j][k] * A[k][i]; A[j][i] = w / A[i][i]; } #pragma omp barrier for (j = i+1; j <= _PB_N; j++) { w = A[i+1][j]; for (k = 0; k <= i; k++) w = w - A[i+1][k] * A[k][j]; A[i+1][j] = w; } } y[0] = b[0]; #pragma omp for private (j, w) for (i = 1; i <= _PB_N; i++) { w = b[i]; for (j = 0; j < i; j++) w = w - A[i][j] * y[j]; y[i] = w; } x[_PB_N] = y[_PB_N] / A[_PB_N][_PB_N]; #pragma omp for private (j, w) for (i = 0; i <= _PB_N - 1; i++) { w = y[_PB_N - 1 - (i)]; for (j = _PB_N - i; j <= _PB_N; j++) w = w - A[_PB_N - 1 - i][j] * x[j]; x[_PB_N - 1 - i] = w / A[_PB_N - 1 - (i)][_PB_N - 1-(i)]; } } #pragma endscop
This part of code just doesn't compile, it gives errors about work-sharing region may not be closely nested inside of work-sharing, critical, ordered, master or explicit task region

Is there any way to fix this issue? Is it because of a mistake in implementation or it's just tricks in openmp that I don't know?

[avacoder42]

yes, I have the same error. Did you find a solution ?