I have an inherited Fortran90 program that I'm attempting to parallelize using OpenMP. The main issue is a nested loop that calls a computationally expensive subroutine many times (specifically, it calculates the confluent hypergeometric function with complex parameters and argument). I first did some local testing with the following program:

PROGRAM Parallel_Loop_Test
USE OMP_LIB

INTEGER :: numprod(10, 10)

numprod(1,1) = 0

OPEN(10,file='mptest.dat',status='unknown')
    
! Enclose nestable loop in parallel, then !$OMP DO commands
!$OMP PARALLEL
PRINT *, "Hello from process: ", OMP_GET_THREAD_NUM()
    !$OMP DO
    DO i=1,10
        DO j = 1,10
        numprod(i,j) = i*j
        ENDDO
    ENDDO
    !$OMP ENDDO
!$OMP END PARALLEL

! array structure preserves ordering, can do serial write
DO i = 1, 10
    DO j = 1, 10
    write(10,*) i,j, numprod(i,j)
    ENDDO
ENDDO

CLOSE(10)
END

My actual code has 5 nested loops and runs on up to 112 cores; when I attempt to implement the above framework with that code, it never leaves the loop. I can tell this because I made it write to file in the parallelized loop for testing, and even though it writes all the values I expect it to, it never hits the print statement after the loop saying that all the values have been calculated. I suspect I just don't understand something about how OpenMP behaves with nested loops, but I'm having a hard time finding a clear explanation on that front.