OpenMP parallel loop incorrect iteration distribution?

[samlor]

G'day All,

The simple program enclosed seems to indicate a problem in ifort OpenMP compilation of parallel DO loop work-sharing directives that distributes increasing numbers of iterations instead of spreading them evenly amongst threads when the loop index variable 'I' is re-used in later loop statements inside a SINGLE directive. The problem does not occur when the program is modified to use a different variable 'J' instead for the 2nd loop. Nor does this problem occur in equivalent C code.

Am I doing something wrong or is this a bug? And, where/how should it be reported?

Regards,
Sam

Code:

$ cat incorrect.f
      PROGRAM MAIN
      INTEGER I, J, OMP_GET_THREAD_NUM, OMP_GET_NUM_THREADS
!$OMP PARALLEL PRIVATE(I)
!$OMP DO
      do I=1,OMP_GET_NUM_THREADS()
        PRINT *, 'TID =', OMP_GET_THREAD_NUM(), 'I =', I
      enddo
!$OMP END DO
!$OMP BARRIER
!$OMP SINGLE
      do I=1,OMP_GET_NUM_THREADS()
        PRINT *, 'TID =', OMP_GET_THREAD_NUM(), 'I =', I
      enddo
!$OMP END SINGLE
!$OMP END PARALLEL
      END
$ cat correct.f
      PROGRAM MAIN
      INTEGER I, J, OMP_GET_THREAD_NUM, OMP_GET_NUM_THREADS
!$OMP PARALLEL PRIVATE(I)
!$OMP DO
      do I=1,OMP_GET_NUM_THREADS()
        PRINT *, 'TID =', OMP_GET_THREAD_NUM(), 'I =', I
      enddo
!$OMP END DO
!$OMP BARRIER
!$OMP SINGLE
      do J=1,OMP_GET_NUM_THREADS()
        PRINT *, 'TID =', OMP_GET_THREAD_NUM(), 'J =', J
      enddo
!$OMP END SINGLE
!$OMP END PARALLEL
      END
$ uname -a
Linux swan.ac3.com.au 2.4.21-sgi303r2 #1 SMP Mon Nov 29 15:29:38 PST 2004 ia64 i
a64 ia64 GNU/Linux
$ ifort -v
Version 9.0
$ ifort -openmp incorrect.f -o incorrect
incorrect.f(10) : (col. 6) remark: OpenMP DEFINED LOOP WAS PARALLELIZED.
incorrect.f(9) : (col. 6) remark: OpenMP DEFINED REGION WAS PARALLELIZED.
$ echo $OMP_NUM_THREADS
8
$ ./incorrect
 TID =           0 I =           1
 TID =           2 I =           1
 TID =           2 I =           2
 TID =           2 I =           3
 TID =           1 I =           1
 TID =           1 I =           2
 TID =           7 I =           1
 TID =           7 I =           2
 TID =           7 I =           3
 TID =           7 I =           4
 TID =           7 I =           5
 TID =           7 I =           6
 TID =           7 I =           7
 TID =           7 I =           8
 TID =           3 I =           1
 TID =           3 I =           2
 TID =           3 I =           3
 TID =           3 I =           4
 TID =           4 I =           1
 TID =           4 I =           2
 TID =           4 I =           3
 TID =           4 I =           4
 TID =           4 I =           5
 TID =           5 I =           1
 TID =           5 I =           2
 TID =           5 I =           3
 TID =           5 I =           4
 TID =           5 I =           5
 TID =           5 I =           6
 TID =           6 I =           1
 TID =           6 I =           2
 TID =           6 I =           3
 TID =           6 I =           4
 TID =           6 I =           5
 TID =           6 I =           6
 TID =           6 I =           7
 TID =           4 I =           1
 TID =           4 I =           2
 TID =           4 I =           3
 TID =           4 I =           4

 TID =           4 I =           5
 TID =           4 I =           6
 TID =           4 I =           7
 TID =           4 I =           8
$ ifort -openmp correct.f -o correct
correct.f(7) : (col. 6) remark: OpenMP DEFINED LOOP WAS PARALLELIZED.
correct.f(6) : (col. 6) remark: OpenMP DEFINED REGION WAS PARALLELIZED.
$ ./correct
 TID =           1 I =           2
 TID =           0 I =           1
 TID =           2 I =           3
 TID =           6 I =           7
 TID =           3 I =           4
 TID =           7 I =           8
 TID =           4 I =           5
 TID =           5 I =           6
 TID =           4 J =           1
 TID =           4 J =           2
 TID =           4 J =           3
 TID =           4 J =           4
 TID =           4 J =           5
 TID =           4 J =           6
 TID =           4 J =           7
 TID =           4 J =           8
$ 

[Steven_L_Intel1]

I'm not familiar with the issues so can't comment on your code, but one reports issues through Intel Premier Support. See this page to get started.