------- Comment #3 from tkoenig at gcc dot gnu dot org 2009-11-21 18:31 -------
(In reply to comment #2)
> Sorry, Steve - my mistake.
>
> The original message should have been:
>
> To illustrate this with a simple example:
>
> DO I = M1, M2, M3
> B(I) = A(I)
> ENDDO
>
> would be most easily, and straightforwardly, implemented as follows:
>
> IF (M3 > 0 .AND. M1 > M2) GOTO 200 ! Loop executed zero times
> IF (M3 < 0 .AND. M1 < M2) GOTO 200 ! Ditto
> ITEMP = (M2 - M1 + M3) / M3 ! Temporary loop counter
> I = M1
> 100 CONTINUE
> B(I) = A(I)
> ITEMP = ITEMP - 1 ! Adjust internal loop counter
> I = I + M3 ! Adjust DO loop variable
> IF (ITEMP >= 0) GOTO 100
> 200 CONTINUE
As written, this executes the assignment one time too many.
The last but one line should read:
IF (ITEMP > 0) GOTO 100
What we generate has the test at the bottom of the loop,
whereas 8.1.6.4.2 puts the test of the iteration count first.
What we could generate, IMHO, is
read (*,*) m1, m2, m3
ITEMP = (M2 - M1 + M3) / M3 ! Temporary loop counter
print *,'itemp=',itemp
I = M1
100 CONTINUE
IF (ITEMP <= 0) GOTO 200
print *,i
I = I + M3 ! Adjust DO loop variable
ITEMP = ITEMP - 1 ! Adjust internal loop counter
GOTO 100
200 CONTINUE
print *,"final : i = ", i
end
Does this have any drawbacks? Would the middle end recognize this for
vectorization and loop unrolling?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42131
