[PATCH 1/4] [GOMP4] [Fortran] OpenACC 1.0+ support in fortran front-end

[Tobias Burnus]

Hi Ilmir,

thanks for the update.

Ilmir Usmanov wrote:
>> I'd use "integer" instead of "INTEGER" as it is not a 'reserved' word 
> OpenMP implementation uses capital letters, so, perhaps, we also 
> should do this, for consistency?

Fine with me - I don't feel strong about it.

> OpenACC 1.0 support to fortran FE -- core.
>
> --- a/gcc/fortran/match.c
> +++ b/gcc/fortran/match.c
> @@ -2595,6 +2595,33 @@ match_exit_cycle (gfc_statement st, gfc_exec_op op)
>     if (cnt > 0
>         && o != NULL
>         && o->state == COMP_OMP_STRUCTURED_BLOCK
> +      && (o->head->op == EXEC_OACC_LOOP
> +          || o->head->op == EXEC_OACC_PARALLEL_LOOP))
> +    {
> +      int collapse = 1;
> +      gcc_assert (o->head->next != NULL
> +                  && (o->head->next->op == EXEC_DO
> +                      || o->head->next->op == EXEC_DO_WHILE)
> +                  && o->previous != NULL
> +                  && o->previous->tail->op == o->head->op);
> +      if (o->previous->tail->ext.omp_clauses != NULL
> +          && o->previous->tail->ext.omp_clauses->collapse > 1)
> +        collapse = o->previous->tail->ext.omp_clauses->collapse;
> +      if (st == ST_EXIT && cnt <= collapse)
> +        {
> +          gfc_error ("EXIT statement at %C terminating !$ACC LOOP loop");
> +          return MATCH_ERROR;
> +        }
> +      if (st == ST_CYCLE && cnt < collapse)
> +        {
> +          gfc_error ("CYCLE statement at %C to non-innermost collapsed"
> +                     " !$ACC LOOP loop");
> +          return MATCH_ERROR;
> +        }
> +    }

Can you  - if you want as follow up patch - also handle the following 
for both your OpenACC code and for the existing OpenMP code:

!$acc loop
outer_loop: do i = 1, n
    inner_loop: do j = 1,m
       ! ...
       cycle outer_loop
       exit outer_loop
    end do inner_loop
end do outer_loop

That's a new Fortran 2008 feature that CYCLE and EXIT can leave the 
inner DO loop.


For DO CONCURRENT, that's already handled via the code in 
"match_exit_cycle". If you want to try the do-concurrent code, replace 
"inner_loop: do j = 1,m" by "inner_loop: do_concurrent(j=1:m)".


Otherwise, it looks good do me. I don't know whether Thomas has 
additional comments.

Tobias