Re: [WIP] [PR fortran/72741] Rework Fortran OpenACC routine clause handling

[Thomas Schwinge <thomas@codesourcery.com>]

Hi!

On Mon, 15 Aug 2016 18:54:49 -0700, Cesar Philippidis <cesar@codesourcery.com> wrote:
> For the moment, I'm ignoring the
> device_type problem and handling all of the matching errors in
> gfc_match_oacc_routine.

OK for the moment; my idea has been to do it generally enough already
now, using generic infrastructure I have been/will be adding for C/C++,
so that device_type support will later be simple to implement for all
three front ends.  But, let's leave that aside for the moment.

> You're patch was handling those errors in
> add_attributes_to_decls, which I think is too late.

I can't tell why that's "too late".  Anyway, we can save this discussion
for later.  ;-)

> Thomas, does this patch ok to you for gomp4?

Yes, please commit, so that we can move this whole thing forward.  :-)

A few quick comments anyway:

> --- a/gcc/fortran/openmp.c
> +++ b/gcc/fortran/openmp.c
> @@ -1993,19 +2002,24 @@ gfc_match_oacc_routine (void)

>    dims = gfc_oacc_routine_dims (c);
>    if (dims == OACC_FUNCTION_NONE)
>      {
>        gfc_error ("Multiple loop axes specified in !$ACC ROUTINE at %C");
> -      goto cleanup;
> +
> +      /* Don't abort early, because it's important to let the user
> +	 know of any potential duplicate routine directives.  */
> +      seen_error = true;
>      }

Hmm, I don't know if that's really important?  I mean, if we run into
"Multiple loop axes specified", that is a hard semantic error already?
Anyway, this can be reconsidered later.

>    if (isym != NULL)
>      {
>        if (c && (c->gang || c->worker || c->vector))
>  	{
> -	  gfc_error ("Intrinsic function specified in !$ACC ROUTINE ( NAME )"
> -		     " at %C, with incompatible GANG, WORKER, or VECTOR clause");
> +	  gfc_error ("Intrinsic symbol specified in !$ACC ROUTINE ( NAME ) "
> +		     "at %C, with incompatible clauses specifying the level "
> +		     "of parallelism");
>  	  goto cleanup;
>  	}

You're re-introducing the wording I had used earlier, before I changed
that to the more specific one mentioning the clause names.  Why change
that again?  Also something the can be reconsidered later.  (Goes
together with the gcc/testsuite/gfortran.dg/goacc/pr72741-intrinsic-2.f
changes.)

> --- a/gcc/testsuite/gfortran.dg/goacc/pr72741-intrinsic-1.f
> +++ b/gcc/testsuite/gfortran.dg/goacc/pr72741-intrinsic-1.f
> @@ -1,17 +1,13 @@
> -! Check for valid clauses with intrinsic function specified in !$ACC ROUTINE ( NAME ).
> -
>        SUBROUTINE sub_1
>        IMPLICIT NONE
> -!$ACC ROUTINE (ABORT)
> -!$ACC ROUTINE (ABORT) SEQ
> +!$ACC ROUTINE (ABORT) SEQ VECTOR ! { dg-error "Intrinsic symbol specified in \\!\\\$ACC ROUTINE \\( NAME \\) at \\(1\\), with incompatible clauses specifying the level of parallelism" }
>  
>        CALL ABORT
>        END SUBROUTINE sub_1
>  
>        MODULE m_w_1
>        IMPLICIT NONE
> -!$ACC ROUTINE (ABORT) SEQ
> -!$ACC ROUTINE (ABORT)
> +!$ACC ROUTINE (ABORT) VECTOR GANG ! { dg-error "Intrinsic symbol specified in \\!\\\$ACC ROUTINE \\( NAME \\) at \\(1\\), with incompatible clauses specifying the level of parallelism" }

This changes the intention of this test file?  Another thing that can be
reconsidered later.

So, please commit as-is, and I'll then base my other changes on top of
that.


Grüße
 Thomas