Drop support for Fortran OpenACC 'openacc_lib.h'?

[Thomas Schwinge]

Hi Jakub!

On 2020-01-27T13:32:05+0100, Jakub Jelinek <jakub@redhat.com> wrote:
> On Mon, Jan 27, 2020 at 01:29:48PM +0100, Thomas Schwinge wrote:
>> Until OpenACC 2.5, in addition to the 'libgomp/openacc.f90' module-based
>> API, there also was a Fortran header file 'openacc_lib.h' described in
>> the specification.  GCC still somewhat supports this, but it's beginning
>> to show signs of bit-rot.  Is there any point in continuing to support
>> this in GCC, or should we get rid of it (now?), changing any libgomp
>> 'include "openacc_lib.h"' test cases to 'use openacc' instead?  (Would it
>> be useful/is there a way to ship a stub 'openacc_lib.h' file that fails
>> compilation with an error "change your code to 'use openacc'"?)

Sorry, that was inaccurate: 'openacc_lib.h' already got removed *in*
OpenACC 2.5, not *after*.  That is, OpenACC 2.0a is the last version to
describe 'openacc_lib.h'.


> I guess primary question is, does GCC claim now full OpenACC 2.6 support,
> or is it still say a mixture of 2.5 and 2.6?

There are some missing pieces, including missing pieces dating back to
OpenACC 2.0a, so:

> If the latter, I'd say it
> would be premature to remove it.

... I don't think these two items ('openacc_lib.h' removal vs. missing
pieces) relate together very much.


My question (suggestion?) was more regarding the Fortran OpenACC user
experience point of view.


Grüße
 Thomas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 658 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20200127/a1d6ec3f/attachment.sig>