From: Tobias Burnus <burnus@net-b.de>
To: Ilmir Usmanov <i.usmanov@samsung.com>
Cc: Thomas Schwinge <thomas@codesourcery.com>,
Evgeny Gavrin <e.gavrin@samsung.com>,
GarbuzovViacheslav <v.garbuzov@samsung.com>,
Dmitri Botcharnikov <dmitry.b@samsung.com>,
gcc-patches@gcc.gnu.org, jakub@redhat.com, fortran@gcc.gnu.org
Subject: Re: [PATCH 2/4] [GOMP4] [Fortran] OpenACC 1.0+ support in fortran front-end
Date: Sat, 08 Mar 2014 19:55:00 -0000 [thread overview]
Message-ID: <531B73F8.1030206@net-b.de> (raw)
In-Reply-To: <5319A358.7040207@samsung.com>
[Resend as it was initially HTML - sorry for those who are CCed.]
Ilmir Usmanov wrote:
> OpenACC 1.0 fortran FE support -- matching and resolving.
> +gfc_match_oacc_cache (void)
> +{
...
> + if (gfc_current_state() != COMP_DO)
> {
> - gfc_free_omp_clauses (c);
> + gfc_error ("ACC CACHE directive must be inside of loop %C");
> + gfc_free_omp_clauses(c);
> return MATCH_ERROR;
> }
Shouldn't it also be supported in DO CONCURRENT? The following is
currently rejected:
real :: b
!$acc loop
outer: do concurrent(i=1:5)
!$acc cache(b)
end do outer
end
(Side question: Is !$acc permitted in DO ... WHILE? If so, you need to
also add EXEC_DO_WHILE.)
> +static void
> +resolve_oacc_positive_int_expr (gfc_expr *expr, const char *clause)
> +{
> + resolve_oacc_scalar_int_expr (expr, clause);
> + if (expr->expr_type == EXPR_CONSTANT && expr->ts.type == BT_INTEGER
> + && expr->value.integer->_mp_size <= 0)
> + gfc_warning ("INTEGER expression of %s clause at %L must be positive",
> + clause, &expr->where);
You shouldn't access internal variables of mpz_t. Use mpz_sgn() instead:
https://gmplib.org/manual/Integer-Comparisons.html
> + if ((sym->ts.type == BT_ASSUMED && sym->attr.pointer)
> + || (sym->ts.type == BT_ASSUMED && CLASS_DATA (sym)
> + && CLASS_DATA (sym)->attr.pointer))
The second line should use BT_CLASS instead of BT_ASSUMED.
> + gfc_error ("POINTER object '%s' of polymorphic type in %s clause at %L",
> + sym->name, name, &loc);
> + if ((sym->ts.type == BT_ASSUMED && sym->attr.cray_pointer)
> + || (sym->ts.type == BT_ASSUMED && CLASS_DATA (sym)
> + && CLASS_DATA (sym)->attr.cray_pointer))
Ditto.
> + gfc_error ("Cray pointer object of polymorphic type '%s' in %s clause at %L",
> + sym->name, name, &loc);
> + if ((sym->ts.type == BT_ASSUMED && sym->attr.cray_pointee)
> + || (sym->ts.type == BT_ASSUMED && CLASS_DATA (sym)
> + && CLASS_DATA (sym)->attr.cray_pointee))
> + gfc_error ("Cray pointee object of polymorphic type '%s' in %s clause at %L",
> + sym->name, name, &loc);
Ditto.
> +static void
> +check_array_not_assumed (gfc_symbol *sym, locus loc, const char *name)
> +{
> + if (sym->as && sym->as->type == AS_ASSUMED_SIZE)
> + gfc_error ("Assumed size array '%s' in %s clause at %L",
> + sym->name, name, &loc);
> + if (sym->as && sym->as->type == AS_ASSUMED_SHAPE)
> + gfc_error ("Assumed shape array '%s' in %s clause at %L",
> + sym->name, name, &loc);
> + if (sym->as && sym->as->type == AS_ASSUMED_RANK)
> + gfc_error ("Assumed rank array '%s' in %s clause at %L",
> + sym->name, name, &loc);
> +}
Actually, I wonder whether one needs to reject assumed-shape: I don't
know what OpenACC says, but my impression is that the problem is that
those can be noncontiguous. However, if they are marked as contiguous
["attr.contiguous"] Â…
On the other hand, your code seems to permit deferred-shape arrays like:
real, pointer :: b(:)
!$acc data copyin(b)
end
The problem is that pointers to deferred-shape arrays can be
noncontiguous. But deferred-shape array are always contiguous when they
are either attr.allocatable or have the "attr.contiguous" attribute.
> + if ((sym->ts.type == BT_ASSUMED && sym->attr.allocatable)
> + || (sym->ts.type == BT_ASSUMED && CLASS_DATA (sym)
> + && CLASS_DATA (sym)->attr.allocatable))
As above: BT_CLASS in the second line.
> +resolve_oacc_deviceptr_clause (gfc_symbol *sym, locus loc, const char *name)
> +{
> + if (sym->ts.type == BT_DERIVED && sym->attr.allocatable)
> + gfc_error ("ALLOCATABLE object '%s' of derived type in %s clause at %L",
> + sym->name, name, &loc);
> + if ((sym->ts.type == BT_ASSUMED && sym->attr.allocatable)
> + || (sym->ts.type == BT_ASSUMED && CLASS_DATA (sym)
> + && CLASS_DATA (sym)->attr.allocatable))
Ditto.
> + gfc_error ("ALLOCATABLE object '%s' of polymorphic type "
> + "in %s clause at %L", sym->name, name, &loc);
> + if (sym->attr.pointer)
> + gfc_error ("POINTER object '%s' in %s clause at %L",
> + sym->name, name, &loc);
Shouldn't you also add
|| (sym->ts.type == BT_CLASS && CLASS_DATA (sym) + && CLASS_DATA
(sym)->attr.class_pointer)
here?
> + case OMP_LIST_USE_DEVICE:
> + if (n->sym->attr.allocatable)
> + gfc_error ("ALLOCATABLE object '%s' in %s clause at %L",
> + n->sym->name, name, &code->loc);
> + if (n->sym->attr.pointer)
> + gfc_error ("POINTER object '%s' in %s clause at %L",
> + n->sym->name, name, &code->loc);
Do you also need to handle BT_CLASS here for allocatable/pointer?
Otherwise, it looks good to me.
Tobias
next prev parent reply other threads:[~2014-03-08 19:48 UTC|newest]
Thread overview: 89+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-01-23 18:01 [PATCH] [GOMP4] " Ilmir Usmanov
2014-01-23 18:03 ` [PATCH 1/6] " Ilmir Usmanov
2014-01-23 18:03 ` [PATCH 2/6] " Ilmir Usmanov
2014-01-23 18:04 ` [PATCH 3/6] " Ilmir Usmanov
2014-01-23 18:05 ` [PATCH 4/6] " Ilmir Usmanov
2014-01-23 18:05 ` [PATCH 5/6] " Ilmir Usmanov
2014-01-23 18:06 ` [PATCH 6/6] " Ilmir Usmanov
2014-01-23 18:09 ` [PATCH 7/6] " Ilmir Usmanov
2014-01-24 19:33 ` [PATCH 4/6] " Thomas Schwinge
2014-11-05 16:29 ` [gomp4] OpenACC cache directive for C Thomas Schwinge
2014-11-05 16:36 ` [gomp4] OpenACC cache directive maintenance (was: [PATCH 4/6] [GOMP4] OpenACC 1.0+ support in fortran front-end) Thomas Schwinge
2014-11-05 16:45 ` [gomp4] OpenACC cache directive maintenance Thomas Schwinge
2015-10-27 15:26 ` [PR fortran/63865] OpenACC cache directive: match Fortran support with C/C++ (was: [gomp4] OpenACC cache directive maintenance) Thomas Schwinge
2015-10-27 15:30 ` Jakub Jelinek
2015-10-27 17:03 ` [PR fortran/63865] OpenACC cache directive: match Fortran support with C/C++ Thomas Schwinge
2014-11-05 16:49 ` [gomp4] Testing of C/C++ OpenACC cache directive (was: OpenACC cache directive for C) Thomas Schwinge
2016-06-02 11:47 ` [PR c/71381] C/C++ OpenACC cache directive rejects valid syntax (was: [gomp4] OpenACC cache directive for C.) Thomas Schwinge
2016-06-08 13:29 ` [PING] [PR c/71381] C/C++ OpenACC cache directive rejects valid syntax Thomas Schwinge
2016-06-08 14:07 ` Jakub Jelinek
2016-06-10 10:32 ` Thomas Schwinge
2016-06-10 13:14 ` Thomas Schwinge
2016-06-10 20:40 ` Gerald Pfeifer
2014-01-24 20:47 ` [PATCH 2/6] [GOMP4] OpenACC 1.0+ support in fortran front-end Thomas Schwinge
2014-01-24 20:31 ` [PATCH 1/6] " Thomas Schwinge
2014-01-27 19:37 ` Tobias Burnus
2014-01-24 18:04 ` [PATCH] " Thomas Schwinge
2014-01-27 13:12 ` Ilmir Usmanov
2014-01-27 15:49 ` Thomas Schwinge
2014-01-27 16:35 ` Ilmir Usmanov
2014-01-31 11:14 ` Ilmir Usmanov
2014-01-31 11:16 ` [PATCH 1/6] " Ilmir Usmanov
2014-01-31 11:17 ` [PATCH 2/6] " Ilmir Usmanov
2014-01-31 11:18 ` [PATCH 3/6] " Ilmir Usmanov
2014-01-31 11:22 ` [PATCH 4/6] " Ilmir Usmanov
2014-01-31 11:34 ` [PATCH 5/6] " Ilmir Usmanov
2014-01-31 11:45 ` [PATCH 6/6] " Ilmir Usmanov
2014-02-09 23:43 ` [PATCH 5/6] " Tobias Burnus
2014-02-10 8:52 ` Thomas Schwinge
2014-02-10 9:34 ` Ilmir Usmanov
2014-02-10 23:13 ` Tobias Burnus
2014-02-10 9:45 ` Ilmir Usmanov
2014-02-10 10:52 ` Thomas Schwinge
2014-02-11 16:51 ` [PATCH 4/6] " Thomas Schwinge
2014-02-13 13:15 ` Ilmir Usmanov
2014-02-13 14:57 ` Thomas Schwinge
2014-02-14 5:45 ` Ilmir Usmanov
2014-02-21 19:29 ` [GOMP4] gimple_code_is_oacc -> is_gimple_omp_oacc_specifically (was: [PATCH 4/6] [GOMP4] OpenACC 1.0+ support in fortran front-end) Thomas Schwinge
2014-02-09 23:24 ` [PATCH 3/6] [GOMP4] OpenACC 1.0+ support in fortran front-end Tobias Burnus
2014-02-09 23:10 ` [PATCH 2/6] " Tobias Burnus
2014-02-10 9:10 ` Thomas Schwinge
2014-02-19 16:43 ` Ilmir Usmanov
2014-01-31 12:00 ` [PATCH 1/6] " Jakub Jelinek
2014-01-31 12:33 ` Ilmir Usmanov
2014-02-03 15:21 ` [PING] " Ilmir Usmanov
2014-02-09 22:22 ` Tobias Burnus
2014-02-19 15:34 ` Ilmir Usmanov
2014-02-19 23:52 ` Tobias Burnus
2014-02-20 8:19 ` Ilmir Usmanov
2014-03-04 7:56 ` [PATCH 1/4] [GOMP4] [Fortran] " Ilmir Usmanov
2014-03-04 7:57 ` Ilmir Usmanov
2014-03-04 7:57 ` [PATCH 2/4] " Ilmir Usmanov
2014-03-04 7:58 ` [PATCH 3/4] " Ilmir Usmanov
2014-03-04 7:59 ` [PATCH 4/4] " Ilmir Usmanov
2014-03-04 22:56 ` Tobias Burnus
2014-03-04 22:52 ` [PATCH 3/4] " Tobias Burnus
2014-03-04 17:20 ` [PATCH 1/4] " Tobias Burnus
2014-03-07 10:44 ` Ilmir Usmanov
2014-03-07 10:45 ` Ilmir Usmanov
2014-03-07 10:46 ` [PATCH 2/4] " Ilmir Usmanov
2014-03-07 10:46 ` [PATCH 3/4] " Ilmir Usmanov
2014-03-07 10:47 ` [PATCH 4/4] " Ilmir Usmanov
2014-03-08 17:55 ` Tobias Burnus
2014-03-20 10:53 ` Thomas Schwinge
2014-03-20 12:48 ` Ilmir Usmanov
2014-03-20 14:43 ` Jakub Jelinek
2014-03-08 17:19 ` [PATCH 3/4] " Tobias Burnus
2014-03-08 19:55 ` Tobias Burnus [this message]
2014-03-11 12:04 ` [PATCH 2/4] " Ilmir Usmanov
2014-03-12 18:46 ` Tobias Burnus
2014-03-12 18:27 ` [PATCH 1/4] " Tobias Burnus
2014-03-13 9:41 ` Ilmir Usmanov
2014-03-13 11:43 ` Thomas Schwinge
2014-03-13 13:24 ` Ilmir Usmanov
2014-03-13 14:13 ` Ilmir Usmanov
2014-03-16 19:46 ` Tobias Burnus
2014-03-16 20:44 ` Thomas Schwinge
2014-04-05 10:40 ` Thomas Schwinge
2014-03-10 15:44 ` Thomas Schwinge
2014-03-04 17:42 ` Tobias Burnus
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=531B73F8.1030206@net-b.de \
--to=burnus@net-b.de \
--cc=dmitry.b@samsung.com \
--cc=e.gavrin@samsung.com \
--cc=fortran@gcc.gnu.org \
--cc=gcc-patches@gcc.gnu.org \
--cc=i.usmanov@samsung.com \
--cc=jakub@redhat.com \
--cc=thomas@codesourcery.com \
--cc=v.garbuzov@samsung.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).