From: Julian Brown <julian@codesourcery.com>
To: <gcc-patches@gcc.gnu.org>
Cc: Thomas Schwinge <thomas@codesourcery.com>, <fortran@gcc.gnu.org>,
Jakub Jelinek <jakub@redhat.com>,
Tobias Burnus <tobias@codesourcery.com>,
"Moore, Catherine" <Catherine_Moore@mentor.com>
Subject: [PATCH 2/9] [OpenACC] GOMP_MAP_ATTACH handling in find_group_last
Date: Tue, 16 Jun 2020 15:38:32 -0700 [thread overview]
Message-ID: <037fa9b35dfbe459776be8ef7b01eca95e3dc7a3.1592343756.git.julian@codesourcery.com> (raw)
In-Reply-To: <cover.1592343756.git.julian@codesourcery.com>
Later patches in this series assume that GOMP_MAP_ATTACH will be grouped
together with a preceding GOMP_MAP_TO_PSET or other "to" data movement
clause, except in cases where an explicit "attach" clause is used.
This patch arranges for that to be so.
OK?
Julian
ChangeLog
libgomp/
* oacc-mem.c (find_group_last): Group data-movement clauses
(GOMP_MAP_TO_PSET, GOMP_MAP_TO, etc.) together with a subsequent
GOMP_MAP_ATTACH. Allow standalone GOMP_MAP_ATTACH also.
---
libgomp/oacc-mem.c | 22 +++++++++++++++++++---
1 file changed, 19 insertions(+), 3 deletions(-)
diff --git a/libgomp/oacc-mem.c b/libgomp/oacc-mem.c
index 936ae649dd9..be7f8d600eb 100644
--- a/libgomp/oacc-mem.c
+++ b/libgomp/oacc-mem.c
@@ -985,9 +985,15 @@ find_group_last (int pos, size_t mapnum, size_t *sizes, unsigned short *kinds)
switch (kind0)
{
case GOMP_MAP_TO_PSET:
- while (pos + 1 < mapnum && (kinds[pos + 1] & 0xff) == GOMP_MAP_POINTER)
+ if (pos + 1 < mapnum
+ && (kinds[pos + 1] & 0xff) == GOMP_MAP_ATTACH)
+ return pos + 1;
+
+ while (pos + 1 < mapnum
+ && (kinds[pos + 1] & 0xff) == GOMP_MAP_POINTER)
pos++;
- /* We expect at least one GOMP_MAP_POINTER after a GOMP_MAP_TO_PSET. */
+ /* We expect at least one GOMP_MAP_POINTER (if not a single
+ GOMP_MAP_ATTACH) after a GOMP_MAP_TO_PSET. */
assert (pos > first_pos);
break;
@@ -1002,6 +1008,9 @@ find_group_last (int pos, size_t mapnum, size_t *sizes, unsigned short *kinds)
gomp_fatal ("unexpected mapping");
break;
+ case GOMP_MAP_ATTACH:
+ return pos;
+
default:
/* GOMP_MAP_ALWAYS_POINTER can only appear directly after some other
mapping. */
@@ -1012,9 +1021,16 @@ find_group_last (int pos, size_t mapnum, size_t *sizes, unsigned short *kinds)
return pos + 1;
}
+ /* We can have a single GOMP_MAP_ATTACH mapping after a to/from
+ mapping. */
+ if (pos + 1 < mapnum
+ && (kinds[pos + 1] & 0xff) == GOMP_MAP_ATTACH)
+ return pos + 1;
+
/* We can have zero or more GOMP_MAP_POINTER mappings after a to/from
(etc.) mapping. */
- while (pos + 1 < mapnum && (kinds[pos + 1] & 0xff) == GOMP_MAP_POINTER)
+ while (pos + 1 < mapnum
+ && (kinds[pos + 1] & 0xff) == GOMP_MAP_POINTER)
pos++;
}
--
2.23.0
next prev parent reply other threads:[~2020-06-16 22:38 UTC|newest]
Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-06-16 22:38 [PATCH 0/9] [OpenACC] Refcounting and manual deep copy improvements Julian Brown
2020-06-16 22:38 ` [PATCH 1/9] [OpenACC] Fortran derived-type mapping fix Julian Brown
2020-06-16 22:38 ` Julian Brown [this message]
2020-06-30 12:42 ` [PATCH 2/9] [OpenACC] GOMP_MAP_ATTACH handling in find_group_last Thomas Schwinge
2020-06-16 22:38 ` [PATCH 3/9] [OpenACC] Adjust dynamic reference count semantics Julian Brown
2020-06-30 13:51 ` Thomas Schwinge
2020-07-03 15:41 ` Thomas Schwinge
2020-07-10 12:08 ` Julian Brown
2020-06-16 22:38 ` [PATCH 4/9] [OpenACC] Don't pass kind array via pointer to goacc_enter_datum Julian Brown
2020-06-16 22:39 ` [PATCH 5/9] [OpenACC] Fix incompatible copyout for acc_map_data (PR92843) Julian Brown
2020-06-16 22:39 ` [PATCH 6/9] [OpenACC] Set bias to zero for explicit attach/detach clauses in C and C++ Julian Brown
2020-06-25 11:36 ` Thomas Schwinge
2020-07-09 21:06 ` Thomas Schwinge
2020-07-09 21:32 ` Julian Brown
2020-06-16 22:39 ` [PATCH 7/9] [OpenACC] Do not strip GOMP_MAP_TO_PSET/GOMP_MAP_POINTER for enter/exit data directives Julian Brown
2020-07-06 16:19 ` Thomas Schwinge
2020-06-16 22:39 ` [PATCH 8/9] [OpenACC] Fix standalone attach for Fortran assumed-shape array pointers Julian Brown
2020-07-14 11:43 ` Thomas Schwinge
2020-07-15 10:28 ` Thomas Schwinge
2020-07-17 11:16 ` Thomas Schwinge
2020-07-27 14:33 ` Julian Brown
2020-07-30 9:53 ` Thomas Schwinge
2020-07-30 20:15 ` Julian Brown
2020-06-16 22:39 ` [PATCH 9/9] [OpenACC] Don't detach for no-op exit data with zero dynamic refcount Julian Brown
2020-07-24 14:18 ` Thomas Schwinge
2020-07-24 22:53 ` Julian Brown
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=037fa9b35dfbe459776be8ef7b01eca95e3dc7a3.1592343756.git.julian@codesourcery.com \
--to=julian@codesourcery.com \
--cc=Catherine_Moore@mentor.com \
--cc=fortran@gcc.gnu.org \
--cc=gcc-patches@gcc.gnu.org \
--cc=jakub@redhat.com \
--cc=thomas@codesourcery.com \
--cc=tobias@codesourcery.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).