From: Martin Jambor <mjambor@suse.cz>
To: GCC Patches <gcc-patches@gcc.gnu.org>
Subject: [hsa 4/10] Merge of HSA branch
Date: Mon, 07 Dec 2015 11:21:00 -0000 [thread overview]
Message-ID: <20151207112122.GE24234@virgil.suse.cz> (raw)
In-Reply-To: <20151207111758.GA24234@virgil.suse.cz>
Subject: Make copy_gimple_seq_and_replace_locals copy seqs in omp clauses
Hi,
this is https://gcc.gnu.org/ml/gcc-patches/2015-12/msg00477.html with
the early return requested by Jakub. Please refer to that previous
email for explanation why it is necessary.
Thanks,
2015-12-03 Martin Jambor <mjambor@suse.cz>
* tree-inline.c (duplicate_remap_omp_clause_seq): New function.
(replace_locals_op): Duplicate gimple sequences in OMP clauses.
diff --git a/gcc/tree-inline.c b/gcc/tree-inline.c
index ebab189..dea23c7 100644
--- a/gcc/tree-inline.c
+++ b/gcc/tree-inline.c
@@ -5116,6 +5116,8 @@ mark_local_labels_stmt (gimple_stmt_iterator *gsip,
return NULL_TREE;
}
+static gimple_seq duplicate_remap_omp_clause_seq (gimple_seq seq,
+ struct walk_stmt_info *wi);
/* Called via walk_gimple_seq by copy_gimple_seq_and_replace_local.
Using the splay_tree pointed to by ST (which is really a `splay_tree'),
@@ -5160,6 +5162,35 @@ replace_locals_op (tree *tp, int *walk_subtrees, void *data)
TREE_OPERAND (expr, 3) = NULL_TREE;
}
}
+ else if (TREE_CODE (expr) == OMP_CLAUSE)
+ {
+ /* Before the omplower pass completes, some OMP clauses can contain
+ sequences that are neither copied by gimple_seq_copy nor walked by
+ walk_gimple_seq. To make copy_gimple_seq_and_replace_locals work even
+ in those situations, we have to copy and process them explicitely. */
+
+ if (OMP_CLAUSE_CODE (expr) == OMP_CLAUSE_LASTPRIVATE)
+ {
+ gimple_seq seq = OMP_CLAUSE_LASTPRIVATE_GIMPLE_SEQ (expr);
+ seq = duplicate_remap_omp_clause_seq (seq, wi);
+ OMP_CLAUSE_LASTPRIVATE_GIMPLE_SEQ (expr) = seq;
+ }
+ else if (OMP_CLAUSE_CODE (expr) == OMP_CLAUSE_LINEAR)
+ {
+ gimple_seq seq = OMP_CLAUSE_LINEAR_GIMPLE_SEQ (expr);
+ seq = duplicate_remap_omp_clause_seq (seq, wi);
+ OMP_CLAUSE_LINEAR_GIMPLE_SEQ (expr) = seq;
+ }
+ else if (OMP_CLAUSE_CODE (expr) == OMP_CLAUSE_REDUCTION)
+ {
+ gimple_seq seq = OMP_CLAUSE_REDUCTION_GIMPLE_INIT (expr);
+ seq = duplicate_remap_omp_clause_seq (seq, wi);
+ OMP_CLAUSE_REDUCTION_GIMPLE_INIT (expr) = seq;
+ seq = OMP_CLAUSE_REDUCTION_GIMPLE_MERGE (expr);
+ seq = duplicate_remap_omp_clause_seq (seq, wi);
+ OMP_CLAUSE_REDUCTION_GIMPLE_MERGE (expr) = seq;
+ }
+ }
/* Keep iterating. */
return NULL_TREE;
@@ -5200,6 +5231,21 @@ replace_locals_stmt (gimple_stmt_iterator *gsip,
return NULL_TREE;
}
+/* Create a copy of SEQ and remap all decls in it. */
+
+static gimple_seq
+duplicate_remap_omp_clause_seq (gimple_seq seq, struct walk_stmt_info *wi)
+{
+ if (!seq)
+ return NULL;
+
+ /* If there are any labels in OMP sequences, they can be only referred to in
+ the sequence itself and therefore we can do both here. */
+ walk_gimple_seq (seq, mark_local_labels_stmt, NULL, wi);
+ gimple_seq copy = gimple_seq_copy (seq);
+ walk_gimple_seq (copy, replace_locals_stmt, replace_locals_op, wi);
+ return copy;
+}
/* Copies everything in SEQ and replaces variables and labels local to
current_function_decl. */
next prev parent reply other threads:[~2015-12-07 11:21 UTC|newest]
Thread overview: 71+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-12-07 11:18 [hsa 0/10] " Martin Jambor
2015-12-07 11:19 ` [hsa 1/10] Configury changes and new options Martin Jambor
2015-12-08 22:43 ` Richard Sandiford
2015-12-10 17:52 ` Martin Jambor
2015-12-11 17:42 ` Jakub Jelinek
2015-12-10 17:52 ` Martin Jambor
2015-12-07 11:20 ` [hsa 3/10] HSA libgomp plugin Martin Jambor
2015-12-09 12:16 ` Jakub Jelinek
2015-12-10 12:06 ` [hsa 3/10] HSA libgomp plugin [part 1/2] Martin Liška
2015-12-10 12:06 ` [hsa 3/10] HSA libgomp plugin [part 2/2] Martin Liška
2015-12-07 11:20 ` [hsa 2/10] Modifications to libgomp proper Martin Jambor
2015-12-09 11:55 ` Jakub Jelinek
2015-12-10 17:52 ` Martin Jambor
2015-12-11 18:05 ` Jakub Jelinek
2016-01-12 13:46 ` Martin Jambor
2016-01-12 14:23 ` Jakub Jelinek
2016-01-15 20:00 ` Jakub Jelinek
2016-01-12 13:00 ` Alexander Monakov
2016-01-12 13:10 ` Jakub Jelinek
[not found] ` <20160112132905.GM3060@virgil.suse.cz>
2016-01-12 13:38 ` Jakub Jelinek
2016-01-12 18:51 ` Martin Jambor
2015-12-07 11:21 ` Martin Jambor [this message]
2015-12-09 12:17 ` [hsa 4/10] Merge of HSA branch Jakub Jelinek
2015-12-07 11:23 ` [hsa 5/10] OpenMP lowering/expansion changes (gridification) Martin Jambor
2015-12-09 13:19 ` Jakub Jelinek
2015-12-09 16:25 ` Splitting up gcc/omp-low.c? (was: [hsa 5/10] OpenMP lowering/expansion changes (gridification)) Thomas Schwinge
2015-12-09 17:23 ` Splitting up gcc/omp-low.c? Bernd Schmidt
2015-12-10 8:08 ` Jakub Jelinek
2015-12-10 11:26 ` Bernd Schmidt
2015-12-10 11:34 ` Jakub Jelinek
2015-12-15 18:28 ` Nathan Sidwell
2016-04-08 9:36 ` Thomas Schwinge
2016-04-08 10:46 ` Martin Jambor
2016-04-08 11:08 ` Jakub Jelinek
2016-04-13 16:01 ` Thomas Schwinge
2016-04-13 17:38 ` Bernd Schmidt
2016-04-13 17:56 ` Thomas Schwinge
2016-04-13 18:20 ` Bernd Schmidt
2016-04-14 13:36 ` Thomas Schwinge
2016-04-14 16:01 ` Bernd Schmidt
2016-04-14 16:01 ` Thomas Schwinge
2016-04-14 20:28 ` Split out OMP constructs' SIMD clone supporting code (was: Splitting up gcc/omp-low.c?) Thomas Schwinge
2016-04-15 6:25 ` Split out OMP constructs' SIMD clone supporting code Thomas Schwinge
2016-04-15 11:15 ` Split out OMP constructs' SIMD clone supporting code (was: Splitting up gcc/omp-low.c?) Jakub Jelinek
2016-04-15 11:53 ` Thomas Schwinge
2016-04-15 11:57 ` Jakub Jelinek
2016-04-15 12:11 ` Thomas Schwinge
2016-04-15 12:15 ` Jakub Jelinek
2016-04-15 14:33 ` Thomas Schwinge
2016-05-03 9:34 ` Splitting up gcc/omp-low.c? Thomas Schwinge
2016-05-11 13:44 ` Thomas Schwinge
2016-05-18 11:42 ` Thomas Schwinge
2016-05-25 9:17 ` Thomas Schwinge
2016-05-25 12:54 ` Martin Jambor
2015-12-18 14:29 ` [hsa 5/10] OpenMP lowering/expansion changes (gridification) Martin Jambor
2015-12-07 11:24 ` [hsa 6/10] Pass manager changes Martin Jambor
2015-12-09 13:20 ` Jakub Jelinek
2015-12-09 13:47 ` Richard Biener
2015-12-07 11:25 ` [hsa 7/10] IPA-HSA pass Martin Jambor
2015-12-07 11:27 ` [hsa 8/10] HSAIL BRIG description header file (and a steering committee request) Martin Jambor
2015-12-07 11:29 ` [hsa 9/10] Majority of the HSA back-end Martin Jambor
2015-12-07 11:30 ` [hsa 10/10] HSA register allocator Martin Jambor
2015-12-07 11:46 ` [hsa 0/10] Merge of HSA branch Jakub Jelinek
2015-12-10 17:51 ` Martin Jambor
2016-01-26 10:46 ` (Non-)offloading diagnostics (was: [hsa 0/10] Merge of HSA branch) Thomas Schwinge
2016-01-26 11:18 ` Alexander Monakov
2016-01-26 11:38 ` (Non-)offloading diagnostics Thomas Schwinge
2016-02-26 16:46 ` Thomas Schwinge
2016-02-26 17:18 ` Martin Jambor
2016-02-26 17:51 ` Jakub Jelinek
2016-02-26 18:48 ` Martin Jambor
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=20151207112122.GE24234@virgil.suse.cz \
--to=mjambor@suse.cz \
--cc=gcc-patches@gcc.gnu.org \
/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).