public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Jakub Jelinek <jakub@redhat.com>
To: gcc-patches@gcc.gnu.org
Subject: [committed] libgomp: Ensure that either gomp_team is properly aligned [PR102838]
Date: Thu, 18 Nov 2021 09:21:48 +0100	[thread overview]
Message-ID: <20211118082148.GQ2710@tucnak> (raw)

Hi!

struct gomp_team has struct gomp_work_share array inside of it.
If that latter structure has 64-byte aligned member in the middle,
the whole struct gomp_team needs to be 64-byte aligned, but we weren't
allocating it using gomp_aligned_alloc.

This patch fixes that, except that on gcn team_malloc is special, so
I've instead decided at least for now to avoid using aligned member
and use the padding instead on gcn.

Bootstrapped/regtested on x86_64-linux and i686-linux and in the PR Rainer
mentioned testing on Solaris, committed to trunk.

2021-11-18  Jakub Jelinek  <jakub@redhat.com>

	PR libgomp/102838
	* libgomp.h (GOMP_USE_ALIGNED_WORK_SHARES): Define if
	GOMP_HAVE_EFFICIENT_ALIGNED_ALLOC is defined and __AMDGCN__ is not.
	(struct gomp_work_share): Use GOMP_USE_ALIGNED_WORK_SHARES instead of
	GOMP_HAVE_EFFICIENT_ALIGNED_ALLOC.
	* work.c (alloc_work_share, gomp_work_share_start): Likewise.
	* team.c (gomp_new_team): If GOMP_USE_ALIGNED_WORK_SHARES, use
	gomp_aligned_alloc instead of team_malloc.

--- libgomp/libgomp.h.jj	2021-11-11 14:35:37.699347142 +0100
+++ libgomp/libgomp.h	2021-11-16 11:57:26.657271188 +0100
@@ -95,6 +95,10 @@ enum memmodel
 #define GOMP_HAVE_EFFICIENT_ALIGNED_ALLOC 1
 #endif
 
+#if defined(GOMP_HAVE_EFFICIENT_ALIGNED_ALLOC) && !defined(__AMDGCN__)
+#define GOMP_USE_ALIGNED_WORK_SHARES 1
+#endif
+
 extern void *gomp_malloc (size_t) __attribute__((malloc));
 extern void *gomp_malloc_cleared (size_t) __attribute__((malloc));
 extern void *gomp_realloc (void *, size_t);
@@ -348,7 +352,7 @@ struct gomp_work_share
      are in a different cache line.  */
 
   /* This lock protects the update of the following members.  */
-#ifdef GOMP_HAVE_EFFICIENT_ALIGNED_ALLOC
+#ifdef GOMP_USE_ALIGNED_WORK_SHARES
   gomp_mutex_t lock __attribute__((aligned (64)));
 #else
   char pad[64 - offsetof (struct gomp_work_share_1st_cacheline, pad)];
--- libgomp/work.c.jj	2021-10-20 09:34:47.027331304 +0200
+++ libgomp/work.c	2021-11-16 11:58:10.136662003 +0100
@@ -78,7 +78,7 @@ alloc_work_share (struct gomp_team *team
   team->work_share_chunk *= 2;
   /* Allocating gomp_work_share structures aligned is just an
      optimization, don't do it when using the fallback method.  */
-#ifdef GOMP_HAVE_EFFICIENT_ALIGNED_ALLOC
+#ifdef GOMP_USE_ALIGNED_WORK_SHARES
   ws = gomp_aligned_alloc (__alignof (struct gomp_work_share),
 			   team->work_share_chunk
 			   * sizeof (struct gomp_work_share));
@@ -191,7 +191,7 @@ gomp_work_share_start (size_t ordered)
   /* Work sharing constructs can be orphaned.  */
   if (team == NULL)
     {
-#ifdef GOMP_HAVE_EFFICIENT_ALIGNED_ALLOC
+#ifdef GOMP_USE_ALIGNED_WORK_SHARES
       ws = gomp_aligned_alloc (__alignof (struct gomp_work_share),
 			       sizeof (*ws));
 #else
--- libgomp/team.c.jj	2021-11-11 14:35:37.699347142 +0100
+++ libgomp/team.c	2021-11-16 11:59:46.401311440 +0100
@@ -177,7 +177,12 @@ gomp_new_team (unsigned nthreads)
     {
       size_t extra = sizeof (team->ordered_release[0])
 		     + sizeof (team->implicit_task[0]);
+#ifdef GOMP_USE_ALIGNED_WORK_SHARES
+      team = gomp_aligned_alloc (__alignof (struct gomp_team),
+				 sizeof (*team) + nthreads * extra);
+#else
       team = team_malloc (sizeof (*team) + nthreads * extra);
+#endif
 
 #ifndef HAVE_SYNC_BUILTINS
       gomp_mutex_init (&team->work_share_list_free_lock);

	Jakub


                 reply	other threads:[~2021-11-18  8:21 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

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=20211118082148.GQ2710@tucnak \
    --to=jakub@redhat.com \
    --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).