public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r14-3846] libstdc++: Remove unconditional use of atomics in Debug Mode
@ 2023-09-11 16:07 Jonathan Wakely
0 siblings, 0 replies; only message in thread
From: Jonathan Wakely @ 2023-09-11 16:07 UTC (permalink / raw)
To: gcc-cvs, libstdc++-cvs
https://gcc.gnu.org/g:4a2766ed00a47904dc8b85bf0538aa116d8e658b
commit r14-3846-g4a2766ed00a47904dc8b85bf0538aa116d8e658b
Author: Jonathan Wakely <jwakely@redhat.com>
Date: Mon Sep 11 16:42:54 2023 +0100
libstdc++: Remove unconditional use of atomics in Debug Mode
The fix for PR 91910 (r10-3426-gf7a3a382279585) introduced unconditional
uses of atomics into src/c++11/debug.cc, which causes linker errors for
arm4t where GCC emits an unresolved reference to __sync_synchronize.
By making the uses of atomics depend on _GLIBCXX_HAS_GTHREADS we can
avoid those unconditional references to __sync_synchronize for targets
where the atomics are unnecessary. As a minor performance optimization
we can also check the __gnu_cxx::__is_single_threaded function to avoid
atomics for single-threaded programs even where they don't cause linker
errors.
libstdc++-v3/ChangeLog:
* src/c++11/debug.cc (acquire_sequence_ptr_for_lock): New
function.
(reset_sequence_ptr): New function.
(_Safe_iterator_base::_M_detach)
(_Safe_local_iterator_base::_M_detach): Replace bare atomic_load
with acquire_sequence_ptr_for_lock.
(_Safe_iterator_base::_M_reset): Replace bare atomic_store with
reset_sequence_ptr.
Diff:
---
libstdc++-v3/src/c++11/debug.cc | 32 +++++++++++++++++++++++++++++---
1 file changed, 29 insertions(+), 3 deletions(-)
diff --git a/libstdc++-v3/src/c++11/debug.cc b/libstdc++-v3/src/c++11/debug.cc
index f40e995e0f30..bb0d0db66799 100644
--- a/libstdc++-v3/src/c++11/debug.cc
+++ b/libstdc++-v3/src/c++11/debug.cc
@@ -24,6 +24,7 @@
#include <bits/move.h>
#include <bits/stl_iterator_base_types.h>
+#include <ext/atomicity.h> // __is_single_threaded
#include <debug/formatter.h>
#include <debug/safe_base.h>
@@ -156,6 +157,31 @@ namespace
__old->_M_reset();
}
}
+
+ void*
+ acquire_sequence_ptr_for_lock(__gnu_debug::_Safe_sequence_base*& seq)
+ {
+#ifdef __GTHREADS
+ if (!__gnu_cxx::__is_single_threaded())
+ return __atomic_load_n(&seq, __ATOMIC_ACQUIRE);
+#endif
+ return seq;
+ }
+
+ void
+ reset_sequence_ptr(__gnu_debug::_Safe_sequence_base*& seq)
+ {
+#ifdef __GTHREADS
+ if (!__gnu_cxx::__is_single_threaded())
+ {
+ __atomic_store_n(&seq, (__gnu_debug::_Safe_sequence_base*)nullptr,
+ __ATOMIC_RELEASE);
+ return;
+ }
+#endif
+ seq = nullptr;
+ }
+
} // anonymous namespace
namespace __gnu_debug
@@ -439,7 +465,7 @@ namespace __gnu_debug
// If the sequence destructor runs between loading the pointer and
// locking the mutex, it will detach this iterator and set _M_sequence
// to null, and then _M_detach_single() will do nothing.
- if (auto seq = __atomic_load_n(&_M_sequence, __ATOMIC_ACQUIRE))
+ if (auto seq = acquire_sequence_ptr_for_lock(_M_sequence))
{
__gnu_cxx::__scoped_lock sentry(get_safe_base_mutex(seq));
_M_detach_single();
@@ -461,7 +487,7 @@ namespace __gnu_debug
_Safe_iterator_base::
_M_reset() throw ()
{
- __atomic_store_n(&_M_sequence, (_Safe_sequence_base*)0, __ATOMIC_RELEASE);
+ reset_sequence_ptr(_M_sequence);
// Do not reset version, so that a detached iterator does not look like a
// value-initialized one.
// _M_version = 0;
@@ -523,7 +549,7 @@ namespace __gnu_debug
_Safe_local_iterator_base::
_M_detach()
{
- if (auto seq = __atomic_load_n(&_M_sequence, __ATOMIC_ACQUIRE))
+ if (auto seq = acquire_sequence_ptr_for_lock(_M_sequence))
{
__gnu_cxx::__scoped_lock sentry(get_safe_base_mutex(seq));
_M_detach_single();
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2023-09-11 16:07 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-09-11 16:07 [gcc r14-3846] libstdc++: Remove unconditional use of atomics in Debug Mode Jonathan Wakely
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).