public inbox for gcc-bugs@sourceware.org help / color / mirror / Atom feed
From: "thiago at kde dot org" <gcc-bugzilla@gcc.gnu.org> To: gcc-bugs@gcc.gnu.org Subject: [Bug tree-optimization/104475] [12/13 Regression] Wstringop-overflow + atomics incorrect warning on dynamic object Date: Tue, 06 Dec 2022 18:03:33 +0000 [thread overview] Message-ID: <bug-104475-4-fnWzrcxdYh@http.gcc.gnu.org/bugzilla/> (raw) In-Reply-To: <bug-104475-4@http.gcc.gnu.org/bugzilla/> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104475 --- Comment #19 from Thiago Macieira <thiago at kde dot org> --- (In reply to Richard Biener from comment #15) > Thanks, it's still the same reason - we isolate a nullptr case and end up > with > > __atomic_or_fetch_4 (184B, 64, 0); [tail call] > > The path we isolate is d->m_mutex == nullptr && !enable in > > void QFutureInterfaceBase::setThrottled(bool enable) > { > QMutexLocker lock(&d->m_mutex); Thank you for the analysis, Richard. But do note that it's &d->m_mutex, not d->m_mutex that is passed to the locker. C++ says that if you do d-> then d != nullptr, so &d->m_mutex can't be nullptr either. However, I guess GCC thinks it can be because the offset of m_mutex in QFIBP is zero. pahole says: public: void QFutureInterfaceBasePrivate(class QFutureInterfaceBasePrivate *, enum State); void ~QFutureInterfaceBasePrivate(class QFutureInterfaceBasePrivate *, int); class QMutex m_mutex; /* 0 8 */ class QBasicMutex continuationMutex; /* 8 8 */ So there's a missed optimisation here. But it doesn't look like GCC is the only one to miss it, see https://gcc.godbolt.org/z/WW5hbW6sW. Maybe it's an intentional choice? > we predict the path to be unlikely but the adjustment to the threader > covered probably never executed paths (with probability zero). The > threading opportunity arises because the DTOR calls > > inline void unlock() noexcept > { > if (!isLocked) > return; > m->unlock(); > isLocked = false; > } > > and we know isLocked on the nullptr path. We know it can't be true. > I thought we could maybe enhance prediction to look for nullptr based > accesses but at the time we estimate probabilities the QMutexLocker > CTOR isn't yet inlined (the DTOR is partially inlined, exposing the > isLocked check). > > Note the "impossible" path is actually in the sources - so there might > be a missing conditional somewhere. I don't see it, but that's probably because I'm looking at it from the C++ side. If the mutex pointer that was passed is null, then isLocked is never set to true. What you're saying is that the unlock() function above was inlined and that GCC knew m to be nullptr, but didn't know isLocked's value... which makes no sense to me. If the constructor wasn't inlined, it couldn't know the value of m either. If the constructor was inlined, then it should know the value of both. Anyway, this discussion made me realise there's a series of changes to QMutexLocker ending in "QMutexLocker: strenghten the locking operations" (https://code.qt.io/cgit/qt/qtbase.git/commit/?id=1b1456975347b044c11169458b53c9f6083dbc59). This probably did change how the optimiser works, explaining why the warnings went away. But it shouldn't have. We went from inline ~QMutexLocker() { unlock(); } inline void unlock() noexcept { if (!isLocked) return; m->unlock(); isLocked = false; } to inline ~QMutexLocker() { if (m_isLocked) unlock(); } inline void unlock() noexcept { Q_ASSERT(m_isLocked); m_mutex->unlock(); m_isLocked = false; } with the Q_ASSERT expanding to nothing in release builds, it should be effectively identical code.
next prev parent reply other threads:[~2022-12-06 18:03 UTC|newest] Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top 2022-02-09 19:37 [Bug c++/104475] New: " thiago at kde dot org 2022-02-09 20:14 ` [Bug tree-optimization/104475] [12 Regression] " pinskia at gcc dot gnu.org 2022-02-09 23:48 ` msebor at gcc dot gnu.org 2022-02-16 14:56 ` jakub at gcc dot gnu.org 2022-03-03 14:00 ` aldyh at gcc dot gnu.org 2022-03-03 16:09 ` amacleod at redhat dot com 2022-03-03 17:55 ` amacleod at redhat dot com 2022-03-04 14:47 ` amacleod at redhat dot com 2022-03-09 13:13 ` rguenth at gcc dot gnu.org 2022-03-12 9:39 ` aldyh at gcc dot gnu.org 2022-03-23 12:47 ` rguenth at gcc dot gnu.org 2022-03-23 17:23 ` aldyh at gcc dot gnu.org 2022-05-06 8:32 ` [Bug tree-optimization/104475] [12/13 " jakub at gcc dot gnu.org 2022-07-26 12:59 ` rguenth at gcc dot gnu.org 2022-12-05 15:50 ` rguenth at gcc dot gnu.org 2022-12-05 16:36 ` thiago at kde dot org 2022-12-06 8:17 ` rguenth at gcc dot gnu.org 2022-12-06 9:11 ` rguenth at gcc dot gnu.org 2022-12-06 10:22 ` cvs-commit at gcc dot gnu.org 2022-12-06 10:26 ` rguenth at gcc dot gnu.org 2022-12-06 18:03 ` thiago at kde dot org [this message] 2022-12-07 9:46 ` rguenth at gcc dot gnu.org 2022-12-07 9:48 ` rguenth at gcc dot gnu.org 2022-12-07 9:49 ` rguenth at gcc dot gnu.org 2022-12-07 9:54 ` rguenth at gcc dot gnu.org 2022-12-07 11:25 ` rguenth at gcc dot gnu.org 2023-01-17 17:37 ` jason at gcc dot gnu.org 2023-05-08 12:23 ` [Bug tree-optimization/104475] [12/13/14 " rguenth at gcc dot gnu.org 2023-09-18 9:19 ` aph at gcc dot gnu.org 2023-09-18 10:04 ` rguenth at gcc dot gnu.org
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=bug-104475-4-fnWzrcxdYh@http.gcc.gnu.org/bugzilla/ \ --to=gcc-bugzilla@gcc.gnu.org \ --cc=gcc-bugs@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: linkBe 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).