This allows us to get the current value of the QMutex / QBasicMutex
after the testAndSet operation failed. It saves an extra load from
memory.
Change-Id: I4922a8b3df15e342b177b13f56cf4f1184314520
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
sleep_for appears to be unreliable (resulting in infinite waits) when
used in a thread which is also using native waiting methods. This is
also a step in the direction of eliminating std::thread's usage in WinRT.
Change-Id: I58bc4bf9ada25de247849333ef925964676b7239
Reviewed-by: Oliver Wolff <oliver.wolff@digia.com>
Instead of looping unnecessarily on top of testAndSet.
Task-number: QTBUG-37031
Change-Id: I8120f8405eb76dccc9066749cee0a92b0f2da20e
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Sergio Ahumada <sahumada@blackberry.com>
[ChangeLog][QtCore][Atomic support]Added more operations to the atomic
classes, including operator T(), operator=(T), operator++,
operator--. For the QAtomicInteger, bit-manipulation operations are
also provided, both in operator and in fetchAndXxxYyyyyy modes.
Change-Id: I39c07be74e15e0a48f9e931f4342b182004dee1a
Reviewed-by: David Faure <david.faure@kdab.com>
This is extremely useful, since the most common action after a failed
compare-and-swap is to loop around, trying again with the current
value as found in memory.
Code currently written as:
do {
Type value = atomic.load();
...
} while (!atomic.testAndSetRelaxed(value, desired));
Becomes:
Type value = atomic.load();
do {
...
} while (!atomic.testAndSetRelaxed(value, desired, value));
In most CPU architectures, the value that was found in memory is known
to the compare-and-swap code, so this is more efficient than the
previous code. In architectures where the value is not known, the new
code is no worse than before.
The implementation sometimes modified an existing function, sometimes
it added a new one, depending on whether more registers were needed in
the assembly (like ARMv6-7), the code became more complex (ARMv5), the
optimizer failed (C++11), or it was just plain equivalent (MIPS).
Change-Id: I7d6d200ea9746ec8978a0c1e1969dbc3580b9285
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Don't delete the thread object without detaching it, use detach()
instead of CloseHandle(), and avoid a double-delete.
Change-Id: Ia169a96fb32805e06abe099c3c35e97ce485f088
Reviewed-by: Oliver Wolff <oliver.wolff@digia.com>
This simplifies the code a lot and avoids silly mistakes where a
specific integer type is missing (such as char16_t).
Change-Id: Id91dfd1919e783e0a9af7bfa093ca560a01b22d1
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
No need to redefine everywhere, since they're required to be supported.
Change-Id: I2bdbbd0b0c44871e3bd0edcf0289fc58dd50ff31
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
This is already implemented in qatomic_x86.h, qatomic_ia64.h,
qatomic_mips.h, qatomic_armv6.h, and qatomic_cxx11.h. For
qatomic_msvc.h, we've just fixed it.
For qatomic_gcc.h, we know that the compiler supports it, so just add
it. According to the GCC manual, it might print a warning on some
platforms, so we only enable that on 64-bit builds.
For qatomic_unix.h, the support was missing (along with support for
unsigned 32-bit), so this commits adds it.
For qatomic_armv5.h, the platform does not always support 64-bit
atomics, but ARMv5 cannot compile in 64-bit mode anyway.
Change-Id: Ia8b3b5c641f11e5df05937fe7442be0a223174ef
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
QAtomicInteger<T> is to QBasicAtomicInteger<T> what QAtomicInt was to
QBasicAtomicInt: just a little more syntactic sugar. The Basic classes
do not always have a constructor, since they depend on compiler
support. The constructor is always present in the non-Basic class, at
the expense of making it non-POD for C++98 code.
This commit also repurposes most of QAtomicInt's documentation for
QAtomicInteger. It adds only the Q_ATOMIC_INTnn_IS_SUPPORTED macro
that explains whether the given type is supported on this platform.
Change-Id: I58886d6fa49fea4de24015c40dae29c9fa534e00
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Kurt Pattyn <pattyn.kurt@gmail.com>
When building with debug, all SLOT or SIGNAL macros will
expand to a function call, and then function will call
QThreadData::current(), which will set
QCoreApplication::theMainThread if it has not already been
done. Since Qt Widgets has these macros in the static
initialization of the library, we would register the
Android main thread as the main thread of Qt, which would
mean that the actual application object was created on
a different thread than the main thread. This caused warnings
to appear, and also triggered a race condition which
caused widget applications to sometimes show a black screen
instead of content on startup when run with the OpenGL plugin.
Task-number: QTBUG-35048
Change-Id: Ie8979f5e7cd5662f8d7dd276de9f94f27cc120b5
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
the diff -w for this commit is empty.
Started-by: Thiago Macieira <thiago.macieira@intel.com>
Change-Id: I77bb84e71c63ce75e0709e5b94bee18e3ce6ab9e
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
- Move private classes in the .cpp file (they aren't needed outside)
- Conform to Qt style, such as includes and braces
- Use ComPtr where appropriate
- Use foreach where appropriate
- Remove non-functional wake/interrupt leftovers
- Remove redundant timer list
- Make the timer callback a static method, so it won't crash if it
gets called on shutdown
Task-number: QTBUG-35945
Change-Id: I5426fba2735e908a04ea60287f9936f5abde6644
Reviewed-by: Oliver Wolff <oliver.wolff@digia.com>
The new atomic code was introduced in Qt 5.0. The platforms that did not
get ported were announced as deprecated in Qt 5.2. The code is now
removed in Qt 5.3.
The status for the platform/compiler/OS combinations affected is:
* Linux with GCC or Clang: still compiles on all platforms
(via qatomic_cxx11.h or qatomic_gcc.h)
* INTEGRITY with Green Hills compiler: no longer compiles
* Solaris on UltraSPARC, with Sun Studio: no longer compiles
* AIX on POWER5 or 6, with IBM Visual Age: no longer compiles
(probably did not compile Qt 5.0 either)
* VxWorks in kernel mode: no longer compiles
[ChangeLog][General] Support for the following platforms has been
removed, due to lack of interest in updating support: INTEGRITY,
VxWorks, Solaris on UltraSPARC (with the Sun Studio compiler suite), AIX
on POWER processors (with IBM Visual Age compiler suite).
Change-Id: I8a961385fd95011c016b2b1eec52034794dae3e1
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Before, we'd get just an error message that the size of the array was
negative. Now, for C++11 compilers, we get a better error message:
qbasicatomic.h:117:5: error: static assertion failed: Template parameter is not a supported integer on this platform
qbasicatomic.h:119:24: error: invalid use of incomplete type ‘struct QAtomicOps<long long unsigned int>’
Change-Id: I6b0792254c0dc6103a4a7608f2942d59cda07c00
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
The T was missing in QT_BASIC_ATOMIC_HAS_CONSTRUCTORS. This makes
QAtomicInt's constructor become constexpr.
Change-Id: Ibe58ff36517c5d05ce8af9c0f28169fa63521632
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
For the conflicts in msvc_nmake.cpp the ifdefs are extended since we
need to support windows phone in the target branch while it is not there
in the current stable branch (as of Qt 5.2).
Conflicts:
configure
qmake/generators/win32/msvc_nmake.cpp
src/3rdparty/angle/src/libEGL/Surface.cpp
src/angle/src/common/common.pri
src/corelib/global/qglobal.h
src/corelib/io/qstandardpaths.cpp
src/plugins/platforms/qnx/qqnxintegration.cpp
src/plugins/platforms/qnx/qqnxscreeneventhandler.h
src/plugins/platforms/xcb/qglxintegration.h
src/widgets/kernel/win.pri
tests/auto/corelib/thread/qreadwritelock/tst_qreadwritelock.cpp
tests/auto/corelib/tools/qdatetime/tst_qdatetime.cpp
tests/auto/gui/text/qtextdocument/tst_qtextdocument.cpp
tools/configure/configureapp.cpp
Change-Id: I00b579eefebaf61d26ab9b00046d2b5bd5958812
The public documentation for load() and store() says it's atomic,
and it is:
* using _q_value.store(newValue, std::memory_order_relaxed) in the C++11
implementation
* using a simple assignment otherwise, which is atomic (and relaxed, no
memory barriers) on all the existing C++ ABIs.
Change-Id: I40faa47120163225bd11c3a32514ac97ef8bbbd4
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The current synchronization mechanism was racy: decrementing waitingThreads
and then hoping that the wakeOne will wake a thread before its expiry
timeout happens. In other words, on timeout, a just-assigned task would
never run. And then no other task would run, if maxThreadCount is reached.
Fixed by using a queue of waiting threads (rather than just a count), and by
moving the wait condition into the thread itself, so we know precisely
which one we're waking up, and we can remove it from the set of waiting threads
before waking it up, and therefore it can determine on wakeup whether it
has work to do (caller removed it from the queue) or it expired (it's still
in the queue). This is reliable, whereas the return value from QWaitCondition::wait
isn't reliable, when the main thread has already decided that this thread
has work to do.
Task-number: QTBUG-3786
Change-Id: I1eac5d6c309daed7f483ac7a8074297bfda6ee32
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Added/Changed:
- Move content from the Thread Basics overview to the QThread class ref
- Rephrase bits for clarity
- Use more links
Removed:
- (threads-basics.qdoc) Warning against moveToThread(this): This usage
came about when people tried to add slots to a QThread subclass. This
patch adds a warning against the root cause.
- (threads-basics.qdoc) Note on sleep() et al.: They were made public in
Qt 5.0.
- (threads-basics.qdoc) The strategy for managing member variables:
Sounds error-prone. Pushing results through signals is safer.
- (qthread.cpp) The note about GUI classes: Irrelevant to QThread,
and it's already mentioned elsewhere.
Change-Id: I6bc53cc22b929523f9976d2b920f94c02bd7273e
Reviewed-by: Geir Vattekar <geir.vattekar@digia.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Default values should have mark-up to denote that they are code.
This commit changes:
-"property is true" to "property is \c true".
-"Returns true" to "Returns \c true".
-"property is false" to "property is \c false".
-"returns true" to "returns \c true".
-"returns false" to "returns \c false".
src/3rdparty and non-documentation instances were ignored.
Task-number: QTBUG-33360
Change-Id: Ie87eaa57af947caa1230602b61c5c46292a4cf4e
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Jerome Pasion <jerome.pasion@digia.com>
qFlagLocation() uses a global char* array to transport source location
information from the connect() side to the metaobject side. The size
of the array is 2 (two), which just about suffices for a single connect()
statement.
Obviously, if more than one thread makes a (_any_) connection at the same
time, the data is useless and, worse, there's a data race.
The non-reentrancy of qFlagLocations() cannot and need not be fixed,
but use a per-thread flagged_locations array in QThreadData so threads
don't disturb each other.
Task-number: QTBUG-3680
Change-Id: If1797c60751f551694def69afee6fbe295bbe2d2
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Rationale: a wait on a condition-variable is usually a cancellation point.
On Posix, and probably in C++ at some point, a thread cancellation is
done by (a kind of) exception unwinding the stack. To ensure that
we call reserveThread() in all cases, wrap the function pair in a RAII
class. Even if we currently don't seem to support exceptions in QtCore,
this is low-hanging fruit, and no worse than what we had before.
Change-Id: Ifb0f428ea50f9ac12be14e620615f46e00d3dd91
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
The read from 'owner' for comparison with 'self' in QRecursiveMutexPrivate::lock()
is not synchronized with the write to 'owner' in the same function further down,
and neither operation is atomic.
Fix by making 'owner' an atomic pointer.
Change-Id: I186b88575589da0dce5827a1e17ceb4ce599ed02
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
- The "Concurrent Programming" page is an exact duplicate of the Qt Concurrent module landing
page.
- The "qtconcurrent intro" target is not referenced anywhere.
Change-Id: Ice9b4360783013fe972258ca54a0004be43b8766
Reviewed-by: Jerome Pasion <jerome.pasion@digia.com>
To ease interruption of long running tasks, a new method
QThread::setInterruptionRequested() can be called.
The task can check QThread::isInterruptionRequested()
and act upon it by stopping itself.
These methods are designed to replace the use of a global variable
and other hacky ways to stop a task running in another thread.
Change-Id: I17622dd60d2262078210e7e4294ad6c53a6dc179
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Makes it possible for QElapsedTimer to be non-POD.
Change-Id: I5ffc59c7a93c187a4a814e6959f8383fa8d4cc44
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
QTBUG-21051 has a testcase where activeThreadCount() could actually
end up at -1 (converted to an autotest in this commit).
The reason was: start() calls tryStart() which returns false due to
too many active threads (reserveThread() causes this), so it calls
enqueueTask() - which actually wakes up the waiting thread, but
it didn't decrement the number of waiting threads.
Note that tryStart() is "if I can grab a waiting thread, enqueue task and wake it"
while start(), in case tryStart() fails, wants to "enqueue, and then if I can grab
a waiting thread, wake it". This is why enqueue shouldn't wake; waking must happen
only if we can grab a thread (d->waitingThreads > 0).
Task-number: QTBUG-21051
Change-Id: I3d98337103031c9bdf0bf365295f245be0c66aa7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Rather than trying to make it lock-free (which requires double-bookkeeping of
4 atomic ints!), just lock the mutex before calling it.
tst_bench_qthreadpool shows no difference whatsoever between the two
solutions, I get 0.005 msecs per iteration in startRunnables().
Of course looping over calls to activeThreadCount() is a bit slower,
from 0.0002 msecs per iteration to 0.00027 msecs, i.e. 35% more.
But polling activeThreadCount() from the app is a really wrong thing to
do anyway, this benchmark was just for my own curiosity about the
price of a mutex in a function that sums up 4 ints.
What matters is start() performance, which is unchanged (0.00007 msecs
is just noise compared to a 0.005 total, that's 1.4%).
Change-Id: I993444eef8bc68eff9badd581fae3626dfd1cc6d
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
QThreadPool::clear() method removes all queued QRunnable.
When a large number of long-running tasks are queud in a
QThreadPool its destruction, which calls waitForDone(), can
be quite long.
QThreadPool:clear() removes (and deletes when appropriate)
all QRunnable that have yet to be started from the queue
enabling a faster interruption.
Change-Id: Ie5d6028ad3cfe7e439d1db068c8d0936ff818db9
Reviewed-by: David Faure <david.faure@kdab.com>
Flip !Q_OS_IOS conditions to Q_OS_MACX where it seems appropriate,
remove a redundant condition in qtextcodec_p.h.
Change-Id: I21c8c0c490f1eb4a9337a7f2f3e907c125489438
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@digia.com>
Thiago Macieira
Andrew Knight
Frederik Gladhorn
Eskil Abrahamsen Blomfeldt
Oswald Buddenhagen
Friedemann Kleint
Fatih Aşıcı
David Faure
Sze Howe Koh
Jerome Pasion
Marc Mutz
Oliver Wolff
Kamil Trzcinski
Sergio Ahumada
Tor Arne Vestbø
Corentin Jabot
Jake Petroules