It's closer to what we do with in QMutex than pthread_mutex_lock.
Change-Id: I86498a800b69b684bf096912e911bc5bca219727
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
This corrects the mismatch between using floats for internal storage
and qreal in the API of QVector*D which leads to lots of implicit
casts between double and float.
This change also stops users from being surprised by the loss of
precision when using these classes on desktop platforms and removes
the need for the private constructors taking a dummy int as the final
argument.
The QMatrix4x4 and QQuaternion classes have been changed to use float
for their internal storage since these are meant to be used in
conjunction with the QVector*D classes. This is to prevent unexpected
loss of precision and to improve performance.
The on-disk format has also been changed from double to float thereby
reducing the storage required when streaming vectors and matrices. This
is potentially a large saving when working with complex 3D meshes etc.
This also has a significant performance improvement when passing
matrices to QOpenGLShaderProgram (and QGLShaderProgram) as we no
longer have to iterate and convert the data to floats. This is
an operation that could easily be needed many times per frame.
This change also opens the door for further optimisations of these
classes to be implemented by using SIMD intrinsics.
This needs to be applied in conjunction with
https://codereview.qt-project.org/#change,33548
Task-number: QTBUG-21035
Task-number: QTBUG-20661
Change-Id: I9321b06040ffb93ae1cbd72fd2013267ac901b2e
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
When QTextLayout is used in a QTextDocument, many code paths use
special caches and thus greatly outperform the raw QTextLayout version
that operates directly on a QString.
This patch brings some of these optimizations also to the raw version.
We now also use a QFormatCollection in such cases and enable the
functionality of QTextEngine::indexAdditionalFormats() and
QTextEngine::resolveAdditionalFormats(). Thanks to that, we can greatly
speed up QTextEngine::format(), which now uses an amort O(1) hash table
lookup instead of a O(N) linear search.
The added benchmark shows a gain in the order of one magnitude:
./tst_bench_QText formattedLayout:long-many
before applying the patch:
378.19 msecs per iteration (total: 37,820, iterations: 100)
after applying the patch:
25.80 msecs per iteration (total: 2,580, iterations: 100)
Note: This change is source-incompatible for applications using the private
QTextEngine API.
Task-number: QTBUG-8389
Change-Id: Ifcf7a8902a394428979ea06a6d955f886ee739c7
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
QDateTime will attempt to convert unknown types of date to UTC time, which isn't
exactly a fast process. As we don't care about local timezones in the process of
sorting (as this is purely for ordering, not display to the end user), we can
force the dates to use UTC time, avoiding the unnecessary local timezone lookup.
This also adds a benchmark covering this case.
Benchmark results, Qt 5:
- before: 11, 489ms
- after: 273ms
Qt 4.8:
- before: 20, 848ms
- after: 278ms
Change-Id: I87fa6260e820b5b172d3306ff395dafe767c33ff
Reported-by: Thomas Perl <m@thp.io>
Reviewed-by: Alberto Mardegan <mardy@users.sourceforge.net>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Qt 5.0 beta requires changing the default to the 5.0 API, disabling
the deprecated code. However, tests should test (and often do) the
compatibility API too, so turn it back on.
Task-number: QTBUG-25053
Change-Id: I8129c3ef3cb58541c95a32d083850d9e7f768927
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Ensure benchmarks which need QtWidgets are gracefully disabled when
that module is unavailable.
Fixed one unnecessary usage of "QT+=widgets".
Change-Id: I8031b5dca585749f0f4d22e0637adc3f57f4e418
Reviewed-by: Kalle Lehtonen <kalle.ju.lehtonen@nokia.com>
Reviewed-by: Sergio Ahumada <sergio.ahumada@nokia.com>
Using the nullary version has the advantage that multiple calls
during a program run are much more efficient, since an inlined
atomic is used to store the result. It also ensures that
Q_DECLARE_METATYPE(T) has been used, whereas qRegisterMetaType<T>("T")
will happily register anything. So I've added the macro where it
was missing, or moved it to a central place when it existed
hidden.
In tst_qnetworkreply, this became a bit tricky, because a private
header is conditionally included, so moved the Q_DECLARE_METATYPE()
into a conditional section, too.
Change-Id: I71484523e4277f4697b7d4b2ddc3505375162727
Reviewed-by: Stephen Kelly <stephen.kelly@kdab.com>
Change-Id: I19d3b2e9a5180b13deb828b55195404ef20be295
Reviewed-by: Daniel Teske <daniel.teske@nokia.com>
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
Store the signal index in QObjectPrivate::Connection, thereby making
it available in "implicit" disconnect contexts (i.e., receiver
deletion).
This change does not cause the size of QObjectPrivate::Connection
to grow (still 40 bytes on 32-bit Linux, 72 bytes on 64-bit Mac).
Valgrinding the new benchmark indicates that the percentage of the
time spent in the QObject destructor increased from 7.8% to 8.4%
on ia32, for that particular stress test; the increase is the
combined cost of calling metaObject(), QMetaObjectPrivate::signal(),
and disconnectNotify() for one connection. In practice, the measured
wallclock time increased by about 3ms for a 500ms run (which
repeatedly constructs, connects, and destroys an object).
Task-number: QTBUG-4844
Change-Id: I1beb01c753f31542fc0acb62edb4c6d165fcc5b4
Reviewed-by: Olivier Goffart <ogoffart@woboq.com>
Reviewed-by: Bradley T. Hughes <bradley.hughes@nokia.com>
qdatetime.h uses std::min/max and on Windows windows.h (or some subsequent
header file) may under certain circumstances define min/max as macros.
The easiest way to prevent the windows header files from doing that is to
define NOMINMAX in the place right before windows.h is included. The other
way is to define min and max to min/max themselves to prevent windows.h
from doing its evil thing.
If a user of Qt (WebKit in this case) chooses the approach of defining
min/max to themselves and then includes qdatetime.h, then a subsequent
inclusion of windows.h doesn't work because qdatetime.h undefines min/max.
We should not enforce the type of workaround needed, therefore this patch
removes the workaround from qdatetime.h and requires user code that
happens to include windows header files before qdatetime.h (seldom case)
to choose either workaround.
Change-Id: I7347eec7369491a065e894cff557004e069453d5
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
+ QChar::LastValidCodePoint enum value that supercede the UNICODE_LAST_CODEPOINT macro
replace uses of hardcoded values with the new API; remove leftovers
Change-Id: I1395c9840b85fcb6b08e241b131794a98773c952
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
MinGW installations on case-sensitive filesystems expect
lowercase names of include-libraries and (usually) include
files.
When crosscompiling on Debian 6 (targeting MS Windows) linking
fails because mingw is looking for non-existent include-libraries.
Using lowercase names solves this.
Change-Id: Id3454f4ed8ba42b6ea93d65d9c0ce567db6712df
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@nokia.com>
This operation should be a no-op anyway, since at this point in time,
the fromAscii and toAscii functions simply call their fromLatin1 and
toLatin1 counterparts.
Task-number: QTBUG-21872
Change-Id: I052a3412a568ad639f2bf169b4491b56dddff1c7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This operation should be a no-op anyway, since at this point in time,
the fromAscii and toAscii functions simply call their fromLatin1 and
toLatin1 counterparts.
Task-number: QTBUG-21872
Change-Id: I6ea46cd6dfed75afc253fa2b4e3f1789bdad1d4e
Reviewed-by: Gunnar Sletta <gunnar.sletta@nokia.com>
This operation should be a no-op anyway, since at this point in time,
the fromAscii and toAscii functions simply call their fromLatin1 and
toLatin1 counterparts.
Task-number: QTBUG-21872
Change-Id: I94cc301ea75cc689bcb6e2d417120cf14e36808d
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
Reviewed-by: Richard J. Moore <rich@kde.org>
This operation should be a no-op anyway, since at this point in time,
the fromAscii and toAscii functions simply call their fromLatin1 and
toLatin1 counterparts.
Task-number: QTBUG-21872
Change-Id: I2850033159508ebb1ff7564e15b99a146dbee94c
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
This operation should be a no-op anyway, since at this point in time,
the fromAscii and toAscii functions simply call their fromLatin1 and
toLatin1 counterparts.
Task-number: QTBUG-21872
Change-Id: I38f97ad379deafebef02c75d611343ca15640c8a
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@nokia.com>
qVariantValue and qVariantCanConvert are Compatibility members, while in
Qt4.8 they are marked as Qt 3 Support Members.
qVariantFromValue and qVariantSetValue are Obsolete members.
Change-Id: Ie8505cad1e0950e40c6f6710fde9f6fb2ac670fd
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@nokia.com>
I had missed these, apparently.
Task-number: QTBUG-23274
Change-Id: I6455dc34b18ec9cefccfe527b3fd3ad34fb61aa3
Reviewed-by: Lorn Potter <lorn.potter@nokia.com>
Default QDataStream version was changed in Qt5, but the test tried to
load an old dumped file.
Change-Id: I49c06c232ec8a27f33c9da345bae4e03cd0c56fb
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
0xfdef-0xfdd0 is definitely 31 and not 15 :)
also fix all copy-pastes of this code (greping for '0xfdd0' helps ;)
Change-Id: I8f3bd4fd9d85f9de066f0f5df378b9188c12bd48
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Denis Dzyubenko <denis.dzyubenko@nokia.com>
Just like qMalloc/qRealloc/qFree, there is absolutely no reason to wrap these
functions just to avoid an include, except to pay for it with worse runtime
performance.
On OS X, on byte sizes from 50 up to 1000, calling memset directly is 28-15%
faster(!) than adding an additional call to qMemSet. The advantage on sizes
above that is unmeasurable.
For qMemCopy, the benefits are a little more modest: 16-7%.
Change-Id: I98aa92bb765aea0448e3f20af42a039b369af0b3
Reviewed-by: Giuseppe D'Angelo <dangelog@gmail.com>
Reviewed-by: John Brooks <john.brooks@dereferenced.net>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
QT_NO_STL is now no longer defined, so remove the conditionals and
select the STL side.
Change-Id: Ieedd248ae16e5a128b4ac287f850b3ebc8fb6181
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
The use of any broken-down components of the query now needs
QUrlQuery.
The QUrl constructor and toString() are now rehabilitated and the
preferred forms. Use toEncoded() and fromEncoded() now only when we
need to store data in a QByteArray or the data comes from a QByteArray
anyway. Change to toString() or the constructor if the data was in a
QString.
Change-Id: I9d761a628bef9c70185a48e927a61779a1642342
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
QSharedPointer isn't meant to be used as a
base class.
Instead of inheriting from it to add implicit
conversions to and from QNetworkReply*, make
QNetworkReplyPtr a typedef, overload two
oft-used functions to take a QNetworkReplyPtr
in addition to QNetworkReply*, and otherwise
make the conversions explicit.
Change-Id: I1eff1793a19f2d5bad1cce8de74c0786675a50f3
Reviewed-by: Shane Kearns <shane.kearns@accenture.com>
QMap used to use a skiplist in Qt 4.x, which has variable
sized nodes and we can thus not optimise using custom
allocators.
The rewrite now uses a red-black tree, and all allocations
and tree operations happen in the cpp file. This will allow
us to introduce custom allocation schemes in later versions
of Qt.
Added some more tests and a benchmark. Memory consumption
of the new QMap implementation is pretty much the same as before.
Performance of insertion and lookup has increased by 10-30%. iteration
is slower, but still extremely fast and should not matter compared
to the work usually done when iterating.
Change-Id: I8796c0e4b207d01111e2ead7ae55afb464dd88f5
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The code was only allocating memory for the next insertion
leading to a reallocation of the whole data for every
single insertion.
The code now reserves some space and uses a decent growth
strategy to avoid repeated reallocs.
Change-Id: I48b0feab71ba8ca73e7037f8460080f198b2f009
Reviewed-by: Jamey Hicks <jamey.hicks@nokia.com>
- Added toByteArray() and fromByteArray() benchmark tests.
Performance tests to measure QVariantMap to bytearray
and bytearray to QVariantMap.
Use case: Interprocess communications via local socket
Change-Id: If5e94ff870890b2ebb665f3cc38f5c33b34547f4
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
- Changed benchmarks to use TESTDATA and QFINDTESTDATA
- Fixed up targets all use tst_bench_ syntax
Change-Id: I5c2936702e248478f5df225ce38893158ee22d7f
Reviewed-by: Jason McDonald <jason.mcdonald@nokia.com>
Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
The implementation uses QScrollBar, which is no longer
included by <QtGui>.
Change-Id: I2422cfccc427179ca71e9a3195f16bd637925fb3
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
QPointF is in the category of types for which QList
is needlessly inefficient (elements are copy-constructed
onto the heap and held through pointers). Use a vector
instead. This is consistent with the QPainter API.
Change-Id: Id0e910c067a60d12fbc175e7ee7da824834be374
Reviewed-by: hjk <qthjk@ovi.com>
Reviewed-by: Lars Knoll <lars.knoll@nokia.com>
A case of a custom type creation was not covered before.
Change-Id: Icd2a7d63633f8e40d9d4a8a26e0eb0896fc85ec8
Reviewed-by: Kent Hansen <kent.hansen@nokia.com>
Do this regardless of whether the event subclass
is public API or only used in examples. Examples
are examples, used by others as templates or even
copied verbatim, so they should also follow sound
engineering rules.
Anyway, there's only one in examples/...
Change-Id: I586ff16407a956c9e89288fdd4377eed73f45c0f
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>
It's not faster under _any_ metric than the new algorithm, and it loses a lot of
spread which is a bad thing.
Change-Id: Ic87258f1c887822ffea1cb1517355564fabc3c26
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
We must do this the same way we do all other hash algorithms for fair
comparison, as otherwise, the call to the PLT unfairly penalises
QHash<QString>'s results, as it's in a different shared object.
Change-Id: I69c891f5a97dcccdfcfbdbf32796f86242a42963
Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
