| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
A RAII style class which calls a function at end of scope.
Example usage:
auto cleanup = qScopeGuard([] { <my cleanup code> ; });
[ChangeLog][QtCore] Introduced QScopeGuard.
Task-number: QTBUG-62894
Change-Id: Ife67f5c76255a1fafbae03367263da0bac9a0070
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
|
|\|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
src/plugins/platforms/cocoa/qcocoawindow.mm
src/plugins/platforms/xcb/qxcbintegration.cpp
Conflicts git missed:
src/plugins/platforms/qnx/qqnxglcontext.cpp
Change-Id: I0582cdc9e66e43efe79038b9c43d4f9572ac88fc
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
It's unclear when the reference to Unicode 4.0 standard
got added - it certainly predates the qt 4 git repository.
Anyhow, nowadays we're using later versions, and it doesn't
make much sense to highlight one specific version here.
Instead, use the correct technical term - UTF-16 code unit.
Also I added a 'correspond _to_', which sounds more common
to me.
Task-number: QTBUG-56699
Change-Id: I4bdcd9060cb2b11521638019c15ef7ab67aa768b
Reviewed-by: Leena Miettinen <riitta-leena.miettinen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Both delegate to the one that did do the if dirty: init() check, but
only after they've tested whether d->collator is set, which it might
not be when dirty.
Change-Id: I77533d6d32c4a8c9b42797c77003e50f5820775a
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Handle hard zero independently in each coordinate, otherwise hard zero
is never equal to anything but itself.
Task-number: QTBUG-69368
Change-Id: I8b1131472bb92efc706a04e0b067e2211a5ccb0c
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
|
| |
| |
| |
| |
| |
| | |
Change-Id: I94420a3dfa30e4a2a4d8786621bea1dd9309bed8
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Routine update in preparation for 5.12
* omitting Chakma because QLocale can't represent the zero digit,
* de_DE no longer uses vorm./nachm. but AM/PM
[ChangeLog][Third-party code] Updated CLDR to version 33.1
Task-number: QTBUG-67654
Change-Id: If20c47bb030abc3700b4f5a592152e617e2767c2
Reviewed-by: Kai Koehne <kai.koehne@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When I wrote zoneMSecsToEpochMSecs I seem to have copied names from
localMSecsToEpochMSecs, that need a s/local/zone/ change to their
names to make sense in this context. Flipped an if/else for clarity
in the process.
Change-Id: If2b288532b16d999f6ff0b9241e2dbcbc016b010
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| | |
Change-Id: Ic81461899c73c8a68bc3b8bdc1de4be4dd6bdf27
Reviewed-by: Martin Smith <martin.smith@qt.io>
|
| |
| |
| |
| |
| |
| |
| | |
qstring.cpp(6655): error C2065: 'lhs': undeclared identifier
Change-Id: Id59bdd8f1a804b809e22fffd15405ed7607bdaca
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
There was no public API for doing case-insensitive comparisons
of QStringView.
Task-number: QTBUG-69389
Change-Id: I1b021eefec35e135b97fb87704c8dc137232d83d
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| | |
And stick to one style in the same file.
Change-Id: I8a9a15425dadf4b480f63b68860de9276dbb787c
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| | |
We need to use macros because MSVC is a crappy C compiler.
Change-Id: Ieb48f7c0dd0e4e0fb35efffd153b8af62d34ebdf
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
As the comment says, Haswell is a nice divider and is a good
optimization target.
I'm using -march=core-avx2 instead of -march=haswell because the latter
form was only added to GCC 4.9 but we still support 4.7 and that has
support for AVX2.
This commit changes the AVX2-optimized code in QtGui to Haswell-
optimized instead. That means, for example, that qdrawhelper_avx2.cpp
can now use the FMA instructions.
Change-Id: If025d476890745368955fffd153129c1716ba006
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
[ChangeLog][QtCore][QByteArray] QByteArray::toInt(),
QByteArray::toDouble() and the other number conversion functions
now ignore leading and trailing whitespaces, as their QString
counterparts already did. For consistency reasons, the same
behavior was added to qEnvironmentVariableIntValue() also.
Task-number: QTBUG-66187
Change-Id: I8b5e478ea8577b811d969286ea9e269f539c1ea4
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Similar to commit cd64a96b31f57e522ab8d29c8357acf384012ebe we also need
to normalize the strings before comparison in order to be compliant with
the ECMAScript test suite.
This patch also adds the remaining test cases from
built-ins/String/prototype/localeCompare/15.5.4.9_CE.
Since the same tests are also failing with strcoll/qt_compare_strings,
this simplifies the code to always normalize except when using ICU
(which gets it right by default).
Change-Id: I16b32da7fc70dc7e6725c49f66fe9941d0bf3a47
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The compiler was generating some vectorized code for qresource.cpp but
it wasn't very efficient. So improve upon it and make use in other
places where we read UTF-16BE strings.
[ChangeLog][QtCore] Added an overload of q{To,From}{Big,Little}Endian
that operates on a memory region.
Change-Id: I6a540578e810472bb455fffd1531fa2f1d724dfc
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Happens a lot because of the implicit conversions. So I made it inline.
Change-Id: Icc2c231dc2c44abdb087fffd1533f311b95460b8
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|\|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
src/plugins/platforms/cocoa/qnsview.mm
src/plugins/platforms/cocoa/qnsview_dragging.mm
src/plugins/platforms/ios/qiosinputcontext.mm
src/plugins/platforms/xcb/qxcbconnection.cpp
src/plugins/platforms/xcb/qxcbconnection_xi2.cpp
src/plugins/platforms/xcb/qxcbwindow.cpp
src/tools/androiddeployqt/main.cpp
Was moved from qttools into qtbase in 5.11.
So re-apply 32398e4d here.
tests/auto/corelib/global/qlogging/test/test.pro
tests/auto/corelib/global/qlogging/tst_qlogging.cpp
tests/auto/corelib/io/qfile/tst_qfile.cpp
tests/auto/corelib/kernel/qtimer/tst_qtimer.cpp
tests/auto/corelib/thread/qthreadstorage/test/test.pro
tests/auto/widgets/itemviews/qheaderview/tst_qheaderview.cpp
tests/auto/widgets/kernel/qapplication/test/test.pro
Done-with: Gatis Paeglis <gatis.paeglis@qt.io>
Done-with: Mårten Nordheim <marten.nordheim@qt.io>
Done-with: Oliver Wolff <oliver.wolff@qt.io>
Change-Id: Id970486c5315a1718c540f00deb2633533e8fc7b
|
| |
| |
| |
| |
| |
| |
| |
| | |
.qdoc files should only be licensed under FDL (for documentation) or
BSD (if they only contain code).
Change-Id: I2a8a5b2c7bd3f927b8f71506c3672bc4916aca1e
Reviewed-by: Leena Miettinen <riitta-leena.miettinen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
QTimeZonePrivate::isValid() just checks m_id is non-empty; so we have
to leave m_id clear if we don't get a valid time-zone back when we ask
the JNI for one. Unfortunately, JNI gives us a "valid" default zone
if it doesn't recognize the given name; so check the known names of
this zone (or of zones with its offset); if the given ianaId isn't one
of them, assume this is a bogus zone.
Task-number: QTBUG-68842
Change-Id: I6245db18c59c4261ed5fcd4d948dd773365ce61d
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| |
| | |
While at it, fix some more issues in the sentences to
harmonize the description between the different classes.
Change-Id: Iee1c3ffe6fd71e82504bfb003d927c4db3b2a065
Reviewed-by: Martin Smith <martin.smith@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| | |
While at it, add the comment to the toFloat() functions also.
Task-number: QTBUG-55232
Change-Id: I21c06363946f35fb3d89a51e4f75be392b57c0a9
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Both use QLocale::toDouble behind the scenes, so the same limitations
apply. Document them for toFloat() also.
Change-Id: I954362a0db203630685c034df6a921fa6447a509
Reviewed-by: Martin Smith <martin.smith@qt.io>
|
| |
| |
| |
| |
| |
| | |
Task-number: QTBUG-68933
Change-Id: I3f2a9f8c562f9a44bb32bddd31d75abbfe6de04d
Reviewed-by: Topi Reiniö <topi.reinio@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Task-number: QTBUG-68702
Change-Id: I04b94079b6da48f39a82fffd153568f8dab3ef1b
Reviewed-by: Ville Voutilainen <ville.voutilainen@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
QString::sprintf(), like the C printf-family, always includes two
digits in any exponent it outputs. Up to 5.6, number() and arg()
taking a double did the same; but changes at 5.7 to enable opting out
of the leading zero this implies for a single-digit exponent
accidentally opted out of it in args() and number(). This commit
fixes number() and arg() to include QLocaleData::ZeroPadExponent in
the flags they pass down to the C locale's doubleToString(), restoring
the prior behavior, including consistency with sprintf().
[ChangeLog][QtCore][QString] Formatting of doubles with single-digit
exponent, by number() or args(), now includes a leading zero in that
exponent, consistently with sprintf(), as it did up to 5.6.
Task-number: QTBUG-63620
Change-Id: I10c491902b8556e9f19e605177ead8d9fd32abd9
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Ulf Hermann <ulf.hermann@qt.io>
|
| |
| |
| |
| |
| |
| |
| | |
Using SSE 4.1 because of the need for PMINUB.
Change-Id: Ib48364abee9f464c96c6fffd152ebd3f8ea7fe94
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
With ICU and on macOS it appears that the comparison is done on a
canonical form, while CompareString(Ex) does not do that, as the added
test verifies. Explicit normalization fixes that.
As a bonus, this also unifies the code path between regular Windows
and UWP by unconditionally using CompareStringEx (which requires
Vista or later).
This issue surfaced while running the ECMASCript 6 Conformance Test
Suite in QtQml.
This re-uses the existing test for localeAwareCompare, which was
disabled on Windows, macOS and Linux with ICU (the common case).
Change-Id: I52440fce60b54745ead1eff005ec51e98e2a79ec
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
... instead of being the bit numbers. This allows us to test more than
one feature at a time with qCpuHasFeature (see commit about the Haswell
architecture features). The drawback is that we won't be able to handle
more than 63 different CPU features, though we're likely quite far from
it (x86 currently has only 36 features).
Change-Id: Iff4151c519c144d580c4fffd153a0acbfd74c2c6
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
It needs to be after our extra #defines for CPU features, otherwise the
qCompilerCpuFeatures variable won't be set correctly.
Change-Id: Iff4151c519c144d580c4fffd153a0a93ca01bbe6
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
|
| |
| |
| |
| |
| |
| |
| | |
Use nullptr, clarify comments.
Change-Id: Ib5f5879d8281cc455635513f82ff8c4bdb951ea8
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Need to do the same for startsWith() and endsWith(). indexOf() is a lot
harder.
[ChangeLog][QtCore][QByteArray] Added compare(), which takes
Qt::CaseSensitivity as one of the parameters. This function is more
efficient than using toLower() or toUpper() and then comparing.
Change-Id: Ib48364abee9f464c96c6fffd152e69bde4194df7
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Timur Pocheptsov <timur.pocheptsov@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
After the move to PCRE2, optimizing patterns has been a thorn in the
side due to the fact that PCRE2's JIT compiler modifies the pattern
object itself (instead of returning a new set of data, like PCRE1
did). To make this fit with the existing behavior, a read/write
lock was introduced, with the read part locking when matching and
the write when compiling (or JIT-compiling) the pattern.
This locking strategy however introduced a performance issue,
as we needed:
* to acquire a write lock to compile/optimize the pattern (incl. the
common case where the pattern was already compiled, so bailing out
immediately);
* to acquire a read lock during the actual match, to prevent
some other thread from optimizing the pattern under our nose.
This was due to the "lazy" optimization policy of QRegularExpression
-- optimize a pattern after a certain number of usages. The
excessive amount of locking effectively limited scalability.
Simplify the code, and drop that policy altogether: since JIT
compiling in PCRE2 is faster and pretty much "always recommended",
just always do it for any pattern (unless it gets disabled via
env variables) when compiling it.
This allows to go back to a plain QMutex, and now the actual
matching doesn't require acquiring any locks any longer. Of course,
there is still a mutex acquired just before matching for checking
whether the pattern needs recompiling in the first place; this can
probably be further optimized via double-checked locking (using
atomics), but not doing it right now.
This shift makes a couple of pattern options controlling
optimization useless, and allows to centralize the 3
QRegularExpression tests (which were actually the very same test,
just setting slightly different optimizations strategies).
While at it, install a stress-test for threading, with the idea
of running it under TSAN or helgrind to catch bugs in
QRegularExpression's locking.
[ChangeLog][Important Behavior Changes][QRegularExpression] Regular
expressions are now automatically optimized (including JIT
compiling) on their first usage. The pattern options
OptimizeOnFirstUsageOption and DontAutomaticallyOptimizeOption no
longer have any effect, and will get removed in a future version of
Qt. QRegularExpression::optimize() can be still used to compile and
optimize the regular expression in advance (before any match), if
needed.
Task-number: QTBUG-66781
Change-Id: Ia0e97208ae78255fe811b78029ed01c204e47bd2
Reviewed-by: David Faure <david.faure@kdab.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Failed on my Sandybridge Mac with -march=native:
qstring.cpp:363:19: error: redefinition of 'mask' with a different type: 'const __m128i'
Change-Id: I6efb28c3145047559ec0fffd15386aeb8d36d681
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Commit 5dc1e08c8c602738f6a827cea0de44683c7bbd0b changed
QString::toLocal8Bit_helper to use qt_convert_to_latin1 so it became
unused. It was never used in inline functions in the headers.
Change-Id: I6efb28c3145047559ec0fffd15382f9d08efdfeb
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
|
|\|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
.qmake.conf
src/corelib/kernel/qeventdispatcher_cf.mm
src/gui/kernel/qguiapplication_p.h
src/gui/kernel/qwindowsysteminterface.cpp
src/gui/kernel/qwindowsysteminterface.h
src/plugins/platforms/cocoa/qcocoawindow.mm
src/plugins/platforms/cocoa/qnswindowdelegate.mm
src/plugins/platforms/ios/qioseventdispatcher.mm
src/plugins/platforms/windows/qwindowsdrag.h
src/plugins/platforms/windows/qwindowsinternalmimedata.h
src/plugins/platforms/windows/qwindowsmime.cpp
src/plugins/platforms/winrt/qwinrtscreen.cpp
Change-Id: Ic817f265c2386e83839d2bb9ef7419cb29705246
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
qFuzzyIsNull has a fixed range, where qFuzzyCompare can tell if numbers
are different in a more relative range. Without it QPointFs that are
heavily scaled will be interpreted as identical, when they are quite
different at their own scale.
Task-number: QTBUG-60359
Task-number: QTBUG-62161
Change-Id: Ic4ba90e9e994aedff5548d690f053eb309b0a60b
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
They were introduced in commit c416a7f25770563a265cc86e779f2e54c01a85a0.
warning: ‘QString::QString(const QByteArray&)’ is deprecated: Use fromUtf8, QStringLiteral, or QLatin1String [-Wdeprecated-declarations]
Change-Id: I6a540578e810472bb455fffd1532e31736e1edc9
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This reverts commit 6e1190053d28877b2e90375056735429525e2ee7.
We had to bring back qatomic_msvc.h for MSVC 2015, which does
not provide 64-bit atomics on 32-bit platforms.
Task-number: QTBUG-68719
Change-Id: Iea902cf0b01191717e0a640944771fcede7b5ff8
Reviewed-by: Liang Qi <liang.qi@qt.io>
|
|\|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
mkspecs/features/qt_common.prf
src/corelib/tools/qstring.cpp
src/plugins/platforms/windows/qwindowsmousehandler.cpp
src/widgets/widgets/qmainwindowlayout_p.h
Change-Id: I5df613008f6336f69b257d08e49a133d033a9d65
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
If LANGUAGE specified only the language, without any script or
country, and matched the value we'd got from other environment
variables, we were throwing away their knowledge of script and
country, leading to falling back on the default script and country for
that language, which might be at odds with what other environment
variables had told us.
Changed to only use LANGUAGE if it contradicts (or extends) what we
would otherwise have used. Clarified some comments in the process.
[ChangeLog][QLocale][Unix] When using LANGUAGE would lose information
about script or country, without changing language, use the locale
implied by LC_ALL, LC_MESSAGES or LANG.
Prompted-by: Safa AlFulaij <safa1996alfulaij@gmail.com>
Change-Id: Ie433e57ae6b995abafd05c931136cc9796494895
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Turns out that the non-AVX2 code was beating the performance of the AVX2
because the simdTestMask function did a little too much. So just use the
same VPMOVMSKB technique for it.
Change-Id: I0825ff5b5f6f4c85939ffffd152f3b636ab998db
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
[ChangeLog][QtCore][QSharedPointer] Fixed a problem that made create()
on a type with const qualification fail to compile.
Task-number: QTBUG-68300
Change-Id: I0825ff5b5f6f4c85939ffffd152f3e55e5b9caae
Reviewed-by: Ville Voutilainen <ville.voutilainen@qt.io>
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
First of all, this removes the UB that used to try and calculate the
distance between the two strings. That's a valid technique in assembly,
but dangerous in C++ and totally unnecessary. The compiler is perfectly
able to generate loops with a single induction variable all on its own.
Second, this commit makes the main loop use 32-byte comparisons (16
characters at a time), which is a reasonable size for strings. We use
AVX2 if that's available, or an unrolled pair of 16-byte loads
otherwise. After the existing 16-byte comparison, this commit inserts an
8-byte (4-character) comparison and then reduces the final, unrolled
comparison to just 3 characters.
Change-Id: Ib48364abee9f464c96c6fffd152e474b39e1f293
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Instead of two 16-byte loads, if AVX2 is present. Otherwise, it's
exactly the same.
Because of the way the SIMD instructions were extended to 256-bit in
AVX2, we gain nothing doing two 32-byte loads, aside from the loop
unrolling.
Change-Id: Ib48364abee9f464c96c6fffd152e531925814ac2
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This instruction is somewhat slow and requires a lot of inputs to be
correctly set. Instead, use the PMIN trick, which does have unsigned
comparison support.
This commit moves the helper function to a lambda inside qt_to_latin1,
to make it easier to reuse the constants in the next commit and to avoid
warnings of unused static functions.
Change-Id: Ib48364abee9f464c96c6fffd152e9e84f4ad3ae8
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
If the input is already known to be Latin 1, we don't need to check and
merge in question marks. QJsonObject already needed this code, now we
can make it more efficient.
I'll need the same code in CBOR.
Change-Id: Ib48364abee9f464c96c6fffd152e508f078404e5
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We don't have _mm_cvtsi64_si128() (the REX.W expansion of MOVD [0F 6E]),
but we do have _mm_loadl_epi64(), the SSE2 expansion of the MMX MOVQ at
opcode 0F 7E. Ditto for _mm_cvtsi128_si64() and _mm_storel_epi64(). And
those work even in 32-bit mode. By doing this, we can reduce the tail
unrolled loops by half, reducing code size.
I'm not adding these new SIMD sections to -Os builds.
Change-Id: Ib48364abee9f464c96c6fffd152e405310ef67be
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Improves performance a little. This is just because I can and the
function is right there for the taking, as this qt_urlRecodeByteArray
function is only used in deprecated QUrl code.
Change-Id: I5d0ee9389a794d80983efffd152d290e570af387
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|