| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The loop for 32 bytes is left unchanged, but the tail operation for 16
to 31 bytes is replaced with an overlapped load-and-scramble. This
should make the operation even faster.
Also updated the key creation back to something similar to what Go does.
This massively improves performance as well as the bit spread. Histogram
for the bits in the hash value for the testcase from QTBUG-91739:
|| Bit || Before || After ||
| 0 | 35.0300% | 50.4800% |
| 1 | 42.5250% | 50.2400% |
| 2 | 46.0100% | 50.0000% |
| 3 | 67.5150% | 49.9400% |
| 4 | 56.5150% | 50.0000% |
| 5 | 51.9950% | 50.0000% |
| 6 | 58.9800% | 50.1400% |
| 7 | 55.9550% | 50.0000% |
| 8 | 41.9850% | 49.9200% |
| 9 | 69.9700% | 49.6400% |
| 10 | 68.4950% | 50.0000% |
| 11 | 37.4950% | 50.3000% |
| 12 | 61.9950% | 49.8200% |
| 13 | 53.4900% | 50.0000% |
| 14 | 63.0200% | 49.9800% |
| 15 | 54.9700% | 50.1000% |
Task-number: QTBUG-91739
Pick-to: 6.2.3 6.2 6.3
Change-Id: Icad7c1bad46a449c8e8afffd16cb7fe7ffd3584f
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Allan Sandfeld Jensen <allan.jensen@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
[ChangeLog][QtCore][QHash] Fixed a bug in the qHashBits() function,
which affected the hashing of QByteArray, QString (and their View
classes), QLatin1String and QBitArray, which caused the hash to not
include the final 32 bytes of the data source. As a result, QHash
containers where the initial string was the same had a serious
performance degradation on x86 CPUs with AES support.
Fixes: QTBUG-91739
Pick-to: 6.2.3 6.2 6.3
Change-Id: Icad7c1bad46a449c8e8afffd16cb74dd43440f6c
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
|
|
|
|
|
|
|
|
|
|
|
| |
We don't need to test for actually crossing a page boundary. It suffices
to check if we're in the upper half or the lower half of the page. In
the upper half, we load ending at the end; in the lower half, we load
starting at the current position. This way, it can never crash.
Pick-to: 6.2.3 6.2 6.3
Change-Id: Icad7c1bad46a449c8e8afffd16cb743e622b3405
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
qSwap() is our wrapper around
using std::swap;
swap(lhs, rhs);
it needn't and shouldn't be overloaded.
ADL swap() should be, though, so qSwap(), std::ranges::swap() and all
the other adl_swap()s out there all find the optimized version.
Qt 5.15 has it correct, Qt 6 wrong. Fix it.
Can't pick to 6.2 because, while backwards-source-compatible, because
the generic qSwap() template provides the name for both qualified and
unqualified calls, it's not forwards-source-compatible: A new user of
ADL swap
// compile error w/o `using std::swap`, pessimization otherwise:
swap(dp1, dp2);
would break or performance-regress when going back to an older
version.
Pick-to: 6.3
Change-Id: I725949a4aa9ae438a182b4b7552ff2dced767e2f
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
(QtGui)
Create macros that wrap the magic developed in
7d63efc16f65f98c657caa90e0d7e9b72a879ade and apply it to all
Q_DECLARE_METATYPE invocations that show up in Clang -ftime-trace for
a PCH'ed QtGui build.
Effects on compile times:
Clang 10 -ftme-trace:
$ ClangBuildAnalyzer --analyze qtgui-before.trace | head -n6
Analyzing build trace from 'qtgui-before.trace'...
**** Time summary:
Compilation (523 times):
Parsing (frontend): 628.3 s
Codegen & opts (backend): 304.5 s
$ ClangBuildAnalyzer --analyze qtgui-after.trace | head -n6
Analyzing build trace from 'qtgui-after.trace'...
**** Time summary:
Compilation (523 times):
Parsing (frontend): 546.0 s
Codegen & opts (backend): 304.4 s
GCC 11 time (bash builtin):
before:
$ time for ((i=0; i < 3; ++i)) do touch src/gui/painting/qpolygon.h ; ninja libQt6Gui.so; done
real 4m13,539s
user 49m24,416s
sys 3m18,177s
after:
$ time for ((i=0; i < 3; ++i)) do touch src/gui/painting/qpolygon.h ; ninja libQt6Gui.so; done
real 3m55,697s
user 45m19,941s
sys 3m7,370s
Task-number: QTBUG-97601
Pick-to: 6.3
Change-Id: Ia8e37a58937568a7ed21cfeb4b27274deca4d53b
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
qSwap() is a monster that looks for ADL overloads of swap() and also
detects the noexcept of the wrapped swap() function, so it should only
be used when the argument type is unknown. In the vast majority of
cases, the type is known to be efficiently std::swap()able or to have
a member-swap. Call either of these.
For the common case of pointer types, circumvent the expensive trait
checks on std::swap() by providing a hand-rolled qt_ptr_swap()
template, the advantage being that it can be unconditionally noexcept,
removing all type traits instantiations. Don't document it, otherwise
we'd be unable to pick it to 6.2.
Effects on Clang -ftime-trace of a PCH'ed libQt6Gui.so build:
before:
**** Template sets that took longest to instantiate:
[...]
27766 ms: qSwap<$> (9073 times, avg 3 ms)
[...]
2806 ms: std::swap<$> (1229 times, avg 2 ms)
(30572ms)
after:
**** Template sets that took longest to instantiate:
[...]
5047 ms: qSwap<$> (641 times, avg 7 ms)
[...]
3371 ms: std::swap<$> (1376 times, avg 2 ms)
[qt_ptr_swap<$> does not appear in the top 400, so < 905ms]
(< 9323ms)
As a drive-by, remove superfluous inline keywords and template
ornaments.
Task-number: QTBUG-97601
Pick-to: 6.3 6.2
Change-Id: I88f9b4e3cbece268c4a1238b6d50e5712a1bab5a
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Qt 5 uses begin() so the fix there will be to use cbegin().
Found by Clang -ftime-trace pin-pointing repeated instantiations
of QList<int>::data().
Pick-to: 6.3 6.2 5.15
Task-number: QTBUG-97601
Change-Id: I6410e5b303766fdbc7e158a9ac1263adec973099
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
|
|
|
|
|
|
|
|
|
| |
Hot function, so help the compiler out. In particular, this removes any
touch to the ref count, so there are no atomic operations or dead code
leading to memory allocations and deallocations.
Change-Id: I0e5f6bec596a4a78bd3bfffd16c9a0fbd8dd2c12
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Use QVarLengthArray instead of QList to avoid allocations in the case
where the result fits into inline storage.
Of course, we now perform one allocation more when we need the QList
backend, but a) that should be the rarer case and b) we use 32 (more
than InlineSegmentsCount) Prealloc for the QVarLengthArray to reliably
skip the first few QList non-reserved capacity jumps (measured to be
{4, 12, 28} for back-insertion, not all of which is probably available
for append()s) to come out ahead either way.
Reviewers may object, saying that we could just reserve(32) the QList,
too, but while that would skip over the first few QList reallocations
alright, it means we'd be carrying the extra capacity around for the
duration of the QVersionNumber's lifetime (unless we'd shrink it at
the end, bringing back the additional allocation and that solution
back to par compared to this one).
As a consequence:
[ChangeLog][QtCore][QVersionNumber] Can now be also be constructed
from QVarLengthArray.
Change-Id: I4016367f64f6cefa6ed9147d33b06636f36b02cb
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We can handle the UTF-8 case by reinterpreting it as Latin-1.
This way, the suffixIndex stays valid as a return value.
As a drive-by, optimize away toLatin1() calls by using a QVLA.
We really need a better way of converting UTF-16 -> L1 than
qt_to_latin1()...
[ChangeLog][QtCore][QVersionNumber] fromString() now takes
QAnyStringView (was: QString, QStringView, QLatin1String)
and a qsizetype pointer (was: int pointer).
Change-Id: I86abaadba2792658fbf93ccd1e0b86e3302c697c
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
|
|
|
|
|
|
|
| |
This completes the update to qsizetype in this class, adding a couple of
methods that need to be removed in Qt 7. They're only required where int
is not qsizetype (i.e., 64-bit platforms).
Change-Id: I0e5f6bec596a4a78bd3bfffd16c9de29bec4c637
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
QDoc made some assumptions about the module a class/header belongs to,
based on the source file path. This feature is rather error-prone and
unnecessarily complex and will be removed from QDoc.
Define modules explicitly to avoid documentation warnings when this
removal happens.
Pick-to: 6.2 6.3
Change-Id: I7947d197db5ac36c12e816caa19bb2f74eda8849
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
Reviewed-by: Venugopal Shivashankar <Venugopal.Shivashankar@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This class is not exported, so we can change the non-exported
methods. None of the exported methods required change; notably,
QVersionNumber::Segments::setVector didn't because it's only called with
values 1, 2, and 3.
[ChangeLog][Potentially source-incompatible changes] Updated the
QVersionNumber API to use qsizetype where length and index values were
used. This change retains binary compatibility and the vast majority of
users will not experience a source compatibility problem. It could occur
with ambiguous overloads when passing results from QVersionNumber to
other API not using either int or qsizetype. There could also be new
warnings from compilers about converting 64-bit types to 32-bit ones.
Change-Id: I0e5f6bec596a4a78bd3bfffd16c9984b61c9b55b
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
|
|
|
|
|
|
|
| |
Pick-to: 6.3 6.2
Change-Id: Id7ab2fe09c01500ca5bd23751ba29ed1394bb9b6
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The recommended way to detect a contiguous iterator isn't to check
the iterator_category; it's to use the iterator concepts.
Similarly, the recommendation set in place by P2259 (for being
backwards-compatible) is to declare a iterator_concept member,
not to change iterator_category to a C++20 category, (also)
because legacy code may be checking for equality against a specific
category, rather than for convertibility. This is erroneous, but
such code exists, alas.
This is enshrined in C++20's stdlib: for instance, iterator_traits<Foo*>
has random_access_category_tag as iterator_category, but
contiguous_iterator_tag as its iterator_concept.
Hence:
1) in QArrayDataOps use the concept, and not the category, to do
the check
2) when declaring iterators, keep the category as random access,
and introduce the concept alias (if supported).
Pick-to: 6.2 6.3
Change-Id: Ib600da7331d687a15082becaa6be06aefc24bb9c
Reviewed-by: Andrei Golubev <andrei.golubev@qt.io>
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
Suppresses GCC's -Wconversion, which is not enabled by default.
error: conversion from ‘int’ to ‘quint8’ {aka ‘unsigned char’} may change value [-Werror=conversion]
Change-Id: I0e5f6bec596a4a78bd3bfffd16c998102bd51f7c
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
|
|
|
|
|
|
|
|
| |
Avoids code duplication.
Pick-to: 6.3 6.2
Change-Id: Ic69e46108baf97a0dc9215866d6c707136ee40b2
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This does exactly what insert() on Qt associative containers does, but
allows to express the intent of using the STL-incompatible Qt insert()
semantics, in an STL-compatible way, instead of leaving the reader of
the code wondering what semantics are expected.
This is part of a very-long-term goal of fixing Qt associative
container's insert() behavior, in which QFlatMap, being an affected,
but private-API type, is used for proof-of-concept purposes.
Task-number: QTBUG-99651
Pick-to: 6.3 6.2
Change-Id: I69010285438259918aef659d3235180c1b5be696
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We use raw pointers to the Nodes in the QHash which is
inherently fine, but we are then subject to invalidation when
nodes are moved around during deletion.
In trim() we don't actually need to iterate the linked-list
since the node we are interested in is always chain.prev
Pick-to: 6.3 6.2 6.2.3
Fixes: QTBUG-99710
Task-number: QTBUG-99224
Task-number: QTBUG-99240
Change-Id: I9c2ed69b29e3cadca013113a3553deb44d7382fc
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Jarek Kobus <jaroslaw.kobus@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
... in an attempt to foster the use of this data structure by making
it less onerous to spell.
Pick-to: 6.3 6.2
Change-Id: Ib9d17029c75278edde6ba90f65f68af179a6d230
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
QFlatMap, like its public brethren, features the broken Qt-style
insert() behavior (what the STL calls insert_or_assign()), which
makes its insert() unusable for actual STL-style insert() work,
with no replacement except the size-check-and-index-operator trick:
const auto oldSize = c.size();
auto &e = c[key];
if (c.size() != oldSize) {
// inserted
}
Even though QFlatMap::insert() appears to return the correct info,
it's useless, because the old value has been assigned over by the
time insert() returns.
Pick-to: 6.3 6.2
Change-Id: If4173c42523a128dfd22ab496dde0089ba73f41c
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|
|
|
|
|
|
|
|
|
| |
Use {copy,move} ctors instead of default-ctor, followed by
(move|copy)-assignment.
Pick-to: 6.3 6.2
Change-Id: Id2fd53050cd353a9374fd065ac25d753d42d1be9
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
| |
When looping over range-insert(), the repeated shrink_to_fit() calls
would cause cause reserved (or geometrically-grown) capacity to be
shed, breaking the underlying container's growth strategy.
Fix by not shedding excess capacity.
Pick-to: 6.3 6.2
Change-Id: I10915a06fc9442039c192486a55e48083da7c839
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|
|
|
|
|
|
|
|
| |
Iterators model pointer, so const must be shallow.
Pick-to: 6.3 6.2
Change-Id: I90494c98762f1494efcca4965ee739540333f5d7
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
|
|
|
|
|
|
|
|
| |
They never worked.
Pick-to: 6.3 6.2
Change-Id: I9a15c848416419823f28ea580248fbe93a4365dd
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
|
|
|
|
|
|
|
|
|
| |
For RAII objects, it's natural to have move semantics enabled these
days. So if a RAII type does _not_ offer it (and, as in this case,
cannot be made to offer it), be explicit and disable moves, too.
Change-Id: I5636a0c7ff2fc51982ce9a406f975b07d51af01a
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
The header only uses std::destroy(), which is in <memory>, not
<algorithm>.
Doesn't change anything, because <algorithm> is still transitively
included via qcontainertools_impl.h.
Pick-to: 6.3 6.2
Change-Id: Ib6a94ea56d5819f916c80a2e307c99b328f3d61d
Reviewed-by: Fabian Kosmale <fabian.kosmale@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- use sliced() instead of mid(), since we know we call sliced()
in-contract
- replace the manual memmove() with calls to QBA::remove(), but first
resize() to tailOffset, and only then remove(0, headOffset), so we
only move Chunk::size() bytes, as the old code did
- don't bother updating the offset members, as we're restricted to
rvalues now, and we only need to leave behind a partially-formed
object
This code shows that we desperately need QBA::slice() and/or an rvalue
overload of QBA::sliced(), cf. QTBUG-99218. Add a comment to use these
functions when they become available.
Change-Id: I5d8c7363d443dff69338f6f6a7b4bff9957917ec
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Alex Trotsenko <alex1973tr@gmail.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
|
|
|
|
|
|
|
|
| |
QByteArray::resize() is now documented to not shrink capacity(), so
remove the extra reserve(0) work-around.
Change-Id: I3d5d944b9f01f8f84b31599548773198903aad4c
Reviewed-by: Alex Trotsenko <alex1973tr@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This removes some calculations that are not required in
80% of the cases where d->erase() is being called (as
the return value is ignored). When removing lots of
elements from the hash, the ++it could loop quite a
bit until it found the next valid item in the hash.
This chain of changes combined improve the overall performance of
QHash by 10-50% depending on the operation. Deletes are twice
as fast, reads around 20% faster, inserts around 10% faster.
Task-number: QTBUG-91739
Fixes: QTBUG-98436
Change-Id: I2d82a7c9dd1dd0a4da8402e6d95bfd620caeff3a
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Introduces a QHashPrivate::Data::Bucket class. This class
helps avoid repeated bitshift and masking operations when
locating bucket entries in the hash.
Change signature of some internal methods to use Bucket
instead of the private iterator, so we can avoid repeated
encoding/decoding steps for the bucket index.
Task-number: QTBUG-91739
Task-number: QTBUG-98436
Change-Id: I9ed2205bf886f9c20a5be109fd88456eec4d1540
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Use NSDMI and let the compiler generate the default ctor, too.
Statically assert that it's still noexcept (we marked it as such,
because MSVC still doesn't noexcept(auto) when an SMF is =default'ed.
Simplify the other ctors, too, relying on NSDMI. Keep one instance of
tailOffset(0), in the 'reserving' ctor, because it's inconsistent with
the other ctors which all do tailOffset{chunk.size()}, so it's best to
leave this explicit, lest the next reader wonders whether it was
forgotten.
Pick-to: 6.3
Change-Id: I0d9f91aa5bc89377c4144589df8786b502e956a4
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
QHash::operator[] could grow the hash even if the key being
looked up already existed. This in turn invalidated all iterators.
Avoid this by refactoring findOrInsert() to not grow if the key
already exists.
Added advantage is that this should make lookups of existing keys
slightly faster.
Fixes: QTBUG-97752
Pick-to: 6.3 6.2
Change-Id: I9df30459797b42c434ba0ee299fd1d55af8d2313
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
|
|
|
|
|
|
|
|
|
| |
The Node::replace overloads and the Node::create method are not used
anywhere (including in QHash).
Pick-to: 6.2 6.3
Change-Id: I1f1f8c847c8f1181cd31d95d01997226f94b5fc4
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
Let the compiler generate the copy and move SMFs.
Statically assert that the move SMFs are still noexcept.
Pick-to: 6.3
Change-Id: I1c569bdf893a5f2cda972c0dd8196cab62494fcb
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
|
|
|
|
|
|
|
|
|
| |
It's only user, QRingBuffer::read(), calls it on an rvalue, so we
don't need the lvalue overload, and we can devil-may-care-like just
std::move(chunk) in the non-isShared() case.
Change-Id: I99c16862f5586125c6346ce5f969dc735de738b8
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
|
|
|
|
|
|
|
|
|
| |
Just use the QByteArray(ptr, n) ctor instead of the (n,
Qt::Uninitialized) one + memcpy() + std::move().
Pick-to: 6.3
Change-Id: I127219c21556e683d15136f7e6f7b3576b7b2444
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
| |
There's no sense in copying a ring buffer. Moving is enough. This
marks an important step on the way to preventing accidental copies of
ring buffer content, because the 'QList buffers' member can now no
longer be implicitly shared. While the compiler will still emit the
code for detach()ing, it will now never be executed.
Pick-to: 6.3
Change-Id: I968bfe3e50c46720ed4baca55c99c1f9c518f653
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
The majority of append() callers in QtBase pass rvalues, so overload
append() to avoid the need for manipulating QBA's atomic ref counts.
Also adjust a caller that could pass by rvalue, but didn't, to do so.
Pick-to: 6.3
Change-Id: I3d9e60b0d04ef837bfdc526e1f0f691a151006f9
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@gmx.de>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
|
|
|
|
|
|
|
|
|
|
|
| |
Clear should bring the QRingChunk into the default-constructe state,
so just assign {} instead of calling assign(QByteArray()), which is
equivalent, but has to manipulate QBA's atomic ref-count, which makes
it unlikely the compiler will optimize this call.
Pick-to: 6.3
Change-Id: Idab9c92ec84392d484395042db2427c668756efa
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add some scaffolding to prevent us from running into BC issues due to
the underspecified nature of std::aligned_storage.
Qt 5.15 uses a union { char[], double, qint64 } instead of
std::aligned_storage, so doesn't need the fix.
References:
- https://github.com/cplusplus/papers/issues/197
- wg21.link/p1413
Task-number: QTBUG-99122
Pick-to: 6.3 6.2
Change-Id: I212be7000376c2db33b8cb244a6e862cc4dad544
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
|
|
|
|
|
|
| |
Check, in tst_qvarlengtharray, that the forwarding header still works.
Pick-to: 6.3
Change-Id: Ia03cf48457f538287880bb676aea3fa44aeb255f
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Be smarter when allocating memory for the real data in
addStorage(). As the amount of entries in there follow
a binominal distribution, we know pretty well, how many
there will be at least and at most. This avoids most of
the reallocations of the storage when not rehashing, while
Tessil's tests show that the total memory consumption has
not changed.
Task-number: QTBUG-91739
Task-number: QTBUG-98436
Change-Id: I98854bfbde8b2a16e788bfa1890c694d38fd09b5
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Change the minimum amount of buckets to be at least
128, ie. one full Span. This will simplify some assumptions
in the code.
Regeneration of rcc test-data needed because the extra buckets
causes the order of the keys to change.
Task-number: QTBUG-91739
Task-number: QTBUG-98436
Change-Id: Ic0c7da03570cc4c4e6aacc9645e536aec3667a98
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
|
|
|
|
|
|
|
| |
To avoid needing a templated Span to access them
Task-number: QTBUG-98436
Change-Id: I5c3b16f9a806986b14ed794a3818990aa3ead0f7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Marc Mutz <marc.mutz@qt.io>
|
|
|
|
|
|
|
|
|
|
| |
QHash changes some of its preconditions, so we must not call
findNode without verifying !isEmpty()
Task-number: QTBUG-91739
Task-number: QTBUG-98436
Change-Id: I2701b9a01187530f541a7c9a12db56c92f856d87
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
|
|
|
|
|
|
|
|
| |
It was added in 5.15 when QHash could have multiple entries per key.
Pick-to: 6.2 6.3
Change-Id: Ifb855b05dd65bbc67bb64bf2567229394525b2aa
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We already checked that the two iterators, indvidually, are valid, but
we didn't check that the range formed by them is valid, namely that
the end iterator is reachable from the start iterator.
Add an assert, because if the range isn't valid, we run into UB in the
std::move() algorithm two lines later.
Qt 5.15 uses std::copy() here, which has the same precondition, so the
assertion would make sense there, too.
Pick-to: 6.2 5.15
Change-Id: I90b7e846455ff86383a8971bea908036684961d8
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A previous attempt to make qHash() a hidden friend ran into an ICE on
Clang, so to simplify the implementation, QVLABase::hash() is public
for now.
Use QtPrivate::QNothrowHashable_v to simplify the noexcept clauses.
Fixes: QTBUG-84785
Change-Id: Ibb520bde7b1d6dc45846bd8e1fa27c39c401b96a
Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
|
|
|
|
|
|
|
|
|
| |
Also added a QVLABase::resize_impl() because insert(it, n, v) was the
only function moved down into QVLABase to call resize().
Task-number: QTBUG-84785
Change-Id: I5dd0092216d73b28b957a01845325d744a5c0ba9
Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
|