Do a check first if we need to transform before doing the transform.
This means we won't detach when transforming data that is already
correct.
And instead of using QChar, use our own hand-rolled table. In a proper
LTO build, the QChar calls would be resolved to a lookup of the Unicode
data, but not many people do LTO builds, Therefore, this means a great
speed-up is achieved by simply avoiding the function call. The extra
gain in performance comes from the simpler translation table instead of
the more complex full-Unicode data.
Also as a consequence, this changes the handling of two characters in
Latin 1: 'ß' should be uppercased to "SS" but we won't do it, and 'ÿ'
can't be uppercased in Latin 1 ('Ÿ' is outside the range).
Benchmarking is included. Comparing the Qt 5.4 algorithm to the new code
is almost 20x faster. Other alternatives are included in the benchmark
and are all faster than the current code, though slower than the new
one. While all of them could compress the tables to be smaller or shared
between uppercasing and lowercasing, they would also expand to more code
(though probably less than the extra bytes required in the full
translation table). In the trade-off, I decided to go with simplicity
and most efficient code.
Change-Id: I002d98318d236de0d27ffbea39d662cbed359985
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
The comparison, Latin 1 and UTF-8 benchmarks contained in this file are
stale. The implementation changed in Qt 5.3 and this benchmark couldn't
be updated (test data too large for Qt).
Please contact Thiago Macieira to obtain the benchmarks and test data.
Change-Id: I48c19b1f1711eb73c953a30ed4da510e97a62472
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Erik Verbruggen <erik.verbruggen@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
The "rbit" instruction requires ARMv6T2 or higher. This was found in the
CI when building the imx6 target:
Compiler: arm-poky-linux-gnueabi-g++
Flags: -mfloat-abi=hard -mfpu=neon
Errors from the assembler:
{standard input}:3078: Error: selected processor does not support ARM mode `rbit r3,r3'
{standard input}:7341: Error: selected processor does not support ARM mode `rbit ip,ip'
That compiler defaults to ARMv5T. That's obviously wrong for an i.MX 6,
which is a Cortex-A9 (ARMv7), but the correction applies for older
processors.
Change-Id: I56c276fa411977dd7cd867d62adf021e4909302c
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
QRingBuffer is a fully inlined class used in many I/O classes.
So, it must be as fast and small as possible. To this end, a lot of
unnecessary special cases were replaced by generic structures.
Change-Id: Ic189ced3b200924da158ce511d69d324337d01b6
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
main.cpp(332) : warning C4307: '*' : integral constant overflow
tst_qpainter.cpp(1293) : warning C4305: '+=' : truncation from 'double' to 'float'
tst_qpainter.cpp(1474) : warning C4305: '+=' : truncation from 'double' to 'float'
tst_qtbench.cpp(155) : warning C4267: 'initializing' : conversion from 'size_t' to 'int', possible loss of data
main.cpp(68) : warning C4189: 'fontHeight' : local variable is initialized but not referenced
Change-Id: If6aadd50df7c5cf7d0f33791c9247730a47ddd27
Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com>
The check has to detect if boost header is present in the system we are
building for.
Change-Id: I700a11df208c8852ba094d8bff387ad21fa309b2
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Some quick benchmarks against GNU coreutils 8.21 and OpenSSL 1.0.1e
(time in µs; time for coreutils and OpenSSL include the loading of the
executable):
Qt Coreutils OpenSSL
n SHA-1 SHA-224 SHA-512 SHA-1 SHA-224 SHA-512 SHA-1 SHA-224 SHA-512
0 0 0 0 717 716 700 2532 2553 2522
64k 120 484 381 927 1074 966 2618 2782 2694
Diff 120 484 381 210 358 266 86 229 172
The numbers for Qt are pretty stable and vary very little; the numbers
for the other two vary quite a bit, since they involve launching and
executing separate processes. We can take the lesson that we're in the
same ballpark for SHA-1 and we should investigate whether our SHA2
implementation is sufficiently optimized.
Change-Id: Ib081d002ed57c4f43741eca45ff5cd13b97b6276
Reviewed-by: Richard J. Moore <rich@kde.org>
According to my profiling of Qt Creator, qHash and the SHA-1 calculation
are the hottest spots remaining in QtCore. The current qHash function is
not really vectorizable. We could come up with a different algorithm
that is more SIMD-friendly, but since we have the CRC32 instruction that
can read 32- and 64-bit entities, we're set.
This commit also updates the benchmark for QHash and benchmarks both
the hashing function itself and the QHash class. The updated
benchmarks for the CRC32 on my machine shows that the hashing function
is *always* improved, but the hashing isn't always. In particular, the
current algorithm is better for the "numbers" case, for which the data
sample differs in very few bits. The new code is 33% slower for that
particular case.
On average, the improvement (including the "numbers" case) is:
compared to qHash only QHash
Qt 5.0 function 2.54x 1.06x
Qt 4.x function 4.34x 1.34x
Java function 2.71x 1.11x
Test machine: Sandybridge Core i7-2620M @ 2.66 GHz with turbo disabled
for the benchmarks
Change-Id: Ia80b98c0e20d785816f7a7f6ddf40b4b302c7297
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Changed the processing of non-character code handling in the UTF8 codec.
Non-character codes are now accepted in QStrings, QUrls and QJson strings.
Unit tests were adapted accordingly.
For more info about non-character codes,
see: http://www.unicode.org/versions/corrigendum9.html
[ChangeLog][QtCore][QUtf8]
UTF-8 now accepts non-character unicode points; these are not replaced
by the replacement character anymore
[ChangeLog][QtCore][QUrl]
QUrl now fully accepts non-character unicode points; they are encoded as
percent characters; they can also be pretty decoded
[ChangeLog][QtCore][QJson]
The Writer and the Parser now fully accept non-character unicode points.
Change-Id: I77cf4f0e6210741eac8082912a0b6118eced4f77
Task-number: QTBUG-33229
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Remove benchmark tests that are no longer required as they are simple
overloads of other methods.
Change-Id: I610211543d17c077f482fa2145ac3da7d0767282
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Add support to QDateTime for time zones using the new QTimeZone class.
[ChangeLog][QtCore][QDateTime] Add support for a new Qt::TimeZone
spec to be used with QTimeZone to define times in a specific
time zone.
Change-Id: I21bfa52a8ba8989b55bb74e025d1f2b2b623b2a7
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
When calling intersect() on a large (1000000 items) QSet, with a small
(1000 items) QSet as the argument, the function takes signifcantly
longer than when the operand and the argument are reversed. This is
because the operand set is always iterated over in its entirety.
This patch changes intersect() to iterate over the smaller set. This
reduces the large operand scenario's benchmark to ~0.000063
milliseconds, compared to the current ~134 milliseconds:
1000000.intersect(1000) = empty: 0.000063 (was 134)
1000.intersect(1000000) = empty: 0.000039 (was 0.000036)
1000000.intersect(1000) = 500: 0.10 vs (was 130)
1000.intersect(1000000) = 500: 0.023 vs (was 0.093)
1000000.intersect(1000) = 1000: 0.20 vs (was 139)
1000.intersect(1000000) = 1000: 0.017 vs (was 0.016)
Task-number: QTBUG-22026
Change-Id: I54b25c49c78c458fef355e9c6222da8a64c7681f
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
It was only used for toUpper/toLower but always computed in the
constructor, including QString::toLatin1 conversion and allocations.
This needlessly slows down all other uses, including supposedly "cheap"
operations QString::toDouble, or accesses inside QResourceFileEngine.
The benchmarks indicates that doing it always when needed is bearable.
There's still a lot of improvement potential on these code paths.
Change-Id: I88b637ee11f9f7ea614f8da4ec5df0bf40664fce
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Remove all trailing whitespace from the following list of files:
*.cpp *.h *.conf *.qdoc *.pro *.pri *.mm *.rc *.pl *.qps *.xpm *.txt *README
excluding 3rdparty, test-data and auto generated code.
Note A): the only non 3rdparty c++-files that still
have trailing whitespace after this change are:
* src/corelib/codecs/cp949codetbl_p.h
* src/corelib/codecs/qjpunicode.cpp
* src/corelib/codecs/qbig5codec.cpp
* src/corelib/xml/qxmlstream_p.h
* src/tools/qdoc/qmlparser/qqmljsgrammar.cpp
* src/tools/uic/ui4.cpp
* tests/auto/other/qtokenautomaton/tokenizers/*
* tests/benchmarks/corelib/tools/qstring/data.cpp
* util/lexgen/tokenizer.cpp
Note B): in about 30 files some overlapping 'leading tab' and
'TAB character in non-leading whitespace' issues have been fixed
to make the sanity bot happy. Plus some general ws-fixes here
and there as asked for during review.
Change-Id: Ia713113c34d82442d6ce4d93d8b1cf545075d11d
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
In case somebody uses QVector as a stack, it is not fair to have
takeLast, removeLast and pop_back to do way too much work.
This is still very slow compared to std::vector::pop_back
(mostly due implicit sharing), however it is more than a
factor faster than before.
Change-Id: I636872675e80c8ca0c8ebc94b04f587a2dcd6d8d
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
This adds a fast insert on QMap when providing a correct hint.
Change-Id: I256bba342932c1d4f24c6e65074e1bf47b519537
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
The macro was made empty in ba3dc5f3b5
and is no longer necessary or used.
Discussed-on: http://lists.qt-project.org/pipermail/development/2013-January/009284.html
Change-Id: Id2bb2e2cabde059305d4af5f12593344ba30f001
Reviewed-by: Laszlo Papp <lpapp@kde.org>
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
Reviewed-by: hjk <hjk121@nokiamail.com>
These types are either built-in or 'automatically declared' and so
don't need to be explicitly declared as metatypes.
Change-Id: Iba4b7f8ff7a1c7974d144b955cbf064e43b36ec7
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
this is much more elegant than the so far propagated !isEmpty(QT.foo.name).
also replace feature-specific tests (no-gui and no-widgets) and the
obsolete contains(QT_CONFIG, foo) syntax.
Change-Id: Ia4b3c8febcabf9eeca67b1f9173a523820b1038b
Reviewed-by: Sergio Ahumada <sergio.ahumada@digia.com>
Reviewed-by: Tasuku Suzuki <stasuku@gmail.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@digia.com>
CONFIG += pcre is enabled if we're using the Qt PCRE, which isn't
compiled for 8-bit. If it isn't set, then we have a system PCRE.
Change-Id: I29d043b9d3f4d3223dcbb41eadc9f859e710eb88
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>
it compiles just fine without it.
if this was meant to inject a newer version of JSC than what is in
QtScript, it can be redone without creating a bizarre hybrid.
Change-Id: I61fe60bfa6a9bdb6423e8a7135250e332a5835ec
Reviewed-by: Joerg Bornemann <joerg.bornemann@digia.com>
Also change Trolltech for QtProject in other places
Task-number: QTBUG-23269
Change-Id: Ie4e344f23cab77c575562d18b481b3369ce30491
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
We can insert directly on the most left-most Node.
We always enforce an insert here (unlike the insert call),
but that is not a problem since the keys in a std::map are unique.
Change-Id: Ib409b90ffc57a5a43dab4a4b08d34f6fdabd057f
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
This suggestion keeps track of the most left node.
The point is that constBegin() becomes a lot faster.
That speeds up iteration a bit, and makes it O(1) to get the
first element. The penalty in insert and remove is very small.
On large trees it seems to be less than 1%.
It should be noticed that constBegin() is a very common hint
on my planned change to 5.1, and this opperation will without
this patch cost 2 x log N. One when the user calls the hint
with begin - and one where it is compared with begin.
Other std::maps has a very fast begin(). E.g
http://www.cplusplus.com/reference/stl/map/begin/
(begin with constant time)
Change-Id: I221f6755aa8bd16a5189771c5bc8ae56c8ee0fb4
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Giving the std-map a hint (normally) improves insert performance.
There seems to be no reason not to provide this hint.
Change-Id: I4344607ebf54574a3ae9666d87a41a3c14762361
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Change copyrights and license headers from Nokia to Digia
Change-Id: If1cc974286d29fd01ec6c19dd4719a67f4c3f00e
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Sergio Ahumada <sergio.ahumada@digia.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>
Change-Id: I19d3b2e9a5180b13deb828b55195404ef20be295
Reviewed-by: Daniel Teske <daniel.teske@nokia.com>
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.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>
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>
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>
- 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>
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>
