summaryrefslogtreecommitdiffstats
path: root/tests/benchmarks/corelib/tools/qhash/outofline.cpp
Commit message (Collapse)AuthorAgeFilesLines
* qHash: implement an AES hasher for QLatin1StringViewThiago Macieira2024-03-121-2/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | It's the same aeshash() as before, except we're passing a template parameter to indicate whether to read half and then zero-extend the data. That is, it will perform a conversion from Latin1 on the fly. When running in zero-extending mode, the length parameters are actually doubled (counting the number of UTF-16 code units) and we then divide again by 2 when advancing. The implementation should have the following performance characteristics: * QLatin1StringView now will be roughly half as fast as Qt 6.7 * QLatin1StringView now will be roughly as fast as QStringView For the aeshash128() in default builds of QtCore (will use SSE4.1), the long loop (32 characters or more) is: QStringView QLatin1StringView movdqu -0x20(%rax),%xmm4 | pmovzxbw -0x10(%rdx),%xmm2 movdqu -0x10(%rax),%xmm5 | pmovzxbw -0x8(%rdx),%xmm3 add $0x20,%rax | add $0x10,%rdx pxor %xmm4,%xmm0 | pxor %xmm2,%xmm0 pxor %xmm5,%xmm1 | pxor %xmm3,%xmm1 aesenc %xmm0,%xmm0 aesenc %xmm0,%xmm0 aesenc %xmm1,%xmm1 aesenc %xmm1,%xmm1 aesenc %xmm0,%xmm0 aesenc %xmm0,%xmm0 aesenc %xmm1,%xmm1 aesenc %xmm1,%xmm1 The number of instructions is identical, but there are actually 2 more uops per iteration. LLVM-MCA simulation shows this should execute in the same number of cycles on older CPUs that do not have support for VAES (see <https://analysis.godbolt.org/z/x95Mrfrf7>). For the VAES version in aeshash256() and the AVX10 version in aeshash256_256(): QStringView QLatin1StringView vpxor -0x40(%rax),%ymm1,%ym | vpmovzxbw -0x20(%rax),%ymm3 vpxor -0x20(%rax),%ymm0,%ym | vpmovzxbw -0x10(%rax),%ymm2 add $0x40,%rax | add $0x20,%rax | vpxor %ymm3,%ymm0,%ymm0 | vpxor %ymm2,%ymm1,%ymm1 vaesenc %ymm1,%ymm1,%ymm1 < vaesenc %ymm0,%ymm0,%ymm0 vaesenc %ymm0,%ymm0,%ymm0 vaesenc %ymm1,%ymm1,%ymm1 vaesenc %ymm1,%ymm1,%ymm1 vaesenc %ymm0,%ymm0,%ymm0 vaesenc %ymm0,%ymm0,%ymm0 > vaesenc %ymm1,%ymm1,%ymm1 In this case, the increase in number of instructions matches the increase in number of uops. The LLVM-MCA simulation says that the QLatin1StringView version is faster at 11 cycles/iteration vs 14 cyc/it (see <https://analysis.godbolt.org/z/1Gv1coz13>), but that can't be right. Measured performance of CPU cycles, on an Intel Core i9-7940X (Skylake, no VAES support), normalized on the QString performance (QByteArray is used as a stand-in for the performance in Qt 6.7): aeshash | siphash QByteArray QL1SV QString QByteArray QString dictionary 94.5% 79.7% 100.0% 150.5%* 159.8% paths-small 90.2% 93.2% 100.0% 202.8% 290.3% uuids 81.8% 100.7% 100.0% 215.2% 350.7% longstrings 42.5% 100.8% 100.0% 185.7% 353.2% numbers 95.5% 77.9% 100.0% 155.3%* 164.5% On an Intel Core i7-1165G7 (Tiger Lake, capable of VAES and AVX512VL): aeshash | siphash QByteArray QL1SV QString QByteArray QString dictionary 90.0% 91.1% 100.0% 103.3%* 157.1% paths-small 99.4% 104.8% 100.0% 237.5% 358.0% uuids 88.5% 117.6% 100.0% 274.5% 461.7% longstrings 57.4% 111.2% 100.0% 503.0% 974.3% numbers 90.6% 89.7% 100.0% 98.7%* 149.9% On an Intel 4th Generation Xeon Scalable Platinum (Sapphire Rapids, same Golden Cove core as Alder Lake): aeshash | siphash QByteArray QL1SV QString QByteArray QString dictionary 89.9% 102.1% 100.0% 158.1%* 172.7% paths-small 78.0% 89.4% 100.0% 159.4% 258.0% uuids 109.1% 107.9% 100.0% 279.0% 496.3% longstrings 52.1% 112.4% 100.0% 564.4% 1078.3% numbers 85.8% 98.9% 100.0% 152.6%* 190.4% * dictionary contains very short entries (6 characters) * paths-small contains strings of varying length, but very few over 32 * uuids-list contains fixed-length strings (38 characters) * longstrings is the same but 304 characters * numbers also a lot contains very short strings (1 to 6 chars) What this shows: * For short strings, the performance difference is negligible between all three * For longer strings, QLatin1StringView now costs between 7 and 17% more than QString on the tested machines instead of up to ~50% less, except on the older machine (where I think the main QString hashing is suffering from memory bandwidth limitations) * The AES hash implementation is anywhere from 1.6 to 11x faster than Siphash * Murmurhash (marked with asterisk) is much faster than Siphash, but it only managed to beat the AES hash in one test Change-Id: I664b9f014ffc48cbb49bfffd17b045c1811ac0ed Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org> Reviewed-by: Mårten Nordheim <marten.nordheim@qt.io>
* Change license for tests filesLucie Gérard2024-02-041-1/+1
| | | | | | | | | | | | According to QUIP-18 [1], all tests file should be LicenseRef-Qt-Commercial OR GPL-3.0-only [1]: https://contribute.qt-project.org/quips/18 Pick-to: 6.7 Task-number: QTBUG-121787 Change-Id: I9657df5d660820e56c96d511ea49d321c54682e8 Reviewed-by: Christian Ehrlicher <ch.ehrlicher@gmx.de>
* Replace ushort*/uint* with char16_t*/char32_t* in private API [1]Ahmad Samir2023-03-151-1/+1
| | | | | | | Task-number: QTBUG-110403 Pick-to: 6.5 Change-Id: Ie20a831f22212d56659cf3c6940d17134ab5f2c5 Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
* Port from container::count() and length() to size() - V5Marc Mutz2022-11-031-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This is a semantic patch using ClangTidyTransformator as in qtbase/df9d882d41b741fef7c5beeddb0abe9d904443d8, but extended to handle typedefs and accesses through pointers, too: const std::string o = "object"; auto hasTypeIgnoringPointer = [](auto type) { return anyOf(hasType(type), hasType(pointsTo(type))); }; auto derivedFromAnyOfClasses = [&](ArrayRef<StringRef> classes) { auto exprOfDeclaredType = [&](auto decl) { return expr(hasTypeIgnoringPointer(hasUnqualifiedDesugaredType(recordType(hasDeclaration(decl))))).bind(o); }; return exprOfDeclaredType(cxxRecordDecl(isSameOrDerivedFrom(hasAnyName(classes)))); }; auto renameMethod = [&] (ArrayRef<StringRef> classes, StringRef from, StringRef to) { return makeRule(cxxMemberCallExpr(on(derivedFromAnyOfClasses(classes)), callee(cxxMethodDecl(hasName(from), parameterCountIs(0)))), changeTo(cat(access(o, cat(to)), "()")), cat("use '", to, "' instead of '", from, "'")); }; renameMethod(<classes>, "count", "size"); renameMethod(<classes>, "length", "size"); except that the on() matcher has been replaced by one that doesn't ignoreParens(). a.k.a qt-port-to-std-compatible-api V5 with config Scope: 'Container'. Added two NOLINTNEXTLINEs in tst_qbitarray and tst_qcontiguouscache, to avoid porting calls that explicitly test count(). Change-Id: Icfb8808c2ff4a30187e9935a51cad26987451c22 Reviewed-by: Ivan Solovev <ivan.solovev@qt.io> Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org>
* Update tst_bench_qhash hash functions to use size_tMårten Nordheim2022-09-271-7/+7
| | | | | | | | | | | Pre-requisite for a fix for qHash. The Qt50String inherits from QString and becomes ambiguous once it no longer goes through a catch-all template function because qHash(QString, size_t) has a better match for the second argument. Pick-to: 6.2 6.4 Change-Id: I23c7afb1b3aa167d40dc4838e82b7763de015f6b Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
* Use SPDX license identifiersLucie Gérard2022-05-161-27/+2
| | | | | | | | | | | | | Replace the current license disclaimer in files by a SPDX-License-Identifier. Files that have to be modified by hand are modified. License files are organized under LICENSES directory. Task-number: QTBUG-67283 Change-Id: Id880c92784c40f3bbde861c0d93f58151c18b9f1 Reviewed-by: Qt CI Bot <qt_ci_bot@qt-project.org> Reviewed-by: Lars Knoll <lars.knoll@qt.io> Reviewed-by: Jörg Bornemann <joerg.bornemann@qt.io>
* Standardise layout and naming in corelib benchmarksEdward Welbourne2021-08-301-1/+1
| | | | | | | | | | | | | | | Make file names match CMake's test names (and those follow dir-name) and class names follow tst_ClassName pattern when testing ClassName. Purge comments about the qmake configs the CMakeLists.txt are generated from. Purge empty constructors and init/cleanup methods of classes. Fix petty coding style violations. Add qdir/tree/, qurl, qbench and qset benchmarks to their parent directories' lists of subdirs. Fix unused return error from qurl benchmark. Change-Id: Ifc15a3a46e71cf82ad0637753517e0df34049763 Reviewed-by: Andrei Golubev <andrei.golubev@qt.io> Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
* Updated license headersJani Heikkinen2016-01-211-17/+12
| | | | | | | | | | | | | From Qt 5.7 -> tools & applications are lisenced under GPL v3 with some exceptions, see http://blog.qt.io/blog/2016/01/13/new-agreement-with-the-kde-free-qt-foundation/ Updated license headers to use new GPL-EXCEPT header instead of LGPL21 one (in those files which will be under GPL 3 with exceptions) Change-Id: I42a473ddc97101492a60b9287d90979d9eb35ae1 Reviewed-by: Thiago Macieira <thiago.macieira@intel.com> Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
* Update copyright headersJani Heikkinen2015-02-111-7/+7
| | | | | | | | | | | | | | | | | | Qt copyrights are now in The Qt Company, so we could update the source code headers accordingly. In the same go we should also fix the links to point to qt.io. Outdated header.LGPL removed (use header.LGPL21 instead) Old header.LGPL3 renamed to header.LGPL3-COMM to match actual licensing combination. New header.LGPL-COMM taken in the use file which were using old header.LGPL3 (src/plugins/platforms/android/extract.cpp) Added new header.LGPL3 containing Commercial + LGPLv3 + GPLv2 license combination Change-Id: I6f49b819a8a20cc4f88b794a8f6726d975e8ffbe Reviewed-by: Matti Paaso <matti.paaso@theqtcompany.com>
* Update license headers and add new license filesMatti Paaso2014-09-241-19/+11
| | | | | | | | | - Renamed LICENSE.LGPL to LICENSE.LGPLv21 - Added LICENSE.LGPLv3 - Removed LICENSE.GPL Change-Id: Iec3406e3eb3f133be549092015cefe33d259a3f2 Reviewed-by: Iikka Eklund <iikka.eklund@digia.com>
* Update the qHash function for strings to use the CRC32 instructionThiago Macieira2014-01-231-0/+10
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 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>
* Update copyright year in Digia's license headersSergio Ahumada2013-01-181-1/+1
| | | | | Change-Id: Ic804938fc352291d011800d21e549c10acac66fb Reviewed-by: Lars Knoll <lars.knoll@digia.com>
* Change copyrights from Nokia to DigiaIikka Eklund2012-09-221-24/+24
| | | | | | | | 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>
* Remove qhash_faster.Robin Burchell2012-02-291-44/+0
| | | | | | | | 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>
* Copy Qt 4's QString hash algorithm.Robin Burchell2012-02-291-2/+16
| | | | | | | | | | 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>
* Remove "All rights reserved" line from license headers.Jason McDonald2012-01-301-1/+1
| | | | | | | | | | As in the past, to avoid rewriting various autotests that contain line-number information, an extra blank line has been inserted at the end of the license text to ensure that this commit does not change the total number of lines in the license header. Change-Id: I311e001373776812699d6efc045b5f742890c689 Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
* QHash benchmark: improve Java's hashGiuseppe D'Angelo2012-01-271-0/+11
| | | | | | | | Added a bit of documentation to the Java-like hashing function. Change-Id: I3f44eee305d91b76f0f89cd1acf21f6430b9482b Reviewed-by: João Abecasis <joao.abecasis@nokia.com> Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
* Move test files around.Robin Burchell2012-01-241-1/+1
| | | | | | | | | main.cpp is easier for my muscle memory to remember than qhash_string as it is used by a number of other tests. Change-Id: I044f995d55a4ff1328dde0ae27b6e36a80114c38 Reviewed-by: Richard J. Moore <rich@kde.org> Reviewed-by: João Abecasis <joao.abecasis@nokia.com>
* Update contact information in license headers.Jason McDonald2012-01-231-1/+1
| | | | | | | Replace Nokia contact email address with Qt Project website. Change-Id: I431bbbf76d7c27d8b502f87947675c116994c415 Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
* Add a string hash implementation similar to the one in Java.Robin Burchell2012-01-201-0/+13
| | | | | | | | | | | | | | | | | | | | | This uses a similar runtime to the approach of sampling part of the string, with the benefit that it doesn't reduce the sampling to subsections of the string. Ironically, Java used to only sample parts of the string as well, but found that it produced too many collisions with certain string types, so they moved to use this method. RESULT : tst_QHash::qhash_qt4(): 0.0537 msecs per iteration (total: 110, iterations: 2048) PASS : tst_QHash::qhash_qt4() RESULT : tst_QHash::qhash_faster(): 0.015 msecs per iteration (total: 62, iterations: 4096) PASS : tst_QHash::qhash_faster() RESULT : tst_QHash::javaString(): 0.016 msecs per iteration (total: 66, iterations: 4096) Change-Id: Icb5da341ab6445163f4217650a0bdb3903e50210 Reviewed-by: hjk <qthjk@ovi.com>
* Update copyright year in license headers.Jason McDonald2012-01-051-1/+1
| | | | | Change-Id: I02f2c620296fcd91d4967d58767ea33fc4e1e7dc Reviewed-by: Rohan McGovern <rohan.mcgovern@nokia.com>
* Update licenseheader text in source files for qtbase Qt moduleJyri Tahtela2011-05-241-17/+17
| | | | | | | Updated version of LGPL and FDL licenseheaders. Apply release phase licenseheaders for all source files. Reviewed-by: Trust Me
* Initial import from the monolithic Qt.Qt by Nokia2011-04-271-0/+90
This is the beginning of revision history for this module. If you want to look at revision history older than this, please refer to the Qt Git wiki for how to use Git history grafting. At the time of writing, this wiki is located here: http://qt.gitorious.org/qt/pages/GitIntroductionWithQt If you have already performed the grafting and you don't see any history beyond this commit, try running "git log" with the "--follow" argument. Branched from the monolithic repo, Qt master branch, at commit 896db169ea224deb96c59ce8af800d019de63f12
generated by cgit v1.2.3 (git 2.25.1) at 2024-04-26 17:55:34 +0000