20097 2008-07-18 11:19:05 -0700 [Qt] 20% Sunspider slow-down 2009-04-20 13:30:54 -0700 1 1 1 Unclassified WebKit JavaScriptCore 528+ (Nightly build) PC Linux RESOLVED FIXED Qt P2 Normal --- 1 zherczeg webkit-unassigned ap ariya.hidayat ddkilzer ggaren hausmann laszlo.gombos loki mjs mrowe vestbo zecke zwarich oldest_to_newest 86345 0 zherczeg 2008-07-18 11:19:05 -0700 I noticed a major slow-down for sunspider between r35202-r35208 Repository : Root: http://svn.webkit.org/repository/webkit UUID: 268f45cc-cd09-0410-ab3c-d52691b4dbfc System: Linux, Qt-4.3.4, gcc-4.1.2 (20061115) CPU: AMD opteron, Dual-core 2.2 GHz (dedicated for benchmark measurement) Revision: 35033: 3168.0ms +/- 0.8% 35123: 3172.0ms +/- 0.7% 35170: 3197.8ms +/- 0.9% 35190: 3191.5ms +/- 0.6% 35202: 3193.7ms +/- 0.7% <<< HERE ~20% slow >>> 35208: 3782.8ms +/- 0.6% 35231: 3779.2ms +/- 0.6% 86346 1 22366 zherczeg 2008-07-18 11:25:25 -0700 Created attachment 22366 Detailed sunspider results (.tgz) 86347 2 ap 2008-07-18 11:34:56 -0700 The only JavaScriptCore change in this range is <http://trac.webkit.org/projects/webkit/changeset/35203>. 86366 3 ddkilzer 2008-07-18 13:13:03 -0700 Could you attach all of the additional tests you're running with SunSpider? 86374 4 zwarich 2008-07-18 14:19:46 -0700 I don't see a negative impact on performance from r35203 on my machine, a 2.16 GHz MacBook Pro with Apple GCC 4.0. 86377 5 mrowe 2008-07-18 14:31:22 -0700 I see r35203 as being around 2% faster than r35202 on my Mac with GCC 4.0. If I force JavaScriptCore to build with GCC 4.2, I see r35203 as being around 2% slower than r35202. This regression may be specific to either your compiler version or perhaps even to the manner in which the Qt port builds JavaScriptCore. 86379 6 ggaren 2008-07-18 14:36:17 -0700 Can you verify that the ALWAYS_INLINE macro works in your port? 86388 7 zherczeg 2008-07-18 15:06:45 -0700 (In reply to comment #6) > Can you verify that the ALWAYS_INLINE macro works in your port? Yes, both ALWAYS_INLINE and NEVER_INLINE works (I checked an assembly dump) Maybe it was not clear, but our extended benchmark set is separated from Sunspider. Those tests are not finalized and may changed in the near future. However, if you are still interested, I can attach them. Is it possible, that the AMD cpu, which has a slower fpu than the intel counterpart, is responsible for the slow? Maybe I should try out my binaries on other computers. 86391 8 zherczeg 2008-07-18 15:31:41 -0700 I simply copied my binaries to a P4 2.40 GHz (jsc-35202, jsc-35208 and libQtCore.so.4 because it was missing from that system), and it seems the results are the same as on the other archtecture. By the way, the size is also increased: jsc-35202 - 1820153 jsc-35208 - 1878409 86394 9 ddkilzer 2008-07-18 18:02:35 -0700 (In reply to comment #7) > Maybe it was not clear, but our extended benchmark set is separated from > Sunspider. Those tests are not finalized and may changed in the near future. > However, if you are still interested, I can attach them. Please attach them. The only way to reproduce the results on other platforms is to use the same tests. You should also try the original SunSpider tests to see how they perform on your platform, e.g., if you see a slowdown or a speedup. 86409 10 zherczeg 2008-07-19 00:04:52 -0700 I am sorry, it seems it was a bad thing to mention the other benchmark set. Naturally, the results I sent you are measured by the original sunsipder tests without any kind of tweaks. In the meantime, I have compiled the 35203 revision as well. Now I can enclose the comparistation between r35202 and r35203. The worst case: access-fannkuch: *1.67x as slow* The reason why I mentioned bubble sort, because it was more than 70% slower, and that is bigger than access-fannkuch. The file sizes as also interesting: jsc-35123: 1814130 bytes jsc-35170: 1820097 bytes jsc-35190: 1820143 bytes jsc-35202: 1820153 bytes << BIG JUMP >> jsc-35203: 1878409 bytes jsc-35208: 1878409 bytes (you mentionad there was no changes here) jsc-35231: 1862421 bytes (little better) 86410 11 22382 zherczeg 2008-07-19 00:06:02 -0700 Created attachment 22382 compare r35202 and r35203 86411 12 mrowe 2008-07-19 00:18:38 -0700 It may be worth testing a different compiler version or a different port to see what result you get. The GTK port should compile and run on Linux, and since it uses a different build system it may give different performance figures. it would be interesting to see whether it also shows the same regression. 86413 13 zherczeg 2008-07-19 05:55:28 -0700 I downloaded WebKit to my home computer: Core 2 Duo, Ubuntu-Linux, GTK port, gcc-4.2.1 sunspider r35202 - 2614.9ms +/- 0.2% sunspider r35203 - 2601.3ms +/- 0.2% It seems a little bit faster! To my shock, the jsc binaries are only 63% of the QT port: r35202: GTK: 1147693 bytes, QT: 1820153 bytes r35203: GTK: 1168479 bytes, QT: 1878409 bytes Do you have any idea? Is the QT port causes the difference? 86437 14 mrowe 2008-07-19 22:32:41 -0700 I would look into the compiler flags that are used when building JavaScriptCore. Is the Qt build system building JavaScriptCore with "-fstrict-aliasing"? Are they both using "-O3"? 86483 15 22392 ddkilzer 2008-07-20 11:01:17 -0700 Created attachment 22392 bubbleSort.js Additional bubbleSort.js test metioned in Zoltan's email to webkit-dev. 86504 16 zherczeg 2008-07-20 14:24:26 -0700 I have successfully compiled the QT port on my home computer: QT-port-35202: sunspider: 2665.1ms +/- 0.3% binary size: 1841368 QT-port-35203: sunspider: 3238.2ms +/- 0.2% binary size: 1904808 As you can see, this is exactly the same thing what we saw before! So this thing is QT port specific. These are the CFLAGS for the QT-port: g++ [...] -pipe -Wreturn-type -fno-strict-aliasing -O2 -fvisibility=hidden -fvisibility-inlines-hidden [...] I will try to change these to -O3 and -f-strict-aliasing, and will send you the results. 86537 17 zherczeg 2008-07-21 01:59:09 -0700 It seems the QT port applies the same CXXFLAGS to all projects. We only need -O3 and -fstrict-aliasing flags for JavaScriptCore. Is there a simple way to do that, or do I need to write a new "GENERATOR" rule in JavaScriptCore.pri? Any other idea (perhaps change the global CXXFLAGS)? 86539 18 ddkilzer 2008-07-21 02:26:23 -0700 Copying some Qt folks regarding Comment #17. 86548 19 hausmann 2008-07-21 03:54:56 -0700 (In reply to comment #17) > It seems the QT port applies the same CXXFLAGS to all projects. We only need > -O3 and -fstrict-aliasing flags for JavaScriptCore. Is there a simple way to do > that, or do I need to write a new "GENERATOR" rule in JavaScriptCore.pri? Any > other idea (perhaps change the global CXXFLAGS)? With the way JavaScriptCore is built into the same library the CXXFLAGS are the same for the entire QtWebKit library. But for testing purposes you could try something along the lines of QMAKE_CXXFLAGS_RELEASE ~= s/-O2/-O3 -fstrict-aliasing/ in JavaScriptCore.pri or WebCore.pro. 86550 20 mrowe 2008-07-21 04:26:30 -0700 It is almost certainly unsafe for WebCore to be built with strict aliasing enabled. 86551 21 zherczeg 2008-07-21 05:06:48 -0700 > QMAKE_CXXFLAGS_RELEASE ~= s/-O2/-O3 -fstrict-aliasing/ I did something similar, just to see what happend with jsc. The resulted binary was a bit bigger, and a bit faster than the version with -O2, but the difference is small. I looked at the resulting binaries, and there are strange strings in QT-r35203 (-O2 and -O3 as well). They seems kinda debug info : [...]void WTF::HashTable<Key, Value, Extractor, HashFunctions, Traits, KeyTraits>::checkKey(const T&) [with T = KJS::UString::Rep*, HashTranslator = WTF::RefPtrHashMapRawKeyTranslator<KJS::UString::Rep*, std::pair<WTF::RefPtr<KJS::UString::Rep>, KJS::SymbolTableEntry>, WTF::PairHash[...] These lines are neither in QT-r35202 nor in GTK binaries. Is it possible that some recently triggered macros made these lines? 86554 22 zherczeg 2008-07-21 07:20:35 -0700 We belive the jsc always compiles in debug mode for QT (and the asserts in Registers.h slows down the sunspider). It seems WebKit/JavaScriptCore/kjs/jsc.pro does not have 'release' mode. 86557 23 hausmann 2008-07-21 07:28:14 -0700 (In reply to comment #22) > We belive the jsc always compiles in debug mode for QT (and the asserts in > Registers.h slows down the sunspider). It seems > WebKit/JavaScriptCore/kjs/jsc.pro does not have 'release' mode. If you build with build-webkit --release for example then jsc is built in release mode. If you run just qmake in JavaScriptCore/kjs then qmake will pick whatever Qt was built with by default, which would normally be a release build, unless you configure your version of Qt with --debug. 86569 24 22404 loki 2008-07-21 08:22:56 -0700 Created attachment 22404 Add missing NDEBUG The problem was that the jsc.pro was not contain any release specific rule. So there was no chance to pass the NDEBUG define to the compiler. The patch contains the similar rule which is in WebKit.pri file. Note: I have to include USE_SYSTEM_MALLOC define in case of release, because there was linking problem (WebKit.pri uses USE_SYSTEM_MALLOC in case of debug also). 86571 25 mrowe 2008-07-21 08:27:17 -0700 Does the patch address the performance issue? 86595 26 loki 2008-07-21 12:01:28 -0700 (In reply to comment #25) > Does the patch address the performance issue? Of course. Here is the SunSpider result: ** TOTAL **: 1.37x as fast 2897.3ms +/- 1.9% 2107.5ms +/- 2.6% significant 86624 27 zherczeg 2008-07-21 23:43:05 -0700 Here is the summarization of the above bug in one sentence: In the QT jsc build, the jsc.pro qmake file never defines the NDEBUG symbol, so the releases contains all asserts and other debug related codes, which heavily slows down all measurements. 86625 28 22404 hausmann 2008-07-22 00:27:45 -0700 Comment on attachment 22404 Add missing NDEBUG Patch looks good to me 86730 29 ddkilzer 2008-07-23 08:00:33 -0700 $ git svn dcommit Committing to http://svn.webkit.org/repository/webkit/trunk ... M JavaScriptCore/ChangeLog M JavaScriptCore/kjs/jsc.pro Committed r35294 22366 2008-07-18 11:25:25 -0700 2008-07-18 13:12:40 -0700 Detailed sunspider results (.tgz) results.tgz application/octet-stream 3897 zherczeg H4sIAHjbgEgAA+2d3XLbOBKFc52ncM31aIsg+Ltb8yRTeyHbSqIaR3JJcibZrX33lfwjWTTOaaIB kIpD3qTGdBLb8+Xg9OkGsFtsd7P1Zvl5ZsvM2n/svu8+xH6y/VMVxeFXW9vs9a/7x1SZsR+MyWxR ZHluqg/7/85t+eEqi/6VOJ6H7W6+ubr68OU/67sl+Tzp/U/6fNt/8+uH3f3D7uqPqz8//vfj1f75 zd7Obh6uF7/988rY+vfjx76uN/dfDh8sq9MHN/Mfu8385umT7fPH5zc3i+12dr1czTc/ZrvNYrHd v6+K89ef5qvVXw83j39k1vmtq+v17Y/9iz0QnRfb5eLb4W+zzfOL6+Vufb+d2f2v+79xt50tV7Pr H7vD5xTV+ed0Xpf5m9d/L7eL2Xx1u3/b1udvn/7mxz/j8Pblm7lZr3ab9d2nu/Xfs83i5mGzXT59 feblEzY/7nfr2fzxZ3D80T1/9Ott+fon8/zR7Ze5OXz45Y+4ne8Ws0/rzdf5brZbf3r84vMsc7z9 fj/fXB9+oPnL791/9MvsZr25Xd4cPm6a1x/ff/ZuOb+bbR++Hr4609jXb7f3i5v9/9y72Wr/h+9f 1y9f/GbxefH9fna7mh++0fzlC9nuNsvV59n1fLuoise/zJy/+TTf/3N7/OrK8xe7+eebu/XD4cdu 2s67h9X9/Oav/bdw+/ht5/n562/zu+XjT2C52mP89C1+/N/vfVmucsByzlmuNSzXCVk2lOWGstxK LOfhLOeU5TIByxVluXSzXECWM8RyAVnOu/8AOiwXmcSyzQDLjYNlqMsCyzlhuUAslxfKci2wXPjo stWwzHU5U7GcKVg2LWI5qxHLFrJs2oqy3JX0/rpcWAfLUJcNZ1mly0Es1yEsB3oMH11Wscx1uQAs 15TlWsEy1uWs9Wc57/6mDsvH/++Q5bzy0eUWsCzosiUsQ12+VL9cDcCyS3lPLLtIlz0GZ1nlMaBf TqLLRmTZAF02DpYrA1gO0GXIsknIclK/7PYYtQ/LXJeLBCznGpYVupxDliPoculR+2lzDKbL7y/H iKDLPMfgupyrWDYalrEuN4hlXPuJOYaoy6YFupw7WD6ufV2WC85yQ1iGutyF/IXlouzBcslZzijL XJfLAVh2KW9fXbZuls9rwjcslxqW4+YY6XS5dLBcFoBlwWOocowcsGzrcJZDaj87AMtMl83R5Xix nEKXDWI5qxDLLF9uKMtyvgz9sleOMWi+3IazTD3G6f+EE2ZzzMTGS5grSjNyzC2juW41NGNltohm lmRwmvslGR///a+PY7cip2eEZ/eq/7/HbKT+vzn0/4ts/6+wKOqn/n8x9f+HeOT+f+VY06A/y/ia FtefJc/m6ZIm+zOfDCh3r2gp6uaKrWitKgNS1M3Mnwl5pryi5R7ZPPRnQgakqpuD+kxBtYalLCsz IC+WuTtrKMstYJn3TFW1hqJuriDLefc3dd1ZIbIMMiB3nlkCloW62WhYbkJYDsrmA1m2TpbdlYZK l11vA3umrabPpOn/B8xlySxbwHLrYLmygGVBlysNy+Nl8zzPVM4YevWZOMs1ZRnp8pB9JjxjWEOW I+gymjF05ZnaWRYVy6hnWmThLI/gMbx0OSQBqjQst0VUlrEuE5albF6cy4IeI+Ysi8pjBOlyQr8c tc+kYlnlMc4z+zcsc79cu1lWzLIEeAwrsozmZZ2z36hnKrAc12P0yeaDPMaAuqzyGLxninSZ1n4q v0x0WZFjiB5D1GUL+kxuXVbWfhc3+837/wPmGCqW+ew36pnSTE41+016plCXcY4hZnK1xDLUZWcm p9Rl1YwhZDmCLnOPUaZgOaLH4LWfTpc5y2B/WYlYVmVykl9uRZbrAfbkML+cZh9DEMu8VzJEvszn snivpAQs8xyj0bCMdTlFr0TOMczU/f+Fn7P+f52N1P8vnvv/xV5ay8f+f5FP/f8hHrH/75w11vb/ M7KmIX/WzfuHm888Da1FDecjHgBg6aKGho1Ltqi1JVvUjqVI30UNGzQDF7VTuAUMmrhp2u8AALSh SYC5VMAMB+fTG7QkxYaXQeOb8/gAADJovHAe6gAAFgIVlOUeBwCgYRbnAAAS5pqyfMz0fYqNHDWa kjdNTWYozGI67yPMKph5taGbAFDtAsHpvGp3Hk80c7ly9prMQhMAAek8gtmOVzkXYSz7TLMAk8Er Z9UJADkVZp7OI5OBUyA4AUBMhskpy/Lc/CDpPHPMlyjMgW3TCMKcwDGnEGZFDMSOAOCRZo943gdm rTAzxwxh7lrp4UazBhwBUAkzj+drN8tUmE9iH0mYoWMmG5okYRZZtj5jhjCeDxBmWP21CVkOOc5C yXK04yy0W00TtJrwUYaKJEPUZXkrA9Jl58isciuDZpwFG+Y+sVyQLgca5gi67HIRgeMsKQyzYqNp 0qMMEcvOJKMGLCfQ5e4RGVFZ5rrchrE8csIMtpgJCTNn2bhZViTMLMjgLAcEGc7tkqj4C2iXwBGA IL+c8ChDJctuj6FimY/MIl0uKMtxt0t2rW8flsXjLOTxbzuNAPyqz1n/vx2r//+y/7+oc2Of9v/X U/9/iEfe/+8MgBqwoAlnjWmS+RH3mfKZNumcaa9Co3QvaCEBEJrPpCfaxJ7PhMk8C4AsXdDkvXl+ hYZuPlPVMh0xAOLmLCrLlZtlXjTzQgOxzItmynLEQiMgmO9lzpKfZaHS5ZQs8wCI67LyLgt3oTGg LvMwk+/NA7oc96QxIQCKrcsoABKaTKqi+Vf2GFyXefdfNTevY1kRzLOzeWvKcnf7iEOXC48ACDZM hTCT3ZcFm0zjhZncYzQSyz5hpoplvv9fd5YFPQHSP5iHARAL5kP98hBnDEUeMUzOMtdlkWUfj6Fq MnG/jCYM+XlZdD8T8suKCUOmy4JfllkeYsJQdY8hPJt3ZJaVHiOiX+bN/0bDsu4siwqxnOLutx77 /9GeaecgC8rkAvwy8hh5hliOcC6LpSwPeJcFYNnliPvWfrrJb95kGsYvC7osN0x9aj/tHUMqv5xy S07IvSxRdVnVMOUsD7hnOu7db9KAoV6XnZPfurss3le+/JMOZQWxHG/wO0CXew1+T83/X/Z53f/P s3zk/n9ZZlX21P/Ppv7/EI/y/H/UM23omqYbmk+ZzV/MfaYJNjOhWoP3/1WDxoo1jQ3NB2fzIANy 32eKWOZbpmnPtDsbfwG1xvgbQDjLutuZeAak8meK7f9sl2mqMyALpy6jnmnA9v80m5mEQeOQ/v/4 LPNaA9XNXJdVd6YrsvkAXZZnWZAuO7N5pS5f3LEsIbqs9BgR75nmuqzLM+NmQLrd/4Iuyx7D54gh 2DMV/LLqzPR3tsk0ol/WncrCdZlv/vc+A7JCLGeQZfEMSL0uFw6WYTZfcpY1m/9xz/TS80yfWZYE fll3kAW/zbR1s6yY/SYb86SeqajL0C+7ar8UfnmUGcOktd/I9xjqWFYd/YZ1Gfnl7hVbcc/m9Tn6 Dc5+Cx5DdZfFpfaZRL/swzLo/4fcY4hqv5BZFsAy1GV8/j87klM4M12e/QY9U2ftB/NlofaLe8dQ 8rmsJLWfV8+Us6w6+S2o9vPWZeiXA3KMXGLZ744hNGMYoMtwXjZoLmtMlpPfYxjfL59Ow4zjlzHL zC8H58vZ1P//hZ/z/n8zTv+/sM/9f2PL4mn///6Xqf8/wCP2/2tXBtSCNa3ka1pN5jNrtKZ1hwaO a1rZw5/Vbxet16+bgq1pJ4vjXNROVyOgVa2oXavacWE4W9W6H31e1WpXPXFc1XJaOVvg0CpDVjWh 03Q6eruzrFm0rBlUbhRsRJNfA1kZcVkD2/OcODcgBhJwrkgM1EX9hDPantcL57d3453h/HayPirO lRPnxolzo8DZ0ILD5m6cSxbQn059cuLcuk0avji9m8OfaGbFM9+gV6gvHKtLB80tGDjuCmlXnEnB gcUZbdCLQXNicXbT7CXOro1LJ5ppFKQSZ6HkgOKMe6eo5igqjHNX0Ts4lyLOtgI4WwfOUJyFLIiJ M8Q5RzhXF49zaq9xMeKM0yAszsRqZDmnWTzWNC89rEYLJgEk50y6p5hmdETvAOJMbxu6AJqPqLhp BnlQlVOaqTgjmmEe1LXUJ5pZHsSvG+qhzeC8odp1D5xWmxsNzT+xNjeJaTY03YRWI2M0GzoNgGiG E4dd1e5Fs6DNPYwzmNJqXNMAKNWQjDPTZlQGdvV+yDIwVJs9aD5uGIiozcBpCNpM57T8Uw1sNdiB AHwgIMA4j51qdCdifv6QLmKqQbdRW+PGWTDOQh0I9ungMwHQfEtArFHKc4dAnZ1eowXzLV30uups FThjr5FenemtQ1qcuzocEmvQ01ps5saZe42MTmspMmeIM7upMzjWQJmzK9Y4Hk7qZ53rjOCMrHOY OodZ5yQ4A+usUmdunVEhyHAWhlygOuOTZ9EuSuqdhUpQ3uHQTpMB7/456/9bM1L/P3vu/2dlURXT +f8DPsr+P1i8pEqZLV7drR5HL4Yq5cr0WLyE0iKsUpYm590tpsa4Fi9Vpcy9GJrQpKWFsAvkuLT1 zn1goVyTtYu3/+WrxlBlUbkmNLWVBZtm6Q4Viw3TY8kxWqEszWgmp/n4BXpdaBpEM9ifh2MfmMkH 9EvFeWOLaHalmMdTqTxpJrPz3YTzVFigDtMA2kwPARhGm2mdzFNMdHWe0P63jGZYJytCeVInS+Is z2aB8Xl3ndwCnIU6mYXyCOeuPznVyT3G59+51Tgqn9fh6SEppr84Fwqau/84uuIsb2xCs1leDVO+ eZqKM7IaZfdFTKsRNmko0uwRYurEmd8FiRqmvMVEh1nMsZcZPs3CxJm3mKzcMQXTLE6vAefABa+h EecSZfLvyzknmANHu04FcaZeA4kznDRUWY1uju/vnBHNzsFZFMkLNLNUA4ozOnEohnN+u5FvAKvh 1WEKSTWQc+YdJj5p6H2GuqbBFJ5qAKvh1maUavC74Kg2Q5phqtGH5qSzWTGtBqC5cVV6fVMNnTbT GwGOO3h6T4FraJaMs5VoRsa5bF00ozIwYJjlAp2GFdqlMcvABN1/aJwDtFmRamhms8KNM5rNcg6z KDdQ1gRnJM7d4jFqqvH2HE8fqzFE5OyavgoVZ3ZcS2xxTmKc5Yyunpr/0zM90zM90zM90zM90zM9 0zM90zM90zM97+H5P3OYxoQA8AAA 22382 2008-07-19 00:06:02 -0700 2008-07-19 00:06:02 -0700 compare r35202 and r35203 compare.txt text/plain 3730 zherczeg DQpURVNUICAgICAgICAgICAgICAgICAgIENPTVBBUklTT04gICAgICAgICAgICBGUk9NICAgICAg ICAgICAgICAgICBUTyAgICAgICAgICAgICBERVRBSUxTDQoNCj09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 DQoNCioqIFRPVEFMICoqOiAgICAgICAgICAgKjEuMTg0eCBhcyBzbG93KiAgMzE5My43bXMgKy8t IDAuNyUgICAzNzgyLjJtcyArLy0gMC40JSAgICAgc2lnbmlmaWNhbnQNCg0KPT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT0NCg0KICAzZDogICAgICAgICAgICAgICAgICAqMS4xODl4IGFzIHNsb3cqICAgNDM4 LjZtcyArLy0gMS4wJSAgICA1MjEuNG1zICsvLSAwLjMlICAgICBzaWduaWZpY2FudA0KICAgIGN1 YmU6ICAgICAgICAgICAgICAqMS4yMnggYXMgc2xvdyogICAgMTQxLjVtcyArLy0gMi40JSAgICAx NzMuMm1zICsvLSAwLjUlICAgICBzaWduaWZpY2FudA0KICAgIG1vcnBoOiAgICAgICAgICAgICAq MS4xODh4IGFzIHNsb3cqICAgMTU5LjltcyArLy0gMS45JSAgICAxOTAuMG1zICsvLSAxLjAlICAg ICBzaWduaWZpY2FudA0KICAgIHJheXRyYWNlOiAgICAgICAgICAqMS4xNTN4IGFzIHNsb3cqICAg MTM3LjJtcyArLy0gMC41JSAgICAxNTguMm1zICsvLSAwLjQlICAgICBzaWduaWZpY2FudA0KDQog IGFjY2VzczogICAgICAgICAgICAgICoxLjMzeCBhcyBzbG93KiAgICA0MjEuN21zICsvLSAwLjUl ICAgIDU2Mi43bXMgKy8tIDAuMyUgICAgIHNpZ25pZmljYW50DQogICAgYmluYXJ5LXRyZWVzOiAg ICAgICoxLjE0MnggYXMgc2xvdyogICAgNjIuOG1zICsvLSAzLjUlICAgICA3MS43bXMgKy8tIDIu NyUgICAgIHNpZ25pZmljYW50DQogICAgZmFubmt1Y2g6ICAgICAgICAgICoxLjY3eCBhcyBzbG93 KiAgICAxMDMuOG1zICsvLSAwLjQlICAgIDE3My43bXMgKy8tIDAuMyUgICAgIHNpZ25pZmljYW50 DQogICAgbmJvZHk6ICAgICAgICAgICAgICoxLjE5OHggYXMgc2xvdyogICAyMTcuMm1zICsvLSAw LjglICAgIDI2MC4zbXMgKy8tIDAuNiUgICAgIHNpZ25pZmljYW50DQogICAgbnNpZXZlOiAgICAg ICAgICAgICoxLjUweCBhcyBzbG93KiAgICAgMzcuOW1zICsvLSAwLjYlICAgICA1Ny4wbXMgKy8t IDAuMCUgICAgIHNpZ25pZmljYW50DQoNCiAgYml0b3BzOiAgICAgICAgICAgICAgKjEuNDV4IGFz IHNsb3cqICAgIDI5MC42bXMgKy8tIDAuMiUgICAgNDIwLjdtcyArLy0gMC4yJSAgICAgc2lnbmlm aWNhbnQNCiAgICAzYml0LWJpdHMtaW4tYnl0ZTogKjEuNTJ4IGFzIHNsb3cqICAgICA0Ni4xbXMg Ky8tIDAuNSUgICAgIDcwLjJtcyArLy0gMC40JSAgICAgc2lnbmlmaWNhbnQNCiAgICBiaXRzLWlu LWJ5dGU6ICAgICAgKjEuNjR4IGFzIHNsb3cqICAgICA1MS41bXMgKy8tIDAuNyUgICAgIDg0LjVt cyArLy0gMC40JSAgICAgc2lnbmlmaWNhbnQNCiAgICBiaXR3aXNlLWFuZDogICAgICAgKjEuMjd4 IGFzIHNsb3cqICAgICA5Ny41bXMgKy8tIDAuNSUgICAgMTI0LjFtcyArLy0gMC40JSAgICAgc2ln bmlmaWNhbnQNCiAgICBuc2lldmUtYml0czogICAgICAgKjEuNDl4IGFzIHNsb3cqICAgICA5NS41 bXMgKy8tIDAuNCUgICAgMTQxLjltcyArLy0gMC4zJSAgICAgc2lnbmlmaWNhbnQNCg0KICBjb250 cm9sZmxvdzogICAgICAgICAqMS40OHggYXMgc2xvdyogICAgIDMxLjZtcyArLy0gMS4yJSAgICAg NDYuOG1zICsvLSAxLjYlICAgICBzaWduaWZpY2FudA0KICAgIHJlY3Vyc2l2ZTogICAgICAgICAq MS40OHggYXMgc2xvdyogICAgIDMxLjZtcyArLy0gMS4yJSAgICAgNDYuOG1zICsvLSAxLjYlICAg ICBzaWduaWZpY2FudA0KDQogIGNyeXB0bzogICAgICAgICAgICAgICoxLjMxeCBhcyBzbG93KiAg ICAxODAuOW1zICsvLSAwLjYlICAgIDIzNi43bXMgKy8tIDAuMiUgICAgIHNpZ25pZmljYW50DQog ICAgYWVzOiAgICAgICAgICAgICAgICoxLjQweCBhcyBzbG93KiAgICAgNTYuNm1zICsvLSAwLjcl ICAgICA3OS4wbXMgKy8tIDAuMCUgICAgIHNpZ25pZmljYW50DQogICAgbWQ1OiAgICAgICAgICAg ICAgICoxLjI2eCBhcyBzbG93KiAgICAgNjMuMG1zICsvLSAxLjElICAgICA3OS4zbXMgKy8tIDAu NCUgICAgIHNpZ25pZmljYW50DQogICAgc2hhMTogICAgICAgICAgICAgICoxLjI4eCBhcyBzbG93 KiAgICAgNjEuM21zICsvLSAwLjYlICAgICA3OC40bXMgKy8tIDAuNSUgICAgIHNpZ25pZmljYW50 DQoNCiAgZGF0ZTogICAgICAgICAgICAgICAgKjEuMDY2eCBhcyBzbG93KiAgIDMyOC40bXMgKy8t IDAuNSUgICAgMzUwLjJtcyArLy0gMC43JSAgICAgc2lnbmlmaWNhbnQNCiAgICBmb3JtYXQtdG9m dGU6ICAgICAgKjEuMDYyeCBhcyBzbG93KiAgIDIwNC44bXMgKy8tIDAuNyUgICAgMjE3LjRtcyAr Ly0gMC43JSAgICAgc2lnbmlmaWNhbnQNCiAgICBmb3JtYXQteHBhcmI6ICAgICAgKjEuMDc0eCBh cyBzbG93KiAgIDEyMy42bXMgKy8tIDAuNyUgICAgMTMyLjhtcyArLy0gMC45JSAgICAgc2lnbmlm aWNhbnQNCg0KICBtYXRoOiAgICAgICAgICAgICAgICAqMS4xODZ4IGFzIHNsb3cqICAgMzkxLjFt cyArLy0gMC43JSAgICA0NjMuOW1zICsvLSAwLjclICAgICBzaWduaWZpY2FudA0KICAgIGNvcmRp YzogICAgICAgICAgICAqMS4zNnggYXMgc2xvdyogICAgMTE3LjhtcyArLy0gMC42JSAgICAxNTku OG1zICsvLSAwLjYlICAgICBzaWduaWZpY2FudA0KICAgIHBhcnRpYWwtc3VtczogICAgICAqMS4w Mzh4IGFzIHNsb3cqICAgMTk2LjhtcyArLy0gMC45JSAgICAyMDQuM21zICsvLSAxLjQlICAgICBz aWduaWZpY2FudA0KICAgIHNwZWN0cmFsLW5vcm06ICAgICAqMS4zMHggYXMgc2xvdyogICAgIDc2 LjVtcyArLy0gMS45JSAgICAgOTkuOG1zICsvLSAwLjYlICAgICBzaWduaWZpY2FudA0KDQogIHJl Z2V4cDogICAgICAgICAgICAgIC0gICAgICAgICAgICAgICAgICAzMjIuN21zICsvLSAwLjUlICAg IDMyMS44bXMgKy8tIDAuMyUgDQogICAgZG5hOiAgICAgICAgICAgICAgIC0gICAgICAgICAgICAg ICAgICAzMjIuN21zICsvLSAwLjUlICAgIDMyMS44bXMgKy8tIDAuMyUgDQoNCiAgc3RyaW5nOiAg ICAgICAgICAgICAgKjEuMDg5eCBhcyBzbG93KiAgIDc4OC4xbXMgKy8tIDEuNyUgICAgODU4LjBt cyArLy0gMS40JSAgICAgc2lnbmlmaWNhbnQNCiAgICBiYXNlNjQ6ICAgICAgICAgICAgKjEuMTAx eCBhcyBzbG93KiAgIDEwOC4zbXMgKy8tIDEuMyUgICAgMTE5LjJtcyArLy0gMS4xJSAgICAgc2ln bmlmaWNhbnQNCiAgICBmYXN0YTogICAgICAgICAgICAgKjEuMTg0eCBhcyBzbG93KiAgIDEyMC45 bXMgKy8tIDAuNSUgICAgMTQzLjJtcyArLy0gMC43JSAgICAgc2lnbmlmaWNhbnQNCiAgICB0YWdj bG91ZDogICAgICAgICAgPz8gICAgICAgICAgICAgICAgIDIwMy4zbXMgKy8tIDIuNSUgICAgMjA5 LjNtcyArLy0gMi41JSAgICAgbm90IGNvbmNsdXNpdmU6IG1pZ2h0IGJlICoxLjAzMHggYXMgc2xv dyoNCiAgICB1bnBhY2stY29kZTogICAgICAgKjEuMTExeCBhcyBzbG93KiAgIDIyOS44bXMgKy8t IDEuOCUgICAgMjU1LjJtcyArLy0gMS42JSAgICAgc2lnbmlmaWNhbnQNCiAgICB2YWxpZGF0ZS1p bnB1dDogICAgKjEuMDQyeCBhcyBzbG93KiAgIDEyNS44bXMgKy8tIDIuNiUgICAgMTMxLjFtcyAr Ly0gMi4yJSAgICAgc2lnbmlmaWNhbnQNCg== 22392 2008-07-20 11:01:17 -0700 2008-07-20 11:01:17 -0700 bubbleSort.js bubbleSort.js text/javascript 1836 ddkilzer Ly8gaHR0cDovL3d3dy5qcy14LmNvbS9wYWdlL2phdmFzY3JpcHRzX19leGFtcGxlLmh0bWw/dmll dz0xODYNCg0KZnVuY3Rpb24gQnViYmxlU29ydFVwKGFycmF5TmFtZSxsZW5ndGgpIA0KeyANCiAg ICBmb3IgKHZhciBpPTA7IGk8KGxlbmd0aC0xKTsgaSsrKSANCiAgICB7IA0KICAgICAgICBmb3Ig KHZhciBiPWkrMTsgYjxsZW5ndGg7IGIrKykgDQogICAgICAgIHsgDQogICAgICAgICAgICBpZiAo YXJyYXlOYW1lW2JdIDwgYXJyYXlOYW1lW2ldKSANCiAgICAgICAgICAgIHsgDQogICAgICAgICAg ICAgICAgdmFyIHRlbXAgPSBhcnJheU5hbWVbaV07IA0KICAgICAgICAgICAgICAgIGFycmF5TmFt ZVtpXSA9IGFycmF5TmFtZVtiXTsgDQogICAgICAgICAgICAgICAgYXJyYXlOYW1lW2JdID0gdGVt cDsgDQogICAgICAgICAgICB9IC8vZW5kLWlmIA0KICAgICAgICB9IC8vZW5kIGZvci1sb29wIA0K ICAgIH0gLy9lbmQgZm9yLWxvb3AgDQp9IC8vIGVuZCBmbiBCdWJibGVTb3J0VXAgDQoNCmZ1bmN0 aW9uIEJ1YmJsZVNvcnREb3duKGFycmF5TmFtZSxsZW5ndGgpIA0KeyANCiAgICBmb3IgKHZhciBp PTA7IGk8KGxlbmd0aC0xKTsgaSsrKSANCiAgICB7IA0KICAgICAgICBmb3IgKHZhciBiPWkrMTsg YjxsZW5ndGg7IGIrKykgDQogICAgICAgIHsgDQogICAgICAgICAgICBpZiAoYXJyYXlOYW1lW2Jd ID4gYXJyYXlOYW1lW2ldKSANCiAgICAgICAgICAgIHsgDQogICAgICAgICAgICAgICAgdmFyIHRl bXAgPSBhcnJheU5hbWVbaV07IA0KICAgICAgICAgICAgICAgIGFycmF5TmFtZVtpXSA9IGFycmF5 TmFtZVtiXTsgDQogICAgICAgICAgICAgICAgYXJyYXlOYW1lW2JdID0gdGVtcDsgDQogICAgICAg ICAgICB9IC8vZW5kLWlmIA0KICAgICAgICB9IC8vZW5kIGZvci1sb29wIA0KICAgIH0gLy9lbmQg Zm9yLWxvb3AgDQp9IC8vIGVuZCBmbiBCdWJibGVTb3J0RG93biANCg0KZnVuY3Rpb24gc2hvd0Fy cmF5KHRleHQsYXJyYXlOYW1lLGxlbmd0aCkgDQp7IA0KICAgIHByaW50KHRleHQgKyAnOiAnKTsg DQogICAgZm9yICh2YXIgaT0wOyBpPGxlbmd0aDsgaSsrKSANCiAgICAgICAge3ByaW50KGFycmF5 TmFtZVtpXSk7fSANCn0gLy9lbmQgZm4gc2hvd0FycmF5IA0KDQovLyBGcm9tIEJyZWF0aCBvZiBG aXJlIElJDQp2YXIgc2VlZCA9IDB4YmJkMTsNCg0KZnVuY3Rpb24gcmFuZCgpDQp7DQogIHZhciBh ID0gKHNlZWQgKiA1KSAmIDB4RkZGRjsNCiAgLy8gU3dhcCBieXRlcw0KICBhID0gKChhICYgMHhG RikgPDwgOCkgKyAoYSA+PiA4KTsNCiAgc2VlZCA9ICgoKHNlZWQgKyBhKSAmIDB4ZmYpPDw4KSAr IChhICYgMHhmZik7DQogIC8vIFJldHVybiBsb3cgYnl0ZQ0KICByZXR1cm4gKGEgJiAweGZmKTsN Cn0NCg0KdmFyIGFycmF5MSA9IG5ldyBBcnJheSgpDQp2YXIgYXJyYXkyID0gbmV3IEFycmF5KCkN Cg0KLy8gPCAwLjAxIHNlYw0KZm9yICh2YXIgaSA9IDA7IGkgPCA0NTUwOyBpKyspIHsNCiAgYXJy YXkxW2ldID0gcmFuZCgpIHwgKHJhbmQoKSA8PCA4KQ0KICBhcnJheTJbaV0gPSByYW5kKCkgfCAo cmFuZCgpIDw8IDgpDQp9DQoNCi8vc2hvd0FycmF5KCdVbnNvcnRlZCcsYXJyYXkxLGFycmF5MS5s ZW5ndGgpOyANCkJ1YmJsZVNvcnREb3duKGFycmF5MSxhcnJheTEubGVuZ3RoKTsgDQovL3Nob3dB cnJheSgnU29ydGVkLURvd24nLGFycmF5MSxhcnJheTEubGVuZ3RoKTsgDQoNCi8vc2hvd0FycmF5 KCdVbnNvcnRlZCcsYXJyYXkyLGFycmF5Mi5sZW5ndGgpOyANCkJ1YmJsZVNvcnRVcChhcnJheTIs YXJyYXkyLmxlbmd0aCk7IA0KLy9zaG93QXJyYXkoJ1NvcnRlZC1VcCcsYXJyYXkyLGFycmF5Mi5s ZW5ndGgpOyANCg0K 22404 2008-07-21 08:22:56 -0700 2008-07-22 00:27:45 -0700 Add missing NDEBUG qt-slow.patch text/plain 941 loki SW5kZXg6IEphdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBKYXZhU2NyaXB0 Q29yZS9DaGFuZ2VMb2cJKHJldmlzaW9uIDM1MjY1KQorKysgSmF2YVNjcmlwdENvcmUvQ2hhbmdl TG9nCSh3b3JraW5nIGNvcHkpCkBAIC0xLDMgKzEsMTIgQEAKKzIwMDgtMDctMjEgIEdhYm9yIExv a2kgIDxsb2tpQGluZi51LXN6ZWdlZC5odT4KKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkg KE9PUFMhKS4KKyAgICAgICAgPGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9p ZD0yMDA5Nz4KKworICAgICAgICBBZGQgbWlzc2luZyBOREVCVUcgZGVmaW5lIGluIGNhc2Ugb2Yg cmVsZWFzZS4KKworICAgICAgICAqIGtqcy9qc2MucHJvOgorCiAyMDA4LTA3LTE5ICBPbGl2ZXIg SHVudCAgPG9saXZlckBhcHBsZS5jb20+CiAKICAgICAgICAgUmV2aWV3ZWQgYnkgQ2FtZXJvbiBa d2FyaWNoLgpJbmRleDogSmF2YVNjcmlwdENvcmUva2pzL2pzYy5wcm8KPT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0g SmF2YVNjcmlwdENvcmUva2pzL2pzYy5wcm8JKHJldmlzaW9uIDM1MjY1KQorKysgSmF2YVNjcmlw dENvcmUva2pzL2pzYy5wcm8JKHdvcmtpbmcgY29weSkKQEAgLTEzLDYgKzEzLDEwIEBAIElOQ0xV REVQQVRIICs9ICQkUFdELy4uIFwKIENPTkZJRyAtPSBhcHBfYnVuZGxlCiBERUZJTkVTICs9IEJV SUxESU5HX1FUX18KIAorQ09ORklHKHJlbGVhc2UpIHsKKyAgICBERUZJTkVTICs9IE5ERUJVRyBV U0VfU1lTVEVNX01BTExPQworfQorCiBDT05GSUcgKz0gbGlua19wa2djb25maWcKIAogUU1BS0Vf UlBBVEhESVIgKz0gJCRPVVRQVVRfRElSL2xpYgo=