- 03 Mar, 2015 38 commits
-
-
Haim Grosman authored
Summary: EventBase::runAfterDelay to throw an exception in case it fails to schedule a proper execution of the given callback (instead of silently returning false) it appears to be the right thing to do, since: @davejwatson: Digging through the layers of code, it appears this can only fail if epoll_ctl() with EPOLL_CTL_ADD fails. Ignoring libevent bugs, it looks like the only relevant errors could be ENOMEM or ENOSPC. So nonrecoverable Test Plan: Unit tests Reviewed By: anca@fb.com Subscribers: yzhan, haijunz, simpkins, net-systems@, varunk, zeus-diffs@, nli, dfechete, fugalh, atlas2-eng@, alandau, bmatheny, everstore-dev@, zhuohuang, wormhole-diffs@, anca, mwa, jgehring, oujin, alikhtarov, fuegen, mshneer, wch, bil, sanketh, zippydb, maxwellsayles, jsedgwick, trunkagent, fbcode-common-diffs@, chaoyc, search-fbcode-diffs@, andrewcox, unicorn-diffs@, tw-eng@, xie, kennyyu, yfeldblum, folly-diffs@, davejwatson FB internal diff: D1805125 Signature: t1:1805125:1424927912:8bebb4c3b9f1fa189c0ce97b12cdb8f95dba97ae
-
John Ehrhardt authored
Summary: Updating Folly Formatting's use of separators for Decimal Integers 'd' and Numbers 'n'. Updated Decimal Integers 'd' use of separators to directly use commas ',' and grouping of 3 digits. Updated Numbers use of separators and grouping to follow locale settings. Test Plan: Wrote unit tests to validate the insertion of thousandsSeparators works. Note that the numbers unit test does not modify the locale since the test cases are not thread safe. Reviewed By: lesha@fb.com Subscribers: tudorb, trunkagent, folly-diffs@, yfeldblum, aric FB internal diff: D1821156 Tasks: 6087521 Signature: t1:1821156:1424923837:d0fb383a07fd733375b72b1905e6112afa141265
-
Sean Cannella authored
Summary: Manual inspection of AsyncSocket.o shows the constructors are generating ugly (and large) code. Call the smaller constructors instead of duplicating the code. This reduces the debug .o size from ~1.9MB to ~1.6MB. Test Plan: existing tests Reviewed By: benyluo@fb.com Subscribers: mzlee, folly-diffs@, yfeldblum, mathieubaudet, subodh, pgriess, benyluo FB internal diff: D1870896 Signature: t1:1870896:1424885638:d37fc79c0f88d04109c8bb73e632dc506b9f773b
-
Hans Fugal authored
Summary: Unconditionally execute the action, passthrough semantics. Test Plan: new unit test Reviewed By: bmatheny@fb.com Subscribers: trunkagent, exa, folly-diffs@, yfeldblum, jsedgwick FB internal diff: D1868837 Tasks: 6166860 Signature: t1:1868837:1424820181:0e83f54b59d7091dac60ab65feb387992e8ae89c
-
Sven Over authored
Summary: writeFull() returns ssize_t and without proper casting, comparing it with data.size() triggers a compiler warning (which is treated as an error) in the gcc-4.9-glibc-2.20 toolchain. Test Plan: fbmake runtests Reviewed By: mhx@fb.com Subscribers: folly-diffs@, yfeldblum FB internal diff: D1870710 Signature: t1:1870710:1424874592:f51026c35196d763ad4b192d43c8ccee0255b41d
-
Nicholas Ormrod authored
Summary: See the github pull request at https://github.com/facebook/folly/issues/135 Important optimizations: remove one of the branches and change ##empty()## to ##capacity()==0## Test Plan: Ran extended fbvector test suite (re-enabled in experimental/njormrod) Also fbconfig -r folly && fbmake runtests Reviewed By: markisaa@fb.com Subscribers: trunkagent, sdwilsh, njormrod, folly-diffs@, yfeldblum FB internal diff: D1869112 Tasks: 6338531 Signature: t1:1869112:1424823901:d2d7331aef82edad1e8c159005cc1c7185550d0c
-
Andrei Alexandrescu authored
Summary: Funnily, google::setThreadName is Facebook specific: it's introduced only in the FB build of glog. This diff removes the call because folly should not depend on FB's glog. Test Plan: made sure the line was deleted Reviewed By: ldbrandy@fb.com Subscribers: net-systems@, folly-diffs@, yfeldblum FB internal diff: D1868249 Tasks: 6345935 Signature: t1:1868249:1424812595:2a084754d614ab9b7b6168454f33638882e3f6db
-
Praveen Kumar authored
Summary: EmulatedFutexAtomic concisely says here that it doesn't copy construct, move construct, copy assign, or move assign. Closes #130 Test Plan: Inspection and all unit tests. ``` Summary (total time 66.97s): PASS: 2029 FAIL: 0 SKIP: 0 FATAL: 0 TIMEOUT: 0 ``` Reviewed By: ngbronson@fb.com Subscribers: folly-diffs@, yfeldblum FB internal diff: D1868033 Signature: t1:1868033:1424808823:d9dd618981b29b5949c00190d69d670d96e60e07
-
Beny Luo authored
Summary: encodeVarintToIOBuf and decodeVarintFromCursor are not used when FOLLY_HAVE_LIBLZMA and FOLLY_HAVE_LIBLZ4 are false, which will trigger -Werror,-Wunused-function on compile on mobile. Test Plan: unit test Reviewed By: seanc@fb.com Subscribers: meisner, trunkagent, seanc, folly-diffs@, yfeldblum FB internal diff: D1864016 Signature: t1:1864016:1424702813:75535ea92ac16385ce5c272d93bba507ccbb67a1
-
Andrii Grynenko authored
Summary: Previously leak-fatal could never happen because we were incorrectly checking singleton state. Sending this diff to see how many tests will actually fail and potentially fix worst offenders. Test Plan: unit tests Reviewed By: stepan@fb.com Subscribers: trunkagent, folly-diffs@, yfeldblum FB internal diff: D1838405 Signature: t1:1838405:1424478983:94cda86ed57f38b0cf626b74804fbc168d182c66
-
Andre Pinto authored
Summary: Add option to disable time measurement in EventBase::loopBody() to improve performances when time measurement is not used. Test Plan: Unit tests, benchmarking and manual tests. Reviewed By: pavlo@fb.com Subscribers: trunkagent, folly-diffs@, yfeldblum FB internal diff: D1846020 Tasks: 6086197 Signature: t1:1846020:1424459703:9686a84027e1137ee43eb0458b0baa4fee3a8abf
-
Andrey Goder authored
Summary: This 'moves' bistro::AutoTimer into folly. I actually changed it a bit, so I'm not migrating any existing uses in this diff, but I can do that as a followup. We're using it in various places now, so it doesn't make sense to have it in bistro. Basically this class lets you conveniently time a series of things while printing log messages, e.g.: { AutoTimer t("Done"); processA(); t.log("processA() complete"); } would print "processA() complete in XXX seconds" "Done in YYY seconds" I changed some things from bistro::AutoTimer: - Use std::chrono::time_point<Clock> to store the start time. - I removed the gflags. It seems weird to rely on those to control the timer, since they would affect all uses of this. If a specific user of the timer wants to actually control timing at runtime, I think it's better to use a custom logger. - I removed the 'min log duration'. As implemented it actually made the constructor conflict: AutoTimer("foo", 1.2) is ambiguous. Also, it should probably be set on the logger, so this likewise feels like a place where you could make a custom logger. I'm not sure that the way I use the default logger as a template argument is best, so please suggest improvements to that. Test Plan: unit test Reviewed By: lesha@fb.com Subscribers: trunkagent, folly-diffs@, yfeldblum FB internal diff: D1860794 Signature: t1:1860794:1424458219:a2dd0d428aedf063a888b2937b1443c7142554e8 -
Hans Fugal authored
Summary: I thought this was a race, but I think now it was something to do with using a value that had been moved out or something. Anyway, this refactor is cleaner and consolidates a few methods so it's all kinds of fuzzy feelings. Test Plan: unit tests Reviewed By: hannesr@fb.com Subscribers: exa, folly-diffs@, yfeldblum, jsedgwick FB internal diff: D1861376 Tasks: 6298004 Signature: t1:1861376:1424465861:736353ab3174656fec9e036e0ebd964970da38d0
-
Ranjeeth Dasineni authored
Summary: a couple of files were catching this exception and included the .cpp file :O. So we moved to .h before realizing this was not intended to be caught. reverting the move to .h. codemod will follow. Test Plan: fbmake -r folly && fbmake && fbmake runtests Reviewed By: delong.j@fb.com Subscribers: trunkagent, doug, seanc, folly-diffs@, yfeldblum FB internal diff: D1854301 Tasks: 6245781 Signature: t1:1854301:1424375745:7e226196315fac99835752896cc5e549e91df2da
-
Hans Fugal authored
Summary: Instead of returning a deactivated future, have `via` just set the executor. Propagate the executor from `then`. This fixes the `via().get()` problem, and has semantics similar to before for `via().then().then()`. However, the semantics are now slightly different - each `then` goes back through the executor. This adds some overhead and tweaks the semantics (e.g. if the executor is a threadpool it might execute subsequent `then`s in another thread). However, with `futures::chain` recently introduced, and any other convenience methods that you can dream up and make a case for, we can reasonably get the old once-through-the-executor behavior when performance or other concerns demand it. e.g. `via().then(futures::chain(a, b, c))`. Test Plan: unit tests Reviewed By: hannesr@fb.com Subscribers: zeus-diffs@, mmandal, steveo, rituraj, trunkagent, exa, folly-diffs@, yfeldblum, jsedgwick, davejwatson FB internal diff: D1839691 Tasks: 6048744 Signature: t1:1839691:1424397180:ca0b0ea7b3867769ab8abd254a510059df67011e
-
Hannes Roth authored
Summary: This reverts commit 1e407b48d379a41f32e7a980285dbdf4dadb2e33. Test Plan: revert-hammer Reviewed By: yitingli@fb.com Subscribers: folly-diffs@, jsedgwick, yfeldblum FB internal diff: D1858994 Signature: t1:1858994:1424386588:5c4608ecbe1f9ab1108ac12e506b157b54d0078b
-
Hannes Roth authored
Summary: These were doing more assignments than necessary. 1) We don't need to set `core_` to `nullptr` and then immediately overwrite it with `other.core_`. 2) Don't need to swap the booleans (this was actually an unitialized memory access). Test Plan: Ran all the tests. Reviewed By: hans@fb.com Subscribers: trunkagent, folly-diffs@, jsedgwick, yfeldblum FB internal diff: D1852386 Signature: t1:1852386:1424375154:25997209e76ccd16169125597e868932a5143ffb
-
Michael Lee authored
Summary: This diff should let us include folly/io:compression into iOS and Android projects without pulling in a lot of dependencies and should allow the use of zlib. Test Plan: fbconfig -r folly/io/tests && fbmake runtests Reviewed By: seanc@fb.com Subscribers: trunkagent, bmatheny, folly-diffs@, yfeldblum, seanc, benyluo FB internal diff: D1853058 Tasks: 6245912 Signature: t1:1853058:1424370881:2ea243d93e6041502e5e356fa430c2483f652b40
-
Dave Watson authored
Summary: Use server bootstrap. * code reuse * Allows multiple accept threads easily. Accept now doesn't happen in the server's eventBase, but a new thread. I think I hit all the corner cases involving existingSocket, duplex, stopListening(), etc. There are a lot of minor changes here to support all the cases, let me know if I should split anything to separate diffs Test Plan: fbconfig -r thrift/lib/cpp2; fbmake runtests Reviewed By: hans@fb.com Subscribers: haijunz, yfeldblum, ruibalp, fbcode-common-diffs@, davejwatson, jsedgwick, trunkagent, doug, alandau, bmatheny, mshneer, folly-diffs@, mcduff, andrewcox, fugalh FB internal diff: D1732895 Tasks: 5788102 Signature: t1:1732895:1423087631:379bbb131c35ce2221801bc7cec226f87ba0b1d9
-
Dave Watson authored
Summary: from discussion Test Plan: unit tests Reviewed By: hans@fb.com Subscribers: doug, folly-diffs@, jsedgwick, yfeldblum FB internal diff: D1829282 Signature: t1:1829282:1423180907:3630dac1378750b05f316c672fbbd71138d2bc0a
-
Dave Watson authored
Summary: They are functionaly equivalent, no need to have more than one Test Plan: fbconfig -r realtime/voip; fbmake runtests Reviewed By: naizhi@fb.com Subscribers: trunkagent, doug, ps, bmatheny, folly-diffs@, yfeldblum FB internal diff: D1828044 Tasks: 6154007 Signature: t1:1828044:1423165354:f71d2fd28ca76a8f67a597c747f8578d2909823c
-
Andrii Grynenko authored
Summary: After D1827390 regular get and get_weak is comparable to Meyers and static singletons, so there's no need to keey _fast APIs. Test Plan: benchmark && fbmake Reviewed By: mshneer@fb.com Subscribers: trunkagent, configerator-diffs@, fbcode-common-diffs@, mcduff, hitesh, mshneer, fugalh, acampi, alikhtarov, folly-diffs@, jsedgwick, yfeldblum FB internal diff: D1843219 Signature: t1:1843219:1424216566:f2f182a1c86bb5f0fb1f978d8c6b7a4388198f5f
-
Andrei Alexandrescu authored
Summary: This are the open-source-related additions. A separate diff will replace the current implementation in common/concurrency/ to use this one. Test Plan: fbmake runtests Reviewed By: simpkins@fb.com Subscribers: trunkagent, net-systems@, folly-diffs@, yfeldblum FB internal diff: D1845525 Signature: t1:1845525:1424207291:d30e3c5e85222527c2aff39c1250aa1e41b9a2cf
-
Sara Golemon authored
Summary: The continuation at the end of the line with the while was causing gcc to complain about it because it's a single line comment. Reformat it into a multiline comment and wrap it to 80 chars. Test Plan: It's just a comment... Reviewed By: paulbiss@fb.com Subscribers: trunkagent, folly-diffs@, yfeldblum FB internal diff: D1850876 Signature: t1:1850876:1424124973:b2afbc2a8177ac6d11a21bf5a52093bcfd9cbb3e
-
Hannes Roth authored
Summary: I think this makes sense, especially if `interrupt_` is really rare. Test Plan: ============================================================================ folly/futures/test/Benchmark.cpp relative time/iter iters/s ============================================================================ constantFuture 101.25ns 9.88M promiseAndFuture 92.86% 109.03ns 9.17M withThen 28.96% 349.65ns 2.86M ---------------------------------------------------------------------------- oneThen 337.33ns 2.96M twoThens 58.46% 577.06ns 1.73M fourThens 32.42% 1.04us 961.23K hundredThens 1.45% 23.24us 43.04K ---------------------------------------------------------------------------- no_contention 4.45ms 224.48 contention 72.89% 6.11ms 163.61 ---------------------------------------------------------------------------- throwAndCatch 6.34us 157.77K throwAndCatchWrapped 95.65% 6.63us 150.91K throwWrappedAndCatch 142.58% 4.45us 224.95K throwWrappedAndCatchWrapped 488.86% 1.30us 771.27K ---------------------------------------------------------------------------- throwAndCatchContended 1.01s 987.01m throwAndCatchWrappedContended 100.03% 1.01s 987.27m throwWrappedAndCatchContended 148.37% 682.86ms 1.46 throwWrappedAndCatchWrappedContended 18765.19% 5.40ms 185.21 ============================================================================ Compare to D1843129#31 (seems to be an improvement?) Reviewed By: hans@fb.com Subscribers: trunkagent, folly-diffs@, jsedgwick, yfeldblum FB internal diff: D1847843 Signature: t1:1847843:1423855151:3602d49df245eda3f85325ab5efd63514213ca44
-
Jim Meyering authored
Summary: Disable the -Wunevaluated-expression warning for this test: EXPECT_EQ(typeid(ie), typeid(IntException)); Otherwise, clang warns about the unevaluated expression because the expansion of EXPECT_EQ applies sizeof to an expression with side effects: folly/test/ExceptionWrapperTest.cpp:179:122: error: expression with side effects has no effect in an unevaluated context [-Werror,-Wunevaluated-expression] switch (0) case 0: default: if (const ::testing::AssertionResult gtest_ar = (::testing::internal:: EqHelper<(sizeof(::testing::internal::IsNullLiteralHelper(typeid(ie))) == 1)>::Compare("typeid(ie)", "typeid(IntException)", typeid(ie), typeid(IntException)))) ; else ::testing::internal::AssertHelper(::testing::TestPartResult::kNonFatalFailure, "folly/test/ExceptionWrapperTest.cpp", 179, gtest_ar.failure_message()) = ::testing::Message(); Test Plan: Ensure this compiles with clang-3.4, 3.5 and clang:dev. I.e., ensure that this prints PASS at the end. for i in '' 3.5 dev; do test -n "$i" && i=--with-project-version=clang:$i fbconfig -r --clang $i folly/test:exception_wrapper_test && fbmake dbgo done && echo PASS Reviewed By: mpawlowski@fb.com Subscribers: mathieubaudet, folly-diffs@, yfeldblum FB internal diff: D1850778 Tasks: 6244745 Signature: t1:1850778:1424111029:136478e9a3cc3a219047547d501de4c579a1a181 -
Jim Meyering authored
Summary: * folly/wangle/channel/test/MockChannelHandler.h: Avoiding a new warning that rendered this file uncompilable by clang-3.5 and older. Add cpp conditionals to protect against that. Test Plan: Ensure it works with clang-3.4, 3.5 and clang:dev. I.e., ensure that this prints PASS at the end. for i in '' 3.5 dev; do test -n "$i" && i=--with-project-version=clang:$i fbconfig -r --clang $i folly/wangle/channel/test && fbmake runtests_dbgo done && echo PASS Reviewed By: meyering@fb.com Subscribers: mpawlowski, fugalh, mathieubaudet, folly-diffs@, jsedgwick, yfeldblum FB internal diff: D1849474 Tasks: 6244745 Blame Revision: D1848330 @build-break
-
Jim Meyering authored
Summary: Clang's new -Wself-move warning dings this code: b = std::move(b); But in this case, that is a deliberate test of this capability, so we suppress that particular warning for the duration of this small test function. Test Plan: Run this and note that it now compiles: (it passes the tests, too) fbconfig --clang --with-project-version=clang:dev -r folly/test:optional_test && fbmake dbgo Reviewed By: watashi@fb.com Subscribers: folly-diffs@, yfeldblum FB internal diff: D1849689 Tasks: 6244745 Signature: t1:1849689:1423877932:4f479d682bd8ba97dcabed7fc29ce0a22c013969 @build-break (using this to work around bogus push failure: https://phabricator.fb.com/P19730606)
-
Jim Meyering authored
Summary: * folly/ExceptionWrapper.h (class_name): Clang is not yet smart enough to see that there is no harm in dereferencing an "item_" that we've just verified is non-NULL. Accommodate it, to avoid this: ./folly/ExceptionWrapper.h:199:30: error: expression with side effects will be evaluated despite being used as an operand to 'typeid' [-Werror,-Wpotentially-evaluated-expression] return demangle(typeid(*item_)); ^ Test Plan: Run these commands and note there are fewer errors than before: fbconfig --clang --with-project-version=clang:dev -r folly && fbmake dbgo Reviewed By: mhorowitz@fb.com Subscribers: trunkagent, yfeldblum, folly-diffs@ FB internal diff: D1848327 Tasks: 6244745 Signature: t1:1848327:1423869597:a58c9e9e3671befb78ae07fbd274a13d08ffb2a7 Blame Revision:
-
Yedidya Feldblum authored
Summary: [Folly] clang:dev rejects vector<atomic<T>> in EventBaseTest.cpp. Test Plan: Build and run `folly/io/async/test/EventBaseTest.cpp` using the current `clang`; try building with `clang:dev` and verifying that at least the related failures are gone. Reviewed By: meyering@fb.com Subscribers: mathieubaudet, folly-diffs@, yfeldblum, dougw, brettp FB internal diff: D1848749 Tasks: 6244720 Signature: t1:1848749:1423866420:7b7354d4568b6a6d6a824236ae8d271f6855d90b
-
Jim Meyering authored
Summary: This avoids warnings about e.g., sizeof(k++) that result from macro expansion. * folly/futures/test/ExecutorTest.cpp: Move increment out of macro argument list. * folly/test/FBVectorTestBenchmarks.cpp.h: Likewise. * folly/test/LazyTest.cpp: Likewise. Otherwise, we'd get errors like this: folly/test/LazyTest.cpp:49:118: error: expression with side effects has no effect in an unevaluated context [-Werror,-Wunevaluated-expression] switch (0) case 0: default: if (const ::testing::AssertionResult gtest_ar = (::testing::internal:: EqHelper<(sizeof(::testing::internal::IsNullLiteralHelper(++globalCount())) == 1)>::Compare("++globalCount()", "1", ++globalCount(), 1))) ; else ::testing::internal::AssertHelper(::testing::TestPartResult::kNonFatalFailure, "folly/test/LazyTest.cpp", 49, gtest_ar.failure_message()) = ::testing::Message(); ^ Test Plan: Run these commands and note there are fewer errors than before: fbconfig --clang --with-project-version=clang:dev -r folly && fbmake dbgo Reviewed By: njormrod@fb.com Subscribers: folly-diffs@, jsedgwick, yfeldblum FB internal diff: D1848324 Tasks: 6244745 Signature: t1:1848324:1423860890:bce44c5e0895804a21957893ae6b78e76dfbc4d3 -
Jim Meyering authored
Summary: Upgrading to clang:dev (clang >3.6) brought in some new warnings. This change address all of the issues exposed by the new -Winconsistent-missing-override, usually by simply adding the missing "override" keyword. However, in folly/wangle/channel/test/MockChannelHandler.h, I chose to ignore those warnings for the mocked functions. * folly/futures/test/ViaTest.cpp: Add missing "override"(s). * folly/io/async/AsyncSSLSocket.h: Likewise. * folly/io/async/AsyncSocket.h: Likewise. * folly/io/async/EventBase.h: Likewise. * folly/test/ExceptionWrapperTest.cpp: Likewise. * folly/wangle/channel/AsyncSocketHandler.h: Likewise. * folly/wangle/channel/test/MockChannelHandler.h: Ignore the new warning for these functions. Test Plan: Run these commands and note there are fewer errors than before: fbconfig --clang --with-project-version=clang:dev -r folly && fbmake dbgo Reviewed By: hans@fb.com Subscribers: trunkagent, fugalh, folly-diffs@, jsedgwick, yfeldblum FB internal diff: D1848330 Tasks: 6244745 Signature: t1:1848330:1423858909:b167ca220d6c9fe036d3adca08cf3053a7a9de16
-
Hannes Roth authored
Summary: This seems cleaner. I also moved a Core member around and changed state to be a smaller size to save us 16 bytes. Might be neat to try to get this down to 128? Fixed compilation with GCC 4.9, too. Test Plan: Ran all the tests. Will also benchmark. Reviewed By: hans@fb.com Subscribers: trunkagent, folly-diffs@, jsedgwick, yfeldblum FB internal diff: D1843129 Signature: t1:1843129:1423789502:60e56d1af553994991195fcc350d37d644a556ca
-
Jonathan Frank authored
Summary: Pass TransportInfo object through connection setup. This allows us to pass along additional info from the setup if necessary. Test Plan: Run proxygen on my devserver listening on ipv4 and ipv6 for ports 80 and 443. Configure it to run a lua request rule with the following lines: io.stderr:write("jonlog: proxy " .. conn:getClientAddrOriginal() .. "\n") io.stderr:write("jonlog: client " .. request:getClientIPStr() .. "\n") Run haproxy also on my devserver, listening on ports 8555-8558, with each forwarding to one of the four pairs (ipv4-localhost/ipv6-localhost, 80/443) From my laptop, run curls against each of these four endpoints and make sure that we print out "127.0.0.1" as the proxy IP, and my laptop's IP as the client IP. Reviewed By: cgheorghe@fb.com Subscribers: jsedgwick, yfeldblum, trunkagent, ruibalp, bmatheny, folly-diffs@, dsp, nbm FB internal diff: D1746590 Tasks: 5688127 Signature: t1:1746590:1423695880:f647964d95636a69a00304e144aef71ee0213d28 -
Giuseppe Ottaviano authored
Summary: `EliasFanoReader::position()` used to return `size() - 1` both when the reader is positioned on the last element, and after `next()` is called after that (and it return `false`). Now in the latter case `position()` returns `size()` (consistently with the usual behaviour of past-the-end iterators). Also fix the return type of `jumpTo`. Test Plan: fbconfig folly/experimental/test:eliasfano_test && fbmake runtests_opt Reviewed By: philipp@fb.com Subscribers: trunkagent, folly-diffs@, yfeldblum FB internal diff: D1846275 Signature: t1:1846275:1423790264:151f5d2e1e09d4e24dfb758473dfb9cd52c070bd
-
Keith Adams authored
Summary: A colleague at another company started making fun of MicroSpinLock for its x86 assembly and ad hoc compiler memory barriers. Use C++11 (which wasn't really a thing at the time I wrote this). Test Plan: folly's runtests. What else would we like? Reviewed By: andrei.alexandrescu@fb.com Subscribers: folly-diffs@, yfeldblum FB internal diff: D1841563 Signature: t1:1841563:1423780458:a447c081fbd72e3420b23e95dcf26575c9a06798
-
Giuseppe Ottaviano authored
Summary: Short skips have been optimized by adding special cases that use simple iteration when it is convenient. Large skips have been optimized by using the broadword selection algorithm by Vigna (improved with ideas by Gog&Petri) instead of iterating on the zeros/ones of the upper bits. The benchmarks had to be made more granular to measure the differences, in particular they used to test skipping with a fixed skip length for each test, while now we average over a range of skips to better simulate a random distribution. The improvements are very significant for `skipTo()` on short skips, about 2-3x for skips at distance 1 or 2, which can occur when intersecting dense lists. On large skips the gain is about 17%. `skip()` exhibits slightly smaller improvements. before after ============================================================================ ================== folly/experimental/test/EliasFanoCodingTest.cpp relative time/iter iters/s time/iter iters/s ============================================================================ ================== Next 2.52ns 396.26M 2.52ns 397.28M Skip_ForwardQ128(1) 8.66ns 115.52M 3.92ns 255.28M Skip_ForwardQ128(2) 8.37ns 119.42M 5.08ns 197.04M Skip_ForwardQ128(4_pm_1) 9.67ns 103.41M 7.04ns 142.02M Skip_ForwardQ128(16_pm_4) 21.44ns 46.65M 19.68ns 50.82M Skip_ForwardQ128(64_pm_16) 30.86ns 32.40M 27.58ns 36.26M Skip_ForwardQ128(256_pm_64) 37.80ns 26.45M 32.49ns 30.78M Skip_ForwardQ128(1024_pm_256) 38.99ns 25.65M 33.39ns 29.95M Jump_ForwardQ128 37.91ns 26.37M 34.05ns 29.37M ---------------------------------------------------------------------------- ------------------ SkipTo_SkipQ128(1) 13.87ns 72.10M 4.42ns 226.49M SkipTo_SkipQ128(2) 18.80ns 53.20M 8.58ns 116.55M SkipTo_SkipQ128(4_pm_1) 23.59ns 42.38M 11.43ns 87.50M SkipTo_SkipQ128(16_pm_4) 36.04ns 27.74M 31.19ns 32.06M SkipTo_SkipQ128(64_pm_16) 53.34ns 18.75M 43.88ns 22.79M SkipTo_SkipQ128(256_pm_64) 62.27ns 16.06M 49.08ns 20.37M SkipTo_SkipQ128(1024_pm_256) 65.63ns 15.24M 52.24ns 19.14M JumpTo_SkipQ128 65.89ns 15.18M 54.61ns 18.31M ---------------------------------------------------------------------------- ------------------ Encode_10 111.94ns 8.93M 117.24ns 8.53M Encode 5.35ms 187.02 5.64ms 177.15 ---------------------------------------------------------------------------- ------------------ Select64 8.07ns 123.96M 8.04ns 124.35M ============================================================================ ================== Test Plan: fbconfig folly/experimental/test:eliasfano_test && fbmake runtests_opt Reviewed By: philipp@fb.com Subscribers: yfeldblum, fbcode-common-diffs@, chaoyc, search-fbcode-diffs@, unicorn-diffs@, trunkagent, folly-diffs@ FB internal diff: D1793554 Signature: t1:1793554:1423619344:1b078c0789639f317342ebcc77b11fe91859cd7b
-
Hannes Roth authored
Summary: Saw this while debugging some stuff. Just an oversight, right? Test Plan: Ran all the tests. Reviewed By: hans@fb.com Subscribers: folly-diffs@, jsedgwick, yfeldblum FB internal diff: D1843016 Signature: t1:1843016:1423704557:d9fdb4b879ab4fcfda54a5fcb16a639df15872b5
-
- 12 Feb, 2015 2 commits
-
-
Sara Golemon authored
-
Tom Jackson authored
Summary: That is, for all inputs except r-values. Now with support for GCC. Test Plan: Run tests Reviewed By: ajaymenon@fb.com Subscribers: moharrami, trunkagent, folly-diffs@, yfeldblum FB internal diff: D1822339 Signature: t1:1822339:1423683482:027871549d69017a7a23a675025851a0b876ad77
-
