- 17 Jan, 2018 5 commits
-
-
Stepan Palamarchuk authored
Summary: Currently appending to the tail of the chain would cause the cursor advancing to the beginning of the chain, which is not correct, instead we should advance to the tail. Reviewed By: yfeldblum Differential Revision: D6734999 fbshipit-source-id: b8b2585e0475b656f7b6bf4ed39686e2ccb2e432
-
Petr Lapukhov authored
Summary: Add non-throwing version of createNetwork(), and rework existing code to throw exceptions based on error codes returned by non-throwing version. Reviewed By: yfeldblum Differential Revision: D6705425 fbshipit-source-id: 268ff64c36e7cceeea3463248d18b7b2cb81390e
-
Yedidya Feldblum authored
Summary: [Folly] Fix Build: `folly/synchronization/test/ParkingLotBenchmark.cpp` (`sign-compare`, `unused-variable`). Reviewed By: djwatson Differential Revision: D6733586 fbshipit-source-id: 04b82110e8e0b655e6b3994dbca3bd0b4bd57eef
-
Adam Simpkins authored
Summary: The cmake files refer to this package as "FindGLog", but the file name on disk incorrectly had the 'L' lower cased. This worked on case-insensitive file systems, but caused cmake to fail when used on a case-sensitive file system. This diff fixes the file name to match the name used in the cmake files. This is also more consistent with the existing "FindGFlags" and "FindGMock" packages. Reviewed By: yfeldblum Differential Revision: D6710439 fbshipit-source-id: 19f6824ef6793e505b5a0080c0fefe0b145a41c3
-
Yedidya Feldblum authored
Summary: This reverts commit 9979f39677284b1051cb109b461097495d77ca17 bypass-lint An infra SEV is better than not reverting this diff. If you copy this password, see you in SEV Review! cause_a_sev_many_files Differential Revision: D6725091 fbshipit-source-id: f1e3c80c869aa47684f5fbe79528e68174dee568
-
- 16 Jan, 2018 6 commits
-
-
Dave Watson authored
Summary: Add benchmark vs. futex Reviewed By: yfeldblum Differential Revision: D6639280 fbshipit-source-id: 9735444a7f48011f31603159561675d472cc4411
-
Jim Meyering authored
Summary: Finding ourselves writing too many identical "if (expr) SKIP << ..." statements, this change factors out/encapsulates the functionality. Reviewed By: yfeldblum Differential Revision: D6727531 fbshipit-source-id: 703abcd5d8c30a6ebab94327a12da4f2d1f7ff74
-
Yedidya Feldblum authored
Summary: [Folly] Use thread-local in `RequestContext::getStaticContext`. `folly::SingletonThreadLocal` uses `folly::ThreadLocal`. However, `static FOLLY_TLS` (`static __thread`) is always faster than `folly::ThreadLocal` for thread-local singletons for which iteration is not required. Reviewed By: djwatson Differential Revision: D6725091 fbshipit-source-id: 9979f39677284b1051cb109b461097495d77ca17
-
Yedidya Feldblum authored
Summary: [Folly] Fix missing `WaitOptions` symbol in autotools build. Closes #736. Reviewed By: spalamarchuk Differential Revision: D6724950 fbshipit-source-id: 54ac0d1a743aa38fe1a7a2bdce585fced18c5a44
-
Stepan Palamarchuk authored
Summary: Start tracking the position of the cursor from the head of IOBuf chain. This comes at almost no cost (one arithmetic operation on IOBuf advance). The main use case for this cursor is Thrift deserialization code. It allows us to stop accumulating `xfer` on every single byte/field/element write and instead get it from Cursor in the end (when we're exiting Thrift code). This allows achieving ~10% better performance of deserialization. Reviewed By: yfeldblum Differential Revision: D6646813 fbshipit-source-id: 8f796854a24a411698e96afe037695e816813022
-
Stepan Palamarchuk authored
Summary: This change simplifies the fastpath by reducing it to bare minimum (i.e. check length, load data) and removes indirection to IOBuf. Additionally it adds `skipNoAdvance` method to have 1-instruction skip. Disassembly of `read<signed char>` is over 35 instructions (just hot path). With this change it's doesn to 8. Disassembly after: Dump of assembler code for function folly::io::detail::CursorBase<folly::io::Cursor, folly::IOBuf const>::read<unsigned char>(): 0x000000000041f0f0 <+0>: mov 0x18(%rdi),%rax 0x000000000041f0f4 <+4>: lea 0x1(%rax),%rcx 0x000000000041f0f8 <+8>: cmp 0x10(%rdi),%rcx 0x000000000041f0fc <+12>: ja 0x41f105 <folly::io::detail::CursorBase<folly::io::Cursor, folly::IOBuf const>::read<unsigned char>()+21> 0x000000000041f0fe <+14>: mov (%rax),%al 0x000000000041f100 <+16>: mov %rcx,0x18(%rdi) 0x000000000041f104 <+20>: retq 0x000000000041f105 <+21>: jmpq 0x41f110 <folly::io::detail::CursorBase<folly::io::Cursor, folly::IOBuf const>::readSlow<unsigned char>()> With this diff Thrift deserialization becomes ~20% faster (with prod workloads). Thrift benchmark: Before: ============================================================================ thrift/lib/cpp2/test/ProtocolBench.cpp relative time/iter iters/s ============================================================================ BinaryProtocol_read_Empty 12.98ns 77.03M BinaryProtocol_read_SmallInt 20.94ns 47.76M BinaryProtocol_read_BigInt 20.86ns 47.93M BinaryProtocol_read_SmallString 34.64ns 28.86M BinaryProtocol_read_BigString 185.53ns 5.39M BinaryProtocol_read_BigBinary 67.34ns 14.85M BinaryProtocol_read_LargeBinary 62.23ns 16.07M BinaryProtocol_read_Mixed 58.74ns 17.03M BinaryProtocol_read_SmallListInt 89.99ns 11.11M BinaryProtocol_read_BigListInt 39.92us 25.05K BinaryProtocol_read_BigListMixed 616.20us 1.62K BinaryProtocol_read_LargeListMixed 83.49ms 11.98 CompactProtocol_read_Empty 11.28ns 88.67M CompactProtocol_read_SmallInt 19.15ns 52.22M CompactProtocol_read_BigInt 26.14ns 38.25M CompactProtocol_read_SmallString 31.04ns 32.22M CompactProtocol_read_BigString 184.55ns 5.42M CompactProtocol_read_BigBinary 69.73ns 14.34M CompactProtocol_read_LargeBinary 64.39ns 15.53M CompactProtocol_read_Mixed 58.73ns 17.03M CompactProtocol_read_SmallListInt 76.50ns 13.07M CompactProtocol_read_BigListInt 25.93us 38.56K CompactProtocol_read_BigListMixed 623.15us 1.60K CompactProtocol_read_LargeListMixed 80.57ms 12.41 ============================================================================ After: ============================================================================ thrift/lib/cpp2/test/ProtocolBench.cpp relative time/iter iters/s ============================================================================ BinaryProtocol_read_Empty 10.40ns 96.17M BinaryProtocol_read_SmallInt 15.14ns 66.03M BinaryProtocol_read_BigInt 15.19ns 65.84M BinaryProtocol_read_SmallString 25.19ns 39.70M BinaryProtocol_read_BigString 172.85ns 5.79M BinaryProtocol_read_BigBinary 56.88ns 17.58M BinaryProtocol_read_LargeBinary 56.77ns 17.61M BinaryProtocol_read_Mixed 43.98ns 22.74M BinaryProtocol_read_SmallListInt 58.19ns 17.19M BinaryProtocol_read_BigListInt 19.75us 50.63K BinaryProtocol_read_BigListMixed 440.20us 2.27K BinaryProtocol_read_LargeListMixed 56.94ms 17.56 CompactProtocol_read_Empty 9.35ns 106.93M CompactProtocol_read_SmallInt 13.07ns 76.49M CompactProtocol_read_BigInt 18.23ns 54.87M CompactProtocol_read_SmallString 25.61ns 39.05M CompactProtocol_read_BigString 174.46ns 5.73M CompactProtocol_read_BigBinary 59.77ns 16.73M CompactProtocol_read_LargeBinary 60.81ns 16.44M CompactProtocol_read_Mixed 42.70ns 23.42M CompactProtocol_read_SmallListInt 66.89ns 14.95M CompactProtocol_read_BigListInt 25.08us 39.87K CompactProtocol_read_BigListMixed 427.93us 2.34K CompactProtocol_read_LargeListMixed 56.11ms 17.82 ============================================================================ Reviewed By: yfeldblum Differential Revision: D6635325 fbshipit-source-id: 393fc1005689042977c03f37f5a898ebe7814d44
-
- 15 Jan, 2018 1 commit
-
-
Alvaro Leiva Geisse authored
Summary: Currently NestedCommandLineApp does not support `--` to indicate that the following arguments should be parsed as arguments, not nested commands or options. This diff fix that by whitelisting any argument given after "--" Reviewed By: yfeldblum Differential Revision: D6721144 fbshipit-source-id: 38a850b6ea803dc758c5fe65a21575e5faeac35b
-
- 14 Jan, 2018 2 commits
-
-
Yedidya Feldblum authored
Summary: [Folly] Fix copyright lines based on file histories. Reviewed By: spalamarchuk Differential Revision: D6720312 fbshipit-source-id: c70a667a1977e70e2d4451ea624f96163982f681
-
Yedidya Feldblum authored
Summary: [Folly] Convert newlines in `folly/portability/PThread.cpp`. ``` dos2unix folly/portability/PThread.cpp ``` Reviewed By: meyering Differential Revision: D6720343 fbshipit-source-id: b92122b4a7012d7f8d73d293af51b4fcc868c582
-
- 13 Jan, 2018 6 commits
-
-
David Goldblatt authored
Summary: This adds the Tearable class template, which holds storage for an arbitrarily-sized object that can be concurrently read or written without any external synchronization. Reviewed By: yfeldblum, djwatson Differential Revision: D6422334 fbshipit-source-id: ee3853bbd393ac8e30dca6439c61606cc5495f92
-
Yedidya Feldblum authored
Summary: [Folly] Promote `aligned_malloc` and `aligned_free` from `namespace folly::detail` to `namespace folly`. And move them from `folly/portability/Memory.h` to `folly/Memory.h`. Differential Revision: D6153394 fbshipit-source-id: eef314d2bc171910ea3c8403da9e9e1d1858ce15
-
Adam Simpkins authored
Summary: A number of the files in CMake/ had inconsistent line-endings. This updates files using DOS-style CRLF line endings to just use CR instead. On Windows, git is capable of automatically changing CR to CRLF when checking out the repository working directory. Reviewed By: meyering Differential Revision: D6714717 fbshipit-source-id: 82adccf4e522d38fd1cb420869f62e52dbd6c5f1
-
Adam Simpkins authored
Summary: Add proper checks to detect and set the values in folly-config.h correctly. Previously the code simply hard-coded values that were appropriate for Windows. This does not yet define all of the settings produced by the autoconf build, but it lets the CMake-based build largely succeed on Linux systems. Reviewed By: meyering Differential Revision: D6710437 fbshipit-source-id: 80490080ee6322995b740ac2a15181d220c6874a
-
Adam Simpkins authored
Summary: Update CMakeLists.txt to check the current platform, and to set compiler flags correctly. It now uses flags for Microsoft Visual Studio on Windows, and flags for gcc or clang on all other platforms. Previously it unconditionally used MSVC flags. Reviewed By: meyering Differential Revision: D6710435 fbshipit-source-id: dbae3097bcadf1ee4a25879dd7770603387c0e4d
-
Adam Simpkins authored
Summary: Fix rules in the CMakeLists.txt file that generate source files to first create their output directory if necessary. This allows the build to succeed when building with a separate build output directory, rather than placing build artifacts in the source tree itself. Reviewed By: meyering Differential Revision: D6710436 fbshipit-source-id: 786a65a37a70833e7e4a5affe4df292626dbb591
-
- 12 Jan, 2018 5 commits
-
-
Eric Niebler authored
Summary: Multi-dispatch in Poly was handled by treating arguments of type Poly<I&> as special, unwrapping them. That's a problem if the underlying API actually wants the Poly<I&> unmolested. Reviewed By: yfeldblum Differential Revision: D6713975 fbshipit-source-id: 18a90fa701fab14c3d3d46c247efe09ea5903b11
-
Sergey Makarenko authored
Summary: scheduleDestroyInstances function is called from createInstance function when new instance of specific singleton type is created thus marking a point in static objects order when all singltons will be destructed. This does not work well for situations when singleton is loaded as part of python extension which is a shared object. In this case static objects of this shared object would be constructed first and singleton from this extension will be created after that. Since destruction order is reversed and all singletons will be destructed together, static objects of python extension will be destroyed before singleton from this extension. Which leads to heap-after-free ASAN exceptions. In other words, lifetime of all folly singletons is aligned to the lifetime of the first created folly singleton. Using std::atexit to register singleton's destruction from singleton constructor will align lifetime of singletons to the most recent singleton which matters for python extensions and does not matter for other use cases. Reviewed By: andriigrynenko Differential Revision: D6705644 fbshipit-source-id: 5c933886ceae649e3c75f8e7e7936d5a7ed04539
-
Enji Cooper authored
Summary: While here, pet the linter with respect to the warning around the indentation with the comment containing the licensing tort. Signed-off-by: Enji Cooper <yaneurabeya@gmail.com> Reviewed By: yfeldblum Differential Revision: D6710314 fbshipit-source-id: fc0b971e6300af9c63a690c54c08fc70e0313b70
-
Yedidya Feldblum authored
Summary: [Folly] Unsafe pre-sorted construction for sorted-vector containers. If the backing container type can be constructed directly in sorted order or can be determined in advance to be in sorted order, then a special constructor can help code take advantage of this condition by avoiding an extra invocation of `std::sort`. Reviewed By: spalamarchuk Differential Revision: D6708379 fbshipit-source-id: 25d886b0814dc9230c6046ed1e7f199fac47754e
-
Adam Simpkins authored
Summary: Building folly with cmake is only supported on Windows for now. This fixes cmake on non-Windows platforms to fail with a helpful message telling people to use autoconf. This message was in place before, but was after an MSVC version check preventing it from appearing. Reviewed By: meyering Differential Revision: D6707328 fbshipit-source-id: a28a07ab0da41d605b11d93bba40f33520c5f57e
-
- 11 Jan, 2018 5 commits
-
-
Yedidya Feldblum authored
Summary: [Folly] `MemoryIdler::futexWaitUntil`. Adds `MemoryIdler::futexWaitUntil` that works like `Futex::futexWaitUntil`, in a similar way that `MemoryIdler::futexWait` works like `Futex::futexWait`. Removes the ability to customize the idle-timeout clock for `MemoryIdler::futexWait` as a side-effect; the idle-timeout is now a pure duration. Now, the clock used with the idle-timeout is the same as the normal deadline clock, so the idle-timeout clock can be set for `MemoryIdler::futexWaitUntil` by providing a deadline with that clock type. This normally would not matter, but it affects the unit-tests. Reviewed By: djwatson Differential Revision: D6681679 fbshipit-source-id: e3cf6e71d7530c5877a834b318b423eb91f71eb9
-
Dave Watson authored
Summary: stop_ can be relaxed everywhere, it implies only an asynchronous signal, not any sort of memory barrier. Reviewed By: davidtgoldblatt Differential Revision: D6690079 fbshipit-source-id: 6f193204e1619f9a8adf81db2d46d05954bfbf85
-
Yedidya Feldblum authored
Summary: [Folly] Cut the `ScopeGuard` alias now that we have `auto`. This form works because of hidden lifetime extension: ```lang=c++ folly::ScopeGuard guard = folly::makeGuard([] { /*...*/ }); // ... // guard falls out of scope ``` But this form would not work correctly: ```lang=c++ folly::ScopeGuard guard = folly::makeGuard([] { /*...*/ }); std::async(std::launch::async, [guard = std::move(guard)] {}); ``` Because `folly::ScopeGuard` is an rvalue-reference-to-base. We have `auto`, so just remove `folly::ScopeGuard`. This form works correctly: ```lang=c++ auto guard = folly::makeGuard([] { /*...*/ }); std::async(std::launch::async, [guard = std::move(guard)] {}); ``` Reviewed By: igorsugak Differential Revision: D6690070 fbshipit-source-id: 54e32b300d36fce4eb95a59f1828819afe312ec0
-
Yedidya Feldblum authored
Summary: [Folly] Move `ScopeGuardImpl` and `ScopeGuardImplBase` into the `detail` namespace. Let them be marked as private implementation details. Reviewed By: andrewjcg Differential Revision: D6665317 fbshipit-source-id: 03e8fee6a16338395ec92c582613b053bd9f74ec
-
Yedidya Feldblum authored
Summary: [Folly] Stop using `ScopeGuardImpl` in `DynamicParser`. `ScopeGuardImpl` is an impl type that should not be treated as public. Reviewed By: igorsugak Differential Revision: D6689835 fbshipit-source-id: aea6c985e40887594c0aeb0c0948fa77c149a89b
-
- 10 Jan, 2018 10 commits
-
-
Yedidya Feldblum authored
Summary: [Folly] `folly::Init`, RAII variant of `folly::init`. Use it in `main` used by unit-tests. Reviewed By: ot Differential Revision: D6566358 fbshipit-source-id: fb8e5a18fc43eb65e2cbeb070d97094bd413bb96
-
Maged Michael authored
Summary: Manage retirement of removed segments using hazptr_obj_batch in order to reduce the chances of fragmenting related segments across thread local lists of retired objects of many threads, which could lead to unnecessarily high memory usage. Reviewed By: djwatson Differential Revision: D6686697 fbshipit-source-id: 0d786c0f9e0bac2c44183ed3da21619e1feb3d52
-
Adam Simpkins authored
Summary: Since std::array is actually a struct containing an array it technically requires double braces around its initializer. The language allows eliding these braces, and clang doesn't complain about only using a single brace, but gcc does when using `-Wmissing-braces`. Reviewed By: yfeldblum Differential Revision: D6695289 fbshipit-source-id: 913fcfbea4164a02d001bd2344e340c0b6ee62aa
-
Yedidya Feldblum authored
Summary: [Folly] Let `Futex` import base-class ctors. Rather than needing to define ctors and inherit `boost::noncopyable`. Reviewed By: WillerZ Differential Revision: D6674054 fbshipit-source-id: 59e0a4815682b227346954fe47c6eda49e3ad62f
-
Maged Michael authored
Summary: Provide capability for custom batching retirement of objects, for example object that belong to the same data structure, to avoid the risk fragmenting related objects across thread local lists of retired objects of many threads, which could lead to unnecessarily high memory usage. Reviewed By: djwatson Differential Revision: D6686603 fbshipit-source-id: fadcade73e71170ef1bcec221c4da6f4ddeecff5
-
Yedidya Feldblum authored
Summary: [Folly] Tweak `FutexResult` comments. Reviewed By: nbronson Differential Revision: D6673979 fbshipit-source-id: 1777311cd93d5a83432c4ebb48a8432a1c504ca9
-
Yedidya Feldblum authored
Summary: [Folly] `Futex::futexWait` returns `FutexResult`. Reviewed By: nbronson Differential Revision: D6673871 fbshipit-source-id: 378c69d8362970e985da31e31d8e9b0179d2917f
-
Yedidya Feldblum authored
Summary: [Folly] Support for all clock types in `Futex`. Only `system_clock` and `steady_clock` remain optimal as before, but at least let `Futex` work, even if non-optimally, for all clock types. Reviewed By: nbronson Differential Revision: D6673741 fbshipit-source-id: 0a0f778f61b71bea76e12b7fab478e33ce3bbaae
-
Yedidya Feldblum authored
Summary: [Folly] Use ptr-to-`const` in `Futex`. Reviewed By: igorsugak Differential Revision: D6673723 fbshipit-source-id: b828c2284b40ec8166e823eca7725beccd330f87
-
Yedidya Feldblum authored
Summary: [Folly] Fix some copyright lines in `folly/detail/` and `folly/test/`. Differential Revision: D6673878 fbshipit-source-id: 6c060a974bb4d40d0f24a44ebddddf892805c65e
-