summaryrefslogtreecommitdiff
path: root/absl/container/btree_benchmark.cc
Commit message (Collapse)AuthorAge
* Export of internal Abseil changesGravatar Abseil Team2021-12-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | -- e7f53dfbf809812e84770217777f81b6308a3084 by Abseil Team <absl-team@google.com>: Add a parameter pack to absl profile to allow profiles to separate dynamic data from static data that is available at constructor-time. Background: `inline_element_size` is effectively constant, but there is a data race between its initialization and its access. We had fixed that race by making inline_element_size atomic. This CL changes `inline_element_size` back to a non-atomic integer, and provides a way for all profiles to provide Register()-time values. PiperOrigin-RevId: 413960559 -- 70234c5943f8e37e17c1d9c54d8ed61d39880abf by Chris Kennelly <ckennelly@google.com>: Document that absl::FunctionRef does not allocate. PiperOrigin-RevId: 413946831 -- 3308ae571412c4be3cc32d088c6edac98ff2d1ed by Samuel Benzaquen <sbenza@google.com>: Internal change PiperOrigin-RevId: 413933619 -- 1617093a730d055edcf7bc04fdd6509783f5f75d by Martijn Vels <mvels@google.com>: Internal Change PiperOrigin-RevId: 413778735 -- 03ad683f059c806a6c8b04f5b79b2662c3df8c73 by Evan Brown <ezb@google.com>: Unify btree erase_if definitions and optimize them so that we only do rebalancing once per leaf node. PiperOrigin-RevId: 413757280 -- 5ba402f70801938178e486617063f01c7862525d by Martijn Vels <mvels@google.com>: Cleanup up cord sampling internals PiperOrigin-RevId: 413755011 -- 522da8f9d3e0f11630d89fb41952004742bc335a by Evan Brown <ezb@google.com>: Add b-tree benchmark for erase_if. Since this benchmark doesn't work for std:: containers before C++20, disable it for them. PiperOrigin-RevId: 413740844 -- a690ea42de8ed4a761d00235d8b2fb7548ba9732 by Andy Getzendanner <durandal@google.com>: Import of CCTZ from GitHub. PiperOrigin-RevId: 413735737 GitOrigin-RevId: e7f53dfbf809812e84770217777f81b6308a3084 Change-Id: I4f9f9039ba92831bc48971964aa063244c9fed72
* Export of internal Abseil changesGravatar Abseil Team2021-03-10
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | -- 5ed5dc9e17c66c298ee31cefc941a46348d8ad34 by Abseil Team <absl-team@google.com>: Fix typo. PiperOrigin-RevId: 362040582 -- ac704b53a49becc42f77e4529d3952f8e7d18ce4 by Abseil Team <absl-team@google.com>: Fix a typo in a comment. PiperOrigin-RevId: 361576641 -- d20ccb27b7e9b53481e9192c1aae5202c06bfcb1 by Derek Mauro <dmauro@google.com>: Remove the inline keyword from functions that aren't defined in the header. This may fix #910. PiperOrigin-RevId: 361551300 -- aed9ae1dffa7b228dcb6ffbeb2fe06a13970c72b by Laramie Leavitt <lar@google.com>: Propagate nice/strict/naggy state on absl::MockingBitGen. Allowing NiceMocks reduces the log spam for un-mocked calls, and it enables nicer setup with ON_CALL, so it is desirable to support it in absl::MockingBitGen. Internally, gmock tracks object "strictness" levels using an internal API; in order to achieve the same results we detect when the MockingBitGen is wrapped in a Nice/Naggy/Strict and wrap the internal implementation MockFunction in the same type. This is achieved by providing overloads to the Call() function, and passing the mock object type down into it's own RegisterMock call, where a compile-time check verifies the state and creates the appropriate mock function. PiperOrigin-RevId: 361233484 -- 96186023fabd13d01d32d60d9c7ac4ead1aeb989 by Abseil Team <absl-team@google.com>: Ensure that trivial types are passed by value rather than reference PiperOrigin-RevId: 361217450 -- e1135944835d27f77e8119b8166d8fb6aa25f906 by Evan Brown <ezb@google.com>: Internal change. PiperOrigin-RevId: 361215882 -- 583fe6c94c1c2ef757ef6e78292a15fbe4030e35 by Evan Brown <ezb@google.com>: Increase the minimum number of slots per node from 3 to 4. We also rename kNodeValues (and related names) to kNodeSlots to make it clear that they are about the number of slots per node rather than the number of values per node - kMinNodeValues keeps the same name because it's actually about the number of values rather than the number of slots. Motivation: I think the expected number of values per node, assuming random insertion order, is the average of the maximum and minimum numbers of values per node (kNodeSlots and kMinNodeValues). For large and/or even kNodeSlots, this is ~75% of kNodeSlots, but for kNodeSlots=3, this is ~67% of kNodeSlots. kMinNodeValues (which corresponds to worst-case occupancy) is ~33% of kNodeSlots, when kNodeSlots=3, compared to 50% for even kNodeSlots. This results in higher memory overhead per value, and since this case (kNodeSlots=3) is used when values are large, it seems worth fixing. PiperOrigin-RevId: 361171495 GitOrigin-RevId: 5ed5dc9e17c66c298ee31cefc941a46348d8ad34 Change-Id: I8e33b5df1f987a77112093821085c410185ab51a
* Export of internal Abseil changesGravatar Abseil Team2021-03-05
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | -- e2de21d54c02b6419c57c0f4e2a16b608deca260 by Evan Brown <ezb@google.com>: Remove the InsertEnd benchmark. This benchmark has significantly different possible behaviors that can result in misleading metrics. Specifically, we can have a case where we are deallocating the last node in the b-tree in the erase and then allocating a new node in the insert call repeatedly, whereas normally, we end up just inserting/erasing a value from the last node. Also, the name of the benchmark is misleading because it involves an erase and an insert, but the name only mentions the insert. PiperOrigin-RevId: 360930639 -- 51f6bb97b9cbdb809c31b77e93ce080ca3cba9ea by Benjamin Barenblat <bbaren@google.com>: Stop testing with double-double random variables On POWER, long double is often represented as a pair of doubles added together (double-double arithmetic). We’ve already special-cased double-double arithmetic in a number of tests, but compiler bugs [1, 2, 3] have now triggered both false positives and false negatives, which suggests testing with double doubles is unlikely to yield useful signal. Remove the special casing and detect if we’re on a double-double system; if so, just don’t test long doubles. [1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99048 [2] https://bugs.llvm.org/show_bug.cgi?id=49131 [3] https://bugs.llvm.org/show_bug.cgi?id=49132 PiperOrigin-RevId: 360793161 -- 07fb4d7932c2f5d711c480f759dacb0be60f975e by Abseil Team <absl-team@google.com>: internal change PiperOrigin-RevId: 360712825 GitOrigin-RevId: e2de21d54c02b6419c57c0f4e2a16b608deca260 Change-Id: I98389b5a8789dcc8f35abc00c767e909181665f0
* Export of internal Abseil changesGravatar Abseil Team2020-03-16
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | -- 91ca367a7548270155721bdda74611aeea2a2153 by Abseil Team <absl-team@google.com>: Replace the only usage of btree_node::swap with simpler logic using transfers and delete btree_node::swap. Add a benchmark for constructing small containers. PiperOrigin-RevId: 301169874 -- ff9d73a7125b7f8ab5733cda877204dfbfac138e by Derek Mauro <dmauro@google.com>: Ensure ABSL_CXX_STANDARD is set. Fixes #640 PiperOrigin-RevId: 301160106 -- 14ca0beee8c109e532134e7e9da7b072da1bf911 by Abseil Team <absl-team@google.com>: Rollback the change to make Cord iterators a fixed size. That change increased the iterator size, which can cause a deep recursion call to hit the stack memory limit, in turn causing a signal 11 failure. PiperOrigin-RevId: 301084915 -- 619e3cd9e56408bdb8b3b5a1e08dda1e95242264 by Matthew Brown <matthewbr@google.com>: Internal Change PiperOrigin-RevId: 300832828 -- 64f8d62ab4c4c78077dbe85a9595a8eeb6d16608 by Gennadiy Rozental <rogeeff@google.com>: Fix for empty braces support. We will call proper aggregate construction in case when {} is used as default value. In other words instead of "new T", we'll call "new T{}". PiperOrigin-RevId: 300715686 -- db3f65594d6db8b104b01262f884dff465b696ef by Abseil Team <absl-team@google.com>: Emscripten supports thread-local storage nowadays. PiperOrigin-RevId: 300675185 GitOrigin-RevId: 91ca367a7548270155721bdda74611aeea2a2153 Change-Id: I3344f745f9c3fc78775532b1808442fabd98e34a
* Export of internal Abseil changesGravatar Abseil Team2020-03-09
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | -- 2c5c118f0615ba90e48ee2f18eccc9f511740f6d by Samuel Benzaquen <sbenza@google.com>: Rename internal macros to follow the convention in absl. PiperOrigin-RevId: 299906738 -- 92d84a707c7ebc4ec19bdd92d5765d1b6d218c1e by Derek Mauro <dmauro@google.com>: Import GitHub #629: Skip the .exe suffix in the helpshort filter on Windows PiperOrigin-RevId: 299892396 -- 2a6910d4be6c67a8376628764121b528ff53504d by Abseil Team <absl-team@google.com>: Use unsigned int128 intrinsic when available. It generates better branchless code. PiperOrigin-RevId: 299848585 -- 110c16cf0a739e1df5028fb6fbd03ef5dde1d278 by Derek Mauro <dmauro@google.com>: Import GitHub #594: Avoid reading the registry for Windows UWP apps PiperOrigin-RevId: 299821671 -- d8397d367e88163e5e8a47f379c716352dc91d03 by Greg Falcon <gfalcon@google.com>: Add absl::Hash support for Cord. The hash function is heterogeneous with other string types: a Cord and a string with the same byte sequence will hash to the same value. SwissTable types know about Cord, and will allow heterogeneous lookup (e.g., you can pass a Cord to flat_hash_map<string, T>::find(), and vice versa.) Add a missing dependency to the cmake Cord target. PiperOrigin-RevId: 299443713 GitOrigin-RevId: 2c5c118f0615ba90e48ee2f18eccc9f511740f6d Change-Id: I7b087c7984b0cb52c4b337d49266c467b98ebdf9
* Export of internal Abseil changesGravatar Abseil Team2020-03-03
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | -- a3e58c1870a9626039f4d178d2d599319bd9f8a8 by Matt Kulukundis <kfm@google.com>: Allow MakeCordFromExternal to take a zero arg releaser. PiperOrigin-RevId: 298650274 -- 01897c4a9bb99f3dc329a794019498ad345ddebd by Samuel Benzaquen <sbenza@google.com>: Reduce library bloat for absl::Flag by moving the definition of base virtual functions to a .cc file. This removes the duplicate symbols in user translation units and has the side effect of moving the vtable definition too (re key function) PiperOrigin-RevId: 298617920 -- 190f0d3782c63aed01046886d7fbc1be5bca2de9 by Derek Mauro <dmauro@google.com>: Import GitHub #596: Unbreak stacktrace code for UWP apps PiperOrigin-RevId: 298600834 -- cd5cf6f8c87b35b85a9584e94da2a99057345b73 by Gennadiy Rozental <rogeeff@google.com>: Use union of heap allocated pointer, one word atomic and two word atomic to represent flags value. Any type T, which is trivially copy-able and with with sizeof(T) <= 8, will be stored in atomic int64_t. Any type T, which is trivially copy-able and with with 8 < sizeof(T) <= 16, will be stored in atomic AlignedTwoWords. We also introducing value storage type to distinguish these cases. PiperOrigin-RevId: 298497200 -- f8fe7bd53bfed601f002f521e34ab4bc083fc28b by Matthew Brown <matthewbr@google.com>: Ensure a deep copy and proper equality on absl::Status::ErasePayload PiperOrigin-RevId: 298482742 -- a5c9ccddf4b04f444e3f7e27dbc14faf1fcb5373 by Gennadiy Rozental <rogeeff@google.com>: Change ChunkIterator implementation to use fixed capacity collection of CordRep*. We can now assume that depth never exceeds 91. That makes comparison operator exception safe. I've tested that with this CL we do not observe an overhead of chunk_end. Compiler optimized this iterator completely. PiperOrigin-RevId: 298458472 -- 327ea5e8910bc388b03389c730763f9823abfce5 by Abseil Team <absl-team@google.com>: Minor cleanups in b-tree code: - Rename some variables: fix issues of different param names between definition/declaration, move away from `x` as a default meaningless variable name. - Make init_leaf/init_internal be non-static methods (they already take the node as the first parameter). - In internal_emplace/try_shrink, update root/rightmost the same way as in insert_unique/insert_multi. - Replace a TODO with a comment. PiperOrigin-RevId: 298432836 -- 8020ce9ec8558ee712d9733ae3d660ac1d3ffe1a by Abseil Team <absl-team@google.com>: Guard against unnecessary copy in case the buffer is empty. This is important in cases were the user is explicitly tuning their chunks to match PiecewiseChunkSize(). PiperOrigin-RevId: 298366044 -- 89324441d1c0c697c90ba7d8fc63639805fcaa9d by Abseil Team <absl-team@google.com>: Internal change PiperOrigin-RevId: 298219363 GitOrigin-RevId: a3e58c1870a9626039f4d178d2d599319bd9f8a8 Change-Id: I28dffc684b6fd0292b94807b88ec6664d5d0e183
* Export of internal Abseil changesGravatar Abseil Team2020-01-21
-- c42a234e2c186bf697ce8d77e85628601fa514a6 by Abseil Team <absl-team@google.com>: Enable the assertion in the iterator's operator++ PiperOrigin-RevId: 290134813 -- f8c53ba8e9c5bb16bbcc1e412a5c2519c912c83e by Abseil Team <absl-team@google.com>: Define operator== and operator!= for absl::{weak,strong}_equality and absl::{partial,weak,strong}_ordering types themselves. PiperOrigin-RevId: 290111564 -- 36bc574090cefad74a451719ce2761982647a51d by Tom Manshreck <shreck@google.com>: Specify Time library flag formats PiperOrigin-RevId: 289928010 -- 26dd40281add260baab2b60fec05dfb9c5304aaa by Mark Barolak <mbar@google.com>: Delete an extraneous forward declaration of absl::Cord. PiperOrigin-RevId: 289708481 -- e60aea7f33554ff66d7699bb70e7af1d26323f1d by Abseil Team <absl-team@google.com>: Release b-tree benchmarks. PiperOrigin-RevId: 289654429 -- 660aa83fa000d4bae072b2d1c790f81d0939bc7e by Greg Falcon <gfalcon@google.com>: Use https links. Import of https://github.com/abseil/abseil-cpp/pull/586 PiperOrigin-RevId: 289479559 -- 0611ea4482dcf23d6b0a0389fe041eeb9052449a by Derek Mauro <dmauro@google.com>: Removes the static initializer for LookupTables<absl::uint128>::kVmaxOverBase Uses template specialization to hard code the resulting array. Static initializers are problematic for a number of reasons. Not only are they responsible for the static initialization order fiasco, but they are in the critical path during program startup. For these reasons, the Google C++ style guide strongly discourages them (and forbids them when they are not trivially destructible), and Chromium even has a test forbidding them. https://google.github.io/styleguide/cppguide.html#Static_and_Global_Variables https://chromium.googlesource.com/chromium/src.git/+/master/docs/static_initializers.md http://neugierig.org/software/chromium/notes/2011/08/static-initializers.html PiperOrigin-RevId: 289458677 -- c869362f6bb7a872314f74750d38d81bdaa73f95 by Greg Falcon <gfalcon@google.com>: Step 2 of 2 to fix our CCTZ fork to respect inline namespaces. Re-import of CCTZ from GitHub, applying new changes to honor Abseil's optional inline namespace in MSVC. PiperOrigin-RevId: 289454407 -- fdb3474d76c2ee0371ccdf7593a78137c03a3f58 by Greg Falcon <gfalcon@google.com>: Step 1 of 2 to fix our CCTZ fork to respect inline namespaces. CCTZ uses a linker flag to simulate weak symbol support in MSVC. This takes the form of a #pragma that includes the mangled names of two types: the symbol to treat as weak, and the symbol to use as its default value if no override is provided. When Abseil is configured to use inline namespaces, the mangled names of these symbols change, and the pragma should change to reflect that. Fortunately for us, MSVC name mangling is simple enough that we can generate the needed string literals in the preprocessor. This CL introduces the new macros; the uses will be introduced in a follow-up CL. PiperOrigin-RevId: 289435599 -- 5f152cc36f008acb9ab78f30b5efa40ebaf2754b by Matt Kulukundis <kfm@google.com>: Improve documentation for lazy_emplace PiperOrigin-RevId: 289333112 GitOrigin-RevId: c42a234e2c186bf697ce8d77e85628601fa514a6 Change-Id: I139ce6c7044a70d083af53e428bcb987f0fd88c6