Abseil Common Libraries (C++) (grcp 依赖) https://abseil.io/
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

2192 lines
67 KiB

Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
// Copyright 2018 The Abseil Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// https://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "absl/container/internal/raw_hash_set.h"
#include <atomic>
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#include <cmath>
#include <cstdint>
#include <deque>
#include <functional>
#include <memory>
#include <numeric>
#include <random>
#include <string>
#include <unordered_map>
#include <unordered_set>
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#include "gmock/gmock.h"
#include "gtest/gtest.h"
#include "absl/base/attributes.h"
#include "absl/base/config.h"
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#include "absl/base/internal/cycleclock.h"
#include "absl/base/internal/prefetch.h"
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#include "absl/base/internal/raw_logging.h"
#include "absl/container/internal/container_memory.h"
#include "absl/container/internal/hash_function_defaults.h"
#include "absl/container/internal/hash_policy_testing.h"
#include "absl/container/internal/hashtable_debug.h"
#include "absl/strings/string_view.h"
namespace absl {
ABSL_NAMESPACE_BEGIN
namespace container_internal {
struct RawHashSetTestOnlyAccess {
template <typename C>
static auto GetSlots(const C& c) -> decltype(c.slots_) {
return c.slots_;
}
};
namespace {
using ::testing::ElementsAre;
using ::testing::Eq;
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
using ::testing::Ge;
using ::testing::Lt;
using ::testing::Pair;
using ::testing::UnorderedElementsAre;
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
// Convenience function to static cast to ctrl_t.
ctrl_t CtrlT(int i) { return static_cast<ctrl_t>(i); }
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
TEST(Util, NormalizeCapacity) {
EXPECT_EQ(1, NormalizeCapacity(0));
EXPECT_EQ(1, NormalizeCapacity(1));
EXPECT_EQ(3, NormalizeCapacity(2));
EXPECT_EQ(3, NormalizeCapacity(3));
EXPECT_EQ(7, NormalizeCapacity(4));
EXPECT_EQ(7, NormalizeCapacity(7));
EXPECT_EQ(15, NormalizeCapacity(8));
EXPECT_EQ(15, NormalizeCapacity(15));
EXPECT_EQ(15 * 2 + 1, NormalizeCapacity(15 + 1));
EXPECT_EQ(15 * 2 + 1, NormalizeCapacity(15 + 2));
}
TEST(Util, GrowthAndCapacity) {
// Verify that GrowthToCapacity gives the minimum capacity that has enough
// growth.
for (size_t growth = 0; growth < 10000; ++growth) {
SCOPED_TRACE(growth);
size_t capacity = NormalizeCapacity(GrowthToLowerboundCapacity(growth));
// The capacity is large enough for `growth`.
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
EXPECT_THAT(CapacityToGrowth(capacity), Ge(growth));
// For (capacity+1) < kWidth, growth should equal capacity.
if (capacity + 1 < Group::kWidth) {
EXPECT_THAT(CapacityToGrowth(capacity), Eq(capacity));
} else {
EXPECT_THAT(CapacityToGrowth(capacity), Lt(capacity));
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
if (growth != 0 && capacity > 1) {
// There is no smaller capacity that works.
EXPECT_THAT(CapacityToGrowth(capacity / 2), Lt(growth));
}
}
for (size_t capacity = Group::kWidth - 1; capacity < 10000;
capacity = 2 * capacity + 1) {
SCOPED_TRACE(capacity);
size_t growth = CapacityToGrowth(capacity);
EXPECT_THAT(growth, Lt(capacity));
EXPECT_LE(GrowthToLowerboundCapacity(growth), capacity);
EXPECT_EQ(NormalizeCapacity(GrowthToLowerboundCapacity(growth)), capacity);
}
}
TEST(Util, probe_seq) {
probe_seq<16> seq(0, 127);
auto gen = [&]() {
size_t res = seq.offset();
seq.next();
return res;
};
std::vector<size_t> offsets(8);
std::generate_n(offsets.begin(), 8, gen);
EXPECT_THAT(offsets, ElementsAre(0, 16, 48, 96, 32, 112, 80, 64));
seq = probe_seq<16>(128, 127);
std::generate_n(offsets.begin(), 8, gen);
EXPECT_THAT(offsets, ElementsAre(0, 16, 48, 96, 32, 112, 80, 64));
}
TEST(BitMask, Smoke) {
EXPECT_FALSE((BitMask<uint8_t, 8>(0)));
EXPECT_TRUE((BitMask<uint8_t, 8>(5)));
EXPECT_THAT((BitMask<uint8_t, 8>(0)), ElementsAre());
EXPECT_THAT((BitMask<uint8_t, 8>(0x1)), ElementsAre(0));
EXPECT_THAT((BitMask<uint8_t, 8>(0x2)), ElementsAre(1));
EXPECT_THAT((BitMask<uint8_t, 8>(0x3)), ElementsAre(0, 1));
EXPECT_THAT((BitMask<uint8_t, 8>(0x4)), ElementsAre(2));
EXPECT_THAT((BitMask<uint8_t, 8>(0x5)), ElementsAre(0, 2));
EXPECT_THAT((BitMask<uint8_t, 8>(0x55)), ElementsAre(0, 2, 4, 6));
EXPECT_THAT((BitMask<uint8_t, 8>(0xAA)), ElementsAre(1, 3, 5, 7));
}
TEST(BitMask, WithShift) {
// See the non-SSE version of Group for details on what this math is for.
uint64_t ctrl = 0x1716151413121110;
uint64_t hash = 0x12;
constexpr uint64_t msbs = 0x8080808080808080ULL;
constexpr uint64_t lsbs = 0x0101010101010101ULL;
auto x = ctrl ^ (lsbs * hash);
uint64_t mask = (x - lsbs) & ~x & msbs;
EXPECT_EQ(0x0000000080800000, mask);
BitMask<uint64_t, 8, 3> b(mask);
EXPECT_EQ(*b, 2);
}
TEST(BitMask, LeadingTrailing) {
EXPECT_EQ((BitMask<uint32_t, 16>(0x00001a40).LeadingZeros()), 3);
EXPECT_EQ((BitMask<uint32_t, 16>(0x00001a40).TrailingZeros()), 6);
EXPECT_EQ((BitMask<uint32_t, 16>(0x00000001).LeadingZeros()), 15);
EXPECT_EQ((BitMask<uint32_t, 16>(0x00000001).TrailingZeros()), 0);
EXPECT_EQ((BitMask<uint32_t, 16>(0x00008000).LeadingZeros()), 0);
EXPECT_EQ((BitMask<uint32_t, 16>(0x00008000).TrailingZeros()), 15);
EXPECT_EQ((BitMask<uint64_t, 8, 3>(0x0000008080808000).LeadingZeros()), 3);
EXPECT_EQ((BitMask<uint64_t, 8, 3>(0x0000008080808000).TrailingZeros()), 1);
EXPECT_EQ((BitMask<uint64_t, 8, 3>(0x0000000000000080).LeadingZeros()), 7);
EXPECT_EQ((BitMask<uint64_t, 8, 3>(0x0000000000000080).TrailingZeros()), 0);
EXPECT_EQ((BitMask<uint64_t, 8, 3>(0x8000000000000000).LeadingZeros()), 0);
EXPECT_EQ((BitMask<uint64_t, 8, 3>(0x8000000000000000).TrailingZeros()), 7);
}
TEST(Group, EmptyGroup) {
for (h2_t h = 0; h != 128; ++h) EXPECT_FALSE(Group{EmptyGroup()}.Match(h));
}
TEST(Group, Match) {
if (Group::kWidth == 16) {
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
ctrl_t group[] = {ctrl_t::kEmpty, CtrlT(1), ctrl_t::kDeleted, CtrlT(3),
ctrl_t::kEmpty, CtrlT(5), ctrl_t::kSentinel, CtrlT(7),
CtrlT(7), CtrlT(5), CtrlT(3), CtrlT(1),
CtrlT(1), CtrlT(1), CtrlT(1), CtrlT(1)};
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
EXPECT_THAT(Group{group}.Match(0), ElementsAre());
EXPECT_THAT(Group{group}.Match(1), ElementsAre(1, 11, 12, 13, 14, 15));
EXPECT_THAT(Group{group}.Match(3), ElementsAre(3, 10));
EXPECT_THAT(Group{group}.Match(5), ElementsAre(5, 9));
EXPECT_THAT(Group{group}.Match(7), ElementsAre(7, 8));
} else if (Group::kWidth == 8) {
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
ctrl_t group[] = {ctrl_t::kEmpty, CtrlT(1), CtrlT(2),
ctrl_t::kDeleted, CtrlT(2), CtrlT(1),
ctrl_t::kSentinel, CtrlT(1)};
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
EXPECT_THAT(Group{group}.Match(0), ElementsAre());
EXPECT_THAT(Group{group}.Match(1), ElementsAre(1, 5, 7));
EXPECT_THAT(Group{group}.Match(2), ElementsAre(2, 4));
} else {
FAIL() << "No test coverage for Group::kWidth==" << Group::kWidth;
}
}
TEST(Group, MaskEmpty) {
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
if (Group::kWidth == 16) {
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
ctrl_t group[] = {ctrl_t::kEmpty, CtrlT(1), ctrl_t::kDeleted, CtrlT(3),
ctrl_t::kEmpty, CtrlT(5), ctrl_t::kSentinel, CtrlT(7),
CtrlT(7), CtrlT(5), CtrlT(3), CtrlT(1),
CtrlT(1), CtrlT(1), CtrlT(1), CtrlT(1)};
EXPECT_THAT(Group{group}.MaskEmpty().LowestBitSet(), 0);
EXPECT_THAT(Group{group}.MaskEmpty().HighestBitSet(), 4);
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
} else if (Group::kWidth == 8) {
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
ctrl_t group[] = {ctrl_t::kEmpty, CtrlT(1), CtrlT(2),
ctrl_t::kDeleted, CtrlT(2), CtrlT(1),
ctrl_t::kSentinel, CtrlT(1)};
EXPECT_THAT(Group{group}.MaskEmpty().LowestBitSet(), 0);
EXPECT_THAT(Group{group}.MaskEmpty().HighestBitSet(), 0);
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
} else {
FAIL() << "No test coverage for Group::kWidth==" << Group::kWidth;
}
}
TEST(Group, MaskEmptyOrDeleted) {
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
if (Group::kWidth == 16) {
ctrl_t group[] = {ctrl_t::kEmpty, CtrlT(1), ctrl_t::kEmpty, CtrlT(3),
ctrl_t::kDeleted, CtrlT(5), ctrl_t::kSentinel, CtrlT(7),
CtrlT(7), CtrlT(5), CtrlT(3), CtrlT(1),
CtrlT(1), CtrlT(1), CtrlT(1), CtrlT(1)};
EXPECT_THAT(Group{group}.MaskEmptyOrDeleted().LowestBitSet(), 0);
EXPECT_THAT(Group{group}.MaskEmptyOrDeleted().HighestBitSet(), 4);
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
} else if (Group::kWidth == 8) {
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
ctrl_t group[] = {ctrl_t::kEmpty, CtrlT(1), CtrlT(2),
ctrl_t::kDeleted, CtrlT(2), CtrlT(1),
ctrl_t::kSentinel, CtrlT(1)};
EXPECT_THAT(Group{group}.MaskEmptyOrDeleted().LowestBitSet(), 0);
EXPECT_THAT(Group{group}.MaskEmptyOrDeleted().HighestBitSet(), 3);
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
} else {
FAIL() << "No test coverage for Group::kWidth==" << Group::kWidth;
}
}
TEST(Batch, DropDeletes) {
constexpr size_t kCapacity = 63;
constexpr size_t kGroupWidth = container_internal::Group::kWidth;
std::vector<ctrl_t> ctrl(kCapacity + 1 + kGroupWidth);
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
ctrl[kCapacity] = ctrl_t::kSentinel;
std::vector<ctrl_t> pattern = {
ctrl_t::kEmpty, CtrlT(2), ctrl_t::kDeleted, CtrlT(2),
ctrl_t::kEmpty, CtrlT(1), ctrl_t::kDeleted};
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
for (size_t i = 0; i != kCapacity; ++i) {
ctrl[i] = pattern[i % pattern.size()];
if (i < kGroupWidth - 1)
ctrl[i + kCapacity + 1] = pattern[i % pattern.size()];
}
ConvertDeletedToEmptyAndFullToDeleted(ctrl.data(), kCapacity);
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
ASSERT_EQ(ctrl[kCapacity], ctrl_t::kSentinel);
Export of internal Abseil changes -- b1fc72630aaa81c8395c3b22ba267d938fe29a2e by Derek Mauro <dmauro@google.com>: Fix -Wdeprecated-copy warnings from Clang 13. Example: error: definition of implicit copy assignment operator for 'UDT' is deprecated because it has a user-declared copy constructor [-Werror,-Wdeprecated-copy] PiperOrigin-RevId: 380058303 -- 0422744812b1a2010d9eea5b17fbe89f3441b66b by Evan Brown <ezb@google.com>: Change the "full table!" asserts in raw_hash_set to use `<= capacity` instead of `< capacity`. If we add support for non-power-of-two-minus-one capacities, this is the correct thing to assert. For example, consider: Group::kWidth = 8, capacity_ = 8, ctrl_ = {kEmpty, 1, 2, 3, 4, 5, 6, 7, kSentinel, kEmpty, 1, 2, 3, 4, 5, 6}. In this case, if we do an unsuccessful lookup with H2 mapping to slot 1, then the first Group will contain {1, 2, 3, 4, 5, 6, 7, kSentinel} so we need to continue to the second Group (at which point seq.index() == 8 == capacity_) to find a kEmpty. Note: this is a no-op change for now since we never have `capacity % Group::kWidth == 0`. PiperOrigin-RevId: 380033480 -- 40628c34d540356de65fabb16c1439c0ec7a0764 by Abseil Team <absl-team@google.com>: Drop out-of-date documentation about `absl::FixedArray`'s allocator support PiperOrigin-RevId: 379811653 -- e7ad047863ae55c9b7aec0753cfc527a4ea614bc by Evan Brown <ezb@google.com>: Fix a bug in ConvertDeletedToEmptyAndFullToDeleted in which we were copying 1 more cloned control byte than actually exists. When alignof(slot_type)>1, this wouldn't cause a problem because the extra byte is padding. Also change loop bounds to not rely on the fact that capacity_+1 is a multiple of Group::kWidth. PiperOrigin-RevId: 379311830 -- 1a3ba500fb2c33205854eb9258cd6e0fb1061bca by Martijn Vels <mvels@google.com>: Change Ring, EXTERNAL and FLAT tag values to be consecutive values The purpose of this change is to have FLAT = EXTERNAL + 1. Especially in the ring and btree alternative code, there is a common check if a node is a 'plain' edge (EXTERNAL or FLAT), or 'something else'. This change can make that check a single branch, i.e., instead of 'tag == EXTERNAL || tag >= FLAT', we can simply check for 'tag >= EXTERNAL'. Likewise we have some cases where we check for RING, EXTERNAL or FLAT, so we align RING + 1 with EXTERNAL. PiperOrigin-RevId: 379291576 -- 0c78e65ca4d85244b106c3f8e24cf268e09e72a3 by Benjamin Barenblat <bbaren@google.com>: Round a double multiplication before casting it to integer The code static_cast<int>(x * y) (for double x and y) performs a double multiplication into a temporary that, by standard, may have excess precision. The subsequent cast to int discards the excess precision. However, the cast may examine the excess precision during conversion, producing surprising results like static_cast<int>(1.7 * 10) == 16 on certain systems. Correct this case by explicitly rounding 1.7 * 10 before casting it. PiperOrigin-RevId: 378922064 GitOrigin-RevId: b1fc72630aaa81c8395c3b22ba267d938fe29a2e Change-Id: Ica708a006921118673e78d5fd2d61fe0fb0894d1
4 years ago
for (size_t i = 0; i < kCapacity + kGroupWidth; ++i) {
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
ctrl_t expected = pattern[i % (kCapacity + 1) % pattern.size()];
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
if (i == kCapacity) expected = ctrl_t::kSentinel;
if (expected == ctrl_t::kDeleted) expected = ctrl_t::kEmpty;
if (IsFull(expected)) expected = ctrl_t::kDeleted;
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
EXPECT_EQ(ctrl[i], expected)
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
<< i << " " << static_cast<int>(pattern[i % pattern.size()]);
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
}
}
TEST(Group, CountLeadingEmptyOrDeleted) {
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
const std::vector<ctrl_t> empty_examples = {ctrl_t::kEmpty, ctrl_t::kDeleted};
const std::vector<ctrl_t> full_examples = {
CtrlT(0), CtrlT(1), CtrlT(2), CtrlT(3),
CtrlT(5), CtrlT(9), CtrlT(127), ctrl_t::kSentinel};
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
for (ctrl_t empty : empty_examples) {
std::vector<ctrl_t> e(Group::kWidth, empty);
Optimize SwissMap iteration for aarch64 by 5-6% Benchmarks: https://pastebin.com/tZ7dr67W. Works well especially on smaller ranges. After a week on spending optimizing NEON SIMD where I almost managed to make hash tables work with NEON SIMD without performance hits (still 1 cycle to optimize and I gave up a little), I found an interesting optimization for aarch64 to use cls instruction (count leading sign bits). The loop has a property that ctrl_ group is not matched against count when the first slot is empty or deleted. ``` void skip_empty_or_deleted() { while (IsEmptyOrDeleted(*ctrl_)) { uint32_t shift = Group{ctrl_}.CountLeadingEmptyOrDeleted(); ctrl_ += shift; slot_ += shift; } ... } ``` However, `kEmpty` and `kDeleted` have format of `1xxxxxx0` and `~ctrl & (ctrl >> 7)` always sets the lowest bit to 1. In naive implementation, it does +1 to start counting zero bits, however, in aarch64 we may start counting one bits immediately. This saves 1 cycle and 5% of iteration performance. Then it becomes hard to find a supported and sustainable C++ version of it. `__clsll` is not supported by GCC and was supported only since clang 8, `__builtin_clrsb` is not producing optimal codegen for clang. `__rbit` is not supported by GCC and there is no intrinsic to do that, however, in clang we have `__builtin_bitreverse{32,64}`. For now I decided to enable this only for clang, only if they have appropriate builtins. PiperOrigin-RevId: 451168570 Change-Id: I7e9256a60aecdc88ced4e6eb15ebc257281b6664
3 years ago
EXPECT_TRUE(IsEmptyOrDeleted(e[0]));
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
EXPECT_EQ(Group::kWidth, Group{e.data()}.CountLeadingEmptyOrDeleted());
for (ctrl_t full : full_examples) {
Optimize SwissMap iteration for aarch64 by 5-6% Benchmarks: https://pastebin.com/tZ7dr67W. Works well especially on smaller ranges. After a week on spending optimizing NEON SIMD where I almost managed to make hash tables work with NEON SIMD without performance hits (still 1 cycle to optimize and I gave up a little), I found an interesting optimization for aarch64 to use cls instruction (count leading sign bits). The loop has a property that ctrl_ group is not matched against count when the first slot is empty or deleted. ``` void skip_empty_or_deleted() { while (IsEmptyOrDeleted(*ctrl_)) { uint32_t shift = Group{ctrl_}.CountLeadingEmptyOrDeleted(); ctrl_ += shift; slot_ += shift; } ... } ``` However, `kEmpty` and `kDeleted` have format of `1xxxxxx0` and `~ctrl & (ctrl >> 7)` always sets the lowest bit to 1. In naive implementation, it does +1 to start counting zero bits, however, in aarch64 we may start counting one bits immediately. This saves 1 cycle and 5% of iteration performance. Then it becomes hard to find a supported and sustainable C++ version of it. `__clsll` is not supported by GCC and was supported only since clang 8, `__builtin_clrsb` is not producing optimal codegen for clang. `__rbit` is not supported by GCC and there is no intrinsic to do that, however, in clang we have `__builtin_bitreverse{32,64}`. For now I decided to enable this only for clang, only if they have appropriate builtins. PiperOrigin-RevId: 451168570 Change-Id: I7e9256a60aecdc88ced4e6eb15ebc257281b6664
3 years ago
// First is always kEmpty or kDeleted.
for (size_t i = 1; i != Group::kWidth; ++i) {
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
std::vector<ctrl_t> f(Group::kWidth, empty);
f[i] = full;
Optimize SwissMap iteration for aarch64 by 5-6% Benchmarks: https://pastebin.com/tZ7dr67W. Works well especially on smaller ranges. After a week on spending optimizing NEON SIMD where I almost managed to make hash tables work with NEON SIMD without performance hits (still 1 cycle to optimize and I gave up a little), I found an interesting optimization for aarch64 to use cls instruction (count leading sign bits). The loop has a property that ctrl_ group is not matched against count when the first slot is empty or deleted. ``` void skip_empty_or_deleted() { while (IsEmptyOrDeleted(*ctrl_)) { uint32_t shift = Group{ctrl_}.CountLeadingEmptyOrDeleted(); ctrl_ += shift; slot_ += shift; } ... } ``` However, `kEmpty` and `kDeleted` have format of `1xxxxxx0` and `~ctrl & (ctrl >> 7)` always sets the lowest bit to 1. In naive implementation, it does +1 to start counting zero bits, however, in aarch64 we may start counting one bits immediately. This saves 1 cycle and 5% of iteration performance. Then it becomes hard to find a supported and sustainable C++ version of it. `__clsll` is not supported by GCC and was supported only since clang 8, `__builtin_clrsb` is not producing optimal codegen for clang. `__rbit` is not supported by GCC and there is no intrinsic to do that, however, in clang we have `__builtin_bitreverse{32,64}`. For now I decided to enable this only for clang, only if they have appropriate builtins. PiperOrigin-RevId: 451168570 Change-Id: I7e9256a60aecdc88ced4e6eb15ebc257281b6664
3 years ago
EXPECT_TRUE(IsEmptyOrDeleted(f[0]));
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
EXPECT_EQ(i, Group{f.data()}.CountLeadingEmptyOrDeleted());
}
std::vector<ctrl_t> f(Group::kWidth, empty);
f[Group::kWidth * 2 / 3] = full;
f[Group::kWidth / 2] = full;
Optimize SwissMap iteration for aarch64 by 5-6% Benchmarks: https://pastebin.com/tZ7dr67W. Works well especially on smaller ranges. After a week on spending optimizing NEON SIMD where I almost managed to make hash tables work with NEON SIMD without performance hits (still 1 cycle to optimize and I gave up a little), I found an interesting optimization for aarch64 to use cls instruction (count leading sign bits). The loop has a property that ctrl_ group is not matched against count when the first slot is empty or deleted. ``` void skip_empty_or_deleted() { while (IsEmptyOrDeleted(*ctrl_)) { uint32_t shift = Group{ctrl_}.CountLeadingEmptyOrDeleted(); ctrl_ += shift; slot_ += shift; } ... } ``` However, `kEmpty` and `kDeleted` have format of `1xxxxxx0` and `~ctrl & (ctrl >> 7)` always sets the lowest bit to 1. In naive implementation, it does +1 to start counting zero bits, however, in aarch64 we may start counting one bits immediately. This saves 1 cycle and 5% of iteration performance. Then it becomes hard to find a supported and sustainable C++ version of it. `__clsll` is not supported by GCC and was supported only since clang 8, `__builtin_clrsb` is not producing optimal codegen for clang. `__rbit` is not supported by GCC and there is no intrinsic to do that, however, in clang we have `__builtin_bitreverse{32,64}`. For now I decided to enable this only for clang, only if they have appropriate builtins. PiperOrigin-RevId: 451168570 Change-Id: I7e9256a60aecdc88ced4e6eb15ebc257281b6664
3 years ago
EXPECT_TRUE(IsEmptyOrDeleted(f[0]));
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
EXPECT_EQ(
Group::kWidth / 2, Group{f.data()}.CountLeadingEmptyOrDeleted());
}
}
}
Export of internal Abseil changes -- 6e9f93888bbe6718997ed90bbd049f1f3572b066 by Abseil Team <absl-team@google.com>: Fix status to be safe for self move assignment. PiperOrigin-RevId: 346815392 -- 35cae74a977f258e81dfbe925fb5a34cb6632036 by Gennadiy Rozental <rogeeff@google.com>: Eliminate unnecessary access to the global vars. PiperOrigin-RevId: 346777168 -- e7e786c243069060f5d6c1c945decb4b0b83f95b by Andy Getzendanner <durandal@google.com>: Internal change. PiperOrigin-RevId: 346685656 -- 4ccd41c48f1a83cfa20b3ea534f743dd7d788376 by Abseil Team <absl-team@google.com>: Move CordRep Ref() and Unref() logic into cord_internal.h This change moves Ref() and Unref() logic out of cord.cc into cord_internal. The main purpose is to make upcoming ring buffer changes easier to isolate from existing cord.cc code. Notice that this removes the nullptr check from Unref() and now requires it to be non null (which held true most times). We may need to rethink if the 'unref unlikely one' is the common case: are cordreps most likely shared? This may be something between 'root' and non root nodes, i.e., is it more likely for leaf / flat nodes in large cords to be shared than top level cordreps being shared? Vice versa? Benchmarks say that we mostly shouldn't care, the caveat being that atomic ops seem more expensive on upcoming archs (arcadia) so we should error on the side of an extra IsOne() branch saving us single owned atomic ops. PiperOrigin-RevId: 346676121 -- f0babab103b9e60d61ba09482d468985e43eceb3 by Samuel Benzaquen <sbenza@google.com>: Fix iterator based constructor and `.insert` members to only require EmplaceConstructible as the standard specifies. PiperOrigin-RevId: 346616707 -- 8f48eedda02277f9c96a88ed7726e34b557cce20 by Evan Brown <ezb@google.com>: Fix a bug in binary_search_impl when there's a transparent, three-way comparator that has different equivalence classes for different lookup types. Add a new can_have_multiple_equivalent_keys method to share the common logic for these cases. PiperOrigin-RevId: 346605948 -- 649183cb3cc9383431de9c81fb1c0f885d4001ae by Abseil Team <absl-team@google.com>: Add benchmark for accessing a Duration flag. PiperOrigin-RevId: 346594843 -- fefdb046520871af63ce2229e2f7cccfc0483dea by Abseil Team <absl-team@google.com>: Restructure CordReader for upcoming ring buffer changes. PiperOrigin-RevId: 346410642 -- 8b2f50e7da0ebab06ead5f94e366e984ca23cb6a by Abseil Team <absl-team@google.com>: Wire in an internal-only flag to toggle upcoming ring buffer changes on/off for experimentation. PiperOrigin-RevId: 346199111 GitOrigin-RevId: 6e9f93888bbe6718997ed90bbd049f1f3572b066 Change-Id: I8f34866b25a79209cb5448bbb28dd3044111d2e9
4 years ago
template <class T>
struct ValuePolicy {
using slot_type = T;
using key_type = T;
using init_type = T;
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
Export of internal Abseil changes -- 6e9f93888bbe6718997ed90bbd049f1f3572b066 by Abseil Team <absl-team@google.com>: Fix status to be safe for self move assignment. PiperOrigin-RevId: 346815392 -- 35cae74a977f258e81dfbe925fb5a34cb6632036 by Gennadiy Rozental <rogeeff@google.com>: Eliminate unnecessary access to the global vars. PiperOrigin-RevId: 346777168 -- e7e786c243069060f5d6c1c945decb4b0b83f95b by Andy Getzendanner <durandal@google.com>: Internal change. PiperOrigin-RevId: 346685656 -- 4ccd41c48f1a83cfa20b3ea534f743dd7d788376 by Abseil Team <absl-team@google.com>: Move CordRep Ref() and Unref() logic into cord_internal.h This change moves Ref() and Unref() logic out of cord.cc into cord_internal. The main purpose is to make upcoming ring buffer changes easier to isolate from existing cord.cc code. Notice that this removes the nullptr check from Unref() and now requires it to be non null (which held true most times). We may need to rethink if the 'unref unlikely one' is the common case: are cordreps most likely shared? This may be something between 'root' and non root nodes, i.e., is it more likely for leaf / flat nodes in large cords to be shared than top level cordreps being shared? Vice versa? Benchmarks say that we mostly shouldn't care, the caveat being that atomic ops seem more expensive on upcoming archs (arcadia) so we should error on the side of an extra IsOne() branch saving us single owned atomic ops. PiperOrigin-RevId: 346676121 -- f0babab103b9e60d61ba09482d468985e43eceb3 by Samuel Benzaquen <sbenza@google.com>: Fix iterator based constructor and `.insert` members to only require EmplaceConstructible as the standard specifies. PiperOrigin-RevId: 346616707 -- 8f48eedda02277f9c96a88ed7726e34b557cce20 by Evan Brown <ezb@google.com>: Fix a bug in binary_search_impl when there's a transparent, three-way comparator that has different equivalence classes for different lookup types. Add a new can_have_multiple_equivalent_keys method to share the common logic for these cases. PiperOrigin-RevId: 346605948 -- 649183cb3cc9383431de9c81fb1c0f885d4001ae by Abseil Team <absl-team@google.com>: Add benchmark for accessing a Duration flag. PiperOrigin-RevId: 346594843 -- fefdb046520871af63ce2229e2f7cccfc0483dea by Abseil Team <absl-team@google.com>: Restructure CordReader for upcoming ring buffer changes. PiperOrigin-RevId: 346410642 -- 8b2f50e7da0ebab06ead5f94e366e984ca23cb6a by Abseil Team <absl-team@google.com>: Wire in an internal-only flag to toggle upcoming ring buffer changes on/off for experimentation. PiperOrigin-RevId: 346199111 GitOrigin-RevId: 6e9f93888bbe6718997ed90bbd049f1f3572b066 Change-Id: I8f34866b25a79209cb5448bbb28dd3044111d2e9
4 years ago
template <class Allocator, class... Args>
static void construct(Allocator* alloc, slot_type* slot, Args&&... args) {
absl::allocator_traits<Allocator>::construct(*alloc, slot,
std::forward<Args>(args)...);
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
}
Export of internal Abseil changes -- 6e9f93888bbe6718997ed90bbd049f1f3572b066 by Abseil Team <absl-team@google.com>: Fix status to be safe for self move assignment. PiperOrigin-RevId: 346815392 -- 35cae74a977f258e81dfbe925fb5a34cb6632036 by Gennadiy Rozental <rogeeff@google.com>: Eliminate unnecessary access to the global vars. PiperOrigin-RevId: 346777168 -- e7e786c243069060f5d6c1c945decb4b0b83f95b by Andy Getzendanner <durandal@google.com>: Internal change. PiperOrigin-RevId: 346685656 -- 4ccd41c48f1a83cfa20b3ea534f743dd7d788376 by Abseil Team <absl-team@google.com>: Move CordRep Ref() and Unref() logic into cord_internal.h This change moves Ref() and Unref() logic out of cord.cc into cord_internal. The main purpose is to make upcoming ring buffer changes easier to isolate from existing cord.cc code. Notice that this removes the nullptr check from Unref() and now requires it to be non null (which held true most times). We may need to rethink if the 'unref unlikely one' is the common case: are cordreps most likely shared? This may be something between 'root' and non root nodes, i.e., is it more likely for leaf / flat nodes in large cords to be shared than top level cordreps being shared? Vice versa? Benchmarks say that we mostly shouldn't care, the caveat being that atomic ops seem more expensive on upcoming archs (arcadia) so we should error on the side of an extra IsOne() branch saving us single owned atomic ops. PiperOrigin-RevId: 346676121 -- f0babab103b9e60d61ba09482d468985e43eceb3 by Samuel Benzaquen <sbenza@google.com>: Fix iterator based constructor and `.insert` members to only require EmplaceConstructible as the standard specifies. PiperOrigin-RevId: 346616707 -- 8f48eedda02277f9c96a88ed7726e34b557cce20 by Evan Brown <ezb@google.com>: Fix a bug in binary_search_impl when there's a transparent, three-way comparator that has different equivalence classes for different lookup types. Add a new can_have_multiple_equivalent_keys method to share the common logic for these cases. PiperOrigin-RevId: 346605948 -- 649183cb3cc9383431de9c81fb1c0f885d4001ae by Abseil Team <absl-team@google.com>: Add benchmark for accessing a Duration flag. PiperOrigin-RevId: 346594843 -- fefdb046520871af63ce2229e2f7cccfc0483dea by Abseil Team <absl-team@google.com>: Restructure CordReader for upcoming ring buffer changes. PiperOrigin-RevId: 346410642 -- 8b2f50e7da0ebab06ead5f94e366e984ca23cb6a by Abseil Team <absl-team@google.com>: Wire in an internal-only flag to toggle upcoming ring buffer changes on/off for experimentation. PiperOrigin-RevId: 346199111 GitOrigin-RevId: 6e9f93888bbe6718997ed90bbd049f1f3572b066 Change-Id: I8f34866b25a79209cb5448bbb28dd3044111d2e9
4 years ago
template <class Allocator>
static void destroy(Allocator* alloc, slot_type* slot) {
absl::allocator_traits<Allocator>::destroy(*alloc, slot);
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
Export of internal Abseil changes -- 6e9f93888bbe6718997ed90bbd049f1f3572b066 by Abseil Team <absl-team@google.com>: Fix status to be safe for self move assignment. PiperOrigin-RevId: 346815392 -- 35cae74a977f258e81dfbe925fb5a34cb6632036 by Gennadiy Rozental <rogeeff@google.com>: Eliminate unnecessary access to the global vars. PiperOrigin-RevId: 346777168 -- e7e786c243069060f5d6c1c945decb4b0b83f95b by Andy Getzendanner <durandal@google.com>: Internal change. PiperOrigin-RevId: 346685656 -- 4ccd41c48f1a83cfa20b3ea534f743dd7d788376 by Abseil Team <absl-team@google.com>: Move CordRep Ref() and Unref() logic into cord_internal.h This change moves Ref() and Unref() logic out of cord.cc into cord_internal. The main purpose is to make upcoming ring buffer changes easier to isolate from existing cord.cc code. Notice that this removes the nullptr check from Unref() and now requires it to be non null (which held true most times). We may need to rethink if the 'unref unlikely one' is the common case: are cordreps most likely shared? This may be something between 'root' and non root nodes, i.e., is it more likely for leaf / flat nodes in large cords to be shared than top level cordreps being shared? Vice versa? Benchmarks say that we mostly shouldn't care, the caveat being that atomic ops seem more expensive on upcoming archs (arcadia) so we should error on the side of an extra IsOne() branch saving us single owned atomic ops. PiperOrigin-RevId: 346676121 -- f0babab103b9e60d61ba09482d468985e43eceb3 by Samuel Benzaquen <sbenza@google.com>: Fix iterator based constructor and `.insert` members to only require EmplaceConstructible as the standard specifies. PiperOrigin-RevId: 346616707 -- 8f48eedda02277f9c96a88ed7726e34b557cce20 by Evan Brown <ezb@google.com>: Fix a bug in binary_search_impl when there's a transparent, three-way comparator that has different equivalence classes for different lookup types. Add a new can_have_multiple_equivalent_keys method to share the common logic for these cases. PiperOrigin-RevId: 346605948 -- 649183cb3cc9383431de9c81fb1c0f885d4001ae by Abseil Team <absl-team@google.com>: Add benchmark for accessing a Duration flag. PiperOrigin-RevId: 346594843 -- fefdb046520871af63ce2229e2f7cccfc0483dea by Abseil Team <absl-team@google.com>: Restructure CordReader for upcoming ring buffer changes. PiperOrigin-RevId: 346410642 -- 8b2f50e7da0ebab06ead5f94e366e984ca23cb6a by Abseil Team <absl-team@google.com>: Wire in an internal-only flag to toggle upcoming ring buffer changes on/off for experimentation. PiperOrigin-RevId: 346199111 GitOrigin-RevId: 6e9f93888bbe6718997ed90bbd049f1f3572b066 Change-Id: I8f34866b25a79209cb5448bbb28dd3044111d2e9
4 years ago
template <class Allocator>
static void transfer(Allocator* alloc, slot_type* new_slot,
slot_type* old_slot) {
construct(alloc, new_slot, std::move(*old_slot));
destroy(alloc, old_slot);
}
static T& element(slot_type* slot) { return *slot; }
template <class F, class... Args>
static decltype(absl::container_internal::DecomposeValue(
std::declval<F>(), std::declval<Args>()...))
apply(F&& f, Args&&... args) {
return absl::container_internal::DecomposeValue(
std::forward<F>(f), std::forward<Args>(args)...);
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
}
};
Export of internal Abseil changes -- 6e9f93888bbe6718997ed90bbd049f1f3572b066 by Abseil Team <absl-team@google.com>: Fix status to be safe for self move assignment. PiperOrigin-RevId: 346815392 -- 35cae74a977f258e81dfbe925fb5a34cb6632036 by Gennadiy Rozental <rogeeff@google.com>: Eliminate unnecessary access to the global vars. PiperOrigin-RevId: 346777168 -- e7e786c243069060f5d6c1c945decb4b0b83f95b by Andy Getzendanner <durandal@google.com>: Internal change. PiperOrigin-RevId: 346685656 -- 4ccd41c48f1a83cfa20b3ea534f743dd7d788376 by Abseil Team <absl-team@google.com>: Move CordRep Ref() and Unref() logic into cord_internal.h This change moves Ref() and Unref() logic out of cord.cc into cord_internal. The main purpose is to make upcoming ring buffer changes easier to isolate from existing cord.cc code. Notice that this removes the nullptr check from Unref() and now requires it to be non null (which held true most times). We may need to rethink if the 'unref unlikely one' is the common case: are cordreps most likely shared? This may be something between 'root' and non root nodes, i.e., is it more likely for leaf / flat nodes in large cords to be shared than top level cordreps being shared? Vice versa? Benchmarks say that we mostly shouldn't care, the caveat being that atomic ops seem more expensive on upcoming archs (arcadia) so we should error on the side of an extra IsOne() branch saving us single owned atomic ops. PiperOrigin-RevId: 346676121 -- f0babab103b9e60d61ba09482d468985e43eceb3 by Samuel Benzaquen <sbenza@google.com>: Fix iterator based constructor and `.insert` members to only require EmplaceConstructible as the standard specifies. PiperOrigin-RevId: 346616707 -- 8f48eedda02277f9c96a88ed7726e34b557cce20 by Evan Brown <ezb@google.com>: Fix a bug in binary_search_impl when there's a transparent, three-way comparator that has different equivalence classes for different lookup types. Add a new can_have_multiple_equivalent_keys method to share the common logic for these cases. PiperOrigin-RevId: 346605948 -- 649183cb3cc9383431de9c81fb1c0f885d4001ae by Abseil Team <absl-team@google.com>: Add benchmark for accessing a Duration flag. PiperOrigin-RevId: 346594843 -- fefdb046520871af63ce2229e2f7cccfc0483dea by Abseil Team <absl-team@google.com>: Restructure CordReader for upcoming ring buffer changes. PiperOrigin-RevId: 346410642 -- 8b2f50e7da0ebab06ead5f94e366e984ca23cb6a by Abseil Team <absl-team@google.com>: Wire in an internal-only flag to toggle upcoming ring buffer changes on/off for experimentation. PiperOrigin-RevId: 346199111 GitOrigin-RevId: 6e9f93888bbe6718997ed90bbd049f1f3572b066 Change-Id: I8f34866b25a79209cb5448bbb28dd3044111d2e9
4 years ago
using IntPolicy = ValuePolicy<int64_t>;
Export of internal Abseil changes -- b1fc72630aaa81c8395c3b22ba267d938fe29a2e by Derek Mauro <dmauro@google.com>: Fix -Wdeprecated-copy warnings from Clang 13. Example: error: definition of implicit copy assignment operator for 'UDT' is deprecated because it has a user-declared copy constructor [-Werror,-Wdeprecated-copy] PiperOrigin-RevId: 380058303 -- 0422744812b1a2010d9eea5b17fbe89f3441b66b by Evan Brown <ezb@google.com>: Change the "full table!" asserts in raw_hash_set to use `<= capacity` instead of `< capacity`. If we add support for non-power-of-two-minus-one capacities, this is the correct thing to assert. For example, consider: Group::kWidth = 8, capacity_ = 8, ctrl_ = {kEmpty, 1, 2, 3, 4, 5, 6, 7, kSentinel, kEmpty, 1, 2, 3, 4, 5, 6}. In this case, if we do an unsuccessful lookup with H2 mapping to slot 1, then the first Group will contain {1, 2, 3, 4, 5, 6, 7, kSentinel} so we need to continue to the second Group (at which point seq.index() == 8 == capacity_) to find a kEmpty. Note: this is a no-op change for now since we never have `capacity % Group::kWidth == 0`. PiperOrigin-RevId: 380033480 -- 40628c34d540356de65fabb16c1439c0ec7a0764 by Abseil Team <absl-team@google.com>: Drop out-of-date documentation about `absl::FixedArray`'s allocator support PiperOrigin-RevId: 379811653 -- e7ad047863ae55c9b7aec0753cfc527a4ea614bc by Evan Brown <ezb@google.com>: Fix a bug in ConvertDeletedToEmptyAndFullToDeleted in which we were copying 1 more cloned control byte than actually exists. When alignof(slot_type)>1, this wouldn't cause a problem because the extra byte is padding. Also change loop bounds to not rely on the fact that capacity_+1 is a multiple of Group::kWidth. PiperOrigin-RevId: 379311830 -- 1a3ba500fb2c33205854eb9258cd6e0fb1061bca by Martijn Vels <mvels@google.com>: Change Ring, EXTERNAL and FLAT tag values to be consecutive values The purpose of this change is to have FLAT = EXTERNAL + 1. Especially in the ring and btree alternative code, there is a common check if a node is a 'plain' edge (EXTERNAL or FLAT), or 'something else'. This change can make that check a single branch, i.e., instead of 'tag == EXTERNAL || tag >= FLAT', we can simply check for 'tag >= EXTERNAL'. Likewise we have some cases where we check for RING, EXTERNAL or FLAT, so we align RING + 1 with EXTERNAL. PiperOrigin-RevId: 379291576 -- 0c78e65ca4d85244b106c3f8e24cf268e09e72a3 by Benjamin Barenblat <bbaren@google.com>: Round a double multiplication before casting it to integer The code static_cast<int>(x * y) (for double x and y) performs a double multiplication into a temporary that, by standard, may have excess precision. The subsequent cast to int discards the excess precision. However, the cast may examine the excess precision during conversion, producing surprising results like static_cast<int>(1.7 * 10) == 16 on certain systems. Correct this case by explicitly rounding 1.7 * 10 before casting it. PiperOrigin-RevId: 378922064 GitOrigin-RevId: b1fc72630aaa81c8395c3b22ba267d938fe29a2e Change-Id: Ica708a006921118673e78d5fd2d61fe0fb0894d1
4 years ago
using Uint8Policy = ValuePolicy<uint8_t>;
Export of internal Abseil changes -- 6e9f93888bbe6718997ed90bbd049f1f3572b066 by Abseil Team <absl-team@google.com>: Fix status to be safe for self move assignment. PiperOrigin-RevId: 346815392 -- 35cae74a977f258e81dfbe925fb5a34cb6632036 by Gennadiy Rozental <rogeeff@google.com>: Eliminate unnecessary access to the global vars. PiperOrigin-RevId: 346777168 -- e7e786c243069060f5d6c1c945decb4b0b83f95b by Andy Getzendanner <durandal@google.com>: Internal change. PiperOrigin-RevId: 346685656 -- 4ccd41c48f1a83cfa20b3ea534f743dd7d788376 by Abseil Team <absl-team@google.com>: Move CordRep Ref() and Unref() logic into cord_internal.h This change moves Ref() and Unref() logic out of cord.cc into cord_internal. The main purpose is to make upcoming ring buffer changes easier to isolate from existing cord.cc code. Notice that this removes the nullptr check from Unref() and now requires it to be non null (which held true most times). We may need to rethink if the 'unref unlikely one' is the common case: are cordreps most likely shared? This may be something between 'root' and non root nodes, i.e., is it more likely for leaf / flat nodes in large cords to be shared than top level cordreps being shared? Vice versa? Benchmarks say that we mostly shouldn't care, the caveat being that atomic ops seem more expensive on upcoming archs (arcadia) so we should error on the side of an extra IsOne() branch saving us single owned atomic ops. PiperOrigin-RevId: 346676121 -- f0babab103b9e60d61ba09482d468985e43eceb3 by Samuel Benzaquen <sbenza@google.com>: Fix iterator based constructor and `.insert` members to only require EmplaceConstructible as the standard specifies. PiperOrigin-RevId: 346616707 -- 8f48eedda02277f9c96a88ed7726e34b557cce20 by Evan Brown <ezb@google.com>: Fix a bug in binary_search_impl when there's a transparent, three-way comparator that has different equivalence classes for different lookup types. Add a new can_have_multiple_equivalent_keys method to share the common logic for these cases. PiperOrigin-RevId: 346605948 -- 649183cb3cc9383431de9c81fb1c0f885d4001ae by Abseil Team <absl-team@google.com>: Add benchmark for accessing a Duration flag. PiperOrigin-RevId: 346594843 -- fefdb046520871af63ce2229e2f7cccfc0483dea by Abseil Team <absl-team@google.com>: Restructure CordReader for upcoming ring buffer changes. PiperOrigin-RevId: 346410642 -- 8b2f50e7da0ebab06ead5f94e366e984ca23cb6a by Abseil Team <absl-team@google.com>: Wire in an internal-only flag to toggle upcoming ring buffer changes on/off for experimentation. PiperOrigin-RevId: 346199111 GitOrigin-RevId: 6e9f93888bbe6718997ed90bbd049f1f3572b066 Change-Id: I8f34866b25a79209cb5448bbb28dd3044111d2e9
4 years ago
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
class StringPolicy {
template <class F, class K, class V,
class = typename std::enable_if<
std::is_convertible<const K&, absl::string_view>::value>::type>
decltype(std::declval<F>()(
std::declval<const absl::string_view&>(), std::piecewise_construct,
std::declval<std::tuple<K>>(),
std::declval<V>())) static apply_impl(F&& f,
std::pair<std::tuple<K>, V> p) {
const absl::string_view& key = std::get<0>(p.first);
return std::forward<F>(f)(key, std::piecewise_construct, std::move(p.first),
std::move(p.second));
}
public:
struct slot_type {
struct ctor {};
template <class... Ts>
slot_type(ctor, Ts&&... ts) : pair(std::forward<Ts>(ts)...) {}
std::pair<std::string, std::string> pair;
};
using key_type = std::string;
using init_type = std::pair<std::string, std::string>;
template <class allocator_type, class... Args>
static void construct(allocator_type* alloc, slot_type* slot, Args... args) {
std::allocator_traits<allocator_type>::construct(
*alloc, slot, typename slot_type::ctor(), std::forward<Args>(args)...);
}
template <class allocator_type>
static void destroy(allocator_type* alloc, slot_type* slot) {
std::allocator_traits<allocator_type>::destroy(*alloc, slot);
}
template <class allocator_type>
static void transfer(allocator_type* alloc, slot_type* new_slot,
slot_type* old_slot) {
construct(alloc, new_slot, std::move(old_slot->pair));
destroy(alloc, old_slot);
}
static std::pair<std::string, std::string>& element(slot_type* slot) {
return slot->pair;
}
template <class F, class... Args>
static auto apply(F&& f, Args&&... args)
-> decltype(apply_impl(std::forward<F>(f),
PairArgs(std::forward<Args>(args)...))) {
return apply_impl(std::forward<F>(f),
PairArgs(std::forward<Args>(args)...));
}
};
struct StringHash : absl::Hash<absl::string_view> {
using is_transparent = void;
};
struct StringEq : std::equal_to<absl::string_view> {
using is_transparent = void;
};
struct StringTable
: raw_hash_set<StringPolicy, StringHash, StringEq, std::allocator<int>> {
using Base = typename StringTable::raw_hash_set;
StringTable() {}
using Base::Base;
};
struct IntTable
: raw_hash_set<IntPolicy, container_internal::hash_default_hash<int64_t>,
std::equal_to<int64_t>, std::allocator<int64_t>> {
using Base = typename IntTable::raw_hash_set;
using Base::Base;
};
Export of internal Abseil changes -- b1fc72630aaa81c8395c3b22ba267d938fe29a2e by Derek Mauro <dmauro@google.com>: Fix -Wdeprecated-copy warnings from Clang 13. Example: error: definition of implicit copy assignment operator for 'UDT' is deprecated because it has a user-declared copy constructor [-Werror,-Wdeprecated-copy] PiperOrigin-RevId: 380058303 -- 0422744812b1a2010d9eea5b17fbe89f3441b66b by Evan Brown <ezb@google.com>: Change the "full table!" asserts in raw_hash_set to use `<= capacity` instead of `< capacity`. If we add support for non-power-of-two-minus-one capacities, this is the correct thing to assert. For example, consider: Group::kWidth = 8, capacity_ = 8, ctrl_ = {kEmpty, 1, 2, 3, 4, 5, 6, 7, kSentinel, kEmpty, 1, 2, 3, 4, 5, 6}. In this case, if we do an unsuccessful lookup with H2 mapping to slot 1, then the first Group will contain {1, 2, 3, 4, 5, 6, 7, kSentinel} so we need to continue to the second Group (at which point seq.index() == 8 == capacity_) to find a kEmpty. Note: this is a no-op change for now since we never have `capacity % Group::kWidth == 0`. PiperOrigin-RevId: 380033480 -- 40628c34d540356de65fabb16c1439c0ec7a0764 by Abseil Team <absl-team@google.com>: Drop out-of-date documentation about `absl::FixedArray`'s allocator support PiperOrigin-RevId: 379811653 -- e7ad047863ae55c9b7aec0753cfc527a4ea614bc by Evan Brown <ezb@google.com>: Fix a bug in ConvertDeletedToEmptyAndFullToDeleted in which we were copying 1 more cloned control byte than actually exists. When alignof(slot_type)>1, this wouldn't cause a problem because the extra byte is padding. Also change loop bounds to not rely on the fact that capacity_+1 is a multiple of Group::kWidth. PiperOrigin-RevId: 379311830 -- 1a3ba500fb2c33205854eb9258cd6e0fb1061bca by Martijn Vels <mvels@google.com>: Change Ring, EXTERNAL and FLAT tag values to be consecutive values The purpose of this change is to have FLAT = EXTERNAL + 1. Especially in the ring and btree alternative code, there is a common check if a node is a 'plain' edge (EXTERNAL or FLAT), or 'something else'. This change can make that check a single branch, i.e., instead of 'tag == EXTERNAL || tag >= FLAT', we can simply check for 'tag >= EXTERNAL'. Likewise we have some cases where we check for RING, EXTERNAL or FLAT, so we align RING + 1 with EXTERNAL. PiperOrigin-RevId: 379291576 -- 0c78e65ca4d85244b106c3f8e24cf268e09e72a3 by Benjamin Barenblat <bbaren@google.com>: Round a double multiplication before casting it to integer The code static_cast<int>(x * y) (for double x and y) performs a double multiplication into a temporary that, by standard, may have excess precision. The subsequent cast to int discards the excess precision. However, the cast may examine the excess precision during conversion, producing surprising results like static_cast<int>(1.7 * 10) == 16 on certain systems. Correct this case by explicitly rounding 1.7 * 10 before casting it. PiperOrigin-RevId: 378922064 GitOrigin-RevId: b1fc72630aaa81c8395c3b22ba267d938fe29a2e Change-Id: Ica708a006921118673e78d5fd2d61fe0fb0894d1
4 years ago
struct Uint8Table
: raw_hash_set<Uint8Policy, container_internal::hash_default_hash<uint8_t>,
std::equal_to<uint8_t>, std::allocator<uint8_t>> {
using Base = typename Uint8Table::raw_hash_set;
using Base::Base;
};
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
template <typename T>
struct CustomAlloc : std::allocator<T> {
CustomAlloc() {}
template <typename U>
CustomAlloc(const CustomAlloc<U>& other) {}
template<class U> struct rebind {
using other = CustomAlloc<U>;
};
};
struct CustomAllocIntTable
: raw_hash_set<IntPolicy, container_internal::hash_default_hash<int64_t>,
std::equal_to<int64_t>, CustomAlloc<int64_t>> {
using Base = typename CustomAllocIntTable::raw_hash_set;
using Base::Base;
};
struct BadFastHash {
template <class T>
size_t operator()(const T&) const {
return 0;
}
};
struct BadTable : raw_hash_set<IntPolicy, BadFastHash, std::equal_to<int>,
std::allocator<int>> {
using Base = typename BadTable::raw_hash_set;
BadTable() {}
using Base::Base;
};
TEST(Table, EmptyFunctorOptimization) {
static_assert(std::is_empty<std::equal_to<absl::string_view>>::value, "");
static_assert(std::is_empty<std::allocator<int>>::value, "");
struct MockTable {
void* ctrl;
void* slots;
size_t size;
size_t capacity;
size_t growth_left;
void* infoz;
};
Export of internal Abseil changes -- 8e75347c10d85112296811be6ef35761744ad9bc by Derek Mauro <dmauro@google.com>: Big update to LTS release process * Add create_lts.py script to to the LTS modification This is simpler than copybara since very few changes are needed * Use the default installation paths instead of a versioned path. If a versioned path is needed, this is easy to change on the commandline. * Make the integration test use the LTS transformed version * Test both static and dynamic linking (fixes pkg-config dynamic linking) PiperOrigin-RevId: 363566934 -- e00e971a2de3138861f5e1900201c9cc7788f714 by Laramie Leavitt <lar@google.com>: Add a non-compile test to absl::BitGenRef for temporaries. PiperOrigin-RevId: 363437284 -- 3685644ec115d99789de32aceb76c32a00756fea by Derek Mauro <dmauro@google.com>: Make OSS code consistent with internal code by using the forward declaration of absl::Status that contains ABSL_MUST_USE_RESULT. PiperOrigin-RevId: 363426906 -- b85fec142c3aa3f632fa985f9f8f73a253819723 by Evan Brown <ezb@google.com>: Move raw_hash_set::infoz_ into raw_hash_set::settings_. This reduces the size of raw_hash_sets by alignof(size_t) bytes when hashtablez is disabled. PiperOrigin-RevId: 363034264 -- c6fde3b17e5845191eb8b2bfc1760c8bfb9573ff by Mark Barolak <mbar@google.com>: Internal change PiperOrigin-RevId: 362990378 -- 81713cf964905b43d1cbe32ce5fed97539029625 by Abseil Team <absl-team@google.com>: Fix typo in comment (execeptions -> exceptions). PiperOrigin-RevId: 362946191 -- 3ee92ca470feca44da417b03ee45a915c6eb5155 by Abseil Team <absl-team@google.com>: Add absl::FindAndReportLeaks and routes it to the corresponding __lsan_do_recoverable_leak_check. PiperOrigin-RevId: 362622199 -- b95b7194b20e02c20d72289fbc79a0d35b82e256 by Abseil Team <absl-team@google.com>: Add `kWithEverything` to StatusToStringMode PiperOrigin-RevId: 362595218 -- 0a960d96a0014eab7e1c55b479269450ed8e98d7 by Abseil Team <absl-team@google.com>: Accept e.g. ".__uniq" as a valid clone name. Further, bring the implementation on par with libiberty's demangler grammar. Clang introduced option -funique-internal-linkage-names that adds the suffix ".__uniq.[0-9]+" to internal linkage functions to give them a globally unique identifier. The suffix was designed to work with existing demanglers which do recognize a "_" along with the alphanumeric string. This change enhances the demangler to allow "_" with the alphanumeric string. Please refer to libiberty's cp-demangle.c where function d_clone_suffix implements the demangling of clone suffixes : 1. '_' is accepted as a valid character with the alphanumeric sequence. 2. The alphanumberic sequence is optional. 3. The digit sequence is optional. PiperOrigin-RevId: 362557420 -- 2ac5ea212c150afd2f58025a5cab8c45d16949c6 by Abseil Team <absl-team@google.com>: Change variable name 'slots' to 'slot_count' to avoid name-clash with Qt builds. PiperOrigin-RevId: 362556289 -- 934f0f409c9c548716a46363d6e243406fad4028 by Mark Barolak <mbar@google.com>: Clarify the comment on ABSL_CACHELINE_SIZE to indicate that the macro definition itself shouldn't change, but rather that call sites should change when possible. This addresses the request for improved documentation in https://github.com/abseil/abseil-cpp/pull/842. PiperOrigin-RevId: 362354288 GitOrigin-RevId: 8e75347c10d85112296811be6ef35761744ad9bc Change-Id: I33ec8561d8d645c3353e9d2dd447501d0e1825a7
4 years ago
struct MockTableInfozDisabled {
void* ctrl;
void* slots;
size_t size;
size_t capacity;
size_t growth_left;
};
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
struct StatelessHash {
size_t operator()(absl::string_view) const { return 0; }
};
struct StatefulHash : StatelessHash {
size_t dummy;
};
Export of internal Abseil changes -- 8e75347c10d85112296811be6ef35761744ad9bc by Derek Mauro <dmauro@google.com>: Big update to LTS release process * Add create_lts.py script to to the LTS modification This is simpler than copybara since very few changes are needed * Use the default installation paths instead of a versioned path. If a versioned path is needed, this is easy to change on the commandline. * Make the integration test use the LTS transformed version * Test both static and dynamic linking (fixes pkg-config dynamic linking) PiperOrigin-RevId: 363566934 -- e00e971a2de3138861f5e1900201c9cc7788f714 by Laramie Leavitt <lar@google.com>: Add a non-compile test to absl::BitGenRef for temporaries. PiperOrigin-RevId: 363437284 -- 3685644ec115d99789de32aceb76c32a00756fea by Derek Mauro <dmauro@google.com>: Make OSS code consistent with internal code by using the forward declaration of absl::Status that contains ABSL_MUST_USE_RESULT. PiperOrigin-RevId: 363426906 -- b85fec142c3aa3f632fa985f9f8f73a253819723 by Evan Brown <ezb@google.com>: Move raw_hash_set::infoz_ into raw_hash_set::settings_. This reduces the size of raw_hash_sets by alignof(size_t) bytes when hashtablez is disabled. PiperOrigin-RevId: 363034264 -- c6fde3b17e5845191eb8b2bfc1760c8bfb9573ff by Mark Barolak <mbar@google.com>: Internal change PiperOrigin-RevId: 362990378 -- 81713cf964905b43d1cbe32ce5fed97539029625 by Abseil Team <absl-team@google.com>: Fix typo in comment (execeptions -> exceptions). PiperOrigin-RevId: 362946191 -- 3ee92ca470feca44da417b03ee45a915c6eb5155 by Abseil Team <absl-team@google.com>: Add absl::FindAndReportLeaks and routes it to the corresponding __lsan_do_recoverable_leak_check. PiperOrigin-RevId: 362622199 -- b95b7194b20e02c20d72289fbc79a0d35b82e256 by Abseil Team <absl-team@google.com>: Add `kWithEverything` to StatusToStringMode PiperOrigin-RevId: 362595218 -- 0a960d96a0014eab7e1c55b479269450ed8e98d7 by Abseil Team <absl-team@google.com>: Accept e.g. ".__uniq" as a valid clone name. Further, bring the implementation on par with libiberty's demangler grammar. Clang introduced option -funique-internal-linkage-names that adds the suffix ".__uniq.[0-9]+" to internal linkage functions to give them a globally unique identifier. The suffix was designed to work with existing demanglers which do recognize a "_" along with the alphanumeric string. This change enhances the demangler to allow "_" with the alphanumeric string. Please refer to libiberty's cp-demangle.c where function d_clone_suffix implements the demangling of clone suffixes : 1. '_' is accepted as a valid character with the alphanumeric sequence. 2. The alphanumberic sequence is optional. 3. The digit sequence is optional. PiperOrigin-RevId: 362557420 -- 2ac5ea212c150afd2f58025a5cab8c45d16949c6 by Abseil Team <absl-team@google.com>: Change variable name 'slots' to 'slot_count' to avoid name-clash with Qt builds. PiperOrigin-RevId: 362556289 -- 934f0f409c9c548716a46363d6e243406fad4028 by Mark Barolak <mbar@google.com>: Clarify the comment on ABSL_CACHELINE_SIZE to indicate that the macro definition itself shouldn't change, but rather that call sites should change when possible. This addresses the request for improved documentation in https://github.com/abseil/abseil-cpp/pull/842. PiperOrigin-RevId: 362354288 GitOrigin-RevId: 8e75347c10d85112296811be6ef35761744ad9bc Change-Id: I33ec8561d8d645c3353e9d2dd447501d0e1825a7
4 years ago
if (std::is_empty<HashtablezInfoHandle>::value) {
EXPECT_EQ(sizeof(MockTableInfozDisabled),
sizeof(raw_hash_set<StringPolicy, StatelessHash,
std::equal_to<absl::string_view>,
std::allocator<int>>));
EXPECT_EQ(sizeof(MockTableInfozDisabled) + sizeof(StatefulHash),
sizeof(raw_hash_set<StringPolicy, StatefulHash,
std::equal_to<absl::string_view>,
std::allocator<int>>));
} else {
EXPECT_EQ(sizeof(MockTable),
sizeof(raw_hash_set<StringPolicy, StatelessHash,
std::equal_to<absl::string_view>,
std::allocator<int>>));
EXPECT_EQ(sizeof(MockTable) + sizeof(StatefulHash),
sizeof(raw_hash_set<StringPolicy, StatefulHash,
std::equal_to<absl::string_view>,
std::allocator<int>>));
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
}
TEST(Table, Empty) {
IntTable t;
EXPECT_EQ(0, t.size());
EXPECT_TRUE(t.empty());
}
TEST(Table, LookupEmpty) {
IntTable t;
auto it = t.find(0);
EXPECT_TRUE(it == t.end());
}
TEST(Table, Insert1) {
IntTable t;
EXPECT_TRUE(t.find(0) == t.end());
auto res = t.emplace(0);
EXPECT_TRUE(res.second);
EXPECT_THAT(*res.first, 0);
EXPECT_EQ(1, t.size());
EXPECT_THAT(*t.find(0), 0);
}
TEST(Table, Insert2) {
IntTable t;
EXPECT_TRUE(t.find(0) == t.end());
auto res = t.emplace(0);
EXPECT_TRUE(res.second);
EXPECT_THAT(*res.first, 0);
EXPECT_EQ(1, t.size());
EXPECT_TRUE(t.find(1) == t.end());
res = t.emplace(1);
EXPECT_TRUE(res.second);
EXPECT_THAT(*res.first, 1);
EXPECT_EQ(2, t.size());
EXPECT_THAT(*t.find(0), 0);
EXPECT_THAT(*t.find(1), 1);
}
TEST(Table, InsertCollision) {
BadTable t;
EXPECT_TRUE(t.find(1) == t.end());
auto res = t.emplace(1);
EXPECT_TRUE(res.second);
EXPECT_THAT(*res.first, 1);
EXPECT_EQ(1, t.size());
EXPECT_TRUE(t.find(2) == t.end());
res = t.emplace(2);
EXPECT_THAT(*res.first, 2);
EXPECT_TRUE(res.second);
EXPECT_EQ(2, t.size());
EXPECT_THAT(*t.find(1), 1);
EXPECT_THAT(*t.find(2), 2);
}
// Test that we do not add existent element in case we need to search through
// many groups with deleted elements
TEST(Table, InsertCollisionAndFindAfterDelete) {
BadTable t; // all elements go to the same group.
// Have at least 2 groups with Group::kWidth collisions
// plus some extra collisions in the last group.
constexpr size_t kNumInserts = Group::kWidth * 2 + 5;
for (size_t i = 0; i < kNumInserts; ++i) {
auto res = t.emplace(i);
EXPECT_TRUE(res.second);
EXPECT_THAT(*res.first, i);
EXPECT_EQ(i + 1, t.size());
}
// Remove elements one by one and check
// that we still can find all other elements.
for (size_t i = 0; i < kNumInserts; ++i) {
EXPECT_EQ(1, t.erase(i)) << i;
for (size_t j = i + 1; j < kNumInserts; ++j) {
EXPECT_THAT(*t.find(j), j);
auto res = t.emplace(j);
EXPECT_FALSE(res.second) << i << " " << j;
EXPECT_THAT(*res.first, j);
EXPECT_EQ(kNumInserts - i - 1, t.size());
}
}
EXPECT_TRUE(t.empty());
}
TEST(Table, InsertWithinCapacity) {
IntTable t;
t.reserve(10);
const size_t original_capacity = t.capacity();
const auto addr = [&](int i) {
return reinterpret_cast<uintptr_t>(&*t.find(i));
};
// Inserting an element does not change capacity.
t.insert(0);
EXPECT_THAT(t.capacity(), original_capacity);
const uintptr_t original_addr_0 = addr(0);
// Inserting another element does not rehash.
t.insert(1);
EXPECT_THAT(t.capacity(), original_capacity);
EXPECT_THAT(addr(0), original_addr_0);
// Inserting lots of duplicate elements does not rehash.
for (int i = 0; i < 100; ++i) {
t.insert(i % 10);
}
EXPECT_THAT(t.capacity(), original_capacity);
EXPECT_THAT(addr(0), original_addr_0);
// Inserting a range of duplicate elements does not rehash.
std::vector<int> dup_range;
for (int i = 0; i < 100; ++i) {
dup_range.push_back(i % 10);
}
t.insert(dup_range.begin(), dup_range.end());
EXPECT_THAT(t.capacity(), original_capacity);
EXPECT_THAT(addr(0), original_addr_0);
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
TEST(Table, LazyEmplace) {
StringTable t;
bool called = false;
auto it = t.lazy_emplace("abc", [&](const StringTable::constructor& f) {
called = true;
f("abc", "ABC");
});
EXPECT_TRUE(called);
EXPECT_THAT(*it, Pair("abc", "ABC"));
called = false;
it = t.lazy_emplace("abc", [&](const StringTable::constructor& f) {
called = true;
f("abc", "DEF");
});
EXPECT_FALSE(called);
EXPECT_THAT(*it, Pair("abc", "ABC"));
}
TEST(Table, ContainsEmpty) {
IntTable t;
EXPECT_FALSE(t.contains(0));
}
TEST(Table, Contains1) {
IntTable t;
EXPECT_TRUE(t.insert(0).second);
EXPECT_TRUE(t.contains(0));
EXPECT_FALSE(t.contains(1));
EXPECT_EQ(1, t.erase(0));
EXPECT_FALSE(t.contains(0));
}
TEST(Table, Contains2) {
IntTable t;
EXPECT_TRUE(t.insert(0).second);
EXPECT_TRUE(t.contains(0));
EXPECT_FALSE(t.contains(1));
t.clear();
EXPECT_FALSE(t.contains(0));
}
int decompose_constructed;
int decompose_copy_constructed;
int decompose_copy_assigned;
int decompose_move_constructed;
int decompose_move_assigned;
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
struct DecomposeType {
DecomposeType(int i = 0) : i(i) { // NOLINT
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
++decompose_constructed;
}
explicit DecomposeType(const char* d) : DecomposeType(*d) {}
DecomposeType(const DecomposeType& other) : i(other.i) {
++decompose_copy_constructed;
}
DecomposeType& operator=(const DecomposeType& other) {
++decompose_copy_assigned;
i = other.i;
return *this;
}
DecomposeType(DecomposeType&& other) : i(other.i) {
++decompose_move_constructed;
}
DecomposeType& operator=(DecomposeType&& other) {
++decompose_move_assigned;
i = other.i;
return *this;
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
int i;
};
struct DecomposeHash {
using is_transparent = void;
size_t operator()(const DecomposeType& a) const { return a.i; }
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
size_t operator()(int a) const { return a; }
size_t operator()(const char* a) const { return *a; }
};
struct DecomposeEq {
using is_transparent = void;
bool operator()(const DecomposeType& a, const DecomposeType& b) const {
return a.i == b.i;
}
bool operator()(const DecomposeType& a, int b) const { return a.i == b; }
bool operator()(const DecomposeType& a, const char* b) const {
return a.i == *b;
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
};
struct DecomposePolicy {
using slot_type = DecomposeType;
using key_type = DecomposeType;
using init_type = DecomposeType;
template <typename T>
static void construct(void*, DecomposeType* slot, T&& v) {
::new (slot) DecomposeType(std::forward<T>(v));
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
}
static void destroy(void*, DecomposeType* slot) { slot->~DecomposeType(); }
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
static DecomposeType& element(slot_type* slot) { return *slot; }
template <class F, class T>
static auto apply(F&& f, const T& x) -> decltype(std::forward<F>(f)(x, x)) {
return std::forward<F>(f)(x, x);
}
};
template <typename Hash, typename Eq>
void TestDecompose(bool construct_three) {
DecomposeType elem{0};
const int one = 1;
const char* three_p = "3";
const auto& three = three_p;
const int elem_vector_count = 256;
std::vector<DecomposeType> elem_vector(elem_vector_count, DecomposeType{0});
std::iota(elem_vector.begin(), elem_vector.end(), 0);
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
using DecomposeSet =
raw_hash_set<DecomposePolicy, Hash, Eq, std::allocator<int>>;
DecomposeSet set1;
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
decompose_constructed = 0;
int expected_constructed = 0;
EXPECT_EQ(expected_constructed, decompose_constructed);
set1.insert(elem);
EXPECT_EQ(expected_constructed, decompose_constructed);
set1.insert(1);
EXPECT_EQ(++expected_constructed, decompose_constructed);
set1.emplace("3");
EXPECT_EQ(++expected_constructed, decompose_constructed);
EXPECT_EQ(expected_constructed, decompose_constructed);
{ // insert(T&&)
set1.insert(1);
EXPECT_EQ(expected_constructed, decompose_constructed);
}
{ // insert(const T&)
set1.insert(one);
EXPECT_EQ(expected_constructed, decompose_constructed);
}
{ // insert(hint, T&&)
set1.insert(set1.begin(), 1);
EXPECT_EQ(expected_constructed, decompose_constructed);
}
{ // insert(hint, const T&)
set1.insert(set1.begin(), one);
EXPECT_EQ(expected_constructed, decompose_constructed);
}
{ // emplace(...)
set1.emplace(1);
EXPECT_EQ(expected_constructed, decompose_constructed);
set1.emplace("3");
expected_constructed += construct_three;
EXPECT_EQ(expected_constructed, decompose_constructed);
set1.emplace(one);
EXPECT_EQ(expected_constructed, decompose_constructed);
set1.emplace(three);
expected_constructed += construct_three;
EXPECT_EQ(expected_constructed, decompose_constructed);
}
{ // emplace_hint(...)
set1.emplace_hint(set1.begin(), 1);
EXPECT_EQ(expected_constructed, decompose_constructed);
set1.emplace_hint(set1.begin(), "3");
expected_constructed += construct_three;
EXPECT_EQ(expected_constructed, decompose_constructed);
set1.emplace_hint(set1.begin(), one);
EXPECT_EQ(expected_constructed, decompose_constructed);
set1.emplace_hint(set1.begin(), three);
expected_constructed += construct_three;
EXPECT_EQ(expected_constructed, decompose_constructed);
}
decompose_copy_constructed = 0;
decompose_copy_assigned = 0;
decompose_move_constructed = 0;
decompose_move_assigned = 0;
int expected_copy_constructed = 0;
int expected_move_constructed = 0;
{ // raw_hash_set(first, last) with random-access iterators
DecomposeSet set2(elem_vector.begin(), elem_vector.end());
// Expect exactly one copy-constructor call for each element if no
// rehashing is done.
expected_copy_constructed += elem_vector_count;
EXPECT_EQ(expected_copy_constructed, decompose_copy_constructed);
EXPECT_EQ(expected_move_constructed, decompose_move_constructed);
EXPECT_EQ(0, decompose_move_assigned);
EXPECT_EQ(0, decompose_copy_assigned);
}
{ // raw_hash_set(first, last) with forward iterators
std::list<DecomposeType> elem_list(elem_vector.begin(), elem_vector.end());
expected_copy_constructed = decompose_copy_constructed;
DecomposeSet set2(elem_list.begin(), elem_list.end());
// Expect exactly N elements copied into set, expect at most 2*N elements
// moving internally for all resizing needed (for a growth factor of 2).
expected_copy_constructed += elem_vector_count;
EXPECT_EQ(expected_copy_constructed, decompose_copy_constructed);
expected_move_constructed += elem_vector_count;
EXPECT_LT(expected_move_constructed, decompose_move_constructed);
expected_move_constructed += elem_vector_count;
EXPECT_GE(expected_move_constructed, decompose_move_constructed);
EXPECT_EQ(0, decompose_move_assigned);
EXPECT_EQ(0, decompose_copy_assigned);
expected_copy_constructed = decompose_copy_constructed;
expected_move_constructed = decompose_move_constructed;
}
{ // insert(first, last)
DecomposeSet set2;
set2.insert(elem_vector.begin(), elem_vector.end());
// Expect exactly N elements copied into set, expect at most 2*N elements
// moving internally for all resizing needed (for a growth factor of 2).
const int expected_new_elements = elem_vector_count;
const int expected_max_element_moves = 2 * elem_vector_count;
expected_copy_constructed += expected_new_elements;
EXPECT_EQ(expected_copy_constructed, decompose_copy_constructed);
expected_move_constructed += expected_max_element_moves;
EXPECT_GE(expected_move_constructed, decompose_move_constructed);
EXPECT_EQ(0, decompose_move_assigned);
EXPECT_EQ(0, decompose_copy_assigned);
expected_copy_constructed = decompose_copy_constructed;
expected_move_constructed = decompose_move_constructed;
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
}
TEST(Table, Decompose) {
TestDecompose<DecomposeHash, DecomposeEq>(false);
struct TransparentHashIntOverload {
size_t operator()(const DecomposeType& a) const { return a.i; }
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
size_t operator()(int a) const { return a; }
};
struct TransparentEqIntOverload {
bool operator()(const DecomposeType& a, const DecomposeType& b) const {
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
return a.i == b.i;
}
bool operator()(const DecomposeType& a, int b) const { return a.i == b; }
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
};
TestDecompose<TransparentHashIntOverload, DecomposeEq>(true);
TestDecompose<TransparentHashIntOverload, TransparentEqIntOverload>(true);
TestDecompose<DecomposeHash, TransparentEqIntOverload>(true);
}
// Returns the largest m such that a table with m elements has the same number
// of buckets as a table with n elements.
size_t MaxDensitySize(size_t n) {
IntTable t;
t.reserve(n);
for (size_t i = 0; i != n; ++i) t.emplace(i);
const size_t c = t.bucket_count();
while (c == t.bucket_count()) t.emplace(n++);
return t.size() - 1;
}
struct Modulo1000Hash {
size_t operator()(int x) const { return x % 1000; }
};
struct Modulo1000HashTable
: public raw_hash_set<IntPolicy, Modulo1000Hash, std::equal_to<int>,
std::allocator<int>> {};
// Test that rehash with no resize happen in case of many deleted slots.
TEST(Table, RehashWithNoResize) {
Modulo1000HashTable t;
// Adding the same length (and the same hash) strings
// to have at least kMinFullGroups groups
// with Group::kWidth collisions. Then fill up to MaxDensitySize;
const size_t kMinFullGroups = 7;
std::vector<int> keys;
for (size_t i = 0; i < MaxDensitySize(Group::kWidth * kMinFullGroups); ++i) {
int k = i * 1000;
t.emplace(k);
keys.push_back(k);
}
const size_t capacity = t.capacity();
// Remove elements from all groups except the first and the last one.
Export of internal Abseil changes -- 69b5d0b2a5adb49a53e51f9da6848eaa484242fe by Derek Mauro <dmauro@google.com>: Changes the absl::Duration factory functions to disallow types that are convertible to int or double, instead requiring that the argument itself is indeed an integer or a floating-point number. This will prevent callers from passing arguments, such as std::atomic<T>. This change is an API break. Information and a tool to fix issues can be found at https://abseil.io/docs/cpp/tools/upgrades/duration-conversions PiperOrigin-RevId: 387153494 -- 786063e438ab6a55ac4baa88ad4d20a8293be52a by Evan Brown <ezb@google.com>: Make ctrl_t be an enum class. This adds type safety, and also when strict aliasing is enabled, the compiler will know that control bytes can't alias non-control bytes. Also make H2() return h2_t. PiperOrigin-RevId: 387120717 -- 7e537aabec1c255d6e7c9d21232c597c1c2077bf by Evan Brown <ezb@google.com>: Add some missing `const` keywords to ctrl_t* function parameters. PiperOrigin-RevId: 386976062 -- da53ac6d91cabd951e81dd0a145e1e52b918955f by Martijn Vels <mvels@google.com>: Change Seek and InitOffset to return nullptr instead of assert / fail. This makes it consistent with the rest of the API (Next, Previous, Skip) and hardens it against invariants that are harder (or less likey) to be upheld correctly by the caller. PiperOrigin-RevId: 386963283 -- a4d1faac020d5025edf53ce81808e5db68da7d89 by Abseil Team <absl-team@google.com>: PC / Backtrace / Symbolization for Emscripten. PiperOrigin-RevId: 386957724 -- 97f2c47d83ba9d3ac89e1f55bd06897686ffd063 by Martijn Vels <mvels@google.com>: Fix static casts ([-Wimplicit-int-conversion]) PiperOrigin-RevId: 386951646 -- 9530c795248543817cbc4013953baa09c35f5e1a by Abseil Team <absl-team@google.com>: Fix incorrect header guard in cord_rep_btree_navigator.h PiperOrigin-RevId: 386907904 -- 90ce5872406df2b7f4c428683741dc13a572267e by Abseil Team <absl-team@google.com>: Small grammar fixes for some StatusCode descriptions. PiperOrigin-RevId: 386906217 -- b30a2fd777f12a04a4d512f37a34614b0d05ce99 by Derek Mauro <dmauro@google.com>: Skip length checking when constructing absl::string_view from std::string. The length check causes unnecessary code bloat. PiperOrigin-RevId: 386857974 -- fa171536c359bfa2a1b80297e844519bb9ee7791 by Martijn Vels <mvels@google.com>: Introduce CordRepBtreeNavigator CordRepBtreeNavigator implements bi-directional navigation over all data edges stored inside a Cord Btree. PiperOrigin-RevId: 386519102 GitOrigin-RevId: 69b5d0b2a5adb49a53e51f9da6848eaa484242fe Change-Id: I1b35188d66133f8cb73d346bc5564aac4e0b3e80
3 years ago
// All elements removed from full groups will be marked as ctrl_t::kDeleted.
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
const size_t erase_begin = Group::kWidth / 2;
const size_t erase_end = (t.size() / Group::kWidth - 1) * Group::kWidth;
for (size_t i = erase_begin; i < erase_end; ++i) {
EXPECT_EQ(1, t.erase(keys[i])) << i;
}
keys.erase(keys.begin() + erase_begin, keys.begin() + erase_end);
auto last_key = keys.back();
size_t last_key_num_probes = GetHashtableDebugNumProbes(t, last_key);
// Make sure that we have to make a lot of probes for last key.
ASSERT_GT(last_key_num_probes, kMinFullGroups);
int x = 1;
// Insert and erase one element, before inplace rehash happen.
while (last_key_num_probes == GetHashtableDebugNumProbes(t, last_key)) {
t.emplace(x);
ASSERT_EQ(capacity, t.capacity());
// All elements should be there.
ASSERT_TRUE(t.find(x) != t.end()) << x;
for (const auto& k : keys) {
ASSERT_TRUE(t.find(k) != t.end()) << k;
}
t.erase(x);
++x;
}
}
TEST(Table, InsertEraseStressTest) {
IntTable t;
const size_t kMinElementCount = 250;
std::deque<int> keys;
size_t i = 0;
for (; i < MaxDensitySize(kMinElementCount); ++i) {
t.emplace(i);
keys.push_back(i);
}
const size_t kNumIterations = 1000000;
for (; i < kNumIterations; ++i) {
ASSERT_EQ(1, t.erase(keys.front()));
keys.pop_front();
t.emplace(i);
keys.push_back(i);
}
}
TEST(Table, InsertOverloads) {
StringTable t;
// These should all trigger the insert(init_type) overload.
t.insert({{}, {}});
t.insert({"ABC", {}});
t.insert({"DEF", "!!!"});
EXPECT_THAT(t, UnorderedElementsAre(Pair("", ""), Pair("ABC", ""),
Pair("DEF", "!!!")));
}
TEST(Table, LargeTable) {
IntTable t;
for (int64_t i = 0; i != 100000; ++i) t.emplace(i << 40);
for (int64_t i = 0; i != 100000; ++i) ASSERT_EQ(i << 40, *t.find(i << 40));
}
// Timeout if copy is quadratic as it was in Rust.
TEST(Table, EnsureNonQuadraticAsInRust) {
static const size_t kLargeSize = 1 << 15;
IntTable t;
for (size_t i = 0; i != kLargeSize; ++i) {
t.insert(i);
}
// If this is quadratic, the test will timeout.
IntTable t2;
for (const auto& entry : t) t2.insert(entry);
}
TEST(Table, ClearBug) {
IntTable t;
constexpr size_t capacity = container_internal::Group::kWidth - 1;
constexpr size_t max_size = capacity / 2 + 1;
for (size_t i = 0; i < max_size; ++i) {
t.insert(i);
}
ASSERT_EQ(capacity, t.capacity());
intptr_t original = reinterpret_cast<intptr_t>(&*t.find(2));
t.clear();
ASSERT_EQ(capacity, t.capacity());
for (size_t i = 0; i < max_size; ++i) {
t.insert(i);
}
ASSERT_EQ(capacity, t.capacity());
intptr_t second = reinterpret_cast<intptr_t>(&*t.find(2));
// We are checking that original and second are close enough to each other
// that they are probably still in the same group. This is not strictly
// guaranteed.
EXPECT_LT(std::abs(original - second),
capacity * sizeof(IntTable::value_type));
}
TEST(Table, Erase) {
IntTable t;
EXPECT_TRUE(t.find(0) == t.end());
auto res = t.emplace(0);
EXPECT_TRUE(res.second);
EXPECT_EQ(1, t.size());
t.erase(res.first);
EXPECT_EQ(0, t.size());
EXPECT_TRUE(t.find(0) == t.end());
}
TEST(Table, EraseMaintainsValidIterator) {
IntTable t;
const int kNumElements = 100;
for (int i = 0; i < kNumElements; i ++) {
EXPECT_TRUE(t.emplace(i).second);
}
EXPECT_EQ(t.size(), kNumElements);
int num_erase_calls = 0;
auto it = t.begin();
while (it != t.end()) {
t.erase(it++);
num_erase_calls++;
}
EXPECT_TRUE(t.empty());
EXPECT_EQ(num_erase_calls, kNumElements);
}
// Collect N bad keys by following algorithm:
// 1. Create an empty table and reserve it to 2 * N.
// 2. Insert N random elements.
// 3. Take first Group::kWidth - 1 to bad_keys array.
// 4. Clear the table without resize.
// 5. Go to point 2 while N keys not collected
std::vector<int64_t> CollectBadMergeKeys(size_t N) {
static constexpr int kGroupSize = Group::kWidth - 1;
Export of internal Abseil changes -- a5af5874c1c5cc02bd2a748d455321f82b6f2a93 by Andy Getzendanner <durandal@google.com>: fix compile fails with asan and -Wredundant-decls Import of https://github.com/abseil/abseil-cpp/pull/801 PiperOrigin-RevId: 336693223 -- ed9df42ab2b742386c6692c2bed015374c919d9c by Derek Mauro <dmauro@google.com>: Fix integer conversion warning Fixes #814 PiperOrigin-RevId: 336651814 -- 0ab4c23884e72dce17b67c1eb520f9dbb802565d by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 336585378 -- eba0e3dccd52a6e91bcff84075bef0affc650b74 by Matt Kulukundis <kfm@google.com>: Add bitset operations to Futex helper. PiperOrigin-RevId: 336409368 -- 8b0709a8b4500bf5f0af4b602d76a298d81645e8 by Abseil Team <absl-team@google.com>: Fix code indentation in a comment. PiperOrigin-RevId: 336368167 -- bc3961c87a7e7760c10319a5b0349c279f7ae3ad by Samuel Benzaquen <sbenza@google.com>: Improve performance of the registry: - Reduce contention - Reduce memory usage for each flag by `6*sizeof(void*)`. - Replace one immortal allocation per-flag with a single one for all the flags - Slightly improve single-threaded performance by avoiding the std::map indirections. PiperOrigin-RevId: 336365904 -- 264ad9f28f935aad8b6b1437f8bf804fa9104346 by Abseil Team <absl-team@google.com>: Fix typo in comment on absl::Condition. PiperOrigin-RevId: 336311680 -- b5b808a8c75ca0df7b09eff9a423ec171d80f771 by Derek Mauro <dmauro@google.com>: Add missing Apache license headers PiperOrigin-RevId: 336294980 -- 89446c3a4793df8b95060385cf3e219357c3db1d by Andy Soffer <asoffer@google.com>: Internal changes PiperOrigin-RevId: 336287465 -- 57c8be4e294881bc79a6a44b8e4bf7ecbb19b9b9 by Matt Kulukundis <kfm@google.com>: Extract Futex from an implementation detail of Wait to a private interface. PiperOrigin-RevId: 336123209 GitOrigin-RevId: a5af5874c1c5cc02bd2a748d455321f82b6f2a93 Change-Id: Ie5a0ebe28e571814e3e11d4c05ca308523ccf311
4 years ago
auto topk_range = [](size_t b, size_t e,
IntTable* t) -> std::vector<int64_t> {
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
for (size_t i = b; i != e; ++i) {
t->emplace(i);
}
std::vector<int64_t> res;
res.reserve(kGroupSize);
auto it = t->begin();
for (size_t i = b; i != e && i != b + kGroupSize; ++i, ++it) {
res.push_back(*it);
}
return res;
};
std::vector<int64_t> bad_keys;
bad_keys.reserve(N);
IntTable t;
t.reserve(N * 2);
for (size_t b = 0; bad_keys.size() < N; b += N) {
auto keys = topk_range(b, b + N, &t);
bad_keys.insert(bad_keys.end(), keys.begin(), keys.end());
t.erase(t.begin(), t.end());
EXPECT_TRUE(t.empty());
}
return bad_keys;
}
struct ProbeStats {
// Number of elements with specific probe length over all tested tables.
std::vector<size_t> all_probes_histogram;
// Ratios total_probe_length/size for every tested table.
std::vector<double> single_table_ratios;
friend ProbeStats operator+(const ProbeStats& a, const ProbeStats& b) {
ProbeStats res = a;
res.all_probes_histogram.resize(std::max(res.all_probes_histogram.size(),
b.all_probes_histogram.size()));
std::transform(b.all_probes_histogram.begin(), b.all_probes_histogram.end(),
res.all_probes_histogram.begin(),
res.all_probes_histogram.begin(), std::plus<size_t>());
res.single_table_ratios.insert(res.single_table_ratios.end(),
b.single_table_ratios.begin(),
b.single_table_ratios.end());
return res;
}
// Average ratio total_probe_length/size over tables.
double AvgRatio() const {
return std::accumulate(single_table_ratios.begin(),
single_table_ratios.end(), 0.0) /
single_table_ratios.size();
}
// Maximum ratio total_probe_length/size over tables.
double MaxRatio() const {
return *std::max_element(single_table_ratios.begin(),
single_table_ratios.end());
}
// Percentile ratio total_probe_length/size over tables.
double PercentileRatio(double Percentile = 0.95) const {
auto r = single_table_ratios;
auto mid = r.begin() + static_cast<size_t>(r.size() * Percentile);
if (mid != r.end()) {
std::nth_element(r.begin(), mid, r.end());
return *mid;
} else {
return MaxRatio();
}
}
// Maximum probe length over all elements and all tables.
size_t MaxProbe() const { return all_probes_histogram.size(); }
// Fraction of elements with specified probe length.
std::vector<double> ProbeNormalizedHistogram() const {
double total_elements = std::accumulate(all_probes_histogram.begin(),
all_probes_histogram.end(), 0ull);
std::vector<double> res;
for (size_t p : all_probes_histogram) {
res.push_back(p / total_elements);
}
return res;
}
size_t PercentileProbe(double Percentile = 0.99) const {
size_t idx = 0;
for (double p : ProbeNormalizedHistogram()) {
if (Percentile > p) {
Percentile -= p;
++idx;
} else {
return idx;
}
}
return idx;
}
friend std::ostream& operator<<(std::ostream& out, const ProbeStats& s) {
out << "{AvgRatio:" << s.AvgRatio() << ", MaxRatio:" << s.MaxRatio()
<< ", PercentileRatio:" << s.PercentileRatio()
<< ", MaxProbe:" << s.MaxProbe() << ", Probes=[";
for (double p : s.ProbeNormalizedHistogram()) {
out << p << ",";
}
out << "]}";
return out;
}
};
struct ExpectedStats {
double avg_ratio;
double max_ratio;
std::vector<std::pair<double, double>> pecentile_ratios;
std::vector<std::pair<double, double>> pecentile_probes;
friend std::ostream& operator<<(std::ostream& out, const ExpectedStats& s) {
out << "{AvgRatio:" << s.avg_ratio << ", MaxRatio:" << s.max_ratio
<< ", PercentileRatios: [";
for (auto el : s.pecentile_ratios) {
out << el.first << ":" << el.second << ", ";
}
out << "], PercentileProbes: [";
for (auto el : s.pecentile_probes) {
out << el.first << ":" << el.second << ", ";
}
out << "]}";
return out;
}
};
void VerifyStats(size_t size, const ExpectedStats& exp,
const ProbeStats& stats) {
EXPECT_LT(stats.AvgRatio(), exp.avg_ratio) << size << " " << stats;
EXPECT_LT(stats.MaxRatio(), exp.max_ratio) << size << " " << stats;
for (auto pr : exp.pecentile_ratios) {
EXPECT_LE(stats.PercentileRatio(pr.first), pr.second)
<< size << " " << pr.first << " " << stats;
}
for (auto pr : exp.pecentile_probes) {
EXPECT_LE(stats.PercentileProbe(pr.first), pr.second)
<< size << " " << pr.first << " " << stats;
}
}
using ProbeStatsPerSize = std::map<size_t, ProbeStats>;
// Collect total ProbeStats on num_iters iterations of the following algorithm:
// 1. Create new table and reserve it to keys.size() * 2
// 2. Insert all keys xored with seed
// 3. Collect ProbeStats from final table.
Export of internal Abseil changes -- a5af5874c1c5cc02bd2a748d455321f82b6f2a93 by Andy Getzendanner <durandal@google.com>: fix compile fails with asan and -Wredundant-decls Import of https://github.com/abseil/abseil-cpp/pull/801 PiperOrigin-RevId: 336693223 -- ed9df42ab2b742386c6692c2bed015374c919d9c by Derek Mauro <dmauro@google.com>: Fix integer conversion warning Fixes #814 PiperOrigin-RevId: 336651814 -- 0ab4c23884e72dce17b67c1eb520f9dbb802565d by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 336585378 -- eba0e3dccd52a6e91bcff84075bef0affc650b74 by Matt Kulukundis <kfm@google.com>: Add bitset operations to Futex helper. PiperOrigin-RevId: 336409368 -- 8b0709a8b4500bf5f0af4b602d76a298d81645e8 by Abseil Team <absl-team@google.com>: Fix code indentation in a comment. PiperOrigin-RevId: 336368167 -- bc3961c87a7e7760c10319a5b0349c279f7ae3ad by Samuel Benzaquen <sbenza@google.com>: Improve performance of the registry: - Reduce contention - Reduce memory usage for each flag by `6*sizeof(void*)`. - Replace one immortal allocation per-flag with a single one for all the flags - Slightly improve single-threaded performance by avoiding the std::map indirections. PiperOrigin-RevId: 336365904 -- 264ad9f28f935aad8b6b1437f8bf804fa9104346 by Abseil Team <absl-team@google.com>: Fix typo in comment on absl::Condition. PiperOrigin-RevId: 336311680 -- b5b808a8c75ca0df7b09eff9a423ec171d80f771 by Derek Mauro <dmauro@google.com>: Add missing Apache license headers PiperOrigin-RevId: 336294980 -- 89446c3a4793df8b95060385cf3e219357c3db1d by Andy Soffer <asoffer@google.com>: Internal changes PiperOrigin-RevId: 336287465 -- 57c8be4e294881bc79a6a44b8e4bf7ecbb19b9b9 by Matt Kulukundis <kfm@google.com>: Extract Futex from an implementation detail of Wait to a private interface. PiperOrigin-RevId: 336123209 GitOrigin-RevId: a5af5874c1c5cc02bd2a748d455321f82b6f2a93 Change-Id: Ie5a0ebe28e571814e3e11d4c05ca308523ccf311
4 years ago
ProbeStats CollectProbeStatsOnKeysXoredWithSeed(
const std::vector<int64_t>& keys, size_t num_iters) {
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
const size_t reserve_size = keys.size() * 2;
ProbeStats stats;
int64_t seed = 0x71b1a19b907d6e33;
while (num_iters--) {
seed = static_cast<int64_t>(static_cast<uint64_t>(seed) * 17 + 13);
IntTable t1;
t1.reserve(reserve_size);
for (const auto& key : keys) {
t1.emplace(key ^ seed);
}
auto probe_histogram = GetHashtableDebugNumProbesHistogram(t1);
stats.all_probes_histogram.resize(
std::max(stats.all_probes_histogram.size(), probe_histogram.size()));
std::transform(probe_histogram.begin(), probe_histogram.end(),
stats.all_probes_histogram.begin(),
stats.all_probes_histogram.begin(), std::plus<size_t>());
size_t total_probe_seq_length = 0;
for (size_t i = 0; i < probe_histogram.size(); ++i) {
total_probe_seq_length += i * probe_histogram[i];
}
stats.single_table_ratios.push_back(total_probe_seq_length * 1.0 /
keys.size());
t1.erase(t1.begin(), t1.end());
}
return stats;
}
ExpectedStats XorSeedExpectedStats() {
constexpr bool kRandomizesInserts =
#ifdef NDEBUG
false;
#else // NDEBUG
true;
#endif // NDEBUG
// The effective load factor is larger in non-opt mode because we insert
// elements out of order.
switch (container_internal::Group::kWidth) {
case 8:
if (kRandomizesInserts) {
return {0.05,
1.0,
{{0.95, 0.5}},
{{0.95, 0}, {0.99, 2}, {0.999, 4}, {0.9999, 10}}};
} else {
return {0.05,
2.0,
{{0.95, 0.1}},
{{0.95, 0}, {0.99, 2}, {0.999, 4}, {0.9999, 10}}};
}
case 16:
if (kRandomizesInserts) {
return {0.1,
2.0,
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
{{0.95, 0.1}},
{{0.95, 0}, {0.99, 1}, {0.999, 8}, {0.9999, 15}}};
} else {
return {0.05,
1.0,
{{0.95, 0.05}},
{{0.95, 0}, {0.99, 1}, {0.999, 4}, {0.9999, 10}}};
}
}
ABSL_RAW_LOG(FATAL, "%s", "Unknown Group width");
return {};
}
// TODO(b/80415403): Figure out why this test is so flaky, esp. on MSVC
TEST(Table, DISABLED_EnsureNonQuadraticTopNXorSeedByProbeSeqLength) {
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
ProbeStatsPerSize stats;
std::vector<size_t> sizes = {Group::kWidth << 5, Group::kWidth << 10};
for (size_t size : sizes) {
stats[size] =
CollectProbeStatsOnKeysXoredWithSeed(CollectBadMergeKeys(size), 200);
}
auto expected = XorSeedExpectedStats();
for (size_t size : sizes) {
auto& stat = stats[size];
VerifyStats(size, expected, stat);
}
}
// Collect total ProbeStats on num_iters iterations of the following algorithm:
// 1. Create new table
// 2. Select 10% of keys and insert 10 elements key * 17 + j * 13
// 3. Collect ProbeStats from final table
ProbeStats CollectProbeStatsOnLinearlyTransformedKeys(
const std::vector<int64_t>& keys, size_t num_iters) {
ProbeStats stats;
std::random_device rd;
std::mt19937 rng(rd());
auto linear_transform = [](size_t x, size_t y) { return x * 17 + y * 13; };
std::uniform_int_distribution<size_t> dist(0, keys.size()-1);
while (num_iters--) {
IntTable t1;
size_t num_keys = keys.size() / 10;
size_t start = dist(rng);
for (size_t i = 0; i != num_keys; ++i) {
for (size_t j = 0; j != 10; ++j) {
t1.emplace(linear_transform(keys[(i + start) % keys.size()], j));
}
}
auto probe_histogram = GetHashtableDebugNumProbesHistogram(t1);
stats.all_probes_histogram.resize(
std::max(stats.all_probes_histogram.size(), probe_histogram.size()));
std::transform(probe_histogram.begin(), probe_histogram.end(),
stats.all_probes_histogram.begin(),
stats.all_probes_histogram.begin(), std::plus<size_t>());
size_t total_probe_seq_length = 0;
for (size_t i = 0; i < probe_histogram.size(); ++i) {
total_probe_seq_length += i * probe_histogram[i];
}
stats.single_table_ratios.push_back(total_probe_seq_length * 1.0 /
t1.size());
t1.erase(t1.begin(), t1.end());
}
return stats;
}
ExpectedStats LinearTransformExpectedStats() {
constexpr bool kRandomizesInserts =
#ifdef NDEBUG
false;
#else // NDEBUG
true;
#endif // NDEBUG
// The effective load factor is larger in non-opt mode because we insert
// elements out of order.
switch (container_internal::Group::kWidth) {
case 8:
if (kRandomizesInserts) {
return {0.1,
0.5,
{{0.95, 0.3}},
{{0.95, 0}, {0.99, 1}, {0.999, 8}, {0.9999, 15}}};
} else {
return {0.4,
0.6,
{{0.95, 0.5}},
{{0.95, 1}, {0.99, 14}, {0.999, 23}, {0.9999, 26}}};
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
}
case 16:
if (kRandomizesInserts) {
return {0.1,
0.4,
{{0.95, 0.3}},
{{0.95, 1}, {0.99, 2}, {0.999, 9}, {0.9999, 15}}};
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
} else {
return {0.05,
0.2,
{{0.95, 0.1}},
{{0.95, 0}, {0.99, 1}, {0.999, 6}, {0.9999, 10}}};
}
}
ABSL_RAW_LOG(FATAL, "%s", "Unknown Group width");
return {};
}
// TODO(b/80415403): Figure out why this test is so flaky.
TEST(Table, DISABLED_EnsureNonQuadraticTopNLinearTransformByProbeSeqLength) {
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
ProbeStatsPerSize stats;
std::vector<size_t> sizes = {Group::kWidth << 5, Group::kWidth << 10};
for (size_t size : sizes) {
stats[size] = CollectProbeStatsOnLinearlyTransformedKeys(
CollectBadMergeKeys(size), 300);
}
auto expected = LinearTransformExpectedStats();
for (size_t size : sizes) {
auto& stat = stats[size];
VerifyStats(size, expected, stat);
}
}
TEST(Table, EraseCollision) {
BadTable t;
// 1 2 3
t.emplace(1);
t.emplace(2);
t.emplace(3);
EXPECT_THAT(*t.find(1), 1);
EXPECT_THAT(*t.find(2), 2);
EXPECT_THAT(*t.find(3), 3);
EXPECT_EQ(3, t.size());
// 1 DELETED 3
t.erase(t.find(2));
EXPECT_THAT(*t.find(1), 1);
EXPECT_TRUE(t.find(2) == t.end());
EXPECT_THAT(*t.find(3), 3);
EXPECT_EQ(2, t.size());
// DELETED DELETED 3
t.erase(t.find(1));
EXPECT_TRUE(t.find(1) == t.end());
EXPECT_TRUE(t.find(2) == t.end());
EXPECT_THAT(*t.find(3), 3);
EXPECT_EQ(1, t.size());
// DELETED DELETED DELETED
t.erase(t.find(3));
EXPECT_TRUE(t.find(1) == t.end());
EXPECT_TRUE(t.find(2) == t.end());
EXPECT_TRUE(t.find(3) == t.end());
EXPECT_EQ(0, t.size());
}
TEST(Table, EraseInsertProbing) {
BadTable t(100);
// 1 2 3 4
t.emplace(1);
t.emplace(2);
t.emplace(3);
t.emplace(4);
// 1 DELETED 3 DELETED
t.erase(t.find(2));
t.erase(t.find(4));
// 1 10 3 11 12
t.emplace(10);
t.emplace(11);
t.emplace(12);
EXPECT_EQ(5, t.size());
EXPECT_THAT(t, UnorderedElementsAre(1, 10, 3, 11, 12));
}
TEST(Table, Clear) {
IntTable t;
EXPECT_TRUE(t.find(0) == t.end());
t.clear();
EXPECT_TRUE(t.find(0) == t.end());
auto res = t.emplace(0);
EXPECT_TRUE(res.second);
EXPECT_EQ(1, t.size());
t.clear();
EXPECT_EQ(0, t.size());
EXPECT_TRUE(t.find(0) == t.end());
}
TEST(Table, Swap) {
IntTable t;
EXPECT_TRUE(t.find(0) == t.end());
auto res = t.emplace(0);
EXPECT_TRUE(res.second);
EXPECT_EQ(1, t.size());
IntTable u;
t.swap(u);
EXPECT_EQ(0, t.size());
EXPECT_EQ(1, u.size());
EXPECT_TRUE(t.find(0) == t.end());
EXPECT_THAT(*u.find(0), 0);
}
TEST(Table, Rehash) {
IntTable t;
EXPECT_TRUE(t.find(0) == t.end());
t.emplace(0);
t.emplace(1);
EXPECT_EQ(2, t.size());
t.rehash(128);
EXPECT_EQ(2, t.size());
EXPECT_THAT(*t.find(0), 0);
EXPECT_THAT(*t.find(1), 1);
}
TEST(Table, RehashDoesNotRehashWhenNotNecessary) {
IntTable t;
t.emplace(0);
t.emplace(1);
auto* p = &*t.find(0);
t.rehash(1);
EXPECT_EQ(p, &*t.find(0));
}
TEST(Table, RehashZeroDoesNotAllocateOnEmptyTable) {
IntTable t;
t.rehash(0);
EXPECT_EQ(0, t.bucket_count());
}
TEST(Table, RehashZeroDeallocatesEmptyTable) {
IntTable t;
t.emplace(0);
t.clear();
EXPECT_NE(0, t.bucket_count());
t.rehash(0);
EXPECT_EQ(0, t.bucket_count());
}
TEST(Table, RehashZeroForcesRehash) {
IntTable t;
t.emplace(0);
t.emplace(1);
auto* p = &*t.find(0);
t.rehash(0);
EXPECT_NE(p, &*t.find(0));
}
TEST(Table, ConstructFromInitList) {
using P = std::pair<std::string, std::string>;
struct Q {
operator P() const { return {}; }
};
StringTable t = {P(), Q(), {}, {{}, {}}};
}
TEST(Table, CopyConstruct) {
IntTable t;
t.emplace(0);
EXPECT_EQ(1, t.size());
{
IntTable u(t);
EXPECT_EQ(1, u.size());
EXPECT_THAT(*u.find(0), 0);
}
{
IntTable u{t};
EXPECT_EQ(1, u.size());
EXPECT_THAT(*u.find(0), 0);
}
{
IntTable u = t;
EXPECT_EQ(1, u.size());
EXPECT_THAT(*u.find(0), 0);
}
}
TEST(Table, CopyConstructWithAlloc) {
StringTable t;
t.emplace("a", "b");
EXPECT_EQ(1, t.size());
StringTable u(t, Alloc<std::pair<std::string, std::string>>());
EXPECT_EQ(1, u.size());
EXPECT_THAT(*u.find("a"), Pair("a", "b"));
}
struct ExplicitAllocIntTable
: raw_hash_set<IntPolicy, container_internal::hash_default_hash<int64_t>,
std::equal_to<int64_t>, Alloc<int64_t>> {
ExplicitAllocIntTable() {}
};
TEST(Table, AllocWithExplicitCtor) {
ExplicitAllocIntTable t;
EXPECT_EQ(0, t.size());
}
TEST(Table, MoveConstruct) {
{
StringTable t;
t.emplace("a", "b");
EXPECT_EQ(1, t.size());
StringTable u(std::move(t));
EXPECT_EQ(1, u.size());
EXPECT_THAT(*u.find("a"), Pair("a", "b"));
}
{
StringTable t;
t.emplace("a", "b");
EXPECT_EQ(1, t.size());
StringTable u{std::move(t)};
EXPECT_EQ(1, u.size());
EXPECT_THAT(*u.find("a"), Pair("a", "b"));
}
{
StringTable t;
t.emplace("a", "b");
EXPECT_EQ(1, t.size());
StringTable u = std::move(t);
EXPECT_EQ(1, u.size());
EXPECT_THAT(*u.find("a"), Pair("a", "b"));
}
}
TEST(Table, MoveConstructWithAlloc) {
StringTable t;
t.emplace("a", "b");
EXPECT_EQ(1, t.size());
StringTable u(std::move(t), Alloc<std::pair<std::string, std::string>>());
EXPECT_EQ(1, u.size());
EXPECT_THAT(*u.find("a"), Pair("a", "b"));
}
TEST(Table, CopyAssign) {
StringTable t;
t.emplace("a", "b");
EXPECT_EQ(1, t.size());
StringTable u;
u = t;
EXPECT_EQ(1, u.size());
EXPECT_THAT(*u.find("a"), Pair("a", "b"));
}
TEST(Table, CopySelfAssign) {
StringTable t;
t.emplace("a", "b");
EXPECT_EQ(1, t.size());
t = *&t;
EXPECT_EQ(1, t.size());
EXPECT_THAT(*t.find("a"), Pair("a", "b"));
}
TEST(Table, MoveAssign) {
StringTable t;
t.emplace("a", "b");
EXPECT_EQ(1, t.size());
StringTable u;
u = std::move(t);
EXPECT_EQ(1, u.size());
EXPECT_THAT(*u.find("a"), Pair("a", "b"));
}
TEST(Table, Equality) {
StringTable t;
std::vector<std::pair<std::string, std::string>> v = {{"a", "b"},
{"aa", "bb"}};
t.insert(std::begin(v), std::end(v));
StringTable u = t;
EXPECT_EQ(u, t);
}
TEST(Table, Equality2) {
StringTable t;
std::vector<std::pair<std::string, std::string>> v1 = {{"a", "b"},
{"aa", "bb"}};
t.insert(std::begin(v1), std::end(v1));
StringTable u;
std::vector<std::pair<std::string, std::string>> v2 = {{"a", "a"},
{"aa", "aa"}};
u.insert(std::begin(v2), std::end(v2));
EXPECT_NE(u, t);
}
TEST(Table, Equality3) {
StringTable t;
std::vector<std::pair<std::string, std::string>> v1 = {{"b", "b"},
{"bb", "bb"}};
t.insert(std::begin(v1), std::end(v1));
StringTable u;
std::vector<std::pair<std::string, std::string>> v2 = {{"a", "a"},
{"aa", "aa"}};
u.insert(std::begin(v2), std::end(v2));
EXPECT_NE(u, t);
}
TEST(Table, NumDeletedRegression) {
IntTable t;
t.emplace(0);
t.erase(t.find(0));
// construct over a deleted slot.
t.emplace(0);
t.clear();
}
TEST(Table, FindFullDeletedRegression) {
IntTable t;
for (int i = 0; i < 1000; ++i) {
t.emplace(i);
t.erase(t.find(i));
}
EXPECT_EQ(0, t.size());
}
TEST(Table, ReplacingDeletedSlotDoesNotRehash) {
size_t n;
{
// Compute n such that n is the maximum number of elements before rehash.
IntTable t;
t.emplace(0);
size_t c = t.bucket_count();
for (n = 1; c == t.bucket_count(); ++n) t.emplace(n);
--n;
}
IntTable t;
t.rehash(n);
const size_t c = t.bucket_count();
for (size_t i = 0; i != n; ++i) t.emplace(i);
EXPECT_EQ(c, t.bucket_count()) << "rehashing threshold = " << n;
t.erase(0);
t.emplace(0);
EXPECT_EQ(c, t.bucket_count()) << "rehashing threshold = " << n;
}
TEST(Table, NoThrowMoveConstruct) {
ASSERT_TRUE(
std::is_nothrow_copy_constructible<absl::Hash<absl::string_view>>::value);
ASSERT_TRUE(std::is_nothrow_copy_constructible<
std::equal_to<absl::string_view>>::value);
ASSERT_TRUE(std::is_nothrow_copy_constructible<std::allocator<int>>::value);
EXPECT_TRUE(std::is_nothrow_move_constructible<StringTable>::value);
}
TEST(Table, NoThrowMoveAssign) {
ASSERT_TRUE(
std::is_nothrow_move_assignable<absl::Hash<absl::string_view>>::value);
ASSERT_TRUE(
std::is_nothrow_move_assignable<std::equal_to<absl::string_view>>::value);
ASSERT_TRUE(std::is_nothrow_move_assignable<std::allocator<int>>::value);
ASSERT_TRUE(
absl::allocator_traits<std::allocator<int>>::is_always_equal::value);
EXPECT_TRUE(std::is_nothrow_move_assignable<StringTable>::value);
}
TEST(Table, NoThrowSwappable) {
ASSERT_TRUE(
container_internal::IsNoThrowSwappable<absl::Hash<absl::string_view>>());
ASSERT_TRUE(container_internal::IsNoThrowSwappable<
std::equal_to<absl::string_view>>());
ASSERT_TRUE(container_internal::IsNoThrowSwappable<std::allocator<int>>());
EXPECT_TRUE(container_internal::IsNoThrowSwappable<StringTable>());
}
TEST(Table, HeterogeneousLookup) {
struct Hash {
size_t operator()(int64_t i) const { return i; }
size_t operator()(double i) const {
ADD_FAILURE();
return i;
}
};
struct Eq {
bool operator()(int64_t a, int64_t b) const { return a == b; }
bool operator()(double a, int64_t b) const {
ADD_FAILURE();
return a == b;
}
bool operator()(int64_t a, double b) const {
ADD_FAILURE();
return a == b;
}
bool operator()(double a, double b) const {
ADD_FAILURE();
return a == b;
}
};
struct THash {
using is_transparent = void;
size_t operator()(int64_t i) const { return i; }
size_t operator()(double i) const { return i; }
};
struct TEq {
using is_transparent = void;
bool operator()(int64_t a, int64_t b) const { return a == b; }
bool operator()(double a, int64_t b) const { return a == b; }
bool operator()(int64_t a, double b) const { return a == b; }
bool operator()(double a, double b) const { return a == b; }
};
raw_hash_set<IntPolicy, Hash, Eq, Alloc<int64_t>> s{0, 1, 2};
// It will convert to int64_t before the query.
EXPECT_EQ(1, *s.find(double{1.1}));
raw_hash_set<IntPolicy, THash, TEq, Alloc<int64_t>> ts{0, 1, 2};
// It will try to use the double, and fail to find the object.
EXPECT_TRUE(ts.find(1.1) == ts.end());
}
template <class Table>
using CallFind = decltype(std::declval<Table&>().find(17));
template <class Table>
using CallErase = decltype(std::declval<Table&>().erase(17));
template <class Table>
using CallExtract = decltype(std::declval<Table&>().extract(17));
template <class Table>
using CallPrefetch = decltype(std::declval<Table&>().prefetch(17));
template <class Table>
using CallCount = decltype(std::declval<Table&>().count(17));
template <template <typename> class C, class Table, class = void>
struct VerifyResultOf : std::false_type {};
template <template <typename> class C, class Table>
struct VerifyResultOf<C, Table, absl::void_t<C<Table>>> : std::true_type {};
TEST(Table, HeterogeneousLookupOverloads) {
using NonTransparentTable =
raw_hash_set<StringPolicy, absl::Hash<absl::string_view>,
std::equal_to<absl::string_view>, std::allocator<int>>;
EXPECT_FALSE((VerifyResultOf<CallFind, NonTransparentTable>()));
EXPECT_FALSE((VerifyResultOf<CallErase, NonTransparentTable>()));
EXPECT_FALSE((VerifyResultOf<CallExtract, NonTransparentTable>()));
EXPECT_FALSE((VerifyResultOf<CallPrefetch, NonTransparentTable>()));
EXPECT_FALSE((VerifyResultOf<CallCount, NonTransparentTable>()));
using TransparentTable = raw_hash_set<
StringPolicy,
absl::container_internal::hash_default_hash<absl::string_view>,
absl::container_internal::hash_default_eq<absl::string_view>,
std::allocator<int>>;
EXPECT_TRUE((VerifyResultOf<CallFind, TransparentTable>()));
EXPECT_TRUE((VerifyResultOf<CallErase, TransparentTable>()));
EXPECT_TRUE((VerifyResultOf<CallExtract, TransparentTable>()));
EXPECT_TRUE((VerifyResultOf<CallPrefetch, TransparentTable>()));
EXPECT_TRUE((VerifyResultOf<CallCount, TransparentTable>()));
}
// TODO(alkis): Expand iterator tests.
TEST(Iterator, IsDefaultConstructible) {
StringTable::iterator i;
EXPECT_TRUE(i == StringTable::iterator());
}
TEST(ConstIterator, IsDefaultConstructible) {
StringTable::const_iterator i;
EXPECT_TRUE(i == StringTable::const_iterator());
}
TEST(Iterator, ConvertsToConstIterator) {
StringTable::iterator i;
EXPECT_TRUE(i == StringTable::const_iterator());
}
TEST(Iterator, Iterates) {
IntTable t;
for (size_t i = 3; i != 6; ++i) EXPECT_TRUE(t.emplace(i).second);
EXPECT_THAT(t, UnorderedElementsAre(3, 4, 5));
}
TEST(Table, Merge) {
StringTable t1, t2;
t1.emplace("0", "-0");
t1.emplace("1", "-1");
t2.emplace("0", "~0");
t2.emplace("2", "~2");
EXPECT_THAT(t1, UnorderedElementsAre(Pair("0", "-0"), Pair("1", "-1")));
EXPECT_THAT(t2, UnorderedElementsAre(Pair("0", "~0"), Pair("2", "~2")));
t1.merge(t2);
EXPECT_THAT(t1, UnorderedElementsAre(Pair("0", "-0"), Pair("1", "-1"),
Pair("2", "~2")));
EXPECT_THAT(t2, UnorderedElementsAre(Pair("0", "~0")));
}
Export of internal Abseil changes -- 6e9f93888bbe6718997ed90bbd049f1f3572b066 by Abseil Team <absl-team@google.com>: Fix status to be safe for self move assignment. PiperOrigin-RevId: 346815392 -- 35cae74a977f258e81dfbe925fb5a34cb6632036 by Gennadiy Rozental <rogeeff@google.com>: Eliminate unnecessary access to the global vars. PiperOrigin-RevId: 346777168 -- e7e786c243069060f5d6c1c945decb4b0b83f95b by Andy Getzendanner <durandal@google.com>: Internal change. PiperOrigin-RevId: 346685656 -- 4ccd41c48f1a83cfa20b3ea534f743dd7d788376 by Abseil Team <absl-team@google.com>: Move CordRep Ref() and Unref() logic into cord_internal.h This change moves Ref() and Unref() logic out of cord.cc into cord_internal. The main purpose is to make upcoming ring buffer changes easier to isolate from existing cord.cc code. Notice that this removes the nullptr check from Unref() and now requires it to be non null (which held true most times). We may need to rethink if the 'unref unlikely one' is the common case: are cordreps most likely shared? This may be something between 'root' and non root nodes, i.e., is it more likely for leaf / flat nodes in large cords to be shared than top level cordreps being shared? Vice versa? Benchmarks say that we mostly shouldn't care, the caveat being that atomic ops seem more expensive on upcoming archs (arcadia) so we should error on the side of an extra IsOne() branch saving us single owned atomic ops. PiperOrigin-RevId: 346676121 -- f0babab103b9e60d61ba09482d468985e43eceb3 by Samuel Benzaquen <sbenza@google.com>: Fix iterator based constructor and `.insert` members to only require EmplaceConstructible as the standard specifies. PiperOrigin-RevId: 346616707 -- 8f48eedda02277f9c96a88ed7726e34b557cce20 by Evan Brown <ezb@google.com>: Fix a bug in binary_search_impl when there's a transparent, three-way comparator that has different equivalence classes for different lookup types. Add a new can_have_multiple_equivalent_keys method to share the common logic for these cases. PiperOrigin-RevId: 346605948 -- 649183cb3cc9383431de9c81fb1c0f885d4001ae by Abseil Team <absl-team@google.com>: Add benchmark for accessing a Duration flag. PiperOrigin-RevId: 346594843 -- fefdb046520871af63ce2229e2f7cccfc0483dea by Abseil Team <absl-team@google.com>: Restructure CordReader for upcoming ring buffer changes. PiperOrigin-RevId: 346410642 -- 8b2f50e7da0ebab06ead5f94e366e984ca23cb6a by Abseil Team <absl-team@google.com>: Wire in an internal-only flag to toggle upcoming ring buffer changes on/off for experimentation. PiperOrigin-RevId: 346199111 GitOrigin-RevId: 6e9f93888bbe6718997ed90bbd049f1f3572b066 Change-Id: I8f34866b25a79209cb5448bbb28dd3044111d2e9
4 years ago
TEST(Table, IteratorEmplaceConstructibleRequirement) {
struct Value {
explicit Value(absl::string_view view) : value(view) {}
std::string value;
bool operator==(const Value& other) const { return value == other.value; }
};
struct H {
size_t operator()(const Value& v) const {
return absl::Hash<std::string>{}(v.value);
}
};
struct Table : raw_hash_set<ValuePolicy<Value>, H, std::equal_to<Value>,
std::allocator<Value>> {
using Base = typename Table::raw_hash_set;
using Base::Base;
};
std::string input[3]{"A", "B", "C"};
Table t(std::begin(input), std::end(input));
EXPECT_THAT(t, UnorderedElementsAre(Value{"A"}, Value{"B"}, Value{"C"}));
input[0] = "D";
input[1] = "E";
input[2] = "F";
t.insert(std::begin(input), std::end(input));
EXPECT_THAT(t, UnorderedElementsAre(Value{"A"}, Value{"B"}, Value{"C"},
Value{"D"}, Value{"E"}, Value{"F"}));
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
TEST(Nodes, EmptyNodeType) {
using node_type = StringTable::node_type;
node_type n;
EXPECT_FALSE(n);
EXPECT_TRUE(n.empty());
EXPECT_TRUE((std::is_same<node_type::allocator_type,
StringTable::allocator_type>::value));
}
TEST(Nodes, ExtractInsert) {
constexpr char k0[] = "Very long string zero.";
constexpr char k1[] = "Very long string one.";
constexpr char k2[] = "Very long string two.";
StringTable t = {{k0, ""}, {k1, ""}, {k2, ""}};
EXPECT_THAT(t,
UnorderedElementsAre(Pair(k0, ""), Pair(k1, ""), Pair(k2, "")));
auto node = t.extract(k0);
EXPECT_THAT(t, UnorderedElementsAre(Pair(k1, ""), Pair(k2, "")));
EXPECT_TRUE(node);
EXPECT_FALSE(node.empty());
StringTable t2;
StringTable::insert_return_type res = t2.insert(std::move(node));
EXPECT_TRUE(res.inserted);
EXPECT_THAT(*res.position, Pair(k0, ""));
EXPECT_FALSE(res.node);
EXPECT_THAT(t2, UnorderedElementsAre(Pair(k0, "")));
// Not there.
EXPECT_THAT(t, UnorderedElementsAre(Pair(k1, ""), Pair(k2, "")));
node = t.extract("Not there!");
EXPECT_THAT(t, UnorderedElementsAre(Pair(k1, ""), Pair(k2, "")));
EXPECT_FALSE(node);
// Inserting nothing.
res = t2.insert(std::move(node));
EXPECT_FALSE(res.inserted);
EXPECT_EQ(res.position, t2.end());
EXPECT_FALSE(res.node);
EXPECT_THAT(t2, UnorderedElementsAre(Pair(k0, "")));
t.emplace(k0, "1");
node = t.extract(k0);
// Insert duplicate.
res = t2.insert(std::move(node));
EXPECT_FALSE(res.inserted);
EXPECT_THAT(*res.position, Pair(k0, ""));
EXPECT_TRUE(res.node);
EXPECT_FALSE(node);
}
TEST(Nodes, HintInsert) {
IntTable t = {1, 2, 3};
auto node = t.extract(1);
EXPECT_THAT(t, UnorderedElementsAre(2, 3));
auto it = t.insert(t.begin(), std::move(node));
EXPECT_THAT(t, UnorderedElementsAre(1, 2, 3));
EXPECT_EQ(*it, 1);
EXPECT_FALSE(node);
node = t.extract(2);
EXPECT_THAT(t, UnorderedElementsAre(1, 3));
// reinsert 2 to make the next insert fail.
t.insert(2);
EXPECT_THAT(t, UnorderedElementsAre(1, 2, 3));
it = t.insert(t.begin(), std::move(node));
EXPECT_EQ(*it, 2);
// The node was not emptied by the insert call.
EXPECT_TRUE(node);
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
IntTable MakeSimpleTable(size_t size) {
IntTable t;
while (t.size() < size) t.insert(t.size());
return t;
}
std::vector<int> OrderOfIteration(const IntTable& t) {
return {t.begin(), t.end()};
}
// These IterationOrderChanges tests depend on non-deterministic behavior.
// We are injecting non-determinism from the pointer of the table, but do so in
// a way that only the page matters. We have to retry enough times to make sure
// we are touching different memory pages to cause the ordering to change.
// We also need to keep the old tables around to avoid getting the same memory
// blocks over and over.
TEST(Table, IterationOrderChangesByInstance) {
for (size_t size : {2, 6, 12, 20}) {
const auto reference_table = MakeSimpleTable(size);
const auto reference = OrderOfIteration(reference_table);
std::vector<IntTable> tables;
bool found_difference = false;
for (int i = 0; !found_difference && i < 5000; ++i) {
tables.push_back(MakeSimpleTable(size));
found_difference = OrderOfIteration(tables.back()) != reference;
}
if (!found_difference) {
FAIL()
<< "Iteration order remained the same across many attempts with size "
<< size;
}
}
}
TEST(Table, IterationOrderChangesOnRehash) {
std::vector<IntTable> garbage;
for (int i = 0; i < 5000; ++i) {
auto t = MakeSimpleTable(20);
const auto reference = OrderOfIteration(t);
// Force rehash to the same size.
t.rehash(0);
auto trial = OrderOfIteration(t);
if (trial != reference) {
// We are done.
return;
}
garbage.push_back(std::move(t));
}
FAIL() << "Iteration order remained the same across many attempts.";
}
// Verify that pointers are invalidated as soon as a second element is inserted.
// This prevents dependency on pointer stability on small tables.
TEST(Table, UnstablePointers) {
IntTable table;
const auto addr = [&](int i) {
return reinterpret_cast<uintptr_t>(&*table.find(i));
};
table.insert(0);
const uintptr_t old_ptr = addr(0);
// This causes a rehash.
table.insert(1);
EXPECT_NE(old_ptr, addr(0));
}
// Confirm that we assert if we try to erase() end().
TEST(TableDeathTest, EraseOfEndAsserts) {
// Use an assert with side-effects to figure out if they are actually enabled.
bool assert_enabled = false;
assert([&]() {
assert_enabled = true;
return true;
}());
if (!assert_enabled) return;
IntTable t;
// Extra simple "regexp" as regexp support is highly varied across platforms.
Export of internal Abseil changes -- 291f7ef542f73e4801ab5108014bc02344ef31df by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 437835981 Change-Id: I42fd92e74903894533ac9984d7f622e3ba20f468 -- 2e8caf1a57c50b518e05b4bca48e4fe1bb19af82 by Andy Getzendanner <durandal@google.com>: Internal change PiperOrigin-RevId: 437832673 Change-Id: I61b35089418d01a54cecf161b254b68252bebff3 -- b927482ccc399f7e337b60582988b914d9946e4e by Derek Mauro <dmauro@google.com>: Simplify endian intrinsics for modern compilers All modern compilers have either __builtin_bswapN (gcc, clang) or _byteswap_TYPE (MSVC). The other intrinsic definitions are no longer necessary. PiperOrigin-RevId: 437772295 Change-Id: Ifb3d88ba24b9097f87ceb202272b36d2f5e5117f -- b6782a2247a16d5c14706a74ec577c19963d9f97 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 437373174 Change-Id: I0f77e1780dee90d7a3c32a08d96c4aeb624a57b4 -- a53e0c724e37b0b01515a99bd25394b8e21ffdfc by Derek Mauro <dmauro@google.com>: Unify detection of SSE2 and SSSE3 instruction sets and include the proper headers Fix the intrinsic implementation of FastHexToBufferZeroPad16 in numbers.h which only relies on SSSE3, not SSE 4.2. https://godbolt.org/z/Pf5bn1Yv9 Closes #639 PiperOrigin-RevId: 437286940 Change-Id: Ic97948399b61b91e9c0bccd09313b795b904d714 -- f173f597cb2a75ef2a989f45a496334b85e6f40d by Abseil Team <absl-team@google.com>: Change assertion function to enable clearer error messages. PiperOrigin-RevId: 437227057 Change-Id: If420d2f63b51feef6648762f344d5be012cd9c85 GitOrigin-RevId: 291f7ef542f73e4801ab5108014bc02344ef31df
3 years ago
constexpr char kDeathMsg[] = "erase.. called on invalid iterator";
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
EXPECT_DEATH_IF_SUPPORTED(t.erase(t.end()), kDeathMsg);
}
#if defined(ABSL_INTERNAL_HASHTABLEZ_SAMPLE)
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
TEST(RawHashSamplerTest, Sample) {
// Enable the feature even if the prod default is off.
SetHashtablezEnabled(true);
SetHashtablezSampleParameter(100);
auto& sampler = GlobalHashtablezSampler();
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
size_t start_size = 0;
std::unordered_set<const HashtablezInfo*> preexisting_info;
start_size += sampler.Iterate([&](const HashtablezInfo& info) {
preexisting_info.insert(&info);
++start_size;
});
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
std::vector<IntTable> tables;
for (int i = 0; i < 1000000; ++i) {
tables.emplace_back();
const bool do_reserve = (i % 10 > 5);
const bool do_rehash = !do_reserve && (i % 10 > 0);
if (do_reserve) {
// Don't reserve on all tables.
tables.back().reserve(10 * (i % 10));
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
tables.back().insert(1);
tables.back().insert(i % 5);
if (do_rehash) {
// Rehash some other tables.
tables.back().rehash(10 * (i % 10));
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
}
size_t end_size = 0;
std::unordered_map<size_t, int> observed_checksums;
std::unordered_map<ssize_t, int> reservations;
end_size += sampler.Iterate([&](const HashtablezInfo& info) {
if (preexisting_info.count(&info) == 0) {
observed_checksums[info.hashes_bitwise_xor.load(
std::memory_order_relaxed)]++;
reservations[info.max_reserve.load(std::memory_order_relaxed)]++;
}
Export of internal Abseil changes -- 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
3 years ago
EXPECT_EQ(info.inline_element_size, sizeof(int64_t));
++end_size;
});
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
EXPECT_NEAR((end_size - start_size) / static_cast<double>(tables.size()),
0.01, 0.005);
EXPECT_EQ(observed_checksums.size(), 5);
for (const auto& [_, count] : observed_checksums) {
EXPECT_NEAR((100 * count) / static_cast<double>(tables.size()), 0.2, 0.05);
}
EXPECT_EQ(reservations.size(), 10);
for (const auto& [reservation, count] : reservations) {
EXPECT_GE(reservation, 0);
EXPECT_LT(reservation, 100);
EXPECT_NEAR((100 * count) / static_cast<double>(tables.size()), 0.1, 0.05)
<< reservation;
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
}
#endif // ABSL_INTERNAL_HASHTABLEZ_SAMPLE
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
TEST(RawHashSamplerTest, DoNotSampleCustomAllocators) {
// Enable the feature even if the prod default is off.
SetHashtablezEnabled(true);
SetHashtablezSampleParameter(100);
auto& sampler = GlobalHashtablezSampler();
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
size_t start_size = 0;
start_size += sampler.Iterate([&](const HashtablezInfo&) { ++start_size; });
std::vector<CustomAllocIntTable> tables;
for (int i = 0; i < 1000000; ++i) {
tables.emplace_back();
tables.back().insert(1);
}
size_t end_size = 0;
end_size += sampler.Iterate([&](const HashtablezInfo&) { ++end_size; });
EXPECT_NEAR((end_size - start_size) / static_cast<double>(tables.size()),
0.00, 0.001);
}
#ifdef ABSL_HAVE_ADDRESS_SANITIZER
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
TEST(Sanitizer, PoisoningUnused) {
IntTable t;
t.reserve(5);
// Insert something to force an allocation.
int64_t& v1 = *t.insert(0).first;
// Make sure there is something to test.
ASSERT_GT(t.capacity(), 1);
int64_t* slots = RawHashSetTestOnlyAccess::GetSlots(t);
for (size_t i = 0; i < t.capacity(); ++i) {
EXPECT_EQ(slots + i != &v1, __asan_address_is_poisoned(slots + i));
}
}
TEST(Sanitizer, PoisoningOnErase) {
IntTable t;
int64_t& v = *t.insert(0).first;
EXPECT_FALSE(__asan_address_is_poisoned(&v));
t.erase(0);
EXPECT_TRUE(__asan_address_is_poisoned(&v));
}
#endif // ABSL_HAVE_ADDRESS_SANITIZER
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
Export of internal Abseil changes -- b1fc72630aaa81c8395c3b22ba267d938fe29a2e by Derek Mauro <dmauro@google.com>: Fix -Wdeprecated-copy warnings from Clang 13. Example: error: definition of implicit copy assignment operator for 'UDT' is deprecated because it has a user-declared copy constructor [-Werror,-Wdeprecated-copy] PiperOrigin-RevId: 380058303 -- 0422744812b1a2010d9eea5b17fbe89f3441b66b by Evan Brown <ezb@google.com>: Change the "full table!" asserts in raw_hash_set to use `<= capacity` instead of `< capacity`. If we add support for non-power-of-two-minus-one capacities, this is the correct thing to assert. For example, consider: Group::kWidth = 8, capacity_ = 8, ctrl_ = {kEmpty, 1, 2, 3, 4, 5, 6, 7, kSentinel, kEmpty, 1, 2, 3, 4, 5, 6}. In this case, if we do an unsuccessful lookup with H2 mapping to slot 1, then the first Group will contain {1, 2, 3, 4, 5, 6, 7, kSentinel} so we need to continue to the second Group (at which point seq.index() == 8 == capacity_) to find a kEmpty. Note: this is a no-op change for now since we never have `capacity % Group::kWidth == 0`. PiperOrigin-RevId: 380033480 -- 40628c34d540356de65fabb16c1439c0ec7a0764 by Abseil Team <absl-team@google.com>: Drop out-of-date documentation about `absl::FixedArray`'s allocator support PiperOrigin-RevId: 379811653 -- e7ad047863ae55c9b7aec0753cfc527a4ea614bc by Evan Brown <ezb@google.com>: Fix a bug in ConvertDeletedToEmptyAndFullToDeleted in which we were copying 1 more cloned control byte than actually exists. When alignof(slot_type)>1, this wouldn't cause a problem because the extra byte is padding. Also change loop bounds to not rely on the fact that capacity_+1 is a multiple of Group::kWidth. PiperOrigin-RevId: 379311830 -- 1a3ba500fb2c33205854eb9258cd6e0fb1061bca by Martijn Vels <mvels@google.com>: Change Ring, EXTERNAL and FLAT tag values to be consecutive values The purpose of this change is to have FLAT = EXTERNAL + 1. Especially in the ring and btree alternative code, there is a common check if a node is a 'plain' edge (EXTERNAL or FLAT), or 'something else'. This change can make that check a single branch, i.e., instead of 'tag == EXTERNAL || tag >= FLAT', we can simply check for 'tag >= EXTERNAL'. Likewise we have some cases where we check for RING, EXTERNAL or FLAT, so we align RING + 1 with EXTERNAL. PiperOrigin-RevId: 379291576 -- 0c78e65ca4d85244b106c3f8e24cf268e09e72a3 by Benjamin Barenblat <bbaren@google.com>: Round a double multiplication before casting it to integer The code static_cast<int>(x * y) (for double x and y) performs a double multiplication into a temporary that, by standard, may have excess precision. The subsequent cast to int discards the excess precision. However, the cast may examine the excess precision during conversion, producing surprising results like static_cast<int>(1.7 * 10) == 16 on certain systems. Correct this case by explicitly rounding 1.7 * 10 before casting it. PiperOrigin-RevId: 378922064 GitOrigin-RevId: b1fc72630aaa81c8395c3b22ba267d938fe29a2e Change-Id: Ica708a006921118673e78d5fd2d61fe0fb0894d1
4 years ago
TEST(Table, AlignOne) {
// We previously had a bug in which we were copying a control byte over the
// first slot when alignof(value_type) is 1. We test repeated
// insertions/erases and verify that the behavior is correct.
Uint8Table t;
std::unordered_set<uint8_t> verifier; // NOLINT
// Do repeated insertions/erases from the table.
for (int64_t i = 0; i < 100000; ++i) {
SCOPED_TRACE(i);
const uint8_t u = (i * -i) & 0xFF;
auto it = t.find(u);
auto verifier_it = verifier.find(u);
if (it == t.end()) {
ASSERT_EQ(verifier_it, verifier.end());
t.insert(u);
verifier.insert(u);
} else {
ASSERT_NE(verifier_it, verifier.end());
t.erase(it);
verifier.erase(verifier_it);
}
}
EXPECT_EQ(t.size(), verifier.size());
for (uint8_t u : t) {
EXPECT_EQ(verifier.count(u), 1);
}
}
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
} // namespace
} // namespace container_internal
ABSL_NAMESPACE_END
} // namespace absl