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.

1292 lines
49 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.
//
// -----------------------------------------------------------------------------
// File: hash.h
// -----------------------------------------------------------------------------
//
#ifndef ABSL_HASH_INTERNAL_HASH_H_
#define ABSL_HASH_INTERNAL_HASH_H_
#include <algorithm>
#include <array>
#include <bitset>
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 <cstddef>
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 <cstring>
#include <deque>
#include <forward_list>
#include <functional>
#include <iterator>
#include <limits>
#include <list>
#include <map>
#include <memory>
#include <set>
#include <string>
#include <tuple>
#include <type_traits>
#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 <utility>
#include <vector>
Export of internal Abseil changes -- 4ff721439234e91caf6f7b772e5f554e7dd423c8 by Benjamin Barenblat <bbaren@google.com>: Remove endian-sensitivity from hash slow path Prior to this commit, the Abseil hash fast path was endian-agnostic, but the slow path assumed a little-endian platform. Change the slow path to be endian-correct, ensuring that values produced by the fast and slow paths are equal even on big-endian systems. PiperOrigin-RevId: 355424258 -- 7f4fe1aa4de46ad0a2ef19fa9c061fc12a7391ed by Abseil Team <absl-team@google.com>: Directly store CordzInfo in the InlineData data contents of InlineRep This greatly reduces the cost of coping and moving cords. Especially the move constructor and move assignment are now back to lean loads and stores without needing any CordzInfo lookups for tracked cords. PiperOrigin-RevId: 355409161 -- 3ca4ca84ed6d98f1e383ffd8d12c28876e905bb3 by Abseil Team <absl-team@google.com>: Add #include <unordered_map> PiperOrigin-RevId: 355386114 -- 30b0ffad0621971b3135148fcc9e183b0dd2a6bb by Abseil Team <absl-team@google.com>: Optimize Cord copy constructor This change avoids double stores of the Cord copy constructor from the zero init of the InlineData / InlineRep contents followed by the assignment and inlines the copy constructor. PiperOrigin-RevId: 355287939 -- 0c043fa7b6e41ca7cefc5edc1e17ad46223e4e77 by CJ Johnson <johnsoncj@google.com>: Now that the absl::Cleanup example returns absl::Status, since we decided on absl::FailedPreconditionError, the precondition should be a positive statement and then the check should be failure to adhere to that positive statement PiperOrigin-RevId: 355216923 -- 9ed922ca5d28fe8790ec6bc0837cf39fbcc92896 by Gennadiy Rozental <rogeeff@google.com>: Do not set mvsc linker flags for clang-cl (fixes #874) Import of https://github.com/abseil/abseil-cpp/pull/891 PiperOrigin-RevId: 355199380 GitOrigin-RevId: 4ff721439234e91caf6f7b772e5f554e7dd423c8 Change-Id: I3d9d2383549720d7a91f9108dfcd979ad6632fce
4 years ago
#include "absl/base/config.h"
#include "absl/base/internal/unaligned_access.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/port.h"
#include "absl/container/fixed_array.h"
Export of internal Abseil changes -- 506fa3e10b3d8399ad937c32ecea26d1ad4e62bb by Abseil Team <absl-team@google.com>: Disable ABSL_HAVE_STD_IS_TRIVIALLY_ASSIGNABLE when GCC < 8.2.0 is used with libc++ PiperOrigin-RevId: 399707056 -- 656b7c7cee87f46a4bc7953618796f82da08e62c by Derek Mauro <dmauro@google.com>: Remove the MSVC flag implementation from flag.h to help clarify that methods on absl::Flag<T> are not part of the public API PiperOrigin-RevId: 399584678 -- a92a9bc156303bc663b84c4b704891ec8f67e333 by Abseil Team <absl-team@google.com>: Get rid of MemcpyIfAllowed while continuing to suppress erroneous warnings PiperOrigin-RevId: 399468864 -- 5f9a66895f707ba001fb51b88c0c6025f8c872a3 by Abseil Team <absl-team@google.com>: Use feature testing to check for availability of invoke_result. Feature testing should be available by C++20, which removes invoke_result. https://en.cppreference.com/w/cpp/feature_test PiperOrigin-RevId: 399447373 -- 946c0a502b4499dbfcabf1ab93ddde0048288fb4 by CJ Johnson <johnsoncj@google.com>: Add rvalue-reference qualifier to the Commit method on ConstructionTransaction PiperOrigin-RevId: 399442206 -- 726a4d036eff49aeb6fd0ca2b1775699b6844395 by Greg Falcon <gfalcon@google.com>: Internal change PiperOrigin-RevId: 399441870 -- 1df6d3f659b88dbac13c3d8e13db23bb3844ece2 by Abseil Team <absl-team@google.com>: Clang-format whitespace changes PiperOrigin-RevId: 399281271 -- 4a828cde95a07421d699ebac775b37810624214f by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 399234071 -- e520c72b34ba2f98668c889139001f8276243d31 by Greg Falcon <gfalcon@google.com>: Import of CCTZ from GitHub. PiperOrigin-RevId: 399233662 GitOrigin-RevId: 506fa3e10b3d8399ad937c32ecea26d1ad4e62bb Change-Id: I92b9176d2387c08eb167f9268efa78b55b8e09c2
3 years ago
#include "absl/hash/internal/city.h"
#include "absl/hash/internal/low_level_hash.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/meta/type_traits.h"
#include "absl/numeric/int128.h"
#include "absl/strings/string_view.h"
#include "absl/types/optional.h"
#include "absl/types/variant.h"
#include "absl/utility/utility.h"
namespace absl {
ABSL_NAMESPACE_BEGIN
class HashState;
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 hash_internal {
// Internal detail: Large buffers are hashed in smaller chunks. This function
// returns the size of these chunks.
constexpr size_t PiecewiseChunkSize() { return 1024; }
// PiecewiseCombiner
//
// PiecewiseCombiner is an internal-only helper class for hashing a piecewise
// buffer of `char` or `unsigned char` as though it were contiguous. This class
// provides two methods:
//
// H add_buffer(state, data, size)
// H finalize(state)
//
// `add_buffer` can be called zero or more times, followed by a single call to
// `finalize`. This will produce the same hash expansion as concatenating each
// buffer piece into a single contiguous buffer, and passing this to
// `H::combine_contiguous`.
//
// Example usage:
// PiecewiseCombiner combiner;
// for (const auto& piece : pieces) {
// state = combiner.add_buffer(std::move(state), piece.data, piece.size);
// }
// return combiner.finalize(std::move(state));
class PiecewiseCombiner {
public:
PiecewiseCombiner() : position_(0) {}
PiecewiseCombiner(const PiecewiseCombiner&) = delete;
PiecewiseCombiner& operator=(const PiecewiseCombiner&) = delete;
// PiecewiseCombiner::add_buffer()
//
// Appends the given range of bytes to the sequence to be hashed, which may
// modify the provided hash state.
template <typename H>
H add_buffer(H state, const unsigned char* data, size_t size);
template <typename H>
H add_buffer(H state, const char* data, size_t size) {
return add_buffer(std::move(state),
reinterpret_cast<const unsigned char*>(data), size);
}
// PiecewiseCombiner::finalize()
//
// Finishes combining the hash sequence, which may may modify the provided
// hash state.
//
// Once finalize() is called, add_buffer() may no longer be called. The
// resulting hash state will be the same as if the pieces passed to
// add_buffer() were concatenated into a single flat buffer, and then provided
// to H::combine_contiguous().
template <typename H>
H finalize(H state);
private:
unsigned char buf_[PiecewiseChunkSize()];
size_t position_;
};
// is_hashable()
//
// Trait class which returns true if T is hashable by the absl::Hash framework.
// Used for the AbslHashValue implementations for composite types below.
template <typename T>
struct is_hashable;
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
// HashStateBase
//
// An internal implementation detail that contains common implementation details
// for all of the "hash state objects" objects generated by Abseil. This is not
// a public API; users should not create classes that inherit from this.
//
// A hash state object is the template argument `H` passed to `AbslHashValue`.
// It represents an intermediate state in the computation of an unspecified hash
// algorithm. `HashStateBase` provides a CRTP style base class for hash state
// implementations. Developers adding type support for `absl::Hash` should not
// rely on any parts of the state object other than the following member
// functions:
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
//
// * HashStateBase::combine()
// * HashStateBase::combine_contiguous()
// * HashStateBase::combine_unordered()
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
//
// A derived hash state class of type `H` must provide a public member function
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
// with a signature similar to the following:
//
// `static H combine_contiguous(H state, const unsigned char*, size_t)`.
//
// It must also provide a private template method named RunCombineUnordered.
//
// A "consumer" is a 1-arg functor returning void. Its argument is a reference
// to an inner hash state object, and it may be called multiple times. When
// called, the functor consumes the entropy from the provided state object,
// and resets that object to its empty state.
//
// A "combiner" is a stateless 2-arg functor returning void. Its arguments are
// an inner hash state object and an ElementStateConsumer functor. A combiner
// uses the provided inner hash state object to hash each element of the
// container, passing the inner hash state object to the consumer after hashing
// each element.
//
// Given these definitions, a derived hash state class of type H
// must provide a private template method with a signature similar to the
// following:
//
// `template <typename CombinerT>`
// `static H RunCombineUnordered(H outer_state, CombinerT combiner)`
//
// This function is responsible for constructing the inner state object and
// providing a consumer to the combiner. It uses side effects of the consumer
// and combiner to mix the state of each element in an order-independent manner,
// and uses this to return an updated value of `outer_state`.
//
// This inside-out approach generates efficient object code in the normal case,
// but allows us to use stack storage to implement the absl::HashState type
// erasure mechanism (avoiding heap allocations while hashing).
//
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
// `HashStateBase` will provide a complete implementation for a hash state
// object in terms of these two methods.
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
//
// Example:
//
// // Use CRTP to define your derived class.
// struct MyHashState : HashStateBase<MyHashState> {
// static H combine_contiguous(H state, const unsigned char*, size_t);
// using MyHashState::HashStateBase::combine;
// using MyHashState::HashStateBase::combine_contiguous;
// using MyHashState::HashStateBase::combine_unordered;
// private:
// template <typename CombinerT>
// static H RunCombineUnordered(H state, CombinerT combiner);
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 H>
class HashStateBase {
public:
// HashStateBase::combine()
//
// Combines an arbitrary number of values into a hash state, returning the
// updated state.
//
// Each of the value types `T` must be separately hashable by the Abseil
// hashing framework.
//
// NOTE:
//
// state = H::combine(std::move(state), value1, value2, value3);
//
// is guaranteed to produce the same hash expansion as:
//
// state = H::combine(std::move(state), value1);
// state = H::combine(std::move(state), value2);
// state = H::combine(std::move(state), value3);
template <typename T, typename... Ts>
static H combine(H state, const T& value, const Ts&... values);
static H combine(H state) { return state; }
// HashStateBase::combine_contiguous()
//
// Combines a contiguous array of `size` elements into a hash state, returning
// the updated state.
//
// NOTE:
//
// state = H::combine_contiguous(std::move(state), data, size);
//
// is NOT guaranteed to produce the same hash expansion as a for-loop (it may
// perform internal optimizations). If you need this guarantee, use the
// for-loop instead.
template <typename T>
static H combine_contiguous(H state, const T* data, size_t size);
template <typename I>
static H combine_unordered(H state, I begin, I end);
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 AbslInternalPiecewiseCombiner = PiecewiseCombiner;
template <typename T>
using is_hashable = absl::hash_internal::is_hashable<T>;
private:
// Common implementation of the iteration step of a "combiner", as described
// above.
template <typename I>
struct CombineUnorderedCallback {
I begin;
I end;
template <typename InnerH, typename ElementStateConsumer>
void operator()(InnerH inner_state, ElementStateConsumer cb) {
for (; begin != end; ++begin) {
inner_state = H::combine(std::move(inner_state), *begin);
cb(inner_state);
}
}
};
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
};
// is_uniquely_represented
//
// `is_uniquely_represented<T>` is a trait class that indicates whether `T`
// is uniquely represented.
//
// A type is "uniquely represented" if two equal values of that type are
// guaranteed to have the same bytes in their underlying storage. In other
// words, if `a == b`, then `memcmp(&a, &b, sizeof(T))` is guaranteed to be
// zero. This property cannot be detected automatically, so this trait is false
// by default, but can be specialized by types that wish to assert that they are
// uniquely represented. This makes them eligible for certain optimizations.
//
// If you have any doubt whatsoever, do not specialize this template.
// The default is completely safe, and merely disables some optimizations
// that will not matter for most types. Specializing this template,
// on the other hand, can be very hazardous.
//
// To be uniquely represented, a type must not have multiple ways of
// representing the same value; for example, float and double are not
// uniquely represented, because they have distinct representations for
// +0 and -0. Furthermore, the type's byte representation must consist
// solely of user-controlled data, with no padding bits and no compiler-
// controlled data such as vptrs or sanitizer metadata. This is usually
// very difficult to guarantee, because in most cases the compiler can
// insert data and padding bits at its own discretion.
//
// If you specialize this template for a type `T`, you must do so in the file
// that defines that type (or in this file). If you define that specialization
// anywhere else, `is_uniquely_represented<T>` could have different meanings
// in different places.
//
// The Enable parameter is meaningless; it is provided as a convenience,
// to support certain SFINAE techniques when defining specializations.
template <typename T, typename Enable = void>
struct is_uniquely_represented : std::false_type {};
// is_uniquely_represented<unsigned char>
//
// unsigned char is a synonym for "byte", so it is guaranteed to be
// uniquely represented.
template <>
struct is_uniquely_represented<unsigned char> : std::true_type {};
// is_uniquely_represented for non-standard integral types
//
// Integral types other than bool should be uniquely represented on any
// platform that this will plausibly be ported to.
template <typename Integral>
struct is_uniquely_represented<
Integral, typename std::enable_if<std::is_integral<Integral>::value>::type>
: std::true_type {};
// is_uniquely_represented<bool>
//
//
template <>
struct is_uniquely_represented<bool> : std::false_type {};
// hash_bytes()
//
// Convenience function that combines `hash_state` with the byte representation
// of `value`.
template <typename H, typename T>
H hash_bytes(H hash_state, const T& value) {
const unsigned char* start = reinterpret_cast<const unsigned char*>(&value);
return H::combine_contiguous(std::move(hash_state), start, sizeof(value));
}
// -----------------------------------------------------------------------------
// AbslHashValue for Basic Types
// -----------------------------------------------------------------------------
// Note: Default `AbslHashValue` implementations live in `hash_internal`. This
// allows us to block lexical scope lookup when doing an unqualified call to
// `AbslHashValue` below. User-defined implementations of `AbslHashValue` can
// only be found via ADL.
// AbslHashValue() for hashing bool values
//
// We use SFINAE to ensure that this overload only accepts bool, not types that
// are convertible to bool.
template <typename H, typename B>
typename std::enable_if<std::is_same<B, bool>::value, H>::type AbslHashValue(
H hash_state, B value) {
return H::combine(std::move(hash_state),
static_cast<unsigned char>(value ? 1 : 0));
}
// AbslHashValue() for hashing enum values
template <typename H, typename Enum>
typename std::enable_if<std::is_enum<Enum>::value, H>::type AbslHashValue(
H hash_state, Enum e) {
// In practice, we could almost certainly just invoke hash_bytes directly,
// but it's possible that a sanitizer might one day want to
// store data in the unused bits of an enum. To avoid that risk, we
// convert to the underlying type before hashing. Hopefully this will get
// optimized away; if not, we can reopen discussion with c-toolchain-team.
return H::combine(std::move(hash_state),
static_cast<typename std::underlying_type<Enum>::type>(e));
}
// AbslHashValue() for hashing floating-point values
template <typename H, typename Float>
typename std::enable_if<std::is_same<Float, float>::value ||
std::is_same<Float, double>::value,
H>::type
AbslHashValue(H hash_state, Float value) {
return hash_internal::hash_bytes(std::move(hash_state),
value == 0 ? 0 : value);
}
// Long double has the property that it might have extra unused bytes in it.
// For example, in x86 sizeof(long double)==16 but it only really uses 80-bits
// of it. This means we can't use hash_bytes on a long double and have to
// convert it to something else first.
template <typename H, typename LongDouble>
typename std::enable_if<std::is_same<LongDouble, long double>::value, H>::type
AbslHashValue(H hash_state, LongDouble value) {
const int category = std::fpclassify(value);
switch (category) {
case FP_INFINITE:
// Add the sign bit to differentiate between +Inf and -Inf
hash_state = H::combine(std::move(hash_state), std::signbit(value));
break;
case FP_NAN:
case FP_ZERO:
default:
// Category is enough for these.
break;
case FP_NORMAL:
case FP_SUBNORMAL:
// We can't convert `value` directly to double because this would have
// undefined behavior if the value is out of range.
// std::frexp gives us a value in the range (-1, -.5] or [.5, 1) that is
// guaranteed to be in range for `double`. The truncation is
// implementation defined, but that works as long as it is deterministic.
int exp;
auto mantissa = static_cast<double>(std::frexp(value, &exp));
hash_state = H::combine(std::move(hash_state), mantissa, exp);
}
return H::combine(std::move(hash_state), category);
}
// AbslHashValue() for hashing pointers
template <typename H, typename T>
H AbslHashValue(H hash_state, T* ptr) {
auto v = reinterpret_cast<uintptr_t>(ptr);
// Due to alignment, pointers tend to have low bits as zero, and the next few
// bits follow a pattern since they are also multiples of some base value.
// Mixing the pointer twice helps prevent stuck low bits for certain alignment
// values.
return H::combine(std::move(hash_state), v, v);
}
// AbslHashValue() for hashing nullptr_t
template <typename H>
H AbslHashValue(H hash_state, std::nullptr_t) {
return H::combine(std::move(hash_state), static_cast<void*>(nullptr));
}
Export of internal Abseil changes -- f0b7d230a90c82c6fee7adcb46a213d2582b6a7b by Martijn Vels <mvels@google.com>: Optimize substring logic now that CONCAT is removed This CL adds a static Substring() method to CordRepSubstring, and implements substring logic in cord.cc in terms of the new function. This cleans up various helper functions and logic remaining from previous complex CONCAT logic that is no longer needed. PiperOrigin-RevId: 431756805 Change-Id: I39c875b5af119916780e68598c7fc619fb2e8476 -- fa7d1bedf0e1244303844869a332c2a5dbd9ac0f by Derek Mauro <dmauro@google.com>: Allow macro expansion within ABSL_FLAG and ABSL_DECLARE_FLAG args PiperOrigin-RevId: 431721184 Change-Id: I6e19713fb541205d796f940998db5ee25178d55e -- 1b328badd92304ed1c634f23e1c191be57b7bb15 by Laramie Leavitt <lar@google.com>: Add #include for std:: types PiperOrigin-RevId: 431546757 Change-Id: I75efbcd3c77e6f53e4db66494101d30d670d988e -- e25323b299d4d3840218702860f537cdd2a3926f by Thomas Köppe <tkoeppe@google.com>: Add hashing support for pointers to member. Also add tests for function pointers to the existing "pointer" test. PiperOrigin-RevId: 431067588 Change-Id: I3686010635d9fee34c47a418b72402e10737cdbc -- ab27b012a61cf10109fd51932b3b0b05ee78f32f by Laramie Leavitt <lar@google.com>: Avoid use of std::pow in ChiSquare test. PiperOrigin-RevId: 431015830 Change-Id: Idd767ff2f51009ee171de48757207b38330ffea3 -- 28c359135d89061177958580fe4a7493802499cb by Laramie Leavitt <lar@google.com>: Add #include <type_traits> for std::false_type PiperOrigin-RevId: 431005757 Change-Id: I85a6a918778601e19512aaea744424cf39018521 -- a920730f23669479d92e3a696d65d0bc3a5b1de1 by Laramie Leavitt <lar@google.com>: #include <utility> for std::declval PiperOrigin-RevId: 431004934 Change-Id: I295237b2d44e9a15e4083698ea121b68ce0a1bb7 GitOrigin-RevId: f0b7d230a90c82c6fee7adcb46a213d2582b6a7b
3 years ago
// AbslHashValue() for hashing pointers-to-member
template <typename H, typename T, typename C>
H AbslHashValue(H hash_state, T C::* ptr) {
auto salient_ptm_size = [](std::size_t n) -> std::size_t {
#if defined(_MSC_VER)
// Pointers-to-member-function on MSVC consist of one pointer plus 0, 1, 2,
// or 3 ints. In 64-bit mode, they are 8-byte aligned and thus can contain
// padding (namely when they have 1 or 3 ints). The value below is a lower
// bound on the number of salient, non-padding bytes that we use for
// hashing.
if (alignof(T C::*) == alignof(int)) {
// No padding when all subobjects have the same size as the total
// alignment. This happens in 32-bit mode.
return n;
} else {
// Padding for 1 int (size 16) or 3 ints (size 24).
// With 2 ints, the size is 16 with no padding, which we pessimize.
return n == 24 ? 20 : n == 16 ? 12 : n;
}
#else
// On other platforms, we assume that pointers-to-members do not have
// padding.
#ifdef __cpp_lib_has_unique_object_representations
static_assert(std::has_unique_object_representations<T C::*>::value);
Export of internal Abseil changes -- f0b7d230a90c82c6fee7adcb46a213d2582b6a7b by Martijn Vels <mvels@google.com>: Optimize substring logic now that CONCAT is removed This CL adds a static Substring() method to CordRepSubstring, and implements substring logic in cord.cc in terms of the new function. This cleans up various helper functions and logic remaining from previous complex CONCAT logic that is no longer needed. PiperOrigin-RevId: 431756805 Change-Id: I39c875b5af119916780e68598c7fc619fb2e8476 -- fa7d1bedf0e1244303844869a332c2a5dbd9ac0f by Derek Mauro <dmauro@google.com>: Allow macro expansion within ABSL_FLAG and ABSL_DECLARE_FLAG args PiperOrigin-RevId: 431721184 Change-Id: I6e19713fb541205d796f940998db5ee25178d55e -- 1b328badd92304ed1c634f23e1c191be57b7bb15 by Laramie Leavitt <lar@google.com>: Add #include for std:: types PiperOrigin-RevId: 431546757 Change-Id: I75efbcd3c77e6f53e4db66494101d30d670d988e -- e25323b299d4d3840218702860f537cdd2a3926f by Thomas Köppe <tkoeppe@google.com>: Add hashing support for pointers to member. Also add tests for function pointers to the existing "pointer" test. PiperOrigin-RevId: 431067588 Change-Id: I3686010635d9fee34c47a418b72402e10737cdbc -- ab27b012a61cf10109fd51932b3b0b05ee78f32f by Laramie Leavitt <lar@google.com>: Avoid use of std::pow in ChiSquare test. PiperOrigin-RevId: 431015830 Change-Id: Idd767ff2f51009ee171de48757207b38330ffea3 -- 28c359135d89061177958580fe4a7493802499cb by Laramie Leavitt <lar@google.com>: Add #include <type_traits> for std::false_type PiperOrigin-RevId: 431005757 Change-Id: I85a6a918778601e19512aaea744424cf39018521 -- a920730f23669479d92e3a696d65d0bc3a5b1de1 by Laramie Leavitt <lar@google.com>: #include <utility> for std::declval PiperOrigin-RevId: 431004934 Change-Id: I295237b2d44e9a15e4083698ea121b68ce0a1bb7 GitOrigin-RevId: f0b7d230a90c82c6fee7adcb46a213d2582b6a7b
3 years ago
#endif // __cpp_lib_has_unique_object_representations
return n;
#endif
};
return H::combine_contiguous(std::move(hash_state),
reinterpret_cast<unsigned char*>(&ptr),
salient_ptm_size(sizeof ptr));
}
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
// -----------------------------------------------------------------------------
// AbslHashValue for Composite Types
// -----------------------------------------------------------------------------
// AbslHashValue() for hashing pairs
template <typename H, typename T1, typename T2>
typename std::enable_if<is_hashable<T1>::value && is_hashable<T2>::value,
H>::type
AbslHashValue(H hash_state, const std::pair<T1, T2>& p) {
return H::combine(std::move(hash_state), p.first, p.second);
}
// hash_tuple()
//
// Helper function for hashing a tuple. The third argument should
// be an index_sequence running from 0 to tuple_size<Tuple> - 1.
template <typename H, typename Tuple, size_t... Is>
H hash_tuple(H hash_state, const Tuple& t, absl::index_sequence<Is...>) {
return H::combine(std::move(hash_state), std::get<Is>(t)...);
}
// AbslHashValue for hashing tuples
template <typename H, typename... Ts>
#if defined(_MSC_VER)
// This SFINAE gets MSVC confused under some conditions. Let's just disable it
// for now.
H
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
#else // _MSC_VER
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
typename std::enable_if<absl::conjunction<is_hashable<Ts>...>::value, H>::type
#endif // _MSC_VER
AbslHashValue(H hash_state, const std::tuple<Ts...>& t) {
return hash_internal::hash_tuple(std::move(hash_state), t,
absl::make_index_sequence<sizeof...(Ts)>());
}
// -----------------------------------------------------------------------------
// AbslHashValue for Pointers
// -----------------------------------------------------------------------------
// AbslHashValue for hashing unique_ptr
template <typename H, typename T, typename D>
H AbslHashValue(H hash_state, const std::unique_ptr<T, D>& ptr) {
return H::combine(std::move(hash_state), ptr.get());
}
// AbslHashValue for hashing shared_ptr
template <typename H, typename T>
H AbslHashValue(H hash_state, const std::shared_ptr<T>& ptr) {
return H::combine(std::move(hash_state), ptr.get());
}
// -----------------------------------------------------------------------------
// AbslHashValue for String-Like Types
// -----------------------------------------------------------------------------
// AbslHashValue for hashing strings
//
// All the string-like types supported here provide the same hash expansion for
// the same character sequence. These types are:
//
// - `absl::Cord`
// - `std::string` (and std::basic_string<char, std::char_traits<char>, A> for
// any allocator A)
// - `absl::string_view` and `std::string_view`
//
// For simplicity, we currently support only `char` strings. This support may
// be broadened, if necessary, but with some caution - this overload would
// misbehave in cases where the traits' `eq()` member isn't equivalent to `==`
// on the underlying character type.
template <typename H>
H AbslHashValue(H hash_state, absl::string_view str) {
return H::combine(
H::combine_contiguous(std::move(hash_state), str.data(), str.size()),
str.size());
}
// Support std::wstring, std::u16string and std::u32string.
template <typename Char, typename Alloc, typename H,
typename = absl::enable_if_t<std::is_same<Char, wchar_t>::value ||
std::is_same<Char, char16_t>::value ||
std::is_same<Char, char32_t>::value>>
H AbslHashValue(
H hash_state,
const std::basic_string<Char, std::char_traits<Char>, Alloc>& str) {
return H::combine(
H::combine_contiguous(std::move(hash_state), str.data(), str.size()),
str.size());
}
// -----------------------------------------------------------------------------
// AbslHashValue for Sequence Containers
// -----------------------------------------------------------------------------
// AbslHashValue for hashing std::array
template <typename H, typename T, size_t N>
typename std::enable_if<is_hashable<T>::value, H>::type AbslHashValue(
H hash_state, const std::array<T, N>& array) {
return H::combine_contiguous(std::move(hash_state), array.data(),
array.size());
}
// AbslHashValue for hashing std::deque
template <typename H, typename T, typename Allocator>
typename std::enable_if<is_hashable<T>::value, H>::type AbslHashValue(
H hash_state, const std::deque<T, Allocator>& deque) {
// TODO(gromer): investigate a more efficient implementation taking
// advantage of the chunk structure.
for (const auto& t : deque) {
hash_state = H::combine(std::move(hash_state), t);
}
return H::combine(std::move(hash_state), deque.size());
}
// AbslHashValue for hashing std::forward_list
template <typename H, typename T, typename Allocator>
typename std::enable_if<is_hashable<T>::value, H>::type AbslHashValue(
H hash_state, const std::forward_list<T, Allocator>& list) {
size_t size = 0;
for (const T& t : list) {
hash_state = H::combine(std::move(hash_state), t);
++size;
}
return H::combine(std::move(hash_state), size);
}
// AbslHashValue for hashing std::list
template <typename H, typename T, typename Allocator>
typename std::enable_if<is_hashable<T>::value, H>::type AbslHashValue(
H hash_state, const std::list<T, Allocator>& list) {
for (const auto& t : list) {
hash_state = H::combine(std::move(hash_state), t);
}
return H::combine(std::move(hash_state), list.size());
}
// AbslHashValue for hashing std::vector
//
// Do not use this for vector<bool> on platforms that have a working
// implementation of std::hash. It does not have a .data(), and a fallback for
// std::hash<> is most likely faster.
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 H, typename T, typename Allocator>
typename std::enable_if<is_hashable<T>::value && !std::is_same<T, bool>::value,
H>::type
AbslHashValue(H hash_state, const std::vector<T, Allocator>& vector) {
return H::combine(H::combine_contiguous(std::move(hash_state), vector.data(),
vector.size()),
vector.size());
}
// AbslHashValue special cases for hashing std::vector<bool>
#if defined(ABSL_IS_BIG_ENDIAN) && \
(defined(__GLIBCXX__) || defined(__GLIBCPP__))
// std::hash in libstdc++ does not work correctly with vector<bool> on Big
// Endian platforms therefore we need to implement a custom AbslHashValue for
// it. More details on the bug:
// https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102531
template <typename H, typename T, typename Allocator>
typename std::enable_if<is_hashable<T>::value && std::is_same<T, bool>::value,
H>::type
AbslHashValue(H hash_state, const std::vector<T, Allocator>& vector) {
typename H::AbslInternalPiecewiseCombiner combiner;
for (const auto& i : vector) {
unsigned char c = static_cast<unsigned char>(i);
hash_state = combiner.add_buffer(std::move(hash_state), &c, sizeof(c));
}
return H::combine(combiner.finalize(std::move(hash_state)), vector.size());
}
#else
// When not working around the libstdc++ bug above, we still have to contend
// with the fact that std::hash<vector<bool>> is often poor quality, hashing
// directly on the internal words and on no other state. On these platforms,
// vector<bool>{1, 1} and vector<bool>{1, 1, 0} hash to the same value.
//
// Mixing in the size (as we do in our other vector<> implementations) on top
// of the library-provided hash implementation avoids this QOI issue.
template <typename H, typename T, typename Allocator>
typename std::enable_if<is_hashable<T>::value && std::is_same<T, bool>::value,
H>::type
AbslHashValue(H hash_state, const std::vector<T, Allocator>& vector) {
return H::combine(std::move(hash_state),
std::hash<std::vector<T, Allocator>>{}(vector),
vector.size());
}
#endif
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
// -----------------------------------------------------------------------------
// AbslHashValue for Ordered Associative Containers
// -----------------------------------------------------------------------------
// AbslHashValue for hashing std::map
template <typename H, typename Key, typename T, typename Compare,
typename Allocator>
typename std::enable_if<is_hashable<Key>::value && is_hashable<T>::value,
H>::type
AbslHashValue(H hash_state, const std::map<Key, T, Compare, Allocator>& map) {
for (const auto& t : map) {
hash_state = H::combine(std::move(hash_state), t);
}
return H::combine(std::move(hash_state), map.size());
}
// AbslHashValue for hashing std::multimap
template <typename H, typename Key, typename T, typename Compare,
typename Allocator>
typename std::enable_if<is_hashable<Key>::value && is_hashable<T>::value,
H>::type
AbslHashValue(H hash_state,
const std::multimap<Key, T, Compare, Allocator>& map) {
for (const auto& t : map) {
hash_state = H::combine(std::move(hash_state), t);
}
return H::combine(std::move(hash_state), map.size());
}
// AbslHashValue for hashing std::set
template <typename H, typename Key, typename Compare, typename Allocator>
typename std::enable_if<is_hashable<Key>::value, H>::type AbslHashValue(
H hash_state, const std::set<Key, Compare, Allocator>& set) {
for (const auto& t : set) {
hash_state = H::combine(std::move(hash_state), t);
}
return H::combine(std::move(hash_state), set.size());
}
// AbslHashValue for hashing std::multiset
template <typename H, typename Key, typename Compare, typename Allocator>
typename std::enable_if<is_hashable<Key>::value, H>::type AbslHashValue(
H hash_state, const std::multiset<Key, Compare, Allocator>& set) {
for (const auto& t : set) {
hash_state = H::combine(std::move(hash_state), t);
}
return H::combine(std::move(hash_state), set.size());
}
// -----------------------------------------------------------------------------
// AbslHashValue for Unordered Associative Containers
// -----------------------------------------------------------------------------
// AbslHashValue for hashing std::unordered_set
template <typename H, typename Key, typename Hash, typename KeyEqual,
typename Alloc>
typename std::enable_if<is_hashable<Key>::value, H>::type AbslHashValue(
H hash_state, const std::unordered_set<Key, Hash, KeyEqual, Alloc>& s) {
return H::combine(
H::combine_unordered(std::move(hash_state), s.begin(), s.end()),
s.size());
}
// AbslHashValue for hashing std::unordered_multiset
template <typename H, typename Key, typename Hash, typename KeyEqual,
typename Alloc>
typename std::enable_if<is_hashable<Key>::value, H>::type AbslHashValue(
H hash_state,
const std::unordered_multiset<Key, Hash, KeyEqual, Alloc>& s) {
return H::combine(
H::combine_unordered(std::move(hash_state), s.begin(), s.end()),
s.size());
}
// AbslHashValue for hashing std::unordered_set
template <typename H, typename Key, typename T, typename Hash,
typename KeyEqual, typename Alloc>
typename std::enable_if<is_hashable<Key>::value && is_hashable<T>::value,
H>::type
AbslHashValue(H hash_state,
const std::unordered_map<Key, T, Hash, KeyEqual, Alloc>& s) {
return H::combine(
H::combine_unordered(std::move(hash_state), s.begin(), s.end()),
s.size());
}
// AbslHashValue for hashing std::unordered_multiset
template <typename H, typename Key, typename T, typename Hash,
typename KeyEqual, typename Alloc>
typename std::enable_if<is_hashable<Key>::value && is_hashable<T>::value,
H>::type
AbslHashValue(H hash_state,
const std::unordered_multimap<Key, T, Hash, KeyEqual, Alloc>& s) {
return H::combine(
H::combine_unordered(std::move(hash_state), s.begin(), s.end()),
s.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
// -----------------------------------------------------------------------------
// AbslHashValue for Wrapper Types
// -----------------------------------------------------------------------------
// AbslHashValue for hashing std::reference_wrapper
template <typename H, typename T>
typename std::enable_if<is_hashable<T>::value, H>::type AbslHashValue(
H hash_state, std::reference_wrapper<T> opt) {
return H::combine(std::move(hash_state), opt.get());
}
// AbslHashValue for hashing absl::optional
template <typename H, typename T>
typename std::enable_if<is_hashable<T>::value, H>::type AbslHashValue(
H hash_state, const absl::optional<T>& opt) {
if (opt) hash_state = H::combine(std::move(hash_state), *opt);
return H::combine(std::move(hash_state), opt.has_value());
}
// VariantVisitor
template <typename H>
struct VariantVisitor {
H&& hash_state;
template <typename T>
H operator()(const T& t) const {
return H::combine(std::move(hash_state), t);
}
};
// AbslHashValue for hashing absl::variant
template <typename H, typename... T>
typename std::enable_if<conjunction<is_hashable<T>...>::value, H>::type
AbslHashValue(H hash_state, const absl::variant<T...>& v) {
if (!v.valueless_by_exception()) {
hash_state = absl::visit(VariantVisitor<H>{std::move(hash_state)}, v);
}
return H::combine(std::move(hash_state), v.index());
}
// -----------------------------------------------------------------------------
// AbslHashValue for Other Types
// -----------------------------------------------------------------------------
// AbslHashValue for hashing std::bitset is not defined on Little Endian
// platforms, for the same reason as for vector<bool> (see std::vector above):
// It does not expose the raw bytes, and a fallback to std::hash<> is most
// likely faster.
#if defined(ABSL_IS_BIG_ENDIAN) && \
(defined(__GLIBCXX__) || defined(__GLIBCPP__))
// AbslHashValue for hashing std::bitset
//
// std::hash in libstdc++ does not work correctly with std::bitset on Big Endian
// platforms therefore we need to implement a custom AbslHashValue for it. More
// details on the bug: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102531
template <typename H, size_t N>
H AbslHashValue(H hash_state, const std::bitset<N>& set) {
typename H::AbslInternalPiecewiseCombiner combiner;
for (int i = 0; i < N; i++) {
unsigned char c = static_cast<unsigned char>(set[i]);
hash_state = combiner.add_buffer(std::move(hash_state), &c, sizeof(c));
}
return H::combine(combiner.finalize(std::move(hash_state)), N);
}
#endif
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
// -----------------------------------------------------------------------------
// hash_range_or_bytes()
//
// Mixes all values in the range [data, data+size) into the hash state.
// This overload accepts only uniquely-represented types, and hashes them by
// hashing the entire range of bytes.
template <typename H, typename T>
typename std::enable_if<is_uniquely_represented<T>::value, H>::type
hash_range_or_bytes(H hash_state, const T* data, size_t size) {
const auto* bytes = reinterpret_cast<const unsigned char*>(data);
return H::combine_contiguous(std::move(hash_state), bytes, sizeof(T) * size);
}
// hash_range_or_bytes()
template <typename H, typename T>
typename std::enable_if<!is_uniquely_represented<T>::value, H>::type
hash_range_or_bytes(H hash_state, const T* data, size_t size) {
for (const auto end = data + size; data < end; ++data) {
hash_state = H::combine(std::move(hash_state), *data);
}
return hash_state;
}
#if defined(ABSL_INTERNAL_LEGACY_HASH_NAMESPACE) && \
ABSL_META_INTERNAL_STD_HASH_SFINAE_FRIENDLY_
#define ABSL_HASH_INTERNAL_SUPPORT_LEGACY_HASH_ 1
#else
#define ABSL_HASH_INTERNAL_SUPPORT_LEGACY_HASH_ 0
#endif
// HashSelect
//
// Type trait to select the appropriate hash implementation to use.
// HashSelect::type<T> will give the proper hash implementation, to be invoked
// as:
// HashSelect::type<T>::Invoke(state, value)
// Also, HashSelect::type<T>::value is a boolean equal to `true` if there is a
// valid `Invoke` function. Types that are not hashable will have a ::value of
// `false`.
struct HashSelect {
private:
struct State : HashStateBase<State> {
static State combine_contiguous(State hash_state, const unsigned char*,
size_t);
using State::HashStateBase::combine_contiguous;
};
struct UniquelyRepresentedProbe {
template <typename H, typename T>
static auto Invoke(H state, const T& value)
-> absl::enable_if_t<is_uniquely_represented<T>::value, H> {
return hash_internal::hash_bytes(std::move(state), value);
}
};
struct HashValueProbe {
template <typename H, typename T>
static auto Invoke(H state, const T& value) -> absl::enable_if_t<
std::is_same<H,
decltype(AbslHashValue(std::move(state), value))>::value,
H> {
return AbslHashValue(std::move(state), value);
}
};
struct LegacyHashProbe {
#if ABSL_HASH_INTERNAL_SUPPORT_LEGACY_HASH_
template <typename H, typename T>
static auto Invoke(H state, const T& value) -> absl::enable_if_t<
std::is_convertible<
decltype(ABSL_INTERNAL_LEGACY_HASH_NAMESPACE::hash<T>()(value)),
size_t>::value,
H> {
return hash_internal::hash_bytes(
std::move(state),
ABSL_INTERNAL_LEGACY_HASH_NAMESPACE::hash<T>{}(value));
}
#endif // ABSL_HASH_INTERNAL_SUPPORT_LEGACY_HASH_
};
struct StdHashProbe {
template <typename H, typename T>
static auto Invoke(H state, const T& value)
-> absl::enable_if_t<type_traits_internal::IsHashable<T>::value, H> {
return hash_internal::hash_bytes(std::move(state), std::hash<T>{}(value));
}
};
template <typename Hash, typename T>
struct Probe : Hash {
private:
template <typename H, typename = decltype(H::Invoke(
std::declval<State>(), std::declval<const T&>()))>
static std::true_type Test(int);
template <typename U>
static std::false_type Test(char);
public:
static constexpr bool value = decltype(Test<Hash>(0))::value;
};
public:
// Probe each implementation in order.
// disjunction provides short circuiting wrt instantiation.
template <typename T>
using Apply = absl::disjunction< //
Probe<UniquelyRepresentedProbe, T>, //
Probe<HashValueProbe, T>, //
Probe<LegacyHashProbe, T>, //
Probe<StdHashProbe, T>, //
std::false_type>;
};
template <typename T>
struct is_hashable
: std::integral_constant<bool, HashSelect::template Apply<T>::value> {};
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
// MixingHashState
class ABSL_DLL MixingHashState : public HashStateBase<MixingHashState> {
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
// absl::uint128 is not an alias or a thin wrapper around the intrinsic.
// We use the intrinsic when available to improve performance.
#ifdef ABSL_HAVE_INTRINSIC_INT128
using uint128 = __uint128_t;
#else // ABSL_HAVE_INTRINSIC_INT128
using uint128 = absl::uint128;
#endif // ABSL_HAVE_INTRINSIC_INT128
static constexpr uint64_t kMul =
sizeof(size_t) == 4 ? uint64_t{0xcc9e2d51}
: uint64_t{0x9ddfea08eb382d69};
template <typename T>
using IntegralFastPath =
conjunction<std::is_integral<T>, is_uniquely_represented<T>>;
public:
// Move only
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
MixingHashState(MixingHashState&&) = default;
MixingHashState& operator=(MixingHashState&&) = default;
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 -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
// MixingHashState::combine_contiguous()
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
//
// Fundamental base case for hash recursion: mixes the given range of bytes
// into the hash state.
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
static MixingHashState combine_contiguous(MixingHashState hash_state,
const unsigned char* first,
size_t size) {
return MixingHashState(
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
CombineContiguousImpl(hash_state.state_, first, size,
std::integral_constant<int, sizeof(size_t)>{}));
}
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
using MixingHashState::HashStateBase::combine_contiguous;
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 -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
// MixingHashState::hash()
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 performance reasons in non-opt mode, we specialize this for
// integral types.
// Otherwise we would be instantiating and calling dozens of functions for
// something that is just one multiplication and a couple xor's.
// The result should be the same as running the whole algorithm, but faster.
template <typename T, absl::enable_if_t<IntegralFastPath<T>::value, int> = 0>
static size_t hash(T value) {
return static_cast<size_t>(Mix(Seed(), static_cast<uint64_t>(value)));
}
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
// Overload of MixingHashState::hash()
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, absl::enable_if_t<!IntegralFastPath<T>::value, int> = 0>
static size_t hash(const T& value) {
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
return static_cast<size_t>(combine(MixingHashState{}, value).state_);
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
}
private:
// Invoked only once for a given argument; that plus the fact that this is
// move-only ensures that there is only one non-moved-from object.
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
MixingHashState() : state_(Seed()) {}
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
friend class MixingHashState::HashStateBase;
template <typename CombinerT>
static MixingHashState RunCombineUnordered(MixingHashState state,
CombinerT combiner) {
uint64_t unordered_state = 0;
combiner(MixingHashState{}, [&](MixingHashState& inner_state) {
// Add the hash state of the element to the running total, but mix the
// carry bit back into the low bit. This in intended to avoid losing
// entropy to overflow, especially when unordered_multisets contain
// multiple copies of the same value.
auto element_state = inner_state.state_;
unordered_state += element_state;
if (unordered_state < element_state) {
++unordered_state;
}
inner_state = MixingHashState{};
});
return MixingHashState::combine(std::move(state), unordered_state);
}
// Allow the HashState type-erasure implementation to invoke
// RunCombinedUnordered() directly.
friend class absl::HashState;
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
// Workaround for MSVC bug.
// We make the type copyable to fix the calling convention, even though we
// never actually copy it. Keep it private to not affect the public API of the
// type.
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
MixingHashState(const MixingHashState&) = default;
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 -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
explicit MixingHashState(uint64_t state) : state_(state) {}
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
// Implementation of the base case for combine_contiguous where we actually
// mix the bytes into the state.
// Dispatch to different implementations of the combine_contiguous depending
// on the value of `sizeof(size_t)`.
static uint64_t CombineContiguousImpl(uint64_t state,
const unsigned char* first, size_t len,
std::integral_constant<int, 4>
/* sizeof_size_t */);
static uint64_t CombineContiguousImpl(uint64_t state,
const unsigned char* first, size_t len,
std::integral_constant<int, 8>
/* sizeof_size_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
// Slow dispatch path for calls to CombineContiguousImpl with a size argument
// larger than PiecewiseChunkSize(). Has the same effect as calling
// CombineContiguousImpl() repeatedly with the chunk stride size.
static uint64_t CombineLargeContiguousImpl32(uint64_t state,
const unsigned char* first,
size_t len);
static uint64_t CombineLargeContiguousImpl64(uint64_t state,
const unsigned char* first,
size_t len);
// Reads 9 to 16 bytes from p.
Export of internal Abseil changes -- 4ff721439234e91caf6f7b772e5f554e7dd423c8 by Benjamin Barenblat <bbaren@google.com>: Remove endian-sensitivity from hash slow path Prior to this commit, the Abseil hash fast path was endian-agnostic, but the slow path assumed a little-endian platform. Change the slow path to be endian-correct, ensuring that values produced by the fast and slow paths are equal even on big-endian systems. PiperOrigin-RevId: 355424258 -- 7f4fe1aa4de46ad0a2ef19fa9c061fc12a7391ed by Abseil Team <absl-team@google.com>: Directly store CordzInfo in the InlineData data contents of InlineRep This greatly reduces the cost of coping and moving cords. Especially the move constructor and move assignment are now back to lean loads and stores without needing any CordzInfo lookups for tracked cords. PiperOrigin-RevId: 355409161 -- 3ca4ca84ed6d98f1e383ffd8d12c28876e905bb3 by Abseil Team <absl-team@google.com>: Add #include <unordered_map> PiperOrigin-RevId: 355386114 -- 30b0ffad0621971b3135148fcc9e183b0dd2a6bb by Abseil Team <absl-team@google.com>: Optimize Cord copy constructor This change avoids double stores of the Cord copy constructor from the zero init of the InlineData / InlineRep contents followed by the assignment and inlines the copy constructor. PiperOrigin-RevId: 355287939 -- 0c043fa7b6e41ca7cefc5edc1e17ad46223e4e77 by CJ Johnson <johnsoncj@google.com>: Now that the absl::Cleanup example returns absl::Status, since we decided on absl::FailedPreconditionError, the precondition should be a positive statement and then the check should be failure to adhere to that positive statement PiperOrigin-RevId: 355216923 -- 9ed922ca5d28fe8790ec6bc0837cf39fbcc92896 by Gennadiy Rozental <rogeeff@google.com>: Do not set mvsc linker flags for clang-cl (fixes #874) Import of https://github.com/abseil/abseil-cpp/pull/891 PiperOrigin-RevId: 355199380 GitOrigin-RevId: 4ff721439234e91caf6f7b772e5f554e7dd423c8 Change-Id: I3d9d2383549720d7a91f9108dfcd979ad6632fce
4 years ago
// The least significant 8 bytes are in .first, the rest (zero padded) bytes
// are in .second.
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 std::pair<uint64_t, uint64_t> Read9To16(const unsigned char* p,
size_t len) {
Export of internal Abseil changes -- 4ff721439234e91caf6f7b772e5f554e7dd423c8 by Benjamin Barenblat <bbaren@google.com>: Remove endian-sensitivity from hash slow path Prior to this commit, the Abseil hash fast path was endian-agnostic, but the slow path assumed a little-endian platform. Change the slow path to be endian-correct, ensuring that values produced by the fast and slow paths are equal even on big-endian systems. PiperOrigin-RevId: 355424258 -- 7f4fe1aa4de46ad0a2ef19fa9c061fc12a7391ed by Abseil Team <absl-team@google.com>: Directly store CordzInfo in the InlineData data contents of InlineRep This greatly reduces the cost of coping and moving cords. Especially the move constructor and move assignment are now back to lean loads and stores without needing any CordzInfo lookups for tracked cords. PiperOrigin-RevId: 355409161 -- 3ca4ca84ed6d98f1e383ffd8d12c28876e905bb3 by Abseil Team <absl-team@google.com>: Add #include <unordered_map> PiperOrigin-RevId: 355386114 -- 30b0ffad0621971b3135148fcc9e183b0dd2a6bb by Abseil Team <absl-team@google.com>: Optimize Cord copy constructor This change avoids double stores of the Cord copy constructor from the zero init of the InlineData / InlineRep contents followed by the assignment and inlines the copy constructor. PiperOrigin-RevId: 355287939 -- 0c043fa7b6e41ca7cefc5edc1e17ad46223e4e77 by CJ Johnson <johnsoncj@google.com>: Now that the absl::Cleanup example returns absl::Status, since we decided on absl::FailedPreconditionError, the precondition should be a positive statement and then the check should be failure to adhere to that positive statement PiperOrigin-RevId: 355216923 -- 9ed922ca5d28fe8790ec6bc0837cf39fbcc92896 by Gennadiy Rozental <rogeeff@google.com>: Do not set mvsc linker flags for clang-cl (fixes #874) Import of https://github.com/abseil/abseil-cpp/pull/891 PiperOrigin-RevId: 355199380 GitOrigin-RevId: 4ff721439234e91caf6f7b772e5f554e7dd423c8 Change-Id: I3d9d2383549720d7a91f9108dfcd979ad6632fce
4 years ago
uint64_t low_mem = absl::base_internal::UnalignedLoad64(p);
uint64_t high_mem = absl::base_internal::UnalignedLoad64(p + len - 8);
#ifdef ABSL_IS_LITTLE_ENDIAN
uint64_t most_significant = high_mem;
uint64_t least_significant = low_mem;
#else
uint64_t most_significant = low_mem;
uint64_t least_significant = high_mem;
#endif
return {least_significant, most_significant >> (128 - len * 8)};
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
}
// Reads 4 to 8 bytes from p. Zero pads to fill uint64_t.
static uint64_t Read4To8(const unsigned char* p, size_t len) {
Export of internal Abseil changes -- 4ff721439234e91caf6f7b772e5f554e7dd423c8 by Benjamin Barenblat <bbaren@google.com>: Remove endian-sensitivity from hash slow path Prior to this commit, the Abseil hash fast path was endian-agnostic, but the slow path assumed a little-endian platform. Change the slow path to be endian-correct, ensuring that values produced by the fast and slow paths are equal even on big-endian systems. PiperOrigin-RevId: 355424258 -- 7f4fe1aa4de46ad0a2ef19fa9c061fc12a7391ed by Abseil Team <absl-team@google.com>: Directly store CordzInfo in the InlineData data contents of InlineRep This greatly reduces the cost of coping and moving cords. Especially the move constructor and move assignment are now back to lean loads and stores without needing any CordzInfo lookups for tracked cords. PiperOrigin-RevId: 355409161 -- 3ca4ca84ed6d98f1e383ffd8d12c28876e905bb3 by Abseil Team <absl-team@google.com>: Add #include <unordered_map> PiperOrigin-RevId: 355386114 -- 30b0ffad0621971b3135148fcc9e183b0dd2a6bb by Abseil Team <absl-team@google.com>: Optimize Cord copy constructor This change avoids double stores of the Cord copy constructor from the zero init of the InlineData / InlineRep contents followed by the assignment and inlines the copy constructor. PiperOrigin-RevId: 355287939 -- 0c043fa7b6e41ca7cefc5edc1e17ad46223e4e77 by CJ Johnson <johnsoncj@google.com>: Now that the absl::Cleanup example returns absl::Status, since we decided on absl::FailedPreconditionError, the precondition should be a positive statement and then the check should be failure to adhere to that positive statement PiperOrigin-RevId: 355216923 -- 9ed922ca5d28fe8790ec6bc0837cf39fbcc92896 by Gennadiy Rozental <rogeeff@google.com>: Do not set mvsc linker flags for clang-cl (fixes #874) Import of https://github.com/abseil/abseil-cpp/pull/891 PiperOrigin-RevId: 355199380 GitOrigin-RevId: 4ff721439234e91caf6f7b772e5f554e7dd423c8 Change-Id: I3d9d2383549720d7a91f9108dfcd979ad6632fce
4 years ago
uint32_t low_mem = absl::base_internal::UnalignedLoad32(p);
uint32_t high_mem = absl::base_internal::UnalignedLoad32(p + len - 4);
#ifdef ABSL_IS_LITTLE_ENDIAN
uint32_t most_significant = high_mem;
uint32_t least_significant = low_mem;
#else
uint32_t most_significant = low_mem;
uint32_t least_significant = high_mem;
#endif
return (static_cast<uint64_t>(most_significant) << (len - 4) * 8) |
least_significant;
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
}
// Reads 1 to 3 bytes from p. Zero pads to fill uint32_t.
static uint32_t Read1To3(const unsigned char* p, size_t len) {
Export of internal Abseil changes -- 4ff721439234e91caf6f7b772e5f554e7dd423c8 by Benjamin Barenblat <bbaren@google.com>: Remove endian-sensitivity from hash slow path Prior to this commit, the Abseil hash fast path was endian-agnostic, but the slow path assumed a little-endian platform. Change the slow path to be endian-correct, ensuring that values produced by the fast and slow paths are equal even on big-endian systems. PiperOrigin-RevId: 355424258 -- 7f4fe1aa4de46ad0a2ef19fa9c061fc12a7391ed by Abseil Team <absl-team@google.com>: Directly store CordzInfo in the InlineData data contents of InlineRep This greatly reduces the cost of coping and moving cords. Especially the move constructor and move assignment are now back to lean loads and stores without needing any CordzInfo lookups for tracked cords. PiperOrigin-RevId: 355409161 -- 3ca4ca84ed6d98f1e383ffd8d12c28876e905bb3 by Abseil Team <absl-team@google.com>: Add #include <unordered_map> PiperOrigin-RevId: 355386114 -- 30b0ffad0621971b3135148fcc9e183b0dd2a6bb by Abseil Team <absl-team@google.com>: Optimize Cord copy constructor This change avoids double stores of the Cord copy constructor from the zero init of the InlineData / InlineRep contents followed by the assignment and inlines the copy constructor. PiperOrigin-RevId: 355287939 -- 0c043fa7b6e41ca7cefc5edc1e17ad46223e4e77 by CJ Johnson <johnsoncj@google.com>: Now that the absl::Cleanup example returns absl::Status, since we decided on absl::FailedPreconditionError, the precondition should be a positive statement and then the check should be failure to adhere to that positive statement PiperOrigin-RevId: 355216923 -- 9ed922ca5d28fe8790ec6bc0837cf39fbcc92896 by Gennadiy Rozental <rogeeff@google.com>: Do not set mvsc linker flags for clang-cl (fixes #874) Import of https://github.com/abseil/abseil-cpp/pull/891 PiperOrigin-RevId: 355199380 GitOrigin-RevId: 4ff721439234e91caf6f7b772e5f554e7dd423c8 Change-Id: I3d9d2383549720d7a91f9108dfcd979ad6632fce
4 years ago
unsigned char mem0 = p[0];
unsigned char mem1 = p[len / 2];
unsigned char mem2 = p[len - 1];
#ifdef ABSL_IS_LITTLE_ENDIAN
unsigned char significant2 = mem2;
unsigned char significant1 = mem1;
unsigned char significant0 = mem0;
#else
unsigned char significant2 = mem0;
unsigned char significant1 = mem1;
unsigned char significant0 = mem2;
#endif
return static_cast<uint32_t>(significant0 | //
(significant1 << (len / 2 * 8)) | //
(significant2 << ((len - 1) * 8)));
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
}
ABSL_ATTRIBUTE_ALWAYS_INLINE static uint64_t Mix(uint64_t state, uint64_t v) {
// Though the 128-bit product on AArch64 needs two instructions, it is
// still a good balance between speed and hash quality.
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 MultType =
absl::conditional_t<sizeof(size_t) == 4, uint64_t, uint128>;
// We do the addition in 64-bit space to make sure the 128-bit
// multiplication is fast. If we were to do it as MultType the compiler has
// to assume that the high word is non-zero and needs to perform 2
// multiplications instead of one.
MultType m = state + v;
m *= kMul;
return static_cast<uint64_t>(m ^ (m >> (sizeof(m) * 8 / 2)));
}
// An extern to avoid bloat on a direct call to LowLevelHash() with fixed
// values for both the seed and salt parameters.
static uint64_t LowLevelHashImpl(const unsigned char* data, size_t len);
ABSL_ATTRIBUTE_ALWAYS_INLINE static uint64_t Hash64(const unsigned char* data,
size_t len) {
#ifdef ABSL_HAVE_INTRINSIC_INT128
return LowLevelHashImpl(data, len);
#else
Export of internal Abseil changes -- 506fa3e10b3d8399ad937c32ecea26d1ad4e62bb by Abseil Team <absl-team@google.com>: Disable ABSL_HAVE_STD_IS_TRIVIALLY_ASSIGNABLE when GCC < 8.2.0 is used with libc++ PiperOrigin-RevId: 399707056 -- 656b7c7cee87f46a4bc7953618796f82da08e62c by Derek Mauro <dmauro@google.com>: Remove the MSVC flag implementation from flag.h to help clarify that methods on absl::Flag<T> are not part of the public API PiperOrigin-RevId: 399584678 -- a92a9bc156303bc663b84c4b704891ec8f67e333 by Abseil Team <absl-team@google.com>: Get rid of MemcpyIfAllowed while continuing to suppress erroneous warnings PiperOrigin-RevId: 399468864 -- 5f9a66895f707ba001fb51b88c0c6025f8c872a3 by Abseil Team <absl-team@google.com>: Use feature testing to check for availability of invoke_result. Feature testing should be available by C++20, which removes invoke_result. https://en.cppreference.com/w/cpp/feature_test PiperOrigin-RevId: 399447373 -- 946c0a502b4499dbfcabf1ab93ddde0048288fb4 by CJ Johnson <johnsoncj@google.com>: Add rvalue-reference qualifier to the Commit method on ConstructionTransaction PiperOrigin-RevId: 399442206 -- 726a4d036eff49aeb6fd0ca2b1775699b6844395 by Greg Falcon <gfalcon@google.com>: Internal change PiperOrigin-RevId: 399441870 -- 1df6d3f659b88dbac13c3d8e13db23bb3844ece2 by Abseil Team <absl-team@google.com>: Clang-format whitespace changes PiperOrigin-RevId: 399281271 -- 4a828cde95a07421d699ebac775b37810624214f by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 399234071 -- e520c72b34ba2f98668c889139001f8276243d31 by Greg Falcon <gfalcon@google.com>: Import of CCTZ from GitHub. PiperOrigin-RevId: 399233662 GitOrigin-RevId: 506fa3e10b3d8399ad937c32ecea26d1ad4e62bb Change-Id: I92b9176d2387c08eb167f9268efa78b55b8e09c2
3 years ago
return hash_internal::CityHash64(reinterpret_cast<const char*>(data), len);
#endif
}
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
// Seed()
//
// A non-deterministic seed.
//
// The current purpose of this seed is to generate non-deterministic results
// and prevent having users depend on the particular hash values.
// It is not meant as a security feature right now, but it leaves the door
// open to upgrade it to a true per-process random seed. A true random seed
// costs more and we don't need to pay for that right now.
//
// On platforms with ASLR, we take advantage of it to make a per-process
// random value.
// See https://en.wikipedia.org/wiki/Address_space_layout_randomization
//
// On other platforms this is still going to be non-deterministic but most
// probably per-build and not per-process.
ABSL_ATTRIBUTE_ALWAYS_INLINE static uint64_t Seed() {
#if (!defined(__clang__) || __clang_major__ > 11) && \
!defined(__apple_build_version__)
return static_cast<uint64_t>(reinterpret_cast<uintptr_t>(&kSeed));
#else
// Workaround the absence of
// https://github.com/llvm/llvm-project/commit/bc15bf66dcca76cc06fe71fca35b74dc4d521021.
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 static_cast<uint64_t>(reinterpret_cast<uintptr_t>(kSeed));
#endif
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 const void* const kSeed;
uint64_t state_;
};
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
// MixingHashState::CombineContiguousImpl()
inline uint64_t MixingHashState::CombineContiguousImpl(
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
uint64_t state, const unsigned char* first, size_t len,
std::integral_constant<int, 4> /* sizeof_size_t */) {
// For large values we use CityHash, for small ones we just use a
// multiplicative hash.
uint64_t v;
if (len > 8) {
if (ABSL_PREDICT_FALSE(len > PiecewiseChunkSize())) {
return CombineLargeContiguousImpl32(state, first, len);
}
Export of internal Abseil changes -- 506fa3e10b3d8399ad937c32ecea26d1ad4e62bb by Abseil Team <absl-team@google.com>: Disable ABSL_HAVE_STD_IS_TRIVIALLY_ASSIGNABLE when GCC < 8.2.0 is used with libc++ PiperOrigin-RevId: 399707056 -- 656b7c7cee87f46a4bc7953618796f82da08e62c by Derek Mauro <dmauro@google.com>: Remove the MSVC flag implementation from flag.h to help clarify that methods on absl::Flag<T> are not part of the public API PiperOrigin-RevId: 399584678 -- a92a9bc156303bc663b84c4b704891ec8f67e333 by Abseil Team <absl-team@google.com>: Get rid of MemcpyIfAllowed while continuing to suppress erroneous warnings PiperOrigin-RevId: 399468864 -- 5f9a66895f707ba001fb51b88c0c6025f8c872a3 by Abseil Team <absl-team@google.com>: Use feature testing to check for availability of invoke_result. Feature testing should be available by C++20, which removes invoke_result. https://en.cppreference.com/w/cpp/feature_test PiperOrigin-RevId: 399447373 -- 946c0a502b4499dbfcabf1ab93ddde0048288fb4 by CJ Johnson <johnsoncj@google.com>: Add rvalue-reference qualifier to the Commit method on ConstructionTransaction PiperOrigin-RevId: 399442206 -- 726a4d036eff49aeb6fd0ca2b1775699b6844395 by Greg Falcon <gfalcon@google.com>: Internal change PiperOrigin-RevId: 399441870 -- 1df6d3f659b88dbac13c3d8e13db23bb3844ece2 by Abseil Team <absl-team@google.com>: Clang-format whitespace changes PiperOrigin-RevId: 399281271 -- 4a828cde95a07421d699ebac775b37810624214f by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 399234071 -- e520c72b34ba2f98668c889139001f8276243d31 by Greg Falcon <gfalcon@google.com>: Import of CCTZ from GitHub. PiperOrigin-RevId: 399233662 GitOrigin-RevId: 506fa3e10b3d8399ad937c32ecea26d1ad4e62bb Change-Id: I92b9176d2387c08eb167f9268efa78b55b8e09c2
3 years ago
v = hash_internal::CityHash32(reinterpret_cast<const char*>(first), len);
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 (len >= 4) {
v = Read4To8(first, len);
} else if (len > 0) {
v = Read1To3(first, len);
} else {
// Empty ranges have no effect.
return state;
}
return Mix(state, v);
}
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
// Overload of MixingHashState::CombineContiguousImpl()
inline uint64_t MixingHashState::CombineContiguousImpl(
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
uint64_t state, const unsigned char* first, size_t len,
std::integral_constant<int, 8> /* sizeof_size_t */) {
// For large values we use LowLevelHash or CityHash depending on the platform,
// for small ones we just use a multiplicative hash.
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
uint64_t v;
if (len > 16) {
if (ABSL_PREDICT_FALSE(len > PiecewiseChunkSize())) {
return CombineLargeContiguousImpl64(state, first, len);
}
v = Hash64(first, len);
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 (len > 8) {
auto p = Read9To16(first, len);
state = Mix(state, p.first);
v = p.second;
} else if (len >= 4) {
v = Read4To8(first, len);
} else if (len > 0) {
v = Read1To3(first, len);
} else {
// Empty ranges have no effect.
return state;
}
return Mix(state, v);
}
struct AggregateBarrier {};
// HashImpl
// Add a private base class to make sure this type is not an aggregate.
// Aggregates can be aggregate initialized even if the default constructor is
// deleted.
struct PoisonedHash : private AggregateBarrier {
PoisonedHash() = delete;
PoisonedHash(const PoisonedHash&) = delete;
PoisonedHash& operator=(const PoisonedHash&) = delete;
};
template <typename T>
struct HashImpl {
Export of internal Abseil changes -- 9fc37c11b9e46287acef00ee06ed9adcba54dd13 by Greg Falcon <gfalcon@google.com>: Rename absl::hash_internal::HashState to absl::hash_internal::MixingHashState. Before this change, we had two classes named HashState: absl::HashState, the public API used for type erasure, and absl::hash_internal::HashState, the internal concrete implementation ordinarily used. The internal class used to be named `CityHashState`, but we renamed it to `HashState` it when we changed underlying hash implementation to wyhash. This inadvertent naming conflict made the code much harder to read, and this change intends to undo that. PiperOrigin-RevId: 373481959 -- 4aec55ffddebd085c239352a2e20721091f719a1 by Greg Falcon <gfalcon@google.com>: Introduce absl::HashOf(), a convenience wrapper around absl::Hash that calculates hashes from the values of its arguments. PiperOrigin-RevId: 373461406 -- 86b5fd8db50bbc8bd0aa9258523527381fe0445d by Abseil Team <absl-team@google.com>: Improve speed of BlockingCounter by making its most common path lock free. With the new implementation, the fast path of BlockingCounter::DecrementCount() is only a fetch_sub operation. This is most times much more efficient than the previous implementation (full mutex lock/unlock). As a matter of fact, in most actual usecases in practice, the waiter thread is already waiting on the Wait() call when DecrementCount() is called, which makes Mutex::Unlock() take the slow path as there's a waiter thread that it might need to wake up. PiperOrigin-RevId: 373394164 -- 65c876be5eac0cd32583ff8535ede4109d39cf3f by Martijn Vels <mvels@google.com>: Move the 'sample copied cord' logic into MaybeTrackCord(), This changes move the logic for selecting if a cord should remain being sampled from Cord to CordzInfo::MaybeTrackCord, and updates the documentation for the latter method. PiperOrigin-RevId: 373363168 -- e84410bd0aada293a81dfb82656c952e209e21fb by Martijn Vels <mvels@google.com>: Add check for the first call to cordz_should_profile() for each thread. This prevents the first cord of a newly created thread to be always sampled, which is a 'bad' kind of determinism for sampling. PiperOrigin-RevId: 373229768 -- bf09c589dc099ac8f4af780bf7e609c53c27574c by Samuel Benzaquen <sbenza@google.com>: Refactor the Flags structure into an enum. This gives us more control over the representation and allows for easier merging during parsing. PiperOrigin-RevId: 373163038 -- b947b0c51083b7b6508284b5d31819596c91729e by Derek Mauro <dmauro@google.com>: Fixes warnings about shadowed variables Fixes #956 PiperOrigin-RevId: 373158133 GitOrigin-RevId: 9fc37c11b9e46287acef00ee06ed9adcba54dd13 Change-Id: I91f35699f9bf439d1a870c6493946a310afe088c
4 years ago
size_t operator()(const T& value) const {
return MixingHashState::hash(value);
}
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 Hash
: absl::conditional_t<is_hashable<T>::value, HashImpl<T>, PoisonedHash> {};
template <typename H>
template <typename T, typename... Ts>
H HashStateBase<H>::combine(H state, const T& value, const Ts&... values) {
return H::combine(hash_internal::HashSelect::template Apply<T>::Invoke(
std::move(state), value),
values...);
}
// HashStateBase::combine_contiguous()
template <typename H>
template <typename T>
H HashStateBase<H>::combine_contiguous(H state, const T* data, size_t size) {
return hash_internal::hash_range_or_bytes(std::move(state), data, size);
}
// HashStateBase::combine_unordered()
template <typename H>
template <typename I>
H HashStateBase<H>::combine_unordered(H state, I begin, I end) {
return H::RunCombineUnordered(std::move(state),
CombineUnorderedCallback<I>{begin, end});
}
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
// HashStateBase::PiecewiseCombiner::add_buffer()
template <typename H>
H PiecewiseCombiner::add_buffer(H state, const unsigned char* data,
size_t size) {
if (position_ + size < PiecewiseChunkSize()) {
// This partial chunk does not fill our existing buffer
memcpy(buf_ + position_, data, size);
position_ += size;
return state;
}
// If the buffer is partially filled we need to complete the buffer
// and hash it.
if (position_ != 0) {
const size_t bytes_needed = PiecewiseChunkSize() - position_;
memcpy(buf_ + position_, data, bytes_needed);
state = H::combine_contiguous(std::move(state), buf_, PiecewiseChunkSize());
data += bytes_needed;
size -= bytes_needed;
}
// Hash whatever chunks we can without copying
while (size >= PiecewiseChunkSize()) {
state = H::combine_contiguous(std::move(state), data, PiecewiseChunkSize());
data += PiecewiseChunkSize();
size -= PiecewiseChunkSize();
}
// Fill the buffer with the remainder
memcpy(buf_, data, size);
position_ = size;
return state;
}
// HashStateBase::PiecewiseCombiner::finalize()
template <typename H>
H PiecewiseCombiner::finalize(H state) {
// Hash the remainder left in the buffer, which may be empty
return H::combine_contiguous(std::move(state), buf_, position_);
}
} // namespace hash_internal
ABSL_NAMESPACE_END
} // namespace absl
#endif // ABSL_HASH_INTERNAL_HASH_H_