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.

1166 lines
38 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 2017 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: int128.h
// -----------------------------------------------------------------------------
//
// This header file defines 128-bit integer types, `uint128` and `int128`.
//
// TODO(absl-team): This module is inconsistent as many inline `uint128` methods
// are defined in this file, while many inline `int128` methods are defined in
// the `int128_*_intrinsic.inc` files.
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
#ifndef ABSL_NUMERIC_INT128_H_
#define ABSL_NUMERIC_INT128_H_
#include <cassert>
#include <cmath>
#include <cstdint>
#include <cstring>
#include <iosfwd>
#include <limits>
#include <utility>
#include "absl/base/config.h"
#include "absl/base/macros.h"
#include "absl/base/port.h"
#if defined(_MSC_VER)
// In very old versions of MSVC and when the /Zc:wchar_t flag is off, wchar_t is
// a typedef for unsigned short. Otherwise wchar_t is mapped to the __wchar_t
// builtin type. We need to make sure not to define operator wchar_t()
// alongside operator unsigned short() in these instances.
#define ABSL_INTERNAL_WCHAR_T __wchar_t
#if defined(_M_X64) && !defined(_M_ARM64EC)
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 <intrin.h>
#pragma intrinsic(_umul128)
#endif // defined(_M_X64)
#else // defined(_MSC_VER)
#define ABSL_INTERNAL_WCHAR_T wchar_t
#endif // defined(_MSC_VER)
namespace absl {
ABSL_NAMESPACE_BEGIN
class int128;
// uint128
//
// An unsigned 128-bit integer type. The API is meant to mimic an intrinsic type
// as closely as is practical, including exhibiting undefined behavior in
// analogous cases (e.g. division by zero). This type is intended to be a
// drop-in replacement once C++ supports an intrinsic `uint128_t` type; when
// that occurs, existing well-behaved uses of `uint128` will continue to work
// using that new type.
//
// Note: code written with this type will continue to compile once `uint128_t`
// is introduced, provided the replacement helper functions
// `Uint128(Low|High)64()` and `MakeUint128()` are made.
//
// A `uint128` supports the following:
//
// * Implicit construction from integral types
// * Explicit conversion to integral types
//
// Additionally, if your compiler supports `__int128`, `uint128` is
// interoperable with that type. (Abseil checks for this compatibility through
// the `ABSL_HAVE_INTRINSIC_INT128` macro.)
//
// However, a `uint128` differs from intrinsic integral types in the following
// ways:
//
// * Errors on implicit conversions that do not preserve value (such as
// loss of precision when converting to float values).
// * Requires explicit construction from and conversion to floating point
// types.
// * Conversion to integral types requires an explicit static_cast() to
// mimic use of the `-Wnarrowing` compiler flag.
// * The alignment requirement of `uint128` may differ from that of an
// intrinsic 128-bit integer type depending on platform and build
// configuration.
//
// Example:
//
// float y = absl::Uint128Max(); // Error. uint128 cannot be implicitly
// // converted to float.
//
// absl::uint128 v;
// uint64_t i = v; // Error
// uint64_t i = static_cast<uint64_t>(v); // OK
//
class
#if defined(ABSL_HAVE_INTRINSIC_INT128)
alignas(unsigned __int128)
#endif // ABSL_HAVE_INTRINSIC_INT128
uint128 {
public:
uint128() = default;
// Constructors from arithmetic types
constexpr uint128(int v); // NOLINT(runtime/explicit)
constexpr uint128(unsigned int v); // NOLINT(runtime/explicit)
constexpr uint128(long v); // NOLINT(runtime/int)
constexpr uint128(unsigned long v); // NOLINT(runtime/int)
constexpr uint128(long long v); // NOLINT(runtime/int)
constexpr uint128(unsigned long long v); // NOLINT(runtime/int)
#ifdef ABSL_HAVE_INTRINSIC_INT128
constexpr uint128(__int128 v); // NOLINT(runtime/explicit)
constexpr uint128(unsigned __int128 v); // NOLINT(runtime/explicit)
#endif // ABSL_HAVE_INTRINSIC_INT128
constexpr uint128(int128 v); // NOLINT(runtime/explicit)
explicit uint128(float v);
explicit uint128(double v);
explicit uint128(long double v);
// Assignment operators from arithmetic types
uint128& operator=(int v);
uint128& operator=(unsigned int v);
uint128& operator=(long v); // NOLINT(runtime/int)
uint128& operator=(unsigned long v); // NOLINT(runtime/int)
uint128& operator=(long long v); // NOLINT(runtime/int)
uint128& operator=(unsigned long long v); // NOLINT(runtime/int)
#ifdef ABSL_HAVE_INTRINSIC_INT128
uint128& operator=(__int128 v);
uint128& operator=(unsigned __int128 v);
#endif // ABSL_HAVE_INTRINSIC_INT128
uint128& operator=(int128 v);
// Conversion operators to other arithmetic types
constexpr explicit operator bool() const;
constexpr explicit operator char() const;
constexpr explicit operator signed char() const;
constexpr explicit operator unsigned char() const;
constexpr explicit operator char16_t() const;
constexpr explicit operator char32_t() const;
constexpr explicit operator ABSL_INTERNAL_WCHAR_T() const;
constexpr explicit operator short() const; // NOLINT(runtime/int)
// NOLINTNEXTLINE(runtime/int)
constexpr explicit operator unsigned short() const;
constexpr explicit operator int() const;
constexpr explicit operator unsigned int() const;
constexpr explicit operator long() const; // NOLINT(runtime/int)
// NOLINTNEXTLINE(runtime/int)
constexpr explicit operator unsigned long() const;
// NOLINTNEXTLINE(runtime/int)
constexpr explicit operator long long() const;
// NOLINTNEXTLINE(runtime/int)
constexpr explicit operator unsigned long long() const;
#ifdef ABSL_HAVE_INTRINSIC_INT128
constexpr explicit operator __int128() const;
constexpr explicit operator unsigned __int128() const;
#endif // ABSL_HAVE_INTRINSIC_INT128
explicit operator float() const;
explicit operator double() const;
explicit operator long double() const;
// Trivial copy constructor, assignment operator and destructor.
// Arithmetic operators.
uint128& operator+=(uint128 other);
uint128& operator-=(uint128 other);
uint128& operator*=(uint128 other);
// Long division/modulo for uint128.
uint128& operator/=(uint128 other);
uint128& operator%=(uint128 other);
uint128 operator++(int);
uint128 operator--(int);
uint128& operator<<=(int);
uint128& operator>>=(int);
uint128& operator&=(uint128 other);
uint128& operator|=(uint128 other);
uint128& operator^=(uint128 other);
uint128& operator++();
uint128& operator--();
// Uint128Low64()
//
// Returns the lower 64-bit value of a `uint128` value.
friend constexpr uint64_t Uint128Low64(uint128 v);
// Uint128High64()
//
// Returns the higher 64-bit value of a `uint128` value.
friend constexpr uint64_t Uint128High64(uint128 v);
// MakeUInt128()
//
// Constructs a `uint128` numeric value from two 64-bit unsigned integers.
// Note that this factory function is the only way to construct a `uint128`
// from integer values greater than 2^64.
//
// Example:
//
// absl::uint128 big = absl::MakeUint128(1, 0);
friend constexpr uint128 MakeUint128(uint64_t high, uint64_t low);
// Uint128Max()
//
// Returns the highest value for a 128-bit unsigned integer.
friend constexpr uint128 Uint128Max();
// Support for absl::Hash.
template <typename H>
friend H AbslHashValue(H h, uint128 v) {
return H::combine(std::move(h), Uint128High64(v), Uint128Low64(v));
}
private:
constexpr uint128(uint64_t high, uint64_t low);
// TODO(strel) Update implementation to use __int128 once all users of
// uint128 are fixed to not depend on alignof(uint128) == 8. Also add
// alignas(16) to class definition to keep alignment consistent across
// platforms.
#if defined(ABSL_IS_LITTLE_ENDIAN)
uint64_t lo_;
uint64_t hi_;
#elif defined(ABSL_IS_BIG_ENDIAN)
uint64_t hi_;
uint64_t lo_;
#else // byte order
#error "Unsupported byte order: must be little-endian or big-endian."
#endif // byte order
};
// Prefer to use the constexpr `Uint128Max()`.
//
// TODO(absl-team) deprecate kuint128max once migration tool is released.
ABSL_DLL extern const uint128 kuint128max;
// allow uint128 to be logged
std::ostream& operator<<(std::ostream& os, uint128 v);
// TODO(strel) add operator>>(std::istream&, uint128)
constexpr uint128 Uint128Max() {
return uint128((std::numeric_limits<uint64_t>::max)(),
(std::numeric_limits<uint64_t>::max)());
}
ABSL_NAMESPACE_END
} // namespace absl
// Specialized numeric_limits for uint128.
namespace std {
template <>
class numeric_limits<absl::uint128> {
public:
static constexpr bool is_specialized = true;
static constexpr bool is_signed = false;
static constexpr bool is_integer = true;
static constexpr bool is_exact = true;
static constexpr bool has_infinity = false;
static constexpr bool has_quiet_NaN = false;
static constexpr bool has_signaling_NaN = false;
static constexpr float_denorm_style has_denorm = denorm_absent;
static constexpr bool has_denorm_loss = false;
static constexpr float_round_style round_style = round_toward_zero;
static constexpr bool is_iec559 = false;
static constexpr bool is_bounded = true;
static constexpr bool is_modulo = true;
static constexpr int digits = 128;
static constexpr int digits10 = 38;
static constexpr int max_digits10 = 0;
static constexpr int radix = 2;
static constexpr int min_exponent = 0;
static constexpr int min_exponent10 = 0;
static constexpr int max_exponent = 0;
static constexpr int max_exponent10 = 0;
#ifdef ABSL_HAVE_INTRINSIC_INT128
static constexpr bool traps = numeric_limits<unsigned __int128>::traps;
#else // ABSL_HAVE_INTRINSIC_INT128
static constexpr bool traps = numeric_limits<uint64_t>::traps;
#endif // ABSL_HAVE_INTRINSIC_INT128
static constexpr bool tinyness_before = false;
static constexpr absl::uint128 (min)() { return 0; }
static constexpr absl::uint128 lowest() { return 0; }
static constexpr absl::uint128 (max)() { return absl::Uint128Max(); }
static constexpr absl::uint128 epsilon() { return 0; }
static constexpr absl::uint128 round_error() { return 0; }
static constexpr absl::uint128 infinity() { return 0; }
static constexpr absl::uint128 quiet_NaN() { return 0; }
static constexpr absl::uint128 signaling_NaN() { return 0; }
static constexpr absl::uint128 denorm_min() { return 0; }
};
} // namespace std
namespace absl {
ABSL_NAMESPACE_BEGIN
// int128
//
// A signed 128-bit integer type. The API is meant to mimic an intrinsic
// integral type as closely as is practical, including exhibiting undefined
// behavior in analogous cases (e.g. division by zero).
//
// An `int128` supports the following:
//
// * Implicit construction from integral types
// * Explicit conversion to integral types
//
// However, an `int128` differs from intrinsic integral types in the following
// ways:
//
// * It is not implicitly convertible to other integral types.
// * Requires explicit construction from and conversion to floating point
// types.
// Additionally, if your compiler supports `__int128`, `int128` is
// interoperable with that type. (Abseil checks for this compatibility through
// the `ABSL_HAVE_INTRINSIC_INT128` macro.)
//
// The design goal for `int128` is that it will be compatible with a future
// `int128_t`, if that type becomes a part of the standard.
//
// Example:
//
// float y = absl::int128(17); // Error. int128 cannot be implicitly
// // converted to float.
//
// absl::int128 v;
// int64_t i = v; // Error
// int64_t i = static_cast<int64_t>(v); // OK
//
class int128 {
public:
int128() = default;
// Constructors from arithmetic types
constexpr int128(int v); // NOLINT(runtime/explicit)
constexpr int128(unsigned int v); // NOLINT(runtime/explicit)
constexpr int128(long v); // NOLINT(runtime/int)
constexpr int128(unsigned long v); // NOLINT(runtime/int)
constexpr int128(long long v); // NOLINT(runtime/int)
constexpr int128(unsigned long long v); // NOLINT(runtime/int)
#ifdef ABSL_HAVE_INTRINSIC_INT128
constexpr int128(__int128 v); // NOLINT(runtime/explicit)
constexpr explicit int128(unsigned __int128 v);
#endif // ABSL_HAVE_INTRINSIC_INT128
constexpr explicit int128(uint128 v);
explicit int128(float v);
explicit int128(double v);
explicit int128(long double v);
// Assignment operators from arithmetic types
int128& operator=(int v);
int128& operator=(unsigned int v);
int128& operator=(long v); // NOLINT(runtime/int)
int128& operator=(unsigned long v); // NOLINT(runtime/int)
int128& operator=(long long v); // NOLINT(runtime/int)
int128& operator=(unsigned long long v); // NOLINT(runtime/int)
#ifdef ABSL_HAVE_INTRINSIC_INT128
int128& operator=(__int128 v);
#endif // ABSL_HAVE_INTRINSIC_INT128
// Conversion operators to other arithmetic types
constexpr explicit operator bool() const;
constexpr explicit operator char() const;
constexpr explicit operator signed char() const;
constexpr explicit operator unsigned char() const;
constexpr explicit operator char16_t() const;
constexpr explicit operator char32_t() const;
constexpr explicit operator ABSL_INTERNAL_WCHAR_T() const;
constexpr explicit operator short() const; // NOLINT(runtime/int)
// NOLINTNEXTLINE(runtime/int)
constexpr explicit operator unsigned short() const;
constexpr explicit operator int() const;
constexpr explicit operator unsigned int() const;
constexpr explicit operator long() const; // NOLINT(runtime/int)
// NOLINTNEXTLINE(runtime/int)
constexpr explicit operator unsigned long() const;
// NOLINTNEXTLINE(runtime/int)
constexpr explicit operator long long() const;
// NOLINTNEXTLINE(runtime/int)
constexpr explicit operator unsigned long long() const;
#ifdef ABSL_HAVE_INTRINSIC_INT128
constexpr explicit operator __int128() const;
constexpr explicit operator unsigned __int128() const;
#endif // ABSL_HAVE_INTRINSIC_INT128
explicit operator float() const;
explicit operator double() const;
explicit operator long double() const;
// Trivial copy constructor, assignment operator and destructor.
// Arithmetic operators
int128& operator+=(int128 other);
int128& operator-=(int128 other);
int128& operator*=(int128 other);
int128& operator/=(int128 other);
int128& operator%=(int128 other);
int128 operator++(int); // postfix increment: i++
int128 operator--(int); // postfix decrement: i--
int128& operator++(); // prefix increment: ++i
int128& operator--(); // prefix decrement: --i
int128& operator&=(int128 other);
int128& operator|=(int128 other);
int128& operator^=(int128 other);
int128& operator<<=(int amount);
int128& operator>>=(int amount);
// Int128Low64()
//
// Returns the lower 64-bit value of a `int128` value.
friend constexpr uint64_t Int128Low64(int128 v);
// Int128High64()
//
// Returns the higher 64-bit value of a `int128` value.
friend constexpr int64_t Int128High64(int128 v);
// MakeInt128()
//
// Constructs a `int128` numeric value from two 64-bit integers. Note that
// signedness is conveyed in the upper `high` value.
//
// (absl::int128(1) << 64) * high + low
//
// Note that this factory function is the only way to construct a `int128`
// from integer values greater than 2^64 or less than -2^64.
//
// Example:
//
// absl::int128 big = absl::MakeInt128(1, 0);
// absl::int128 big_n = absl::MakeInt128(-1, 0);
friend constexpr int128 MakeInt128(int64_t high, uint64_t low);
// Int128Max()
//
// Returns the maximum value for a 128-bit signed integer.
friend constexpr int128 Int128Max();
// Int128Min()
//
// Returns the minimum value for a 128-bit signed integer.
friend constexpr int128 Int128Min();
// Support for absl::Hash.
template <typename H>
friend H AbslHashValue(H h, int128 v) {
return H::combine(std::move(h), Int128High64(v), Int128Low64(v));
}
private:
constexpr int128(int64_t high, uint64_t low);
#if defined(ABSL_HAVE_INTRINSIC_INT128)
__int128 v_;
#else // ABSL_HAVE_INTRINSIC_INT128
#if defined(ABSL_IS_LITTLE_ENDIAN)
uint64_t lo_;
int64_t hi_;
#elif defined(ABSL_IS_BIG_ENDIAN)
int64_t hi_;
uint64_t lo_;
#else // byte order
#error "Unsupported byte order: must be little-endian or big-endian."
#endif // byte order
#endif // ABSL_HAVE_INTRINSIC_INT128
};
std::ostream& operator<<(std::ostream& os, int128 v);
// TODO(absl-team) add operator>>(std::istream&, int128)
constexpr int128 Int128Max() {
return int128((std::numeric_limits<int64_t>::max)(),
(std::numeric_limits<uint64_t>::max)());
}
constexpr int128 Int128Min() {
return int128((std::numeric_limits<int64_t>::min)(), 0);
}
ABSL_NAMESPACE_END
} // namespace absl
// Specialized numeric_limits for int128.
namespace std {
template <>
class numeric_limits<absl::int128> {
public:
static constexpr bool is_specialized = true;
static constexpr bool is_signed = true;
static constexpr bool is_integer = true;
static constexpr bool is_exact = true;
static constexpr bool has_infinity = false;
static constexpr bool has_quiet_NaN = false;
static constexpr bool has_signaling_NaN = false;
static constexpr float_denorm_style has_denorm = denorm_absent;
static constexpr bool has_denorm_loss = false;
static constexpr float_round_style round_style = round_toward_zero;
static constexpr bool is_iec559 = false;
static constexpr bool is_bounded = true;
static constexpr bool is_modulo = false;
static constexpr int digits = 127;
static constexpr int digits10 = 38;
static constexpr int max_digits10 = 0;
static constexpr int radix = 2;
static constexpr int min_exponent = 0;
static constexpr int min_exponent10 = 0;
static constexpr int max_exponent = 0;
static constexpr int max_exponent10 = 0;
#ifdef ABSL_HAVE_INTRINSIC_INT128
static constexpr bool traps = numeric_limits<__int128>::traps;
#else // ABSL_HAVE_INTRINSIC_INT128
static constexpr bool traps = numeric_limits<uint64_t>::traps;
#endif // ABSL_HAVE_INTRINSIC_INT128
static constexpr bool tinyness_before = false;
static constexpr absl::int128 (min)() { return absl::Int128Min(); }
static constexpr absl::int128 lowest() { return absl::Int128Min(); }
static constexpr absl::int128 (max)() { return absl::Int128Max(); }
static constexpr absl::int128 epsilon() { return 0; }
static constexpr absl::int128 round_error() { return 0; }
static constexpr absl::int128 infinity() { return 0; }
static constexpr absl::int128 quiet_NaN() { return 0; }
static constexpr absl::int128 signaling_NaN() { return 0; }
static constexpr absl::int128 denorm_min() { return 0; }
};
} // namespace std
// --------------------------------------------------------------------------
// Implementation details follow
// --------------------------------------------------------------------------
namespace absl {
ABSL_NAMESPACE_BEGIN
constexpr uint128 MakeUint128(uint64_t high, uint64_t low) {
return uint128(high, low);
}
// Assignment from integer types.
inline uint128& uint128::operator=(int v) { return *this = uint128(v); }
inline uint128& uint128::operator=(unsigned int v) {
return *this = uint128(v);
}
inline uint128& uint128::operator=(long v) { // NOLINT(runtime/int)
return *this = uint128(v);
}
// NOLINTNEXTLINE(runtime/int)
inline uint128& uint128::operator=(unsigned long v) {
return *this = uint128(v);
}
// NOLINTNEXTLINE(runtime/int)
inline uint128& uint128::operator=(long long v) {
return *this = uint128(v);
}
// NOLINTNEXTLINE(runtime/int)
inline uint128& uint128::operator=(unsigned long long v) {
return *this = uint128(v);
}
#ifdef ABSL_HAVE_INTRINSIC_INT128
inline uint128& uint128::operator=(__int128 v) {
return *this = uint128(v);
}
inline uint128& uint128::operator=(unsigned __int128 v) {
return *this = uint128(v);
}
#endif // ABSL_HAVE_INTRINSIC_INT128
inline uint128& uint128::operator=(int128 v) {
return *this = uint128(v);
}
// Arithmetic operators.
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr uint128 operator<<(uint128 lhs, int amount);
constexpr uint128 operator>>(uint128 lhs, int amount);
constexpr uint128 operator+(uint128 lhs, uint128 rhs);
constexpr uint128 operator-(uint128 lhs, uint128 rhs);
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
uint128 operator*(uint128 lhs, uint128 rhs);
uint128 operator/(uint128 lhs, uint128 rhs);
uint128 operator%(uint128 lhs, uint128 rhs);
inline uint128& uint128::operator<<=(int amount) {
*this = *this << amount;
return *this;
}
inline uint128& uint128::operator>>=(int amount) {
*this = *this >> amount;
return *this;
}
inline uint128& uint128::operator+=(uint128 other) {
*this = *this + other;
return *this;
}
inline uint128& uint128::operator-=(uint128 other) {
*this = *this - other;
return *this;
}
inline uint128& uint128::operator*=(uint128 other) {
*this = *this * other;
return *this;
}
inline uint128& uint128::operator/=(uint128 other) {
*this = *this / other;
return *this;
}
inline uint128& uint128::operator%=(uint128 other) {
*this = *this % other;
return *this;
}
constexpr uint64_t Uint128Low64(uint128 v) { return v.lo_; }
constexpr uint64_t Uint128High64(uint128 v) { return v.hi_; }
// Constructors from integer types.
#if defined(ABSL_IS_LITTLE_ENDIAN)
constexpr uint128::uint128(uint64_t high, uint64_t low)
: lo_{low}, hi_{high} {}
constexpr uint128::uint128(int v)
: lo_{static_cast<uint64_t>(v)},
hi_{v < 0 ? (std::numeric_limits<uint64_t>::max)() : 0} {}
constexpr uint128::uint128(long v) // NOLINT(runtime/int)
: lo_{static_cast<uint64_t>(v)},
hi_{v < 0 ? (std::numeric_limits<uint64_t>::max)() : 0} {}
constexpr uint128::uint128(long long v) // NOLINT(runtime/int)
: lo_{static_cast<uint64_t>(v)},
hi_{v < 0 ? (std::numeric_limits<uint64_t>::max)() : 0} {}
constexpr uint128::uint128(unsigned int v) : lo_{v}, hi_{0} {}
// NOLINTNEXTLINE(runtime/int)
constexpr uint128::uint128(unsigned long v) : lo_{v}, hi_{0} {}
// NOLINTNEXTLINE(runtime/int)
constexpr uint128::uint128(unsigned long long v) : lo_{v}, hi_{0} {}
#ifdef ABSL_HAVE_INTRINSIC_INT128
constexpr uint128::uint128(__int128 v)
: lo_{static_cast<uint64_t>(v & ~uint64_t{0})},
hi_{static_cast<uint64_t>(static_cast<unsigned __int128>(v) >> 64)} {}
constexpr uint128::uint128(unsigned __int128 v)
: lo_{static_cast<uint64_t>(v & ~uint64_t{0})},
hi_{static_cast<uint64_t>(v >> 64)} {}
#endif // ABSL_HAVE_INTRINSIC_INT128
constexpr uint128::uint128(int128 v)
: lo_{Int128Low64(v)}, hi_{static_cast<uint64_t>(Int128High64(v))} {}
#elif defined(ABSL_IS_BIG_ENDIAN)
constexpr uint128::uint128(uint64_t high, uint64_t low)
: hi_{high}, lo_{low} {}
constexpr uint128::uint128(int v)
: hi_{v < 0 ? (std::numeric_limits<uint64_t>::max)() : 0},
lo_{static_cast<uint64_t>(v)} {}
constexpr uint128::uint128(long v) // NOLINT(runtime/int)
: hi_{v < 0 ? (std::numeric_limits<uint64_t>::max)() : 0},
lo_{static_cast<uint64_t>(v)} {}
constexpr uint128::uint128(long long v) // NOLINT(runtime/int)
: hi_{v < 0 ? (std::numeric_limits<uint64_t>::max)() : 0},
lo_{static_cast<uint64_t>(v)} {}
constexpr uint128::uint128(unsigned int v) : hi_{0}, lo_{v} {}
// NOLINTNEXTLINE(runtime/int)
constexpr uint128::uint128(unsigned long v) : hi_{0}, lo_{v} {}
// NOLINTNEXTLINE(runtime/int)
constexpr uint128::uint128(unsigned long long v) : hi_{0}, lo_{v} {}
#ifdef ABSL_HAVE_INTRINSIC_INT128
constexpr uint128::uint128(__int128 v)
: hi_{static_cast<uint64_t>(static_cast<unsigned __int128>(v) >> 64)},
lo_{static_cast<uint64_t>(v & ~uint64_t{0})} {}
constexpr uint128::uint128(unsigned __int128 v)
: hi_{static_cast<uint64_t>(v >> 64)},
lo_{static_cast<uint64_t>(v & ~uint64_t{0})} {}
#endif // ABSL_HAVE_INTRINSIC_INT128
constexpr uint128::uint128(int128 v)
: hi_{static_cast<uint64_t>(Int128High64(v))}, lo_{Int128Low64(v)} {}
#else // byte order
#error "Unsupported byte order: must be little-endian or big-endian."
#endif // byte order
// Conversion operators to integer types.
constexpr uint128::operator bool() const { return lo_ || hi_; }
constexpr uint128::operator char() const { return static_cast<char>(lo_); }
constexpr uint128::operator signed char() const {
return static_cast<signed char>(lo_);
}
constexpr uint128::operator unsigned char() const {
return static_cast<unsigned char>(lo_);
}
constexpr uint128::operator char16_t() const {
return static_cast<char16_t>(lo_);
}
constexpr uint128::operator char32_t() const {
return static_cast<char32_t>(lo_);
}
constexpr uint128::operator ABSL_INTERNAL_WCHAR_T() const {
return static_cast<ABSL_INTERNAL_WCHAR_T>(lo_);
}
// NOLINTNEXTLINE(runtime/int)
constexpr uint128::operator short() const { return static_cast<short>(lo_); }
constexpr uint128::operator unsigned short() const { // NOLINT(runtime/int)
return static_cast<unsigned short>(lo_); // NOLINT(runtime/int)
}
constexpr uint128::operator int() const { return static_cast<int>(lo_); }
constexpr uint128::operator unsigned int() const {
return static_cast<unsigned int>(lo_);
}
// NOLINTNEXTLINE(runtime/int)
constexpr uint128::operator long() const { return static_cast<long>(lo_); }
constexpr uint128::operator unsigned long() const { // NOLINT(runtime/int)
return static_cast<unsigned long>(lo_); // NOLINT(runtime/int)
}
constexpr uint128::operator long long() const { // NOLINT(runtime/int)
return static_cast<long long>(lo_); // NOLINT(runtime/int)
}
constexpr uint128::operator unsigned long long() const { // NOLINT(runtime/int)
return static_cast<unsigned long long>(lo_); // NOLINT(runtime/int)
}
#ifdef ABSL_HAVE_INTRINSIC_INT128
constexpr uint128::operator __int128() const {
return (static_cast<__int128>(hi_) << 64) + lo_;
}
constexpr uint128::operator unsigned __int128() const {
return (static_cast<unsigned __int128>(hi_) << 64) + lo_;
}
#endif // ABSL_HAVE_INTRINSIC_INT128
// Conversion operators to floating point types.
inline uint128::operator float() const {
return static_cast<float>(lo_) + std::ldexp(static_cast<float>(hi_), 64);
}
inline uint128::operator double() const {
return static_cast<double>(lo_) + std::ldexp(static_cast<double>(hi_), 64);
}
inline uint128::operator long double() const {
return static_cast<long double>(lo_) +
std::ldexp(static_cast<long double>(hi_), 64);
}
// Comparison operators.
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr bool operator==(uint128 lhs, uint128 rhs) {
#if defined(ABSL_HAVE_INTRINSIC_INT128)
return static_cast<unsigned __int128>(lhs) ==
static_cast<unsigned __int128>(rhs);
#else
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 (Uint128Low64(lhs) == Uint128Low64(rhs) &&
Uint128High64(lhs) == Uint128High64(rhs));
#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
}
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr bool operator!=(uint128 lhs, uint128 rhs) { return !(lhs == rhs); }
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 -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr bool operator<(uint128 lhs, uint128 rhs) {
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
#ifdef ABSL_HAVE_INTRINSIC_INT128
return static_cast<unsigned __int128>(lhs) <
static_cast<unsigned __int128>(rhs);
#else
return (Uint128High64(lhs) == Uint128High64(rhs))
? (Uint128Low64(lhs) < Uint128Low64(rhs))
: (Uint128High64(lhs) < Uint128High64(rhs));
#endif
}
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr bool operator>(uint128 lhs, uint128 rhs) { return rhs < lhs; }
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 -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr bool operator<=(uint128 lhs, uint128 rhs) { return !(rhs < lhs); }
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 -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr bool operator>=(uint128 lhs, uint128 rhs) { return !(lhs < rhs); }
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
// Unary operators.
constexpr inline uint128 operator+(uint128 val) {
return val;
}
constexpr inline int128 operator+(int128 val) {
return val;
}
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr uint128 operator-(uint128 val) {
#if defined(ABSL_HAVE_INTRINSIC_INT128)
return -static_cast<unsigned __int128>(val);
#else
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
return MakeUint128(
~Uint128High64(val) + static_cast<unsigned long>(Uint128Low64(val) == 0),
~Uint128Low64(val) + 1);
#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
}
Export of internal Abseil changes -- 9e0df3dd23da17cd0ff75c93c1493a858032639c by Martijn Vels <mvels@google.com>: Prep work for changing Cordz instrumentation Create CordzInfo::TrackCord() and CordzInfo::MaybeTrackCord() methods on InlineData This follows a suggestion from kfm@ to move where possible Cordz logic out of Cord and Cord::InlineRep, which we can now that InlineData provides us the abstraction of Cords internal data. PiperOrigin-RevId: 369844310 -- 45f39709033bd3bc09fa1a7880a5b3c9eaa617c7 by Abseil Team <absl-team@google.com>: Fix dependence on C++20 constexpr default construction of std::atomic. PiperOrigin-RevId: 369745251 -- 195ae230963c95068406ab0e73b4e711b5f3cd62 by Martijn Vels <mvels@google.com>: Reduce the cost of 'SetCordRep()` This change inlines SetCordRep(), which is currently two stores and a branch, going forward a naked store. AssertHeld is only enforced for debug builds. The intention here is not to reduce the 'actual' (or 'self') cost of SetCordRep() as it should only be called for sampled cords, but to reduce the cost of a branch and out of line call at the call site. The compiler can now 'perfectly inline' the single branch / store. PiperOrigin-RevId: 369696265 -- d722199ed69d413994740624159ac7bd001a9219 by Martijn Vels <mvels@google.com>: Add kDefaultInit initialization This avoids double store on init PiperOrigin-RevId: 369655217 -- 3499aed79e6cc12ce36277063ec37991bba0cccd by Martijn Vels <mvels@google.com>: Introduce CordzUpdateScope PiperOrigin-RevId: 369491326 -- 325c1bc99c7d1aeca7bd1273e51a0900b6faf731 by Abseil Team <absl-team@google.com>: make unary and logical operators constexpr PiperOrigin-RevId: 369476773 -- ad3bed3dea5e5d7d281ff36ee786f802630c3728 by Martijn Vels <mvels@google.com>: Add LOCKABLE attribute to CordzInfo PiperOrigin-RevId: 369476772 GitOrigin-RevId: 9e0df3dd23da17cd0ff75c93c1493a858032639c Change-Id: I00e2859328fe8da46d2e04d3f07dfe70ec6cb1f5
4 years ago
constexpr inline bool operator!(uint128 val) {
#if defined(ABSL_HAVE_INTRINSIC_INT128)
return !static_cast<unsigned __int128>(val);
#else
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 !Uint128High64(val) && !Uint128Low64(val);
#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
}
// Logical operators.
Export of internal Abseil changes -- 9e0df3dd23da17cd0ff75c93c1493a858032639c by Martijn Vels <mvels@google.com>: Prep work for changing Cordz instrumentation Create CordzInfo::TrackCord() and CordzInfo::MaybeTrackCord() methods on InlineData This follows a suggestion from kfm@ to move where possible Cordz logic out of Cord and Cord::InlineRep, which we can now that InlineData provides us the abstraction of Cords internal data. PiperOrigin-RevId: 369844310 -- 45f39709033bd3bc09fa1a7880a5b3c9eaa617c7 by Abseil Team <absl-team@google.com>: Fix dependence on C++20 constexpr default construction of std::atomic. PiperOrigin-RevId: 369745251 -- 195ae230963c95068406ab0e73b4e711b5f3cd62 by Martijn Vels <mvels@google.com>: Reduce the cost of 'SetCordRep()` This change inlines SetCordRep(), which is currently two stores and a branch, going forward a naked store. AssertHeld is only enforced for debug builds. The intention here is not to reduce the 'actual' (or 'self') cost of SetCordRep() as it should only be called for sampled cords, but to reduce the cost of a branch and out of line call at the call site. The compiler can now 'perfectly inline' the single branch / store. PiperOrigin-RevId: 369696265 -- d722199ed69d413994740624159ac7bd001a9219 by Martijn Vels <mvels@google.com>: Add kDefaultInit initialization This avoids double store on init PiperOrigin-RevId: 369655217 -- 3499aed79e6cc12ce36277063ec37991bba0cccd by Martijn Vels <mvels@google.com>: Introduce CordzUpdateScope PiperOrigin-RevId: 369491326 -- 325c1bc99c7d1aeca7bd1273e51a0900b6faf731 by Abseil Team <absl-team@google.com>: make unary and logical operators constexpr PiperOrigin-RevId: 369476773 -- ad3bed3dea5e5d7d281ff36ee786f802630c3728 by Martijn Vels <mvels@google.com>: Add LOCKABLE attribute to CordzInfo PiperOrigin-RevId: 369476772 GitOrigin-RevId: 9e0df3dd23da17cd0ff75c93c1493a858032639c Change-Id: I00e2859328fe8da46d2e04d3f07dfe70ec6cb1f5
4 years ago
constexpr inline uint128 operator~(uint128 val) {
#if defined(ABSL_HAVE_INTRINSIC_INT128)
return ~static_cast<unsigned __int128>(val);
#else
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 MakeUint128(~Uint128High64(val), ~Uint128Low64(val));
#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
}
Export of internal Abseil changes -- 9e0df3dd23da17cd0ff75c93c1493a858032639c by Martijn Vels <mvels@google.com>: Prep work for changing Cordz instrumentation Create CordzInfo::TrackCord() and CordzInfo::MaybeTrackCord() methods on InlineData This follows a suggestion from kfm@ to move where possible Cordz logic out of Cord and Cord::InlineRep, which we can now that InlineData provides us the abstraction of Cords internal data. PiperOrigin-RevId: 369844310 -- 45f39709033bd3bc09fa1a7880a5b3c9eaa617c7 by Abseil Team <absl-team@google.com>: Fix dependence on C++20 constexpr default construction of std::atomic. PiperOrigin-RevId: 369745251 -- 195ae230963c95068406ab0e73b4e711b5f3cd62 by Martijn Vels <mvels@google.com>: Reduce the cost of 'SetCordRep()` This change inlines SetCordRep(), which is currently two stores and a branch, going forward a naked store. AssertHeld is only enforced for debug builds. The intention here is not to reduce the 'actual' (or 'self') cost of SetCordRep() as it should only be called for sampled cords, but to reduce the cost of a branch and out of line call at the call site. The compiler can now 'perfectly inline' the single branch / store. PiperOrigin-RevId: 369696265 -- d722199ed69d413994740624159ac7bd001a9219 by Martijn Vels <mvels@google.com>: Add kDefaultInit initialization This avoids double store on init PiperOrigin-RevId: 369655217 -- 3499aed79e6cc12ce36277063ec37991bba0cccd by Martijn Vels <mvels@google.com>: Introduce CordzUpdateScope PiperOrigin-RevId: 369491326 -- 325c1bc99c7d1aeca7bd1273e51a0900b6faf731 by Abseil Team <absl-team@google.com>: make unary and logical operators constexpr PiperOrigin-RevId: 369476773 -- ad3bed3dea5e5d7d281ff36ee786f802630c3728 by Martijn Vels <mvels@google.com>: Add LOCKABLE attribute to CordzInfo PiperOrigin-RevId: 369476772 GitOrigin-RevId: 9e0df3dd23da17cd0ff75c93c1493a858032639c Change-Id: I00e2859328fe8da46d2e04d3f07dfe70ec6cb1f5
4 years ago
constexpr inline uint128 operator|(uint128 lhs, uint128 rhs) {
#if defined(ABSL_HAVE_INTRINSIC_INT128)
return static_cast<unsigned __int128>(lhs) |
static_cast<unsigned __int128>(rhs);
#else
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 MakeUint128(Uint128High64(lhs) | Uint128High64(rhs),
Uint128Low64(lhs) | Uint128Low64(rhs));
#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
}
Export of internal Abseil changes -- 9e0df3dd23da17cd0ff75c93c1493a858032639c by Martijn Vels <mvels@google.com>: Prep work for changing Cordz instrumentation Create CordzInfo::TrackCord() and CordzInfo::MaybeTrackCord() methods on InlineData This follows a suggestion from kfm@ to move where possible Cordz logic out of Cord and Cord::InlineRep, which we can now that InlineData provides us the abstraction of Cords internal data. PiperOrigin-RevId: 369844310 -- 45f39709033bd3bc09fa1a7880a5b3c9eaa617c7 by Abseil Team <absl-team@google.com>: Fix dependence on C++20 constexpr default construction of std::atomic. PiperOrigin-RevId: 369745251 -- 195ae230963c95068406ab0e73b4e711b5f3cd62 by Martijn Vels <mvels@google.com>: Reduce the cost of 'SetCordRep()` This change inlines SetCordRep(), which is currently two stores and a branch, going forward a naked store. AssertHeld is only enforced for debug builds. The intention here is not to reduce the 'actual' (or 'self') cost of SetCordRep() as it should only be called for sampled cords, but to reduce the cost of a branch and out of line call at the call site. The compiler can now 'perfectly inline' the single branch / store. PiperOrigin-RevId: 369696265 -- d722199ed69d413994740624159ac7bd001a9219 by Martijn Vels <mvels@google.com>: Add kDefaultInit initialization This avoids double store on init PiperOrigin-RevId: 369655217 -- 3499aed79e6cc12ce36277063ec37991bba0cccd by Martijn Vels <mvels@google.com>: Introduce CordzUpdateScope PiperOrigin-RevId: 369491326 -- 325c1bc99c7d1aeca7bd1273e51a0900b6faf731 by Abseil Team <absl-team@google.com>: make unary and logical operators constexpr PiperOrigin-RevId: 369476773 -- ad3bed3dea5e5d7d281ff36ee786f802630c3728 by Martijn Vels <mvels@google.com>: Add LOCKABLE attribute to CordzInfo PiperOrigin-RevId: 369476772 GitOrigin-RevId: 9e0df3dd23da17cd0ff75c93c1493a858032639c Change-Id: I00e2859328fe8da46d2e04d3f07dfe70ec6cb1f5
4 years ago
constexpr inline uint128 operator&(uint128 lhs, uint128 rhs) {
#if defined(ABSL_HAVE_INTRINSIC_INT128)
return static_cast<unsigned __int128>(lhs) &
static_cast<unsigned __int128>(rhs);
#else
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 MakeUint128(Uint128High64(lhs) & Uint128High64(rhs),
Uint128Low64(lhs) & Uint128Low64(rhs));
#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
}
Export of internal Abseil changes -- 9e0df3dd23da17cd0ff75c93c1493a858032639c by Martijn Vels <mvels@google.com>: Prep work for changing Cordz instrumentation Create CordzInfo::TrackCord() and CordzInfo::MaybeTrackCord() methods on InlineData This follows a suggestion from kfm@ to move where possible Cordz logic out of Cord and Cord::InlineRep, which we can now that InlineData provides us the abstraction of Cords internal data. PiperOrigin-RevId: 369844310 -- 45f39709033bd3bc09fa1a7880a5b3c9eaa617c7 by Abseil Team <absl-team@google.com>: Fix dependence on C++20 constexpr default construction of std::atomic. PiperOrigin-RevId: 369745251 -- 195ae230963c95068406ab0e73b4e711b5f3cd62 by Martijn Vels <mvels@google.com>: Reduce the cost of 'SetCordRep()` This change inlines SetCordRep(), which is currently two stores and a branch, going forward a naked store. AssertHeld is only enforced for debug builds. The intention here is not to reduce the 'actual' (or 'self') cost of SetCordRep() as it should only be called for sampled cords, but to reduce the cost of a branch and out of line call at the call site. The compiler can now 'perfectly inline' the single branch / store. PiperOrigin-RevId: 369696265 -- d722199ed69d413994740624159ac7bd001a9219 by Martijn Vels <mvels@google.com>: Add kDefaultInit initialization This avoids double store on init PiperOrigin-RevId: 369655217 -- 3499aed79e6cc12ce36277063ec37991bba0cccd by Martijn Vels <mvels@google.com>: Introduce CordzUpdateScope PiperOrigin-RevId: 369491326 -- 325c1bc99c7d1aeca7bd1273e51a0900b6faf731 by Abseil Team <absl-team@google.com>: make unary and logical operators constexpr PiperOrigin-RevId: 369476773 -- ad3bed3dea5e5d7d281ff36ee786f802630c3728 by Martijn Vels <mvels@google.com>: Add LOCKABLE attribute to CordzInfo PiperOrigin-RevId: 369476772 GitOrigin-RevId: 9e0df3dd23da17cd0ff75c93c1493a858032639c Change-Id: I00e2859328fe8da46d2e04d3f07dfe70ec6cb1f5
4 years ago
constexpr inline uint128 operator^(uint128 lhs, uint128 rhs) {
#if defined(ABSL_HAVE_INTRINSIC_INT128)
return static_cast<unsigned __int128>(lhs) ^
static_cast<unsigned __int128>(rhs);
#else
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 MakeUint128(Uint128High64(lhs) ^ Uint128High64(rhs),
Uint128Low64(lhs) ^ Uint128Low64(rhs));
#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
}
inline uint128& uint128::operator|=(uint128 other) {
*this = *this | other;
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 *this;
}
inline uint128& uint128::operator&=(uint128 other) {
*this = *this & other;
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 *this;
}
inline uint128& uint128::operator^=(uint128 other) {
*this = *this ^ other;
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 *this;
}
// Arithmetic operators.
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr uint128 operator<<(uint128 lhs, int amount) {
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
#ifdef ABSL_HAVE_INTRINSIC_INT128
return static_cast<unsigned __int128>(lhs) << amount;
#else
// uint64_t shifts of >= 64 are undefined, so we will need some
// special-casing.
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
return amount >= 64 ? MakeUint128(Uint128Low64(lhs) << (amount - 64), 0)
: amount == 0 ? lhs
: MakeUint128((Uint128High64(lhs) << amount) |
(Uint128Low64(lhs) >> (64 - amount)),
Uint128Low64(lhs) << amount);
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#endif
}
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr uint128 operator>>(uint128 lhs, int amount) {
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
#ifdef ABSL_HAVE_INTRINSIC_INT128
return static_cast<unsigned __int128>(lhs) >> amount;
#else
// uint64_t shifts of >= 64 are undefined, so we will need some
// special-casing.
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
return amount >= 64 ? MakeUint128(0, Uint128High64(lhs) >> (amount - 64))
: amount == 0 ? lhs
: MakeUint128(Uint128High64(lhs) >> amount,
(Uint128Low64(lhs) >> amount) |
(Uint128High64(lhs) << (64 - amount)));
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
#endif
}
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
#if !defined(ABSL_HAVE_INTRINSIC_INT128)
namespace int128_internal {
constexpr uint128 AddResult(uint128 result, uint128 lhs) {
// check for carry
return (Uint128Low64(result) < Uint128Low64(lhs))
? MakeUint128(Uint128High64(result) + 1, Uint128Low64(result))
: result;
}
} // namespace int128_internal
#endif
constexpr uint128 operator+(uint128 lhs, uint128 rhs) {
#if defined(ABSL_HAVE_INTRINSIC_INT128)
return static_cast<unsigned __int128>(lhs) +
static_cast<unsigned __int128>(rhs);
#else
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
return int128_internal::AddResult(
MakeUint128(Uint128High64(lhs) + Uint128High64(rhs),
Uint128Low64(lhs) + Uint128Low64(rhs)),
lhs);
#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
}
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
#if !defined(ABSL_HAVE_INTRINSIC_INT128)
namespace int128_internal {
constexpr uint128 SubstructResult(uint128 result, uint128 lhs, uint128 rhs) {
// check for carry
return (Uint128Low64(lhs) < Uint128Low64(rhs))
? MakeUint128(Uint128High64(result) - 1, Uint128Low64(result))
: result;
}
} // namespace int128_internal
#endif
constexpr uint128 operator-(uint128 lhs, uint128 rhs) {
#if defined(ABSL_HAVE_INTRINSIC_INT128)
return static_cast<unsigned __int128>(lhs) -
static_cast<unsigned __int128>(rhs);
#else
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
return int128_internal::SubstructResult(
MakeUint128(Uint128High64(lhs) - Uint128High64(rhs),
Uint128Low64(lhs) - Uint128Low64(rhs)),
lhs, rhs);
#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
}
inline uint128 operator*(uint128 lhs, uint128 rhs) {
#if defined(ABSL_HAVE_INTRINSIC_INT128)
// TODO(strel) Remove once alignment issues are resolved and unsigned __int128
// can be used for uint128 storage.
return static_cast<unsigned __int128>(lhs) *
static_cast<unsigned __int128>(rhs);
#elif defined(_MSC_VER) && defined(_M_X64) && !defined(_M_ARM64EC)
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 carry;
uint64_t low = _umul128(Uint128Low64(lhs), Uint128Low64(rhs), &carry);
return MakeUint128(Uint128Low64(lhs) * Uint128High64(rhs) +
Uint128High64(lhs) * Uint128Low64(rhs) + carry,
low);
#else // ABSL_HAVE_INTRINSIC128
uint64_t a32 = Uint128Low64(lhs) >> 32;
uint64_t a00 = Uint128Low64(lhs) & 0xffffffff;
uint64_t b32 = Uint128Low64(rhs) >> 32;
uint64_t b00 = Uint128Low64(rhs) & 0xffffffff;
uint128 result =
MakeUint128(Uint128High64(lhs) * Uint128Low64(rhs) +
Uint128Low64(lhs) * Uint128High64(rhs) + a32 * b32,
a00 * b00);
result += uint128(a32 * b00) << 32;
result += uint128(a00 * b32) << 32;
return result;
#endif // ABSL_HAVE_INTRINSIC128
}
#if defined(ABSL_HAVE_INTRINSIC_INT128)
inline uint128 operator/(uint128 lhs, uint128 rhs) {
return static_cast<unsigned __int128>(lhs) /
static_cast<unsigned __int128>(rhs);
}
inline uint128 operator%(uint128 lhs, uint128 rhs) {
return static_cast<unsigned __int128>(lhs) %
static_cast<unsigned __int128>(rhs);
}
#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
// Increment/decrement operators.
inline uint128 uint128::operator++(int) {
uint128 tmp(*this);
*this += 1;
return tmp;
}
inline uint128 uint128::operator--(int) {
uint128 tmp(*this);
*this -= 1;
return tmp;
}
inline uint128& uint128::operator++() {
*this += 1;
return *this;
}
inline uint128& uint128::operator--() {
*this -= 1;
return *this;
}
constexpr int128 MakeInt128(int64_t high, uint64_t low) {
return int128(high, low);
}
// Assignment from integer types.
inline int128& int128::operator=(int v) {
return *this = int128(v);
}
inline int128& int128::operator=(unsigned int v) {
return *this = int128(v);
}
inline int128& int128::operator=(long v) { // NOLINT(runtime/int)
return *this = int128(v);
}
// NOLINTNEXTLINE(runtime/int)
inline int128& int128::operator=(unsigned long v) {
return *this = int128(v);
}
// NOLINTNEXTLINE(runtime/int)
inline int128& int128::operator=(long long v) {
return *this = int128(v);
}
// NOLINTNEXTLINE(runtime/int)
inline int128& int128::operator=(unsigned long long v) {
return *this = int128(v);
}
// Arithmetic operators.
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr int128 operator-(int128 v);
constexpr int128 operator+(int128 lhs, int128 rhs);
constexpr int128 operator-(int128 lhs, int128 rhs);
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
int128 operator*(int128 lhs, int128 rhs);
int128 operator/(int128 lhs, int128 rhs);
int128 operator%(int128 lhs, int128 rhs);
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
constexpr int128 operator|(int128 lhs, int128 rhs);
constexpr int128 operator&(int128 lhs, int128 rhs);
constexpr int128 operator^(int128 lhs, int128 rhs);
constexpr int128 operator<<(int128 lhs, int amount);
constexpr int128 operator>>(int128 lhs, int amount);
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
inline int128& int128::operator+=(int128 other) {
*this = *this + other;
return *this;
}
inline int128& int128::operator-=(int128 other) {
*this = *this - other;
return *this;
}
inline int128& int128::operator*=(int128 other) {
*this = *this * other;
return *this;
}
inline int128& int128::operator/=(int128 other) {
*this = *this / other;
return *this;
}
inline int128& int128::operator%=(int128 other) {
*this = *this % other;
return *this;
}
inline int128& int128::operator|=(int128 other) {
*this = *this | other;
return *this;
}
inline int128& int128::operator&=(int128 other) {
*this = *this & other;
return *this;
}
inline int128& int128::operator^=(int128 other) {
*this = *this ^ other;
return *this;
}
inline int128& int128::operator<<=(int amount) {
*this = *this << amount;
return *this;
}
inline int128& int128::operator>>=(int amount) {
*this = *this >> amount;
return *this;
}
Export of internal Abseil changes -- 04cb3b22497190170aa5b774e98080c5de2ba60b by Abseil Team <absl-team@google.com>: Alternative bit mixer for LowLevelHash on ARM LowLevelHash's bit-mixer is inefficient on ARM because it calculates a 128-bit product of two 64-bit numbers. On ARM, this requires a sequence of two instructions with a high combined latency and poor throughput. This change provides alternative bit-mixing code for ARM that uses only 64-bit arithmetic (multiplication, xor, and left-shifts) and speeds things up considerably. The bit-mixing code for ARM was inspired by by Woothash[1] and xxh3[1]. Once I landed on a sequence of operations that provided good mixing, I used a test harness to search for the combination of shift / rotate factors that provided the best mixing, as indicated by SMHasher hash quality tests. The new mixing code passes 13 out of 15 of the hash quality test suites in SMHasher, with the two failures being in the noise range: e.g. 1 collision vs. zero expected in a keyset of ~8m keys. [1]: https://github.com/tommyettinger/waterhash/blob/49f5cf0b63b9/woothash.h#L16-L20 [2]: https://github.com/Cyan4973/xxHash/blob/6853ddc36e46/xxhash.h#L3240-L3265 PiperOrigin-RevId: 391833008 -- 17a4de1f9d623155c75b19285d414cd55a487cd6 by Saleem Abdulrasool <abdulras@google.com>: debugging: add support for unwinding on RISCV Linux This adds partial support for unwinding the RISCV call stack. It is largely duplicated from the AArch64 support with alterations for the ELF RISCV psABI. This covers RISCV64 and RISCV32, though not the ILP32E calling convention. PiperOrigin-RevId: 391818522 -- 32c93e449327b2cea32b32f6365e84b420fe1ed3 by Gennadiy Rozental <rogeeff@google.com>: New storage for types smaller than 8 bytes. Also adding new read interface for types smaller than or rqual to 8 bytes to avoid passing the pointer. PiperOrigin-RevId: 391726822 -- e987ac08a7787801cbfc7d7c96649e97fa8cff1a by Abseil Team <absl-team@google.com>: Extern template `find_first_non_full` to reduce linkage size for TU with single not inlined function. PiperOrigin-RevId: 391718862 -- 73af9bfcb5bf045089133e18bbd20eb5bb699172 by Gennadiy Rozental <rogeeff@google.com>: Make most non-mutable most int128 methods and friend free functions constexpr. Some functions are implemented offline (at least in some configurations) and can't be made constexpr. Mutable methods can't be made constexpr until we drop c++11 support. Fixes #978 PiperOrigin-RevId: 391706535 GitOrigin-RevId: 04cb3b22497190170aa5b774e98080c5de2ba60b Change-Id: If051fad5ff004e2e82fa53618fc04a6fe3d2d4be
3 years ago
// Forward declaration for comparison operators.
constexpr bool operator!=(int128 lhs, int128 rhs);
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 int128_internal {
// Casts from unsigned to signed while preserving the underlying binary
// representation.
constexpr int64_t BitCastToSigned(uint64_t v) {
// Casting an unsigned integer to a signed integer of the same
// width is implementation defined behavior if the source value would not fit
// in the destination type. We step around it with a roundtrip bitwise not
// operation to make sure this function remains constexpr. Clang, GCC, and
// MSVC optimize this to a no-op on x86-64.
return v & (uint64_t{1} << 63) ? ~static_cast<int64_t>(~v)
: static_cast<int64_t>(v);
}
} // namespace int128_internal
#if defined(ABSL_HAVE_INTRINSIC_INT128)
#include "absl/numeric/int128_have_intrinsic.inc" // IWYU pragma: export
#else // ABSL_HAVE_INTRINSIC_INT128
#include "absl/numeric/int128_no_intrinsic.inc" // IWYU pragma: export
#endif // ABSL_HAVE_INTRINSIC_INT128
ABSL_NAMESPACE_END
} // namespace absl
#undef ABSL_INTERNAL_WCHAR_T
#endif // ABSL_NUMERIC_INT128_H_