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.

261 lines
10 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.
//
// Each active thread has an ThreadIdentity that may represent the thread in
// various level interfaces. ThreadIdentity objects are never deallocated.
// When a thread terminates, its ThreadIdentity object may be reused for a
// thread created later.
#ifndef ABSL_BASE_INTERNAL_THREAD_IDENTITY_H_
#define ABSL_BASE_INTERNAL_THREAD_IDENTITY_H_
#ifndef _WIN32
#include <pthread.h>
// Defines __GOOGLE_GRTE_VERSION__ (via glibc-specific features.h) when
// supported.
#include <unistd.h>
#endif
#include <atomic>
#include <cstdint>
#include "absl/base/config.h"
#include "absl/base/internal/per_thread_tls.h"
Export of internal Abseil changes -- 730bb88bee556aa11fa19aa33e1434cb6fa78985 by Evan Brown <ezb@google.com>: Support missing allocator-related constructors in b-tree. See [reference](https://en.cppreference.com/w/cpp/container/set/set). Also use allocator_traits::select_on_container_copy_construction() to get allocator for copy construction. PiperOrigin-RevId: 339058322 -- b6cc121689ae3e452d1db2d66122cb198d25142b by Derek Mauro <dmauro@google.com>: Fix more sign-compare warnings PiperOrigin-RevId: 339057920 -- 0e2c62da1dcaf6529abab952bdcc96c6de2d9506 by Abseil Team <absl-team@google.com>: Add missing <limits> include PiperOrigin-RevId: 339054753 -- d5a9ec2d1e40fe6359e720942e4955009ee415ec by Derek Mauro <dmauro@google.com>: Stop disabling sign-compare warnings for non-test targets. Our users complain about these. This does not catch issues in header-only libraries (like btree.h) but we may work on those in the future PiperOrigin-RevId: 338967089 -- 0c062c542a4c61ea0f65d25811827c0858e3adde by Abseil Team <absl-team@google.com>: Improve cache-locality for ThreadIdentity and PerThreadSynch. This is a change based on an observation in RPC benchmarks that shows significant cycles being spent in waking up a thread, 99.8% of which was on cache misses. Investigating this a bit more, it turns out to be due to sharing the cache line with the waiter state. To fix this issue, the following changes are introduced: - Reorder fields in PerThreadSync so that it fits in a single cache line The size of this structure was 80 bytes before this change. Note: Manually inspected all booleans to make sure they are not modified by multiple threads concurrently. PiperOrigin-RevId: 338852058 -- a90d6f2b2346385017e32dd8ae1b5ca691a5863f by Derek Mauro <dmauro@google.com>: Delete GCC 4.9 test script. It is no longer supported PiperOrigin-RevId: 338779452 -- 7274008d4757e88869110be9db39d03d911ae2b5 by Abseil Team <absl-team@google.com>: Fix the usage example in which SetFlag should take a pointer. PiperOrigin-RevId: 338744529 GitOrigin-RevId: 730bb88bee556aa11fa19aa33e1434cb6fa78985 Change-Id: Iff99594c4022e60e482a392d334b376c7ae8883e
4 years ago
#include "absl/base/optimization.h"
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
namespace absl {
ABSL_NAMESPACE_BEGIN
struct SynchLocksHeld;
struct SynchWaitParams;
namespace base_internal {
class SpinLock;
struct ThreadIdentity;
// Used by the implementation of absl::Mutex and absl::CondVar.
struct PerThreadSynch {
// The internal representation of absl::Mutex and absl::CondVar rely
// on the alignment of PerThreadSynch. Both store the address of the
// PerThreadSynch in the high-order bits of their internal state,
// which means the low kLowZeroBits of the address of PerThreadSynch
// must be zero.
static constexpr int kLowZeroBits = 8;
static constexpr int kAlignment = 1 << kLowZeroBits;
// Returns the associated ThreadIdentity.
// This can be implemented as a cast because we guarantee
// PerThreadSynch is the first element of ThreadIdentity.
ThreadIdentity* thread_identity() {
return reinterpret_cast<ThreadIdentity*>(this);
}
PerThreadSynch *next; // Circular waiter queue; initialized to 0.
PerThreadSynch *skip; // If non-zero, all entries in Mutex queue
// up to and including "skip" have same
// condition as this, and will be woken later
bool may_skip; // if false while on mutex queue, a mutex unlocker
// is using this PerThreadSynch as a terminator. Its
// skip field must not be filled in because the loop
// might then skip over the terminator.
Export of internal Abseil changes -- 730bb88bee556aa11fa19aa33e1434cb6fa78985 by Evan Brown <ezb@google.com>: Support missing allocator-related constructors in b-tree. See [reference](https://en.cppreference.com/w/cpp/container/set/set). Also use allocator_traits::select_on_container_copy_construction() to get allocator for copy construction. PiperOrigin-RevId: 339058322 -- b6cc121689ae3e452d1db2d66122cb198d25142b by Derek Mauro <dmauro@google.com>: Fix more sign-compare warnings PiperOrigin-RevId: 339057920 -- 0e2c62da1dcaf6529abab952bdcc96c6de2d9506 by Abseil Team <absl-team@google.com>: Add missing <limits> include PiperOrigin-RevId: 339054753 -- d5a9ec2d1e40fe6359e720942e4955009ee415ec by Derek Mauro <dmauro@google.com>: Stop disabling sign-compare warnings for non-test targets. Our users complain about these. This does not catch issues in header-only libraries (like btree.h) but we may work on those in the future PiperOrigin-RevId: 338967089 -- 0c062c542a4c61ea0f65d25811827c0858e3adde by Abseil Team <absl-team@google.com>: Improve cache-locality for ThreadIdentity and PerThreadSynch. This is a change based on an observation in RPC benchmarks that shows significant cycles being spent in waking up a thread, 99.8% of which was on cache misses. Investigating this a bit more, it turns out to be due to sharing the cache line with the waiter state. To fix this issue, the following changes are introduced: - Reorder fields in PerThreadSync so that it fits in a single cache line The size of this structure was 80 bytes before this change. Note: Manually inspected all booleans to make sure they are not modified by multiple threads concurrently. PiperOrigin-RevId: 338852058 -- a90d6f2b2346385017e32dd8ae1b5ca691a5863f by Derek Mauro <dmauro@google.com>: Delete GCC 4.9 test script. It is no longer supported PiperOrigin-RevId: 338779452 -- 7274008d4757e88869110be9db39d03d911ae2b5 by Abseil Team <absl-team@google.com>: Fix the usage example in which SetFlag should take a pointer. PiperOrigin-RevId: 338744529 GitOrigin-RevId: 730bb88bee556aa11fa19aa33e1434cb6fa78985 Change-Id: Iff99594c4022e60e482a392d334b376c7ae8883e
4 years ago
bool wake; // This thread is to be woken from a Mutex.
// If "x" is on a waiter list for a mutex, "x->cond_waiter" is true iff the
// waiter is waiting on the mutex as part of a CV Wait or Mutex Await.
//
// The value of "x->cond_waiter" is meaningless if "x" is not on a
// Mutex waiter list.
bool cond_waiter;
bool maybe_unlocking; // Valid at head of Mutex waiter queue;
// true if UnlockSlow could be searching
// for a waiter to wake. Used for an optimization
// in Enqueue(). true is always a valid value.
// Can be reset to false when the unlocker or any
// writer releases the lock, or a reader fully
// releases the lock. It may not be set to false
// by a reader that decrements the count to
// non-zero. protected by mutex spinlock
bool suppress_fatal_errors; // If true, try to proceed even in the face
// of broken invariants. This is used within
// fatal signal handlers to improve the
// chances of debug logging information being
// output successfully.
int priority; // Priority of thread (updated every so often).
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
// State values:
// kAvailable: This PerThreadSynch is available.
// kQueued: This PerThreadSynch is unavailable, it's currently queued on a
// Mutex or CondVar waistlist.
//
// Transitions from kQueued to kAvailable require a release
// barrier. This is needed as a waiter may use "state" to
// independently observe that it's no longer queued.
//
// Transitions from kAvailable to kQueued require no barrier, they
// are externally ordered by the Mutex.
enum State {
kAvailable,
kQueued
};
std::atomic<State> state;
Export of internal Abseil changes -- 730bb88bee556aa11fa19aa33e1434cb6fa78985 by Evan Brown <ezb@google.com>: Support missing allocator-related constructors in b-tree. See [reference](https://en.cppreference.com/w/cpp/container/set/set). Also use allocator_traits::select_on_container_copy_construction() to get allocator for copy construction. PiperOrigin-RevId: 339058322 -- b6cc121689ae3e452d1db2d66122cb198d25142b by Derek Mauro <dmauro@google.com>: Fix more sign-compare warnings PiperOrigin-RevId: 339057920 -- 0e2c62da1dcaf6529abab952bdcc96c6de2d9506 by Abseil Team <absl-team@google.com>: Add missing <limits> include PiperOrigin-RevId: 339054753 -- d5a9ec2d1e40fe6359e720942e4955009ee415ec by Derek Mauro <dmauro@google.com>: Stop disabling sign-compare warnings for non-test targets. Our users complain about these. This does not catch issues in header-only libraries (like btree.h) but we may work on those in the future PiperOrigin-RevId: 338967089 -- 0c062c542a4c61ea0f65d25811827c0858e3adde by Abseil Team <absl-team@google.com>: Improve cache-locality for ThreadIdentity and PerThreadSynch. This is a change based on an observation in RPC benchmarks that shows significant cycles being spent in waking up a thread, 99.8% of which was on cache misses. Investigating this a bit more, it turns out to be due to sharing the cache line with the waiter state. To fix this issue, the following changes are introduced: - Reorder fields in PerThreadSync so that it fits in a single cache line The size of this structure was 80 bytes before this change. Note: Manually inspected all booleans to make sure they are not modified by multiple threads concurrently. PiperOrigin-RevId: 338852058 -- a90d6f2b2346385017e32dd8ae1b5ca691a5863f by Derek Mauro <dmauro@google.com>: Delete GCC 4.9 test script. It is no longer supported PiperOrigin-RevId: 338779452 -- 7274008d4757e88869110be9db39d03d911ae2b5 by Abseil Team <absl-team@google.com>: Fix the usage example in which SetFlag should take a pointer. PiperOrigin-RevId: 338744529 GitOrigin-RevId: 730bb88bee556aa11fa19aa33e1434cb6fa78985 Change-Id: Iff99594c4022e60e482a392d334b376c7ae8883e
4 years ago
// The wait parameters of the current wait. waitp is null if the
// thread is not waiting. Transitions from null to non-null must
// occur before the enqueue commit point (state = kQueued in
// Enqueue() and CondVarEnqueue()). Transitions from non-null to
// null must occur after the wait is finished (state = kAvailable in
// Mutex::Block() and CondVar::WaitCommon()). This field may be
// changed only by the thread that describes this PerThreadSynch. A
// special case is Fer(), which calls Enqueue() on another thread,
// but with an identical SynchWaitParams pointer, thus leaving the
// pointer unchanged.
SynchWaitParams* waitp;
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 -- 730bb88bee556aa11fa19aa33e1434cb6fa78985 by Evan Brown <ezb@google.com>: Support missing allocator-related constructors in b-tree. See [reference](https://en.cppreference.com/w/cpp/container/set/set). Also use allocator_traits::select_on_container_copy_construction() to get allocator for copy construction. PiperOrigin-RevId: 339058322 -- b6cc121689ae3e452d1db2d66122cb198d25142b by Derek Mauro <dmauro@google.com>: Fix more sign-compare warnings PiperOrigin-RevId: 339057920 -- 0e2c62da1dcaf6529abab952bdcc96c6de2d9506 by Abseil Team <absl-team@google.com>: Add missing <limits> include PiperOrigin-RevId: 339054753 -- d5a9ec2d1e40fe6359e720942e4955009ee415ec by Derek Mauro <dmauro@google.com>: Stop disabling sign-compare warnings for non-test targets. Our users complain about these. This does not catch issues in header-only libraries (like btree.h) but we may work on those in the future PiperOrigin-RevId: 338967089 -- 0c062c542a4c61ea0f65d25811827c0858e3adde by Abseil Team <absl-team@google.com>: Improve cache-locality for ThreadIdentity and PerThreadSynch. This is a change based on an observation in RPC benchmarks that shows significant cycles being spent in waking up a thread, 99.8% of which was on cache misses. Investigating this a bit more, it turns out to be due to sharing the cache line with the waiter state. To fix this issue, the following changes are introduced: - Reorder fields in PerThreadSync so that it fits in a single cache line The size of this structure was 80 bytes before this change. Note: Manually inspected all booleans to make sure they are not modified by multiple threads concurrently. PiperOrigin-RevId: 338852058 -- a90d6f2b2346385017e32dd8ae1b5ca691a5863f by Derek Mauro <dmauro@google.com>: Delete GCC 4.9 test script. It is no longer supported PiperOrigin-RevId: 338779452 -- 7274008d4757e88869110be9db39d03d911ae2b5 by Abseil Team <absl-team@google.com>: Fix the usage example in which SetFlag should take a pointer. PiperOrigin-RevId: 338744529 GitOrigin-RevId: 730bb88bee556aa11fa19aa33e1434cb6fa78985 Change-Id: Iff99594c4022e60e482a392d334b376c7ae8883e
4 years ago
intptr_t readers; // Number of readers in mutex.
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 -- 730bb88bee556aa11fa19aa33e1434cb6fa78985 by Evan Brown <ezb@google.com>: Support missing allocator-related constructors in b-tree. See [reference](https://en.cppreference.com/w/cpp/container/set/set). Also use allocator_traits::select_on_container_copy_construction() to get allocator for copy construction. PiperOrigin-RevId: 339058322 -- b6cc121689ae3e452d1db2d66122cb198d25142b by Derek Mauro <dmauro@google.com>: Fix more sign-compare warnings PiperOrigin-RevId: 339057920 -- 0e2c62da1dcaf6529abab952bdcc96c6de2d9506 by Abseil Team <absl-team@google.com>: Add missing <limits> include PiperOrigin-RevId: 339054753 -- d5a9ec2d1e40fe6359e720942e4955009ee415ec by Derek Mauro <dmauro@google.com>: Stop disabling sign-compare warnings for non-test targets. Our users complain about these. This does not catch issues in header-only libraries (like btree.h) but we may work on those in the future PiperOrigin-RevId: 338967089 -- 0c062c542a4c61ea0f65d25811827c0858e3adde by Abseil Team <absl-team@google.com>: Improve cache-locality for ThreadIdentity and PerThreadSynch. This is a change based on an observation in RPC benchmarks that shows significant cycles being spent in waking up a thread, 99.8% of which was on cache misses. Investigating this a bit more, it turns out to be due to sharing the cache line with the waiter state. To fix this issue, the following changes are introduced: - Reorder fields in PerThreadSync so that it fits in a single cache line The size of this structure was 80 bytes before this change. Note: Manually inspected all booleans to make sure they are not modified by multiple threads concurrently. PiperOrigin-RevId: 338852058 -- a90d6f2b2346385017e32dd8ae1b5ca691a5863f by Derek Mauro <dmauro@google.com>: Delete GCC 4.9 test script. It is no longer supported PiperOrigin-RevId: 338779452 -- 7274008d4757e88869110be9db39d03d911ae2b5 by Abseil Team <absl-team@google.com>: Fix the usage example in which SetFlag should take a pointer. PiperOrigin-RevId: 338744529 GitOrigin-RevId: 730bb88bee556aa11fa19aa33e1434cb6fa78985 Change-Id: Iff99594c4022e60e482a392d334b376c7ae8883e
4 years ago
// When priority will next be read (cycles).
int64_t next_priority_read_cycles;
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
// Locks held; used during deadlock detection.
// Allocated in Synch_GetAllLocks() and freed in ReclaimThreadIdentity().
SynchLocksHeld *all_locks;
};
Export of internal Abseil changes -- 730bb88bee556aa11fa19aa33e1434cb6fa78985 by Evan Brown <ezb@google.com>: Support missing allocator-related constructors in b-tree. See [reference](https://en.cppreference.com/w/cpp/container/set/set). Also use allocator_traits::select_on_container_copy_construction() to get allocator for copy construction. PiperOrigin-RevId: 339058322 -- b6cc121689ae3e452d1db2d66122cb198d25142b by Derek Mauro <dmauro@google.com>: Fix more sign-compare warnings PiperOrigin-RevId: 339057920 -- 0e2c62da1dcaf6529abab952bdcc96c6de2d9506 by Abseil Team <absl-team@google.com>: Add missing <limits> include PiperOrigin-RevId: 339054753 -- d5a9ec2d1e40fe6359e720942e4955009ee415ec by Derek Mauro <dmauro@google.com>: Stop disabling sign-compare warnings for non-test targets. Our users complain about these. This does not catch issues in header-only libraries (like btree.h) but we may work on those in the future PiperOrigin-RevId: 338967089 -- 0c062c542a4c61ea0f65d25811827c0858e3adde by Abseil Team <absl-team@google.com>: Improve cache-locality for ThreadIdentity and PerThreadSynch. This is a change based on an observation in RPC benchmarks that shows significant cycles being spent in waking up a thread, 99.8% of which was on cache misses. Investigating this a bit more, it turns out to be due to sharing the cache line with the waiter state. To fix this issue, the following changes are introduced: - Reorder fields in PerThreadSync so that it fits in a single cache line The size of this structure was 80 bytes before this change. Note: Manually inspected all booleans to make sure they are not modified by multiple threads concurrently. PiperOrigin-RevId: 338852058 -- a90d6f2b2346385017e32dd8ae1b5ca691a5863f by Derek Mauro <dmauro@google.com>: Delete GCC 4.9 test script. It is no longer supported PiperOrigin-RevId: 338779452 -- 7274008d4757e88869110be9db39d03d911ae2b5 by Abseil Team <absl-team@google.com>: Fix the usage example in which SetFlag should take a pointer. PiperOrigin-RevId: 338744529 GitOrigin-RevId: 730bb88bee556aa11fa19aa33e1434cb6fa78985 Change-Id: Iff99594c4022e60e482a392d334b376c7ae8883e
4 years ago
// The instances of this class are allocated in NewThreadIdentity() with an
// alignment of PerThreadSynch::kAlignment.
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
struct ThreadIdentity {
// Must be the first member. The Mutex implementation requires that
// the PerThreadSynch object associated with each thread is
// PerThreadSynch::kAlignment aligned. We provide this alignment on
// ThreadIdentity itself.
PerThreadSynch per_thread_synch;
// Private: Reserved for absl::synchronization_internal::Waiter.
struct WaiterState {
char data[128];
} waiter_state;
// Used by PerThreadSem::{Get,Set}ThreadBlockedCounter().
std::atomic<int>* blocked_count_ptr;
// The following variables are mostly read/written just by the
// thread itself. The only exception is that these are read by
// a ticker thread as a hint.
std::atomic<int> ticker; // Tick counter, incremented once per second.
std::atomic<int> wait_start; // Ticker value when thread started waiting.
std::atomic<bool> is_idle; // Has thread become idle yet?
ThreadIdentity* next;
};
// Returns the ThreadIdentity object representing the calling thread; guaranteed
// to be unique for its lifetime. The returned object will remain valid for the
// program's lifetime; although it may be re-assigned to a subsequent thread.
// If one does not exist, return nullptr instead.
//
// Does not malloc(*), and is async-signal safe.
// [*] Technically pthread_setspecific() does malloc on first use; however this
// is handled internally within tcmalloc's initialization already.
//
// New ThreadIdentity objects can be constructed and associated with a thread
// by calling GetOrCreateCurrentThreadIdentity() in per-thread-sem.h.
ThreadIdentity* CurrentThreadIdentityIfPresent();
using ThreadIdentityReclaimerFunction = void (*)(void*);
// Sets the current thread identity to the given value. 'reclaimer' is a
// pointer to the global function for cleaning up instances on thread
// destruction.
void SetCurrentThreadIdentity(ThreadIdentity* identity,
ThreadIdentityReclaimerFunction reclaimer);
// Removes the currently associated ThreadIdentity from the running thread.
// This must be called from inside the ThreadIdentityReclaimerFunction, and only
// from that function.
void ClearCurrentThreadIdentity();
// May be chosen at compile time via: -DABSL_FORCE_THREAD_IDENTITY_MODE=<mode
// index>
#ifdef ABSL_THREAD_IDENTITY_MODE_USE_POSIX_SETSPECIFIC
#error ABSL_THREAD_IDENTITY_MODE_USE_POSIX_SETSPECIFIC cannot be direcly set
#else
#define ABSL_THREAD_IDENTITY_MODE_USE_POSIX_SETSPECIFIC 0
#endif
#ifdef ABSL_THREAD_IDENTITY_MODE_USE_TLS
#error ABSL_THREAD_IDENTITY_MODE_USE_TLS cannot be direcly set
#else
#define ABSL_THREAD_IDENTITY_MODE_USE_TLS 1
#endif
#ifdef ABSL_THREAD_IDENTITY_MODE_USE_CPP11
#error ABSL_THREAD_IDENTITY_MODE_USE_CPP11 cannot be direcly set
#else
#define ABSL_THREAD_IDENTITY_MODE_USE_CPP11 2
#endif
#ifdef ABSL_THREAD_IDENTITY_MODE
#error ABSL_THREAD_IDENTITY_MODE cannot be direcly set
#elif defined(ABSL_FORCE_THREAD_IDENTITY_MODE)
#define ABSL_THREAD_IDENTITY_MODE ABSL_FORCE_THREAD_IDENTITY_MODE
#elif defined(_WIN32) && !defined(__MINGW32__)
#define ABSL_THREAD_IDENTITY_MODE ABSL_THREAD_IDENTITY_MODE_USE_CPP11
#elif defined(__APPLE__) && defined(ABSL_HAVE_THREAD_LOCAL)
#define ABSL_THREAD_IDENTITY_MODE ABSL_THREAD_IDENTITY_MODE_USE_CPP11
#elif ABSL_PER_THREAD_TLS && defined(__GOOGLE_GRTE_VERSION__) && \
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
(__GOOGLE_GRTE_VERSION__ >= 20140228L)
// Support for async-safe TLS was specifically added in GRTEv4. It's not
// present in the upstream eglibc.
// Note: Current default for production systems.
#define ABSL_THREAD_IDENTITY_MODE ABSL_THREAD_IDENTITY_MODE_USE_TLS
#else
#define ABSL_THREAD_IDENTITY_MODE \
ABSL_THREAD_IDENTITY_MODE_USE_POSIX_SETSPECIFIC
#endif
#if ABSL_THREAD_IDENTITY_MODE == ABSL_THREAD_IDENTITY_MODE_USE_TLS || \
ABSL_THREAD_IDENTITY_MODE == ABSL_THREAD_IDENTITY_MODE_USE_CPP11
#if ABSL_PER_THREAD_TLS
ABSL_CONST_INIT extern ABSL_PER_THREAD_TLS_KEYWORD ThreadIdentity*
thread_identity_ptr;
#elif defined(ABSL_HAVE_THREAD_LOCAL)
ABSL_CONST_INIT extern thread_local ThreadIdentity* thread_identity_ptr;
#else
#error Thread-local storage not detected on this platform
#endif
// thread_local variables cannot be in headers exposed by DLLs. However, it is
// important for performance reasons in general that
// `CurrentThreadIdentityIfPresent` be inlined. This is not possible across a
// DLL boundary so, with DLLs, we opt to have the function not be inlined. Note
// that `CurrentThreadIdentityIfPresent` is declared above so we can exclude
// this entire inline definition when compiling as a DLL.
#if !defined(ABSL_BUILD_DLL) && !defined(ABSL_CONSUME_DLL)
inline ThreadIdentity* CurrentThreadIdentityIfPresent() {
return thread_identity_ptr;
}
#endif
#elif ABSL_THREAD_IDENTITY_MODE != \
ABSL_THREAD_IDENTITY_MODE_USE_POSIX_SETSPECIFIC
#error Unknown ABSL_THREAD_IDENTITY_MODE
#endif
} // namespace base_internal
ABSL_NAMESPACE_END
} // namespace absl
#endif // ABSL_BASE_INTERNAL_THREAD_IDENTITY_H_