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.

237 lines
8.4 KiB

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
// Copyright 2021 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.
#ifndef ABSL_DEBUGGING_INTERNAL_STACKTRACE_RISCV_INL_H_
#define ABSL_DEBUGGING_INTERNAL_STACKTRACE_RISCV_INL_H_
// Generate stack trace for riscv
#include <sys/ucontext.h>
#include "absl/base/config.h"
#if defined(__linux__)
#include <sys/mman.h>
#include <ucontext.h>
#include <unistd.h>
#endif
#include <atomic>
#include <cassert>
#include <cstdint>
#include <iostream>
#include "absl/base/attributes.h"
#include "absl/debugging/internal/address_is_readable.h"
#include "absl/debugging/internal/vdso_support.h"
#include "absl/debugging/stacktrace.h"
static const uintptr_t kUnknownFrameSize = 0;
#if defined(__linux__)
// Returns the address of the VDSO __kernel_rt_sigreturn function, if present.
static const unsigned char *GetKernelRtSigreturnAddress() {
constexpr uintptr_t kImpossibleAddress = 0;
ABSL_CONST_INIT static std::atomic<uintptr_t> memoized(kImpossibleAddress);
uintptr_t address = memoized.load(std::memory_order_relaxed);
if (address != kImpossibleAddress) {
return reinterpret_cast<const unsigned char *>(address);
}
address = reinterpret_cast<uintptr_t>(nullptr);
#if ABSL_HAVE_VDSO_SUPPORT
absl::debugging_internal::VDSOSupport vdso;
if (vdso.IsPresent()) {
absl::debugging_internal::VDSOSupport::SymbolInfo symbol_info;
// Symbol versioning pulled from arch/riscv/kernel/vdso/vdso.lds at v5.10.
auto lookup = [&](int type) {
return vdso.LookupSymbol("__vdso_rt_sigreturn", "LINUX_4.15", type,
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
&symbol_info);
};
if ((!lookup(STT_FUNC) && !lookup(STT_NOTYPE)) ||
symbol_info.address == nullptr) {
// Unexpected: VDSO is present, yet the expected symbol is missing or
// null.
assert(false && "VDSO is present, but doesn't have expected symbol");
} else {
if (reinterpret_cast<uintptr_t>(symbol_info.address) !=
kImpossibleAddress) {
address = reinterpret_cast<uintptr_t>(symbol_info.address);
} else {
assert(false && "VDSO returned invalid address");
}
}
}
#endif
memoized.store(address, std::memory_order_relaxed);
return reinterpret_cast<const unsigned char *>(address);
}
#endif // __linux__
// Compute the size of a stack frame in [low..high). We assume that low < high.
// Return size of kUnknownFrameSize.
template <typename T>
static inline uintptr_t ComputeStackFrameSize(const T *low, const T *high) {
const char *low_char_ptr = reinterpret_cast<const char *>(low);
const char *high_char_ptr = reinterpret_cast<const char *>(high);
return low < high ? high_char_ptr - low_char_ptr : kUnknownFrameSize;
}
// Given a pointer to a stack frame, locate and return the calling stackframe,
// or return null if no stackframe can be found. Perform sanity checks (the
// strictness of which is controlled by the boolean parameter
// "STRICT_UNWINDING") to reduce the chance that a bad pointer is returned.
template <bool STRICT_UNWINDING, bool WITH_CONTEXT>
ABSL_ATTRIBUTE_NO_SANITIZE_ADDRESS // May read random elements from stack.
ABSL_ATTRIBUTE_NO_SANITIZE_MEMORY // May read random elements from stack.
static void ** NextStackFrame(void **old_frame_pointer, const void *uc) {
// .
// .
// .
// +-> +----------------+
// | | return address |
// | | previous fp |
// | | ... |
// | +----------------+ <-+
// | | return address | |
// +---|- previous fp | |
// | ... | |
// $fp ->|----------------+ |
// | return address | |
// | previous fp -|---+
// $sp ->| ... |
// +----------------+
void **new_frame_pointer = reinterpret_cast<void **>(old_frame_pointer[-2]);
bool check_frame_size = true;
#if defined(__linux__)
if (WITH_CONTEXT && uc != nullptr) {
// Check to see if next frame's return address is __kernel_rt_sigreturn.
if (old_frame_pointer[-1] == GetKernelRtSigreturnAddress()) {
const ucontext_t *ucv = static_cast<const ucontext_t *>(uc);
// old_frame_pointer is not suitable for unwinding, look at ucontext to
// discover frame pointer before signal.
//
// RISCV ELF psABI has the frame pointer at x8/fp/s0.
// -- RISCV psABI Table 18.2
void **const pre_signal_frame_pointer =
reinterpret_cast<void **>(ucv->uc_mcontext.__gregs[8]);
// Check the alleged frame pointer is actually readable. This is to
// prevent "double fault" in case we hit the first fault due to stack
// corruption.
if (!absl::debugging_internal::AddressIsReadable(
pre_signal_frame_pointer))
return nullptr;
// Alleged frame pointer is readable, use it for further unwinding.
new_frame_pointer = pre_signal_frame_pointer;
// Skip frame size check if we return from a signal. We may be using an
// alterate stack for signals.
check_frame_size = false;
}
}
#endif
// The RISCV ELF psABI mandates that the stack pointer is always 16-byte
// aligned.
// FIXME(abdulras) this doesn't hold for ILP32E which only mandates a 4-byte
// alignment.
if ((reinterpret_cast<uintptr_t>(new_frame_pointer) & 15) != 0)
return nullptr;
// Check frame size. In strict mode, we assume frames to be under 100,000
// bytes. In non-strict mode, we relax the limit to 1MB.
if (check_frame_size) {
const uintptr_t max_size = STRICT_UNWINDING ? 100000 : 1000000;
const uintptr_t frame_size =
ComputeStackFrameSize(old_frame_pointer, new_frame_pointer);
if (frame_size == kUnknownFrameSize || frame_size > max_size)
return nullptr;
}
return new_frame_pointer;
}
template <bool IS_STACK_FRAMES, bool IS_WITH_CONTEXT>
ABSL_ATTRIBUTE_NO_SANITIZE_ADDRESS // May read random elements from stack.
ABSL_ATTRIBUTE_NO_SANITIZE_MEMORY // May read random elements from stack.
static int UnwindImpl(void **result, int *sizes, int max_depth, int skip_count,
const void *ucp, int *min_dropped_frames) {
// The `frame_pointer` that is computed here points to the top of the frame.
// The two words preceding the address are the return address and the previous
// frame pointer.
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(__GNUC__)
void **frame_pointer = reinterpret_cast<void **>(__builtin_frame_address(0));
#else
#error reading stack pointer not yet supported on this platform
#endif
int n = 0;
void *return_address = nullptr;
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
while (frame_pointer && n < max_depth) {
return_address = frame_pointer[-1];
// The absl::GetStackFrames routine is called when we are in some
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
// informational context (the failure signal handler for example). Use the
// non-strict unwinding rules to produce a stack trace that is as complete
// as possible (even if it contains a few bogus entries in some rare cases).
void **next_frame_pointer =
NextStackFrame<!IS_STACK_FRAMES, IS_WITH_CONTEXT>(frame_pointer, ucp);
if (skip_count > 0) {
skip_count--;
} else {
result[n] = return_address;
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 (IS_STACK_FRAMES) {
sizes[n] = ComputeStackFrameSize(frame_pointer, next_frame_pointer);
}
n++;
}
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
frame_pointer = next_frame_pointer;
}
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 (min_dropped_frames != nullptr) {
// Implementation detail: we clamp the max of frames we are willing to
// count, so as not to spend too much time in the loop below.
const int kMaxUnwind = 200;
Export of internal Abseil changes -- 389189dbb322df0d0468ab13edf7dc185dc63833 by Abseil Team <absl-team@google.com>: absl str_format.h can compile with -Wconversion and -Wsign-compare PiperOrigin-RevId: 420799960 -- 762e5adc429fc143756c42fe92fe8073c87c075f by Abseil Team <absl-team@google.com>: GetStackTraceWithContext: Fix min_dropped_frames when no frames are recorded Previously, if there were still frames to skip, they would be included in min_dropped_frames. PiperOrigin-RevId: 420766341 -- 7d4374b8eaa410f4f98ec03d6a8997dccadfb271 by Abseil Team <absl-team@google.com>: absl::flags compiles with -Wconversion and -Wsign-compare PiperOrigin-RevId: 420476807 -- 5f00f7805419d725fa1ff57b388e4c0750d1d6b0 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 420282152 -- bd5471fc34956acf3888bf90287b2aee4415c96d by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 420282033 -- 61c78020804e4290e9b2fe151aeaf99b198716ee by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 420281867 -- dbe3aad24b65ea11664401a307ea3d2f28e8a7b9 by Derek Mauro <dmauro@google.com>: Remove the incorrect test for the availability of [[nodiscard]] It should have been ABSL_HAVE_CPP_ATTRIBUTE(nodiscard) instead of ABSL_HAVE_ATTRIBUTE(nodiscard). As a result, some code is not compliant with [[nodiscard]], so we cannot simply correct the availability test. Recommend that C++17-only code use the standard [[nodiscard] directly. PiperOrigin-RevId: 420150702 GitOrigin-RevId: 389189dbb322df0d0468ab13edf7dc185dc63833 Change-Id: Idf6ebae3c4edd945c9032c7db3d0ab32d16e8078
3 years ago
int num_dropped_frames = 0;
for (int j = 0; frame_pointer != nullptr && j < kMaxUnwind; j++) {
if (skip_count > 0) {
skip_count--;
} else {
num_dropped_frames++;
}
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
frame_pointer =
NextStackFrame<!IS_STACK_FRAMES, IS_WITH_CONTEXT>(frame_pointer, ucp);
}
Export of internal Abseil changes -- 389189dbb322df0d0468ab13edf7dc185dc63833 by Abseil Team <absl-team@google.com>: absl str_format.h can compile with -Wconversion and -Wsign-compare PiperOrigin-RevId: 420799960 -- 762e5adc429fc143756c42fe92fe8073c87c075f by Abseil Team <absl-team@google.com>: GetStackTraceWithContext: Fix min_dropped_frames when no frames are recorded Previously, if there were still frames to skip, they would be included in min_dropped_frames. PiperOrigin-RevId: 420766341 -- 7d4374b8eaa410f4f98ec03d6a8997dccadfb271 by Abseil Team <absl-team@google.com>: absl::flags compiles with -Wconversion and -Wsign-compare PiperOrigin-RevId: 420476807 -- 5f00f7805419d725fa1ff57b388e4c0750d1d6b0 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 420282152 -- bd5471fc34956acf3888bf90287b2aee4415c96d by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 420282033 -- 61c78020804e4290e9b2fe151aeaf99b198716ee by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 420281867 -- dbe3aad24b65ea11664401a307ea3d2f28e8a7b9 by Derek Mauro <dmauro@google.com>: Remove the incorrect test for the availability of [[nodiscard]] It should have been ABSL_HAVE_CPP_ATTRIBUTE(nodiscard) instead of ABSL_HAVE_ATTRIBUTE(nodiscard). As a result, some code is not compliant with [[nodiscard]], so we cannot simply correct the availability test. Recommend that C++17-only code use the standard [[nodiscard] directly. PiperOrigin-RevId: 420150702 GitOrigin-RevId: 389189dbb322df0d0468ab13edf7dc185dc63833 Change-Id: Idf6ebae3c4edd945c9032c7db3d0ab32d16e8078
3 years ago
*min_dropped_frames = num_dropped_frames;
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
}
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 n;
}
namespace absl {
ABSL_NAMESPACE_BEGIN
namespace debugging_internal {
bool StackTraceWorksForTest() { return true; }
} // namespace debugging_internal
ABSL_NAMESPACE_END
} // namespace absl
#endif