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.

830 lines
25 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 2019 The Abseil Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// https://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include <array>
#include <string>
#include <vector>
#include "absl/base/internal/raw_logging.h"
#include "absl/base/macros.h"
#include "absl/container/inlined_vector.h"
#include "absl/strings/str_cat.h"
#include "benchmark/benchmark.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 {
void BM_InlinedVectorFill(benchmark::State& state) {
const int len = state.range(0);
absl::InlinedVector<int, 8> v;
v.reserve(len);
for (auto _ : state) {
v.resize(0); // Use resize(0) as InlinedVector releases storage on clear().
for (int i = 0; i < len; ++i) {
v.push_back(i);
}
benchmark::DoNotOptimize(v);
}
}
BENCHMARK(BM_InlinedVectorFill)->Range(1, 256);
void BM_InlinedVectorFillRange(benchmark::State& state) {
const int len = state.range(0);
const std::vector<int> src(len, len);
absl::InlinedVector<int, 8> v;
v.reserve(len);
for (auto _ : state) {
benchmark::DoNotOptimize(src);
v.assign(src.begin(), src.end());
benchmark::DoNotOptimize(v);
}
}
BENCHMARK(BM_InlinedVectorFillRange)->Range(1, 256);
void BM_StdVectorFill(benchmark::State& state) {
const int len = state.range(0);
std::vector<int> v;
v.reserve(len);
for (auto _ : state) {
v.clear();
for (int i = 0; i < len; ++i) {
v.push_back(i);
}
benchmark::DoNotOptimize(v);
}
}
BENCHMARK(BM_StdVectorFill)->Range(1, 256);
// The purpose of the next two benchmarks is to verify that
// absl::InlinedVector is efficient when moving is more efficent than
// copying. To do so, we use strings that are larger than the short
// string optimization.
bool StringRepresentedInline(std::string s) {
const char* chars = s.data();
std::string s1 = std::move(s);
return s1.data() != chars;
}
int GetNonShortStringOptimizationSize() {
for (int i = 24; i <= 192; i *= 2) {
if (!StringRepresentedInline(std::string(i, 'A'))) {
return i;
}
}
ABSL_RAW_LOG(
FATAL,
"Failed to find a string larger than the short string optimization");
return -1;
}
void BM_InlinedVectorFillString(benchmark::State& state) {
const int len = state.range(0);
const int no_sso = GetNonShortStringOptimizationSize();
std::string strings[4] = {std::string(no_sso, 'A'), std::string(no_sso, 'B'),
std::string(no_sso, 'C'), std::string(no_sso, 'D')};
for (auto _ : state) {
absl::InlinedVector<std::string, 8> v;
for (int i = 0; i < len; i++) {
v.push_back(strings[i & 3]);
}
}
state.SetItemsProcessed(static_cast<int64_t>(state.iterations()) * len);
}
BENCHMARK(BM_InlinedVectorFillString)->Range(0, 1024);
void BM_StdVectorFillString(benchmark::State& state) {
const int len = state.range(0);
const int no_sso = GetNonShortStringOptimizationSize();
std::string strings[4] = {std::string(no_sso, 'A'), std::string(no_sso, 'B'),
std::string(no_sso, 'C'), std::string(no_sso, 'D')};
for (auto _ : state) {
std::vector<std::string> v;
for (int i = 0; i < len; i++) {
v.push_back(strings[i & 3]);
}
}
state.SetItemsProcessed(static_cast<int64_t>(state.iterations()) * len);
}
BENCHMARK(BM_StdVectorFillString)->Range(0, 1024);
struct Buffer { // some arbitrary structure for benchmarking.
char* base;
int length;
int capacity;
void* user_data;
};
void BM_InlinedVectorAssignments(benchmark::State& state) {
const int len = state.range(0);
using BufferVec = absl::InlinedVector<Buffer, 2>;
BufferVec src;
src.resize(len);
BufferVec dst;
for (auto _ : state) {
benchmark::DoNotOptimize(dst);
benchmark::DoNotOptimize(src);
dst = src;
}
}
BENCHMARK(BM_InlinedVectorAssignments)
->Arg(0)
->Arg(1)
->Arg(2)
->Arg(3)
->Arg(4)
->Arg(20);
void BM_CreateFromContainer(benchmark::State& state) {
for (auto _ : state) {
absl::InlinedVector<int, 4> src{1, 2, 3};
benchmark::DoNotOptimize(src);
absl::InlinedVector<int, 4> dst(std::move(src));
benchmark::DoNotOptimize(dst);
}
}
BENCHMARK(BM_CreateFromContainer);
struct LargeCopyableOnly {
LargeCopyableOnly() : d(1024, 17) {}
LargeCopyableOnly(const LargeCopyableOnly& o) = default;
LargeCopyableOnly& operator=(const LargeCopyableOnly& o) = default;
std::vector<int> d;
};
struct LargeCopyableSwappable {
LargeCopyableSwappable() : d(1024, 17) {}
LargeCopyableSwappable(const LargeCopyableSwappable& o) = default;
LargeCopyableSwappable& operator=(LargeCopyableSwappable o) {
using std::swap;
swap(*this, o);
return *this;
}
friend void swap(LargeCopyableSwappable& a, LargeCopyableSwappable& b) {
using std::swap;
swap(a.d, b.d);
}
std::vector<int> d;
};
struct LargeCopyableMovable {
LargeCopyableMovable() : d(1024, 17) {}
// Use implicitly defined copy and move.
std::vector<int> d;
};
struct LargeCopyableMovableSwappable {
LargeCopyableMovableSwappable() : d(1024, 17) {}
LargeCopyableMovableSwappable(const LargeCopyableMovableSwappable& o) =
default;
LargeCopyableMovableSwappable(LargeCopyableMovableSwappable&& o) = default;
LargeCopyableMovableSwappable& operator=(LargeCopyableMovableSwappable o) {
using std::swap;
swap(*this, o);
return *this;
}
LargeCopyableMovableSwappable& operator=(LargeCopyableMovableSwappable&& o) =
default;
friend void swap(LargeCopyableMovableSwappable& a,
LargeCopyableMovableSwappable& b) {
using std::swap;
swap(a.d, b.d);
}
std::vector<int> d;
};
template <typename ElementType>
void BM_SwapElements(benchmark::State& state) {
const int len = state.range(0);
using Vec = absl::InlinedVector<ElementType, 32>;
Vec a(len);
Vec b;
for (auto _ : state) {
using std::swap;
benchmark::DoNotOptimize(a);
benchmark::DoNotOptimize(b);
swap(a, b);
}
}
BENCHMARK_TEMPLATE(BM_SwapElements, LargeCopyableOnly)->Range(0, 1024);
BENCHMARK_TEMPLATE(BM_SwapElements, LargeCopyableSwappable)->Range(0, 1024);
BENCHMARK_TEMPLATE(BM_SwapElements, LargeCopyableMovable)->Range(0, 1024);
BENCHMARK_TEMPLATE(BM_SwapElements, LargeCopyableMovableSwappable)
->Range(0, 1024);
// The following benchmark is meant to track the efficiency of the vector size
// as a function of stored type via the benchmark label. It is not meant to
// output useful sizeof operator performance. The loop is a dummy operation
// to fulfill the requirement of running the benchmark.
template <typename VecType>
void BM_Sizeof(benchmark::State& state) {
int size = 0;
for (auto _ : state) {
VecType vec;
size = sizeof(vec);
}
state.SetLabel(absl::StrCat("sz=", size));
}
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<char, 1>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<char, 4>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<char, 7>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<char, 8>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<int, 1>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<int, 4>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<int, 7>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<int, 8>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<void*, 1>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<void*, 4>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<void*, 7>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<void*, 8>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<std::string, 1>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<std::string, 4>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<std::string, 7>);
BENCHMARK_TEMPLATE(BM_Sizeof, absl::InlinedVector<std::string, 8>);
void BM_InlinedVectorIndexInlined(benchmark::State& state) {
absl::InlinedVector<int, 8> v = {1, 2, 3, 4, 5, 6, 7};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v[4]);
}
}
BENCHMARK(BM_InlinedVectorIndexInlined);
void BM_InlinedVectorIndexExternal(benchmark::State& state) {
absl::InlinedVector<int, 8> v = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v[4]);
}
}
BENCHMARK(BM_InlinedVectorIndexExternal);
void BM_StdVectorIndex(benchmark::State& state) {
std::vector<int> v = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v[4]);
}
}
BENCHMARK(BM_StdVectorIndex);
void BM_InlinedVectorDataInlined(benchmark::State& state) {
absl::InlinedVector<int, 8> v = {1, 2, 3, 4, 5, 6, 7};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v.data());
}
}
BENCHMARK(BM_InlinedVectorDataInlined);
void BM_InlinedVectorDataExternal(benchmark::State& state) {
absl::InlinedVector<int, 8> v = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v.data());
}
state.SetItemsProcessed(16 * static_cast<int64_t>(state.iterations()));
}
BENCHMARK(BM_InlinedVectorDataExternal);
void BM_StdVectorData(benchmark::State& state) {
std::vector<int> v = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v.data());
}
state.SetItemsProcessed(16 * static_cast<int64_t>(state.iterations()));
}
BENCHMARK(BM_StdVectorData);
void BM_InlinedVectorSizeInlined(benchmark::State& state) {
absl::InlinedVector<int, 8> v = {1, 2, 3, 4, 5, 6, 7};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v.size());
}
}
BENCHMARK(BM_InlinedVectorSizeInlined);
void BM_InlinedVectorSizeExternal(benchmark::State& state) {
absl::InlinedVector<int, 8> v = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v.size());
}
}
BENCHMARK(BM_InlinedVectorSizeExternal);
void BM_StdVectorSize(benchmark::State& state) {
std::vector<int> v = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v.size());
}
}
BENCHMARK(BM_StdVectorSize);
void BM_InlinedVectorEmptyInlined(benchmark::State& state) {
absl::InlinedVector<int, 8> v = {1, 2, 3, 4, 5, 6, 7};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v.empty());
}
}
BENCHMARK(BM_InlinedVectorEmptyInlined);
void BM_InlinedVectorEmptyExternal(benchmark::State& state) {
absl::InlinedVector<int, 8> v = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v.empty());
}
}
BENCHMARK(BM_InlinedVectorEmptyExternal);
void BM_StdVectorEmpty(benchmark::State& state) {
std::vector<int> v = {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
for (auto _ : state) {
benchmark::DoNotOptimize(v);
benchmark::DoNotOptimize(v.empty());
}
}
BENCHMARK(BM_StdVectorEmpty);
constexpr size_t kInlinedCapacity = 4;
constexpr size_t kLargeSize = kInlinedCapacity * 2;
constexpr size_t kSmallSize = kInlinedCapacity / 2;
constexpr size_t kBatchSize = 100;
#define ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_FunctionTemplate, T) \
BENCHMARK_TEMPLATE(BM_FunctionTemplate, T, kLargeSize); \
BENCHMARK_TEMPLATE(BM_FunctionTemplate, T, kSmallSize)
#define ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_FunctionTemplate, T) \
BENCHMARK_TEMPLATE(BM_FunctionTemplate, T, kLargeSize, kLargeSize); \
BENCHMARK_TEMPLATE(BM_FunctionTemplate, T, kLargeSize, kSmallSize); \
BENCHMARK_TEMPLATE(BM_FunctionTemplate, T, kSmallSize, kLargeSize); \
BENCHMARK_TEMPLATE(BM_FunctionTemplate, T, kSmallSize, kSmallSize)
template <typename T>
using InlVec = absl::InlinedVector<T, kInlinedCapacity>;
struct TrivialType {
size_t val;
};
class NontrivialType {
public:
ABSL_ATTRIBUTE_NOINLINE NontrivialType() : val_() {
benchmark::DoNotOptimize(*this);
}
ABSL_ATTRIBUTE_NOINLINE NontrivialType(const NontrivialType& other)
: val_(other.val_) {
benchmark::DoNotOptimize(*this);
}
ABSL_ATTRIBUTE_NOINLINE NontrivialType& operator=(
const NontrivialType& other) {
val_ = other.val_;
benchmark::DoNotOptimize(*this);
return *this;
}
ABSL_ATTRIBUTE_NOINLINE ~NontrivialType() noexcept {
benchmark::DoNotOptimize(*this);
}
private:
size_t val_;
};
template <typename T, typename PrepareVecFn, typename TestVecFn>
void BatchedBenchmark(benchmark::State& state, PrepareVecFn prepare_vec,
TestVecFn test_vec) {
std::array<InlVec<T>, kBatchSize> vector_batch{};
while (state.KeepRunningBatch(kBatchSize)) {
// Prepare batch
state.PauseTiming();
for (size_t i = 0; i < kBatchSize; ++i) {
prepare_vec(vector_batch.data() + i, i);
}
benchmark::DoNotOptimize(vector_batch);
state.ResumeTiming();
// Test batch
for (size_t i = 0; i < kBatchSize; ++i) {
test_vec(vector_batch.data() + i, i);
}
}
}
template <typename T, size_t ToSize>
void BM_ConstructFromSize(benchmark::State& state) {
using VecT = InlVec<T>;
auto size = ToSize;
BatchedBenchmark<T>(
state,
/* prepare_vec = */ [](InlVec<T>* vec, size_t) { vec->~VecT(); },
/* test_vec = */
[&](void* ptr, size_t) {
benchmark::DoNotOptimize(size);
::new (ptr) VecT(size);
});
}
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_ConstructFromSize, TrivialType);
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_ConstructFromSize, NontrivialType);
template <typename T, size_t ToSize>
void BM_ConstructFromSizeRef(benchmark::State& state) {
using VecT = InlVec<T>;
auto size = ToSize;
auto ref = T();
BatchedBenchmark<T>(
state,
/* prepare_vec = */ [](InlVec<T>* vec, size_t) { vec->~VecT(); },
/* test_vec = */
[&](void* ptr, size_t) {
benchmark::DoNotOptimize(size);
benchmark::DoNotOptimize(ref);
::new (ptr) VecT(size, ref);
});
}
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_ConstructFromSizeRef, TrivialType);
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_ConstructFromSizeRef, NontrivialType);
template <typename T, size_t ToSize>
void BM_ConstructFromRange(benchmark::State& state) {
using VecT = InlVec<T>;
std::array<T, ToSize> arr{};
BatchedBenchmark<T>(
state,
/* prepare_vec = */ [](InlVec<T>* vec, size_t) { vec->~VecT(); },
/* test_vec = */
[&](void* ptr, size_t) {
benchmark::DoNotOptimize(arr);
::new (ptr) VecT(arr.begin(), arr.end());
});
}
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_ConstructFromRange, TrivialType);
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_ConstructFromRange, NontrivialType);
template <typename T, size_t ToSize>
void BM_ConstructFromCopy(benchmark::State& state) {
using VecT = InlVec<T>;
VecT other_vec(ToSize);
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[](InlVec<T>* vec, size_t) { vec->~VecT(); },
/* test_vec = */
[&](void* ptr, size_t) {
benchmark::DoNotOptimize(other_vec);
::new (ptr) VecT(other_vec);
});
}
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_ConstructFromCopy, TrivialType);
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_ConstructFromCopy, NontrivialType);
template <typename T, size_t ToSize>
void BM_ConstructFromMove(benchmark::State& state) {
using VecT = InlVec<T>;
std::array<VecT, kBatchSize> vector_batch{};
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[&](InlVec<T>* vec, size_t i) {
vector_batch[i].clear();
vector_batch[i].resize(ToSize);
vec->~VecT();
},
/* test_vec = */
[&](void* ptr, size_t i) {
benchmark::DoNotOptimize(vector_batch[i]);
::new (ptr) VecT(std::move(vector_batch[i]));
});
}
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_ConstructFromMove, TrivialType);
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_ConstructFromMove, NontrivialType);
Export of internal Abseil changes -- c68f1886f5e8fd90eb0c2d2e68feaf00a7cdacda by CJ Johnson <johnsoncj@google.com>: Introduce absl::Cleanup to the OSS repo PiperOrigin-RevId: 354583156 -- 17030cf388e10f7eb959e3e566326d1072ce392e by Abseil Team <absl-team@google.com>: Internal change only PiperOrigin-RevId: 354574953 -- e979d7236d4f3252e79ddda6739b67a9a326bf6d by CJ Johnson <johnsoncj@google.com>: Internal change PiperOrigin-RevId: 354545297 -- 7ea02b3783f7f49ef97d86a8f6580a19cc57df14 by Abseil Team <absl-team@google.com>: Pre-allocate memory for vectors where the size is known. PiperOrigin-RevId: 354344576 -- 9246c7cb11f1d6444f79ebe25acc69a8a9b870e0 by Matt Kulukundis <kfm@google.com>: Add support for Elbrus 2000 (e2k) Import of https://github.com/abseil/abseil-cpp/pull/889 PiperOrigin-RevId: 354344013 -- 0fc93d359cc1fb307552e917b37b7b2e7eed822f by Abseil Team <absl-team@google.com>: Integrate CordRepRing logic into cord (but do not enable it) PiperOrigin-RevId: 354312238 -- eda05622f7da71466723acb33403f783529df24b by Abseil Team <absl-team@google.com>: Protect ignore diagnostic with "__has_warning". PiperOrigin-RevId: 354112334 -- 47716c5d8fb10efa4fdd801d28bac414c6f8ec32 by Abseil Team <absl-team@google.com>: Rearrange InlinedVector copy constructor and destructor to treat a few special cases inline and then tail-call a non-inlined routine for the rest. In particular, we optimize for empty vectors in both cases. Added a couple of benchmarks that copy either an InlVec<int64> or an InlVec<InlVec<int64>>. Speed difference: ``` BM_CopyTrivial/0 0.92ns +- 0% 0.47ns +- 0% -48.91% (p=0.000 n=11+12) BM_CopyTrivial/1 0.92ns +- 0% 1.15ns +- 0% +25.00% (p=0.000 n=10+9) BM_CopyTrivial/8 8.57ns +- 0% 10.72ns +- 1% +25.16% (p=0.000 n=10+12) BM_CopyNonTrivial/0 3.21ns +- 0% 0.70ns +- 0% -78.23% (p=0.000 n=12+10) BM_CopyNonTrivial/1 5.88ns +- 1% 5.51ns +- 0% -6.28% (p=0.000 n=10+8) BM_CopyNonTrivial/8 21.5ns +- 1% 15.2ns +- 2% -29.23% (p=0.000 n=12+12) ``` Note: the slowdowns are a few cycles which is expected given the procedure call added in that case. We decided this is a good tradeoff given the code size reductions and the more significant speedups for empty vectors. Size difference (as measured by nm): ``` BM_CopyTrivial from 1048 bytes to 326 bytes. BM_CopyNonTrivial from 749 bytes to 470 bytes. ``` Code size for a large binary drops by ~500KB (from 349415719 to 348906015 348906191). All of the benchmarks that showed a significant difference: Ones that improve with this CL: ``` BM_CopyNonTrivial/0 3.21ns +- 0% 0.70ns +- 0% -78.23% (p=0.000 n=12+10) BM_InlinedVectorFillString/0 0.93ns +- 0% 0.24ns +- 0% -74.19% (p=0.000 n=12+10) BM_InlinedVectorAssignments/1 10.5ns +- 0% 4.1ns +- 0% -60.64% (p=0.000 n=11+10) BM_InlinedVectorAssignments/2 10.7ns +- 0% 4.4ns +- 0% -59.08% (p=0.000 n=11+11) BM_CopyTrivial/0 0.92ns +- 0% 0.47ns +- 0% -48.91% (p=0.000 n=11+12) BM_CopyNonTrivial/8 21.5ns +- 1% 15.2ns +- 2% -29.23% (p=0.000 n=12+12) BM_StdVectorEmpty 0.47ns +- 1% 0.35ns +- 0% -24.73% (p=0.000 n=12+12) BM_StdVectorSize 0.46ns +- 2% 0.35ns +- 0% -24.32% (p=0.000 n=12+12) BM_SwapElements<LargeCopyableOnly>/0 3.44ns +- 0% 2.76ns +- 1% -19.83% (p=0.000 n=11+11) BM_InlinedVectorFillRange/256 20.7ns +- 1% 17.8ns +- 0% -14.08% (p=0.000 n=12+9) BM_CopyNonTrivial/1 5.88ns +- 1% 5.51ns +- 0% -6.28% (p=0.000 n=10+8) BM_SwapElements<LargeCopyableMovable>/1 4.19ns +- 0% 3.95ns +- 1% -5.63% (p=0.000 n=11+12) BM_SwapElements<LargeCopyableMovableSwappable>/1 4.18ns +- 0% 3.99ns +- 0% -4.70% (p=0.000 n=9+11) BM_SwapElements<LargeCopyableMovable>/0 2.41ns +- 0% 2.31ns +- 0% -4.45% (p=0.000 n=12+12) BM_InlinedVectorFillRange/64 8.25ns +- 0% 8.04ns +- 0% -2.51% (p=0.000 n=12+11) BM_SwapElements<LargeCopyableOnly>/1 82.4ns +- 0% 81.5ns +- 0% -1.06% (p=0.000 n=12+12) ``` Ones that get worse with this CL: ``` BM_CopyTrivial/1 0.92ns +- 0% 1.15ns +- 0% +25.00% (p=0.000 n=10+9) BM_CopyTrivial/8 8.57ns +- 0% 10.72ns +- 1% +25.16% (p=0.000 n=10+12) BM_SwapElements<LargeCopyableMovableSwappable>/512 1.48ns +- 1% 1.66ns +- 1% +11.88% (p=0.000 n=12+12) BM_InlinedVectorFillString/1 11.5ns +- 0% 12.8ns +- 1% +11.62% (p=0.000 n=12+11) BM_SwapElements<LargeCopyableMovableSwappable>/64 1.48ns +- 2% 1.66ns +- 1% +11.66% (p=0.000 n=12+11) BM_SwapElements<LargeCopyableMovableSwappable>/1k 1.48ns +- 1% 1.65ns +- 2% +11.32% (p=0.000 n=12+12) BM_SwapElements<LargeCopyableMovable>/512 1.48ns +- 2% 1.58ns +- 4% +6.62% (p=0.000 n=11+12) BM_SwapElements<LargeCopyableMovable>/1k 1.49ns +- 2% 1.58ns +- 3% +6.05% (p=0.000 n=12+12) BM_SwapElements<LargeCopyableMovable>/64 1.48ns +- 2% 1.57ns +- 4% +6.04% (p=0.000 n=11+12) BM_InlinedVectorFillRange/1 4.81ns +- 0% 5.05ns +- 0% +4.83% (p=0.000 n=11+11) BM_InlinedVectorFillString/8 79.4ns +- 1% 83.1ns +- 1% +4.64% (p=0.000 n=10+12) BM_StdVectorFillString/1 16.3ns +- 0% 16.6ns +- 0% +2.13% (p=0.000 n=11+8) ``` PiperOrigin-RevId: 353906786 -- 8e26518b3cec9c598e5e9573c46c3bd1b03a67ef by Abseil Team <absl-team@google.com>: Internal change PiperOrigin-RevId: 353737330 -- f206ae0983e58c9904ed8b8f05f9caf564a446be by Matt Kulukundis <kfm@google.com>: Import of CCTZ from GitHub. PiperOrigin-RevId: 353682256 GitOrigin-RevId: c68f1886f5e8fd90eb0c2d2e68feaf00a7cdacda Change-Id: I5790c1036c4f543c701d1039848fabf7ae881ad8
4 years ago
// Measure cost of copy-constructor+destructor.
void BM_CopyTrivial(benchmark::State& state) {
const int n = state.range(0);
InlVec<int64_t> src(n);
for (auto s : state) {
InlVec<int64_t> copy(src);
benchmark::DoNotOptimize(copy);
}
}
BENCHMARK(BM_CopyTrivial)->Arg(0)->Arg(1)->Arg(kLargeSize);
// Measure cost of copy-constructor+destructor.
void BM_CopyNonTrivial(benchmark::State& state) {
const int n = state.range(0);
InlVec<InlVec<int64_t>> src(n);
for (auto s : state) {
InlVec<InlVec<int64_t>> copy(src);
benchmark::DoNotOptimize(copy);
}
}
BENCHMARK(BM_CopyNonTrivial)->Arg(0)->Arg(1)->Arg(kLargeSize);
Export of internal Abseil changes -- f012012ef78234a6a4585321b67d7b7c92ebc266 by Laramie Leavitt <lar@google.com>: Slight restructuring of absl/random/internal randen implementation. Convert round-keys.inc into randen_round_keys.cc file. Consistently use a 128-bit pointer type for internal method parameters. This allows simpler pointer arithmetic in C++ & permits removal of some constants and casts. Remove some redundancy in comments & constexpr variables. Specifically, all references to Randen algorithm parameters use RandenTraits; duplication in RandenSlow removed. PiperOrigin-RevId: 312190313 -- dc8b42e054046741e9ed65335bfdface997c6063 by Abseil Team <absl-team@google.com>: Internal change. PiperOrigin-RevId: 312167304 -- f13d248fafaf206492c1362c3574031aea3abaf7 by Matthew Brown <matthewbr@google.com>: Cleanup StrFormat extensions a little. PiperOrigin-RevId: 312166336 -- 9d9117589667afe2332bb7ad42bc967ca7c54502 by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 312105213 -- 9a12b9b3aa0e59b8ee6cf9408ed0029045543a9b by Abseil Team <absl-team@google.com>: Complete IGNORE_TYPE macro renaming. PiperOrigin-RevId: 311999699 -- 64756f20d61021d999bd0d4c15e9ad3857382f57 by Gennadiy Rozental <rogeeff@google.com>: Switch to fixed bytes specific default value. This fixes the Abseil Flags for big endian platforms. PiperOrigin-RevId: 311844448 -- bdbe6b5b29791dbc3816ada1828458b3010ff1e9 by Laramie Leavitt <lar@google.com>: Change many distribution tests to use pcg_engine as a deterministic source of entropy. It's reasonable to test that the BitGen itself has good entropy, however when testing the cross product of all random distributions x all the architecture variations x all submitted changes results in a large number of tests. In order to account for these failures while still using good entropy requires that our allowed sigma need to account for all of these independent tests. Our current sigma values are too restrictive, and we see a lot of failures, so we have to either relax the sigma values or convert some of the statistical tests to use deterministic values. This changelist does the latter. PiperOrigin-RevId: 311840096 GitOrigin-RevId: f012012ef78234a6a4585321b67d7b7c92ebc266 Change-Id: Ic84886f38ff30d7d72c126e9b63c9a61eb729a1a
5 years ago
template <typename T, size_t FromSize, size_t ToSize>
void BM_AssignSizeRef(benchmark::State& state) {
auto size = ToSize;
auto ref = T();
BatchedBenchmark<T>(
state,
/* prepare_vec = */ [](InlVec<T>* vec, size_t) { vec->resize(FromSize); },
/* test_vec = */
[&](InlVec<T>* vec, size_t) {
benchmark::DoNotOptimize(size);
benchmark::DoNotOptimize(ref);
vec->assign(size, ref);
});
}
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_AssignSizeRef, TrivialType);
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_AssignSizeRef, NontrivialType);
template <typename T, size_t FromSize, size_t ToSize>
void BM_AssignRange(benchmark::State& state) {
std::array<T, ToSize> arr{};
BatchedBenchmark<T>(
state,
/* prepare_vec = */ [](InlVec<T>* vec, size_t) { vec->resize(FromSize); },
/* test_vec = */
[&](InlVec<T>* vec, size_t) {
benchmark::DoNotOptimize(arr);
vec->assign(arr.begin(), arr.end());
});
}
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_AssignRange, TrivialType);
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_AssignRange, NontrivialType);
template <typename T, size_t FromSize, size_t ToSize>
void BM_AssignFromCopy(benchmark::State& state) {
InlVec<T> other_vec(ToSize);
BatchedBenchmark<T>(
state,
/* prepare_vec = */ [](InlVec<T>* vec, size_t) { vec->resize(FromSize); },
/* test_vec = */
[&](InlVec<T>* vec, size_t) {
benchmark::DoNotOptimize(other_vec);
*vec = other_vec;
});
}
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_AssignFromCopy, TrivialType);
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_AssignFromCopy, NontrivialType);
template <typename T, size_t FromSize, size_t ToSize>
void BM_AssignFromMove(benchmark::State& state) {
using VecT = InlVec<T>;
std::array<VecT, kBatchSize> vector_batch{};
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[&](InlVec<T>* vec, size_t i) {
vector_batch[i].clear();
vector_batch[i].resize(ToSize);
vec->resize(FromSize);
},
/* test_vec = */
[&](InlVec<T>* vec, size_t i) {
benchmark::DoNotOptimize(vector_batch[i]);
*vec = std::move(vector_batch[i]);
});
}
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_AssignFromMove, TrivialType);
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_AssignFromMove, NontrivialType);
template <typename T, size_t FromSize, size_t ToSize>
void BM_ResizeSize(benchmark::State& state) {
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[](InlVec<T>* vec, size_t) {
vec->clear();
vec->resize(FromSize);
},
/* test_vec = */
[](InlVec<T>* vec, size_t) { vec->resize(ToSize); });
}
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_ResizeSize, TrivialType);
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_ResizeSize, NontrivialType);
template <typename T, size_t FromSize, size_t ToSize>
void BM_ResizeSizeRef(benchmark::State& state) {
auto t = T();
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[](InlVec<T>* vec, size_t) {
vec->clear();
vec->resize(FromSize);
},
/* test_vec = */
[&](InlVec<T>* vec, size_t) {
benchmark::DoNotOptimize(t);
vec->resize(ToSize, t);
});
}
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_ResizeSizeRef, TrivialType);
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_ResizeSizeRef, NontrivialType);
template <typename T, size_t FromSize, size_t ToSize>
void BM_InsertSizeRef(benchmark::State& state) {
auto t = T();
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[](InlVec<T>* vec, size_t) {
vec->clear();
vec->resize(FromSize);
},
/* test_vec = */
[&](InlVec<T>* vec, size_t) {
benchmark::DoNotOptimize(t);
auto* pos = vec->data() + (vec->size() / 2);
vec->insert(pos, t);
});
}
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_InsertSizeRef, TrivialType);
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_InsertSizeRef, NontrivialType);
template <typename T, size_t FromSize, size_t ToSize>
void BM_InsertRange(benchmark::State& state) {
InlVec<T> other_vec(ToSize);
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[](InlVec<T>* vec, size_t) {
vec->clear();
vec->resize(FromSize);
},
/* test_vec = */
[&](InlVec<T>* vec, size_t) {
benchmark::DoNotOptimize(other_vec);
auto* pos = vec->data() + (vec->size() / 2);
vec->insert(pos, other_vec.begin(), other_vec.end());
});
}
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_InsertRange, TrivialType);
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_InsertRange, NontrivialType);
template <typename T, size_t FromSize>
void BM_EmplaceBack(benchmark::State& state) {
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[](InlVec<T>* vec, size_t) {
vec->clear();
vec->resize(FromSize);
},
/* test_vec = */
[](InlVec<T>* vec, size_t) { vec->emplace_back(); });
}
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_EmplaceBack, TrivialType);
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_EmplaceBack, NontrivialType);
template <typename T, size_t FromSize>
void BM_PopBack(benchmark::State& state) {
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[](InlVec<T>* vec, size_t) {
vec->clear();
vec->resize(FromSize);
},
/* test_vec = */
[](InlVec<T>* vec, size_t) { vec->pop_back(); });
}
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_PopBack, TrivialType);
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_PopBack, NontrivialType);
template <typename T, size_t FromSize>
void BM_EraseOne(benchmark::State& state) {
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[](InlVec<T>* vec, size_t) {
vec->clear();
vec->resize(FromSize);
},
/* test_vec = */
[](InlVec<T>* vec, size_t) {
auto* pos = vec->data() + (vec->size() / 2);
vec->erase(pos);
});
}
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_EraseOne, TrivialType);
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_EraseOne, NontrivialType);
template <typename T, size_t FromSize>
void BM_EraseRange(benchmark::State& state) {
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[](InlVec<T>* vec, size_t) {
vec->clear();
vec->resize(FromSize);
},
/* test_vec = */
[](InlVec<T>* vec, size_t) {
auto* pos = vec->data() + (vec->size() / 2);
vec->erase(pos, pos + 1);
});
}
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_EraseRange, TrivialType);
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_EraseRange, NontrivialType);
template <typename T, size_t FromSize>
void BM_Clear(benchmark::State& state) {
BatchedBenchmark<T>(
state,
/* prepare_vec = */ [](InlVec<T>* vec, size_t) { vec->resize(FromSize); },
/* test_vec = */ [](InlVec<T>* vec, size_t) { vec->clear(); });
}
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_Clear, TrivialType);
ABSL_INTERNAL_BENCHMARK_ONE_SIZE(BM_Clear, NontrivialType);
template <typename T, size_t FromSize, size_t ToCapacity>
void BM_Reserve(benchmark::State& state) {
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[](InlVec<T>* vec, size_t) {
vec->clear();
vec->resize(FromSize);
},
/* test_vec = */
[](InlVec<T>* vec, size_t) { vec->reserve(ToCapacity); });
}
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_Reserve, TrivialType);
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_Reserve, NontrivialType);
template <typename T, size_t FromCapacity, size_t ToCapacity>
void BM_ShrinkToFit(benchmark::State& state) {
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[](InlVec<T>* vec, size_t) {
vec->clear();
vec->resize(ToCapacity);
vec->reserve(FromCapacity);
},
/* test_vec = */ [](InlVec<T>* vec, size_t) { vec->shrink_to_fit(); });
}
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_ShrinkToFit, TrivialType);
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_ShrinkToFit, NontrivialType);
template <typename T, size_t FromSize, size_t ToSize>
void BM_Swap(benchmark::State& state) {
using VecT = InlVec<T>;
std::array<VecT, kBatchSize> vector_batch{};
BatchedBenchmark<T>(
state,
/* prepare_vec = */
[&](InlVec<T>* vec, size_t i) {
vector_batch[i].clear();
vector_batch[i].resize(ToSize);
vec->resize(FromSize);
},
/* test_vec = */
[&](InlVec<T>* vec, size_t i) {
using std::swap;
benchmark::DoNotOptimize(vector_batch[i]);
swap(*vec, vector_batch[i]);
});
}
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_Swap, TrivialType);
ABSL_INTERNAL_BENCHMARK_TWO_SIZE(BM_Swap, NontrivialType);
} // namespace