diff options
author | Abseil Team <absl-team@google.com> | 2021-07-02 20:42:20 -0700 |
---|---|---|
committer | Derek Mauro <dmauro@google.com> | 2021-07-05 18:10:27 -0400 |
commit | 58e042da9210710dc4ac3b320e48b54e2449521e (patch) | |
tree | 91d76560a010a8ce5b5a13c23d922ba2a4bce579 /absl/flags/flag_benchmark.cc | |
parent | 9a7e447c511dae7276ab65fde4d04f6ed52b39c9 (diff) |
Export of internal Abseil changes
--
1620e8ffaa93ef24510ca60c7fff2a07248ac9f6 by Abseil Team <absl-team@google.com>:
Update comment.
PiperOrigin-RevId: 382858259
--
20db116f28469149d10e0f7f8b976cb903dd4879 by Gennadiy Rozental <rogeeff@google.com>:
Add benchmark running on multiple flags.
Update size_tester to include cost of absl::GetFlag call.
Add size_tester invocation for bool flag.
New benchmark better represent GetFlag usage.
PiperOrigin-RevId: 382820341
--
2e097ad3811c4e329f75b98877a5e74c1d3d84fd by Abseil Team <absl-team@google.com>:
Avoid 64x64->128 multiplication in absl::Hash's mix on AArch64
On AArch64, calculating a 128-bit product is inefficient, because it requires a sequence of two instructions to calculate the upper and lower halves of the result. So calculate a 64-bit product instead.
Making MultType 64-bits means the upper 32 bits of the result do not participate in shift/xor, but the add/multiply gives us sufficient mixing.
PiperOrigin-RevId: 382625931
--
f3ae3f32cb53168c8dc91b766f2932dc87cec503 by Abseil Team <absl-team@google.com>:
Remove homegrown Round implementation
absl/time/duration.cc defined a Round implementation to accommodate old
versions of MSVC that lacked std::round(long double). Abseil no longer
supports those MSVCs, so we don’t need the homegrown implementation
anymore. Remove it, and replace calls to it with std::rint.
PiperOrigin-RevId: 382605191
--
a13631c91bf5478289e1a512ce215c85501a26f7 by Martijn Vels <mvels@google.com>:
Move the Consume() conversion functions out of cord_rep_ring into cord_rep_consume.
This makes these functions generic, so we can repurpose these for the new Btree conversion functions.
PiperOrigin-RevId: 382594902
--
7394c737500c2d8371fcf913b21ad1b321ba499d by Benjamin Barenblat <bbaren@google.com>:
Remove homegrown Round implementation
absl/time/duration.cc defined a Round implementation to accommodate old
versions of MSVC that lacked std::round(long double). Abseil no longer
supports those MSVCs, so we don’t need the homegrown implementation
anymore. Remove it, and replace calls to it with std::rint.
PiperOrigin-RevId: 382569900
--
d72a761f43dc5c9b9510c3a1363177ed26646b5d by Abseil Team <absl-team@google.com>:
Prefer `getentropy` for Emscripten.
It needs a different header, so I've separated it out from the GLIBC
check above.
PiperOrigin-RevId: 382332475
--
74e261dbb467741b2ddd8b490e04c531fdd2f559 by Martijn Vels <mvels@google.com>:
Add BTREE tag for CordRepNode implementing a Btree cord.
This change only forward declared the CordRepBtree class (not implemented yet) and defines the enum value BTREE. While RING and BTREE should never co-exist, we define a new value for BTREE so as not to make transitioning between RING and BTREE harder than it needs to be. This changes shifts the FLAT value / computation from FLAT = 4 to FLAT =5
PiperOrigin-RevId: 382326710
GitOrigin-RevId: 1620e8ffaa93ef24510ca60c7fff2a07248ac9f6
Change-Id: Ia8f99dde3874808f56062bd37ab3e63764099734
Diffstat (limited to 'absl/flags/flag_benchmark.cc')
-rw-r--r-- | absl/flags/flag_benchmark.cc | 138 |
1 files changed, 110 insertions, 28 deletions
diff --git a/absl/flags/flag_benchmark.cc b/absl/flags/flag_benchmark.cc index 57584f85..fc572d9c 100644 --- a/absl/flags/flag_benchmark.cc +++ b/absl/flags/flag_benchmark.cc @@ -101,7 +101,39 @@ std::string AbslUnparseFlag(const UDT&) { return ""; } A(AbslDuration) \ A(UDT) -#define FLAG_DEF(T) ABSL_FLAG(T, T##_flag, {}, ""); +#define REPLICATE_0(A, T, name, index) A(T, name, index) +#define REPLICATE_1(A, T, name, index) \ + REPLICATE_0(A, T, name, index##0) REPLICATE_0(A, T, name, index##1) +#define REPLICATE_2(A, T, name, index) \ + REPLICATE_1(A, T, name, index##0) REPLICATE_1(A, T, name, index##1) +#define REPLICATE_3(A, T, name, index) \ + REPLICATE_2(A, T, name, index##0) REPLICATE_2(A, T, name, index##1) +#define REPLICATE_4(A, T, name, index) \ + REPLICATE_3(A, T, name, index##0) REPLICATE_3(A, T, name, index##1) +#define REPLICATE_5(A, T, name, index) \ + REPLICATE_4(A, T, name, index##0) REPLICATE_4(A, T, name, index##1) +#define REPLICATE_6(A, T, name, index) \ + REPLICATE_5(A, T, name, index##0) REPLICATE_5(A, T, name, index##1) +#define REPLICATE_7(A, T, name, index) \ + REPLICATE_6(A, T, name, index##0) REPLICATE_6(A, T, name, index##1) +#define REPLICATE_8(A, T, name, index) \ + REPLICATE_7(A, T, name, index##0) REPLICATE_7(A, T, name, index##1) +#define REPLICATE_9(A, T, name, index) \ + REPLICATE_8(A, T, name, index##0) REPLICATE_8(A, T, name, index##1) +#if defined(_MSC_VER) +#define REPLICATE(A, T, name) \ + REPLICATE_7(A, T, name, 0) REPLICATE_7(A, T, name, 1) +#define SINGLE_FLAG(T) FLAGS_##T##_flag_00000000 +#else +#define REPLICATE(A, T, name) \ + REPLICATE_9(A, T, name, 0) REPLICATE_9(A, T, name, 1) +#define SINGLE_FLAG(T) FLAGS_##T##_flag_0000000000 +#endif +#define REPLICATE_ALL(A, T, name) \ + REPLICATE_9(A, T, name, 0) REPLICATE_9(A, T, name, 1) + +#define COUNT(T, name, index) +1 +constexpr size_t kNumFlags = 0 REPLICATE(COUNT, _, _); #if defined(__clang__) && defined(__linux__) // Force the flags used for benchmarks into a separate ELF section. @@ -110,38 +142,87 @@ std::string AbslUnparseFlag(const UDT&) { return ""; } // benchmark results more reproducible across unrelated code changes. #pragma clang section data = ".benchmark_flags" #endif +#define DEFINE_FLAG(T, name, index) ABSL_FLAG(T, name##_##index, {}, ""); +#define FLAG_DEF(T) REPLICATE(DEFINE_FLAG, T, T##_flag); BENCHMARKED_TYPES(FLAG_DEF) #if defined(__clang__) && defined(__linux__) #pragma clang section data = "" #endif // Register thousands of flags to bloat up the size of the registry. // This mimics real life production binaries. -#define DEFINE_FLAG_0(name) ABSL_FLAG(int, name, 0, ""); -#define DEFINE_FLAG_1(name) DEFINE_FLAG_0(name##0) DEFINE_FLAG_0(name##1) -#define DEFINE_FLAG_2(name) DEFINE_FLAG_1(name##0) DEFINE_FLAG_1(name##1) -#define DEFINE_FLAG_3(name) DEFINE_FLAG_2(name##0) DEFINE_FLAG_2(name##1) -#define DEFINE_FLAG_4(name) DEFINE_FLAG_3(name##0) DEFINE_FLAG_3(name##1) -#define DEFINE_FLAG_5(name) DEFINE_FLAG_4(name##0) DEFINE_FLAG_4(name##1) -#define DEFINE_FLAG_6(name) DEFINE_FLAG_5(name##0) DEFINE_FLAG_5(name##1) -#define DEFINE_FLAG_7(name) DEFINE_FLAG_6(name##0) DEFINE_FLAG_6(name##1) -#define DEFINE_FLAG_8(name) DEFINE_FLAG_7(name##0) DEFINE_FLAG_7(name##1) -#define DEFINE_FLAG_9(name) DEFINE_FLAG_8(name##0) DEFINE_FLAG_8(name##1) -#define DEFINE_FLAG_10(name) DEFINE_FLAG_9(name##0) DEFINE_FLAG_9(name##1) -#define DEFINE_FLAG_11(name) DEFINE_FLAG_10(name##0) DEFINE_FLAG_10(name##1) -#define DEFINE_FLAG_12(name) DEFINE_FLAG_11(name##0) DEFINE_FLAG_11(name##1) -DEFINE_FLAG_12(bloat_flag_); +#define BLOAT_FLAG(_unused1, _unused2, index) \ + ABSL_FLAG(int, bloat_flag_##index, 0, ""); +REPLICATE_ALL(BLOAT_FLAG, _, _) namespace { -#define BM_GetFlag(T) \ - void BM_GetFlag_##T(benchmark::State& state) { \ - for (auto _ : state) { \ - benchmark::DoNotOptimize(absl::GetFlag(FLAGS_##T##_flag)); \ - } \ - } \ - BENCHMARK(BM_GetFlag_##T)->ThreadRange(1, 16); +#define FLAG_PTR(T, name, index) &FLAGS_##name##_##index, +#define FLAG_PTR_ARR(T) \ + static constexpr absl::Flag<T>* FlagPtrs_##T[] = { \ + REPLICATE(FLAG_PTR, T, T##_flag)}; +BENCHMARKED_TYPES(FLAG_PTR_ARR) + +#define BM_SingleGetFlag(T) \ + void BM_SingleGetFlag_##T(benchmark::State& state) { \ + for (auto _ : state) { \ + benchmark::DoNotOptimize(absl::GetFlag(SINGLE_FLAG(T))); \ + } \ + } \ + BENCHMARK(BM_SingleGetFlag_##T)->ThreadRange(1, 16); + +BENCHMARKED_TYPES(BM_SingleGetFlag) + +template <typename T> +struct Accumulator { + using type = T; +}; +template <> +struct Accumulator<String> { + using type = size_t; +}; +template <> +struct Accumulator<VectorOfStrings> { + using type = size_t; +}; +template <> +struct Accumulator<OptionalInt> { + using type = bool; +}; +template <> +struct Accumulator<OptionalString> { + using type = bool; +}; +template <> +struct Accumulator<UDT> { + using type = bool; +}; + +template <typename T> +void Accumulate(typename Accumulator<T>::type& a, const T& f) { + a += f; +} +void Accumulate(bool& a, bool f) { a = a || f; } +void Accumulate(size_t& a, const std::string& f) { a += f.size(); } +void Accumulate(size_t& a, const std::vector<std::string>& f) { a += f.size(); } +void Accumulate(bool& a, const OptionalInt& f) { a |= f.has_value(); } +void Accumulate(bool& a, const OptionalString& f) { a |= f.has_value(); } +void Accumulate(bool& a, const UDT& f) { + a |= reinterpret_cast<int64_t>(&f) & 0x1; +} + +#define BM_ManyGetFlag(T) \ + void BM_ManyGetFlag_##T(benchmark::State& state) { \ + Accumulator<T>::type res = {}; \ + while (state.KeepRunningBatch(kNumFlags)) { \ + for (auto* flag_ptr : FlagPtrs_##T) { \ + Accumulate(res, absl::GetFlag(*flag_ptr)); \ + } \ + } \ + benchmark::DoNotOptimize(res); \ + } \ + BENCHMARK(BM_ManyGetFlag_##T)->ThreadRange(1, 8); -BENCHMARKED_TYPES(BM_GetFlag) +BENCHMARKED_TYPES(BM_ManyGetFlag) void BM_ThreadedFindCommandLineFlag(benchmark::State& state) { char dummy[] = "dummy"; @@ -150,17 +231,18 @@ void BM_ThreadedFindCommandLineFlag(benchmark::State& state) { // is finalized. absl::ParseCommandLine(1, argv); - for (auto s : state) { - benchmark::DoNotOptimize( - absl::FindCommandLineFlag("bloat_flag_010101010101")); + while (state.KeepRunningBatch(kNumFlags)) { + for (auto* flag_ptr : FlagPtrs_bool) { + benchmark::DoNotOptimize(absl::FindCommandLineFlag(flag_ptr->Name())); + } } } BENCHMARK(BM_ThreadedFindCommandLineFlag)->ThreadRange(1, 16); } // namespace -#define InvokeGetFlag(T) \ - T AbslInvokeGetFlag##T() { return absl::GetFlag(FLAGS_##T##_flag); } \ +#define InvokeGetFlag(T) \ + T AbslInvokeGetFlag##T() { return absl::GetFlag(SINGLE_FLAG(T)); } \ int odr##T = (benchmark::DoNotOptimize(AbslInvokeGetFlag##T), 1); BENCHMARKED_TYPES(InvokeGetFlag) |